diff --git a/Database/new.npy b/Database/new.npy
new file mode 100644
index 0000000..1569ad6
Binary files /dev/null and b/Database/new.npy differ
diff --git a/Database/student.npy b/Database/student.npy
new file mode 100644
index 0000000..02d1549
Binary files /dev/null and b/Database/student.npy differ
diff --git a/accuracy.py b/accuracy.py
new file mode 100644
index 0000000..1a4e057
--- /dev/null
+++ b/accuracy.py
@@ -0,0 +1,132 @@
+import os
+import time
+
+import torch
+import cv2
+import numpy as np
+from backbones import iresnet50,iresnet18,iresnet100
+
+def load_image(img_path):
+    #img = cv2.imread(img_path)
+    img = cv2.imdecode(np.fromfile(img_path,dtype=np.uint8),cv2.IMREAD_COLOR)
+    img = img.transpose((2, 0, 1))
+    img = img[np.newaxis, :, :, :]
+    img = np.array(img, dtype=np.float32)
+    img -= 127.5
+    img /= 127.5
+    return img
+
+def findEuclideanDistance(source_representation, test_representation):
+    euclidean_distance = source_representation - test_representation
+    euclidean_distance = np.sum(np.multiply(euclidean_distance, euclidean_distance))
+    euclidean_distance = np.sqrt(euclidean_distance)
+    return euclidean_distance
+
+def l2_normalize(x):
+    return x / np.sqrt(np.sum(np.multiply(x, x)))
+
+def load_npy(path):
+    data = np.load(path,allow_pickle=True)
+    data = data.item()
+    return data
+
+def findmindistance(pred,threshold,k_v):
+    distance = 10
+    most_like = ""
+    for name in k_v.keys():
+        tmp = findEuclideanDistance(k_v[name],pred)
+        if distance > tmp:
+            distance = tmp
+            most_like = name
+    if distance < threshold:
+        return most_like
+    else:
+        return -1
+
+def findOne(img,model,k_v):
+    with torch.no_grad():
+        start_time = time.time()
+        pred = model(img)
+        end_time = time.time()
+        #print("predOne time: " + str(end_time - start_time))
+        pred = pred.numpy()
+        name = findmindistance(l2_normalize(pred),threshold=1.20,k_v=k_v)
+        if name != -1:
+            return name
+        else:
+            return "unknown"
+def findAll(imglist,model,k_v):
+    with torch.no_grad():
+        name_list = []
+        pred = model(imglist)
+        pred = pred.numpy()
+        for pr in pred:
+            name = findmindistance(l2_normalize(pr),threshold=1.20,k_v=k_v)
+            if name != -1:
+                name_list.append(name)
+            else:
+                name_list.append("unknown")
+        return name_list
+
+if __name__=='__main__':
+    model = iresnet100()
+    model.load_state_dict(torch.load("./model/backbone100.pth", map_location="cpu"))
+    model.eval()
+    pred_name = []
+    order_name = []
+    order_path = []
+    unknown = []
+    test_path = "D:\Download\out\cfp_test"
+    name_list = os.listdir(test_path)
+    for name in name_list:
+        img_list = os.listdir(os.path.join(test_path,name))
+        for img in img_list:
+            order_name.append(name)
+            order_path.append(os.path.join(os.path.join(test_path,name),img))
+    order_img = np.zeros((len(order_path), 3, 112, 112), dtype=np.float32)
+    for index,img_path in enumerate(order_path):
+        order_img[index] = load_image(img_path)
+    print(order_img.shape)
+    # for name in order_path:
+    #     print(name)
+
+    k_v = load_npy("cfp.npy")
+    start_time = time.time()
+    order_img = torch.from_numpy(order_img)
+
+    batch = 256
+    now = 0
+    number = len(order_img)
+    #number = 1400
+    for i in range(number):
+        unknown.append("unknown")
+
+    while now < number:
+        if now+batch < number:
+            name = findAll(order_img[now:now+batch],model,k_v)
+        else:
+            name = findAll(order_img[now:number], model, k_v)
+        now = now+batch
+        for na in name:
+            pred_name.append(na)
+        print("batch"+str(now))
+    end_time = time.time()
+    print("findAll time: " + str(end_time - start_time))
+    #print(len(pred_name))
+    right = 0
+    for i,name in enumerate(pred_name):
+        if pred_name[i] == order_name[i]:
+            right += 1
+    filed = 0
+    for i, name in enumerate(pred_name):
+        if pred_name[i] == unknown[i]:
+            filed += 1
+    error = 0
+    for i,name in enumerate(pred_name):
+        if pred_name[i] != order_name[i]:
+            error += 1
+            print(order_name[i]+" "+pred_name[i]+" "+order_path[i])
+    print("total:" + str(number))
+    print("right:" + str(right) + " rate:" + str(right / number))
+    print("filed:" + str(filed) + " rate:" + str(filed / number))
+    print("error:"+str(error-filed)+" rate:"+str((error-filed)/number))
diff --git a/accuracy_GPU.py b/accuracy_GPU.py
new file mode 100644
index 0000000..42c4f9a
--- /dev/null
+++ b/accuracy_GPU.py
@@ -0,0 +1,134 @@
+import os
+import time
+
+import torch
+import cv2
+import numpy as np
+from backbones import iresnet50,iresnet18,iresnet100
+
+def load_image(img_path):
+    #img = cv2.imread(img_path)
+    img = cv2.imdecode(np.fromfile(img_path,dtype=np.uint8),cv2.IMREAD_COLOR)
+    img = img.transpose((2, 0, 1))
+    img = img[np.newaxis, :, :, :]
+    img = np.array(img, dtype=np.float32)
+    img -= 127.5
+    img /= 127.5
+    return img
+
+def findEuclideanDistance(source_representation, test_representation):
+    euclidean_distance = source_representation - test_representation
+    euclidean_distance = np.sum(np.multiply(euclidean_distance, euclidean_distance))
+    euclidean_distance = np.sqrt(euclidean_distance)
+    return euclidean_distance
+
+def l2_normalize(x):
+    return x / np.sqrt(np.sum(np.multiply(x, x)))
+
+def load_npy(path):
+    data = np.load(path,allow_pickle=True)
+    data = data.item()
+    return data
+
+def findmindistance(pred,threshold,k_v):
+    distance = 10
+    most_like = ""
+    for name in k_v.keys():
+        tmp = findEuclideanDistance(k_v[name],pred)
+        if distance > tmp:
+            distance = tmp
+            most_like = name
+    if distance < threshold:
+        return most_like
+    else:
+        return -1
+
+def findOne(img,model,k_v):
+    with torch.no_grad():
+        start_time = time.time()
+        pred = model(img)
+        end_time = time.time()
+        #print("predOne time: " + str(end_time - start_time))
+        pred = pred.numpy()
+        name = findmindistance(l2_normalize(pred),threshold=1.20,k_v=k_v)
+        if name != -1:
+            return name
+        else:
+            return "unknown"
+def findAll(imglist,model,k_v):
+    with torch.no_grad():
+        name_list = []
+        imglist = imglist.to(torch.device("cuda"))
+        pred = model(imglist)
+        pred = pred.cpu().numpy()
+        for pr in pred:
+            name = findmindistance(l2_normalize(pr),threshold=1.20,k_v=k_v)
+            if name != -1:
+                name_list.append(name)
+            else:
+                name_list.append("unknown")
+        return name_list
+
+if __name__=='__main__':
+    model = iresnet100()
+    model.load_state_dict(torch.load("./model/backbone100.pth"))
+    model.to(torch.device("cuda"))
+    model.eval()
+    pred_name = []
+    order_name = []
+    order_path = []
+    unknown = []
+    test_path = "./retinaface_test"
+    name_list = os.listdir(test_path)
+    for name in name_list:
+        img_list = os.listdir(os.path.join(test_path,name))
+        for img in img_list:
+            order_name.append(name)
+            order_path.append(os.path.join(os.path.join(test_path,name),img))
+    order_img = np.zeros((len(order_path), 3, 112, 112), dtype=np.float32)
+    for index,img_path in enumerate(order_path):
+        order_img[index] = load_image(img_path)
+    print(order_img.shape)
+    # for name in order_path:
+    #     print(name)
+
+    k_v = load_npy("retinaface_lfw_myalign.npy")
+    start_time = time.time()
+    order_img = torch.from_numpy(order_img)
+
+    batch = 256
+    now = 0
+    number = len(order_img)
+    #number = 1400
+    for i in range(number):
+        unknown.append("unknown")
+
+    while now < number:
+        if now+batch < number:
+            name = findAll(order_img[now:now+batch],model,k_v)
+        else:
+            name = findAll(order_img[now:number], model, k_v)
+        now = now+batch
+        for na in name:
+            pred_name.append(na)
+        print("batch"+str(now))
+    end_time = time.time()
+    print("findAll time: " + str(end_time - start_time))
+    #print(len(pred_name))
+    right = 0
+    for i,name in enumerate(pred_name):
+        if pred_name[i] == order_name[i]:
+            right += 1
+    filed = 0
+    for i, name in enumerate(pred_name):
+        if pred_name[i] == unknown[i]:
+            filed += 1
+    error = 0
+    for i,name in enumerate(pred_name):
+        if pred_name[i] != order_name[i]:
+            error += 1
+            print(order_name[i]+" "+pred_name[i]+" "+order_path[i])
+    print("total:" + str(number))
+    print("right:" + str(right) + " rate:" + str(right / number))
+    print("filed:" + str(filed) + " rate:" + str(filed / number))
+    print("error:"+str(error-filed)+" rate:"+str((error-filed)/number))
\ No newline at end of file
diff --git a/anti.py b/anti.py
new file mode 100644
index 0000000..5ec59a2
--- /dev/null
+++ b/anti.py
@@ -0,0 +1,150 @@
+import os
+import cv2
+import numpy as np
+import argparse
+import warnings
+import time
+import torch
+import torch.nn.functional as F
+
+from src.generate_patches import CropImage
+from src.model_lib.MiniFASNet import MiniFASNetV1, MiniFASNetV2,MiniFASNetV1SE,MiniFASNetV2SE
+from src.data_io import transform as trans
+from src.utility import get_kernel, parse_model_name
+warnings.filterwarnings('ignore')
+
+MODEL_MAPPING = {
+    'MiniFASNetV1': MiniFASNetV1,
+    'MiniFASNetV2': MiniFASNetV2,
+    'MiniFASNetV1SE':MiniFASNetV1SE,
+    'MiniFASNetV2SE':MiniFASNetV2SE
+}
+
+class AntiSpoofPredict():
+    def __init__(self, cpu_or_cuda):
+        super(AntiSpoofPredict, self).__init__()
+        self.device = torch.device("cuda" if cpu_or_cuda == "cuda" else "cpu")
+
+    def predict(self, img, model):
+        test_transform = trans.Compose([
+            trans.ToTensor(),
+        ])
+        img = test_transform(img)
+        img = img.unsqueeze(0).to(self.device)
+        with torch.no_grad():
+            result = model.forward(img)
+            result = F.softmax(result).cpu().numpy()
+        return result
+
+def load_anti_model(model_dir,cpu_or_cuda):
+    model_list = []
+    for model_path in os.listdir(model_dir):
+        model_list.append(_load_model(os.path.join(model_dir, model_path), cpu_or_cuda))
+    return model_list
+
+def _load_model(model_path,cpu_or_cuda):
+    # define model
+    device = torch.device("cuda" if cpu_or_cuda == "cuda" else "cpu")
+    model_name = os.path.basename(model_path)
+    h_input, w_input, model_type, _ = parse_model_name(model_name)
+    kernel_size = get_kernel(h_input, w_input, )
+    model = MODEL_MAPPING[model_type](conv6_kernel=kernel_size).to(device)
+
+    # load model weight
+    state_dict = torch.load(model_path, map_location=device)
+    keys = iter(state_dict)
+    first_layer_name = keys.__next__()
+    if first_layer_name.find('module.') >= 0:
+        from collections import OrderedDict
+        new_state_dict = OrderedDict()
+        for key, value in state_dict.items():
+            name_key = key[7:]
+            new_state_dict[name_key] = value
+        model.load_state_dict(new_state_dict)
+    else:
+        model.load_state_dict(state_dict)
+    model.eval()
+    return model
+
+
+# 因为安卓端APK获取的视频流宽高比为3:4,为了与之一致，所以将宽高比限制为3:4
+def check_image(image):
+    height, width, channel = image.shape
+    if width/height != 3/4:
+        print("Image is not appropriate!!!\nHeight/Width should be 4/3.")
+        return False
+    else:
+        return True
+
+# 人脸活体检测
+def anti_spoofing(image_name, model_dir, cpu_or_cuda, bbox, model_list):
+    model_test = AntiSpoofPredict(cpu_or_cuda)
+    image_cropper = CropImage()
+    image = cv2.imdecode(np.fromfile(image_name, dtype=np.uint8), cv2.IMREAD_COLOR)
+    h, w = image.shape[:2]
+    factor = h / w
+    if (w > 1000):
+        image = cv2.resize(image, (600, int(600 * factor)))
+    # result = check_image(image)
+    # if result is False:
+    #     return
+    # image_bbox = model_test.get_bbox(image)
+    image_bbox = bbox
+    prediction = np.zeros((1, 3))
+    test_speed = 0
+    # sum the prediction from single model's result
+    for index, model_name in enumerate(os.listdir(model_dir)):
+        h_input, w_input, model_type, scale = parse_model_name(model_name)
+        param = {
+            "org_img": image,
+            "bbox": image_bbox,
+            "scale": scale,
+            "out_w": w_input,
+            "out_h": h_input,
+            "crop": True,
+        }
+        if scale is None:
+            param["crop"] = False
+        img = image_cropper.crop(**param)
+
+        start = time.time()
+        prediction += model_test.predict(img, model_list[index])
+        test_speed += time.time()-start
+
+    label = np.argmax(prediction)
+    # print(prediction)
+    # cv2.rectangle(
+    #     image,
+    #     (image_bbox[0], image_bbox[1]),
+    #     (image_bbox[0] + image_bbox[2], image_bbox[1] + image_bbox[3]),
+    #     (225,0,0), 2)
+    # cv2.imshow("out",image)
+    # cv2.waitKey(0)
+    value = prediction[0][1]/2
+    if value > 0.915:
+        return "real face", '{:.10f}'.format(value)
+    else:
+        return "fake face", '{:.10f}'.format(value)
+
+
+
+if __name__ == "__main__":
+    desc = "test"
+    parser = argparse.ArgumentParser(description=desc)
+    parser.add_argument(
+        "--device_id",
+        type=int,
+        default=0,
+        help="which gpu id, [0/1/2/3]")
+    parser.add_argument(
+        "--model_dir",
+        type=str,
+        default="./resources/anti_spoof_models",
+        help="model_lib used to test")
+    parser.add_argument(
+        "--image_name",
+        type=str,
+        default="000_0.bmp",
+        help="image used to test")
+    args = parser.parse_args()
+    # anti_spoofing(args.image_name, args.model_dir, args.device_id)
diff --git a/app.py b/app.py
new file mode 100644
index 0000000..f820df7
--- /dev/null
+++ b/app.py
@@ -0,0 +1,449 @@
+import time
+
+import faiss
+from flask import Flask, render_template, request, jsonify, send_from_directory
+from markupsafe import escape, escape_silent
+from werkzeug.utils import secure_filename
+
+from anti import anti_spoofing, load_anti_model
+from face_api import load_arcface_model, load_npy, findOne, load_image, face_verification, findAll, add_one_to_database, \
+    get_claster_tmp_file_embedding, cluster, detect_video
+from gender_age import set_gender_conf, gender_age, load_gender_model
+from retinaface_detect import load_retinaface_model, detect_one, set_retinaface_conf
+from werkzeug.exceptions import RequestEntityTooLarge
+import zipfile
+import os
+import shutil
+import re
+import numpy as np
+import torch
+
+ALLOWED_IMG = set(['png', 'jpg', 'jpeg', 'bmp', 'PNG', 'JPG', 'JPEG'])
+# 限制上传的图片最大为10M
+ALLOWED_IMG_SIZE = 10 * 1024 * 1024
+ALLOWED_FILE = set(['zip'])
+ALLOWED_VIDEO = set(['mp4'])
+app = Flask(__name__)
+
+# 限制上传的文件最大为100M
+app.config['MAX_CONTENT_LENGTH'] = 100 * 1024 * 1024
+# 使用jsonify，避免中文乱码
+app.config['JSON_AS_ASCII'] = False
+
+# 设置使用CPU或者GPU（传入cuda）
+cpu_or_cuda = "cuda" if torch.cuda.is_available() else "cpu"
+# 加载人脸识别模型
+arcface_model = load_arcface_model("./model/backbone100.pth", cpu_or_cuda=cpu_or_cuda)
+# 加载人脸检测模型
+retinaface_args = set_retinaface_conf(cpu_or_cuda=cpu_or_cuda)
+retinaface_model = load_retinaface_model(retinaface_args)
+# 加载性别年龄识别模型
+gender_args = set_gender_conf()
+gender_model = load_gender_model(gender_args, 'fc1')
+anti_spoofing_model_path = "model/anti_spoof_models"
+anti_model = load_anti_model(anti_spoofing_model_path, cpu_or_cuda)
+
+
+# 读取人脸库
+
+
+@app.route('/')
+def index():
+    return "model"
+
+
+@app.route('/hello')
+@app.route('/hello/<name>')
+def hello(name=None):
+    return render_template('hello.html', name=name)
+
+
+@app.route('/user', methods=['GET'])
+def show_user_name():
+    return request.args.get('username', '')
+
+
+# 创建返回的json数据
+# 函数参数用是否=None判断，函数中定义的data,result用true，false判断
+def create_response(status, name=None, distance=None, verification=None, gender=None, age=None, num=None, anti=None,
+                    score=None, box_and_point=None, addfile_names=None,fail_names=None,database_name=None,msg=None,
+                    delete_names=None,not_exist_names=None):
+    # res为总的json结构体
+    res = {}
+    res['status'] = status
+
+    data = {}
+    try:
+        data["box_and_point"] = box_and_point.tolist()
+    except AttributeError:
+        pass
+    if anti != None and score != None:
+        liveness = {}
+        liveness["spoofing"] = anti
+        liveness['score'] = score
+        data['liveness'] = liveness
+    if distance!=None:
+        data['distance'] = float(distance)
+    if verification!=None:
+        data['verification'] = verification
+    if num!=None:
+        data['number'] = num
+    if gender!=None:
+        data['gender'] = gender
+    if age!=None:
+        data['age'] = age
+    if name!=None:
+        data['name'] = name
+    if data:
+        res['data'] = data
+
+    # 数据库增删接口返回数据
+    result = {}
+    if msg!=None:
+        res['msg'] = msg
+    if database_name!=None:
+        result['database_name'] = database_name
+    # 增加人脸
+    if addfile_names!=None or fail_names!=None:
+        result['success_names'] = addfile_names
+        result['fail_names'] = fail_names
+    # 删除人脸
+    if delete_names!=None or not_exist_names!=None:
+        result['delete_names'] = delete_names
+        result['not_exist_names'] = not_exist_names
+    if result:
+        res['result'] = result
+
+    return jsonify(res)
+
+
+# 创建cluster接口返回的json数据
+def create_cluster_response(status, all_cluster):
+    res = {}
+    data = {}
+    for index, cluster in enumerate(all_cluster):
+        data['cluster' + str(index)] = cluster
+    res['data'] = data
+    res['status'] = status
+    return res
+
+
+# 检查上传文件格式
+def check_file_format(file_name, format):
+    if '.' in file_name:
+        file_format = file_name.rsplit('.')[1]
+        if file_format in format:
+            return True
+    return False
+
+
+# 检查img大小，大于10M抛出异常
+def check_img_size(img_path):
+    fsize = os.path.getsize(img_path)
+    if fsize > ALLOWED_IMG_SIZE:
+        raise RequestEntityTooLarge
+
+
+# 解压zip文件存到某路径：
+def unzip(zip_src, dst_dir):
+    f = zipfile.is_zipfile(zip_src)
+    if f:
+        fz = zipfile.ZipFile(zip_src, 'r')
+        for file in fz.namelist():
+            fz.extract(file, dst_dir)
+        return True
+    else:
+        return False
+
+
+# 解压文件
+def un_zip(file_path, output_path):
+    zip_file = zipfile.ZipFile(file_path)
+    if os.path.isdir(output_path):
+        pass
+    else:
+        os.mkdir(output_path)
+    zip_file.extractall(output_path)
+    # for names in zip_file.namelist():
+    #     zip_file.extract(names,output_path)
+    zip_file.close()
+
+
+# 人脸识别、性别年龄识别
+@app.route('/recognition', methods=['POST'])
+def recognition():
+    try:
+        f = request.files['file_name']
+        if f and check_file_format(f.filename, ALLOWED_IMG):
+            img_path = './img/recognition/' + secure_filename(f.filename)
+            f.save(img_path)
+            check_img_size(img_path)
+            # img3 = load_image('./file/'+secure_filename(f.filename))
+            # img3 = torch.from_numpy(img3)
+            tic = time.time()
+            img3, box_and_point = detect_one(img_path, retinaface_model, retinaface_args)
+            print('detect time: {:.4f}'.format(time.time() - tic))
+            if len(img3) == 0:
+                return create_response('no face')
+            elif len(img3) > 1:
+                namelist = findAll(img3, arcface_model, index, database_name_list, cpu_or_cuda)
+                gender_list, age_list = [], []
+                # gender_list, age_list = gender_age(img3, gender_model)
+                res = create_response('success', namelist, gender=gender_list, age=age_list,
+                                      box_and_point=box_and_point)
+            else:
+                b = box_and_point[0]
+                w = b[2] - b[0]
+                h = b[3] - b[1]
+                b[2] = w
+                b[3] = h
+                label, value = anti_spoofing(img_path, anti_spoofing_model_path, cpu_or_cuda, np.array(b[:4], int),
+                                             anti_model)
+                # print(index,database_name_list)
+                name, distance = findOne(img3, arcface_model, index, database_name_list, cpu_or_cuda)
+                gender_list, age_list = [], []
+                # gender_list, age_list = gender_age(img3, gender_model)
+                res = create_response('success', name, gender=gender_list, age=age_list, distance=distance,
+                                      anti=label, score=value, box_and_point=box_and_point)
+            return res
+        else:
+            return create_response('png jpg jpeg bmp are allowed')
+    except RequestEntityTooLarge:
+        return create_response('image size should be less than 10M')
+
+
+# 两张图片比对
+@app.route('/compare', methods=['POST'])
+def compare_file():
+    try:
+        file1 = request.files['file1_name']
+        file2 = request.files['file2_name']
+        if file1 and check_file_format(file1.filename, ALLOWED_IMG) and file2 and check_file_format(file2.filename,
+                                                                                                    ALLOWED_IMG):
+            img1_path = './img/compare/' + secure_filename(file1.filename)
+            img2_path = './img/compare/' + secure_filename(file2.filename)
+            file1.save(img1_path)
+            file2.save(img2_path)
+            check_img_size(img1_path)
+            check_img_size(img2_path)
+            img1, box_and_point1 = detect_one(img1_path, retinaface_model,
+                                              retinaface_args)
+            img2, box_and_point2 = detect_one(img2_path, retinaface_model, retinaface_args)
+            if len(img1) == 1 and len(img2) == 1:
+                result,distance = face_verification(img1, img2, arcface_model, cpu_or_cuda)
+                print(result,distance)
+                return create_response('success', verification=result,distance=distance)
+            else:
+                return create_response('image contains no face or more than 1 face')
+        else:
+            return create_response('png jpg jpeg bmp are allowed')
+    except RequestEntityTooLarge:
+        return create_response('image size should be less than 10M')
+
+
+# 数据库增加人脸，可实现向“现有/新建”数据库增加“单张/多张”人脸
+# 增和改
+@app.route('/databaseAdd', methods=['POST'])
+def DB_add_face():
+    try:
+        # 上传人脸图片（>=1）
+        # key都为file_list，value为不同的值可实现批量上传图片
+        upload_files = request.files.getlist("file_list")
+        # '',[],{},0都可以视为False
+        if not upload_files:
+            msg = "上传文件为空"
+            return create_response(0,msg=msg)
+        database_name = request.form.get("database_name")
+        database_path = "./Database/" + database_name + ".npy"
+        if not os.path.exists(database_path):
+            msg = "数据库不存在"
+            return create_response(0,msg=msg)
+        # 数据库中已存在的人名
+        names = load_npy(database_path).keys()
+        # print(names)
+
+        # 这是服务器上用于暂存上传图片的文件夹，每次上传前重建，使用后删除
+        # 后面可根据需要改为定期删除
+        file_temp_path = './img/uploadNew/'
+        if not os.path.exists(file_temp_path):
+            os.makedirs(file_temp_path)
+
+        # 正则表达式用于提取文件名中的中文，用于.npy中的keys
+        r = re.compile('[\u4e00-\u9fa5]+')
+        # 分别存取添加成功或失败的名字
+        success_names = []
+        fail_names = {}
+        # 添加失败的两种情况：格式错误或已经存在
+        format_wrong = []
+        alreadyExist = []
+        # 分别处理每一张图片,先判断格式对不对，再判断是否存在
+        for file in upload_files:
+            filename = file.filename
+            name = r.findall(filename)[0]
+            if file and check_file_format(filename, ALLOWED_IMG):
+                if name in names:
+                    alreadyExist.append(name)
+                    continue
+                save_path = file_temp_path + filename
+                file.save(save_path)
+                check_img_size(save_path)
+                img_file, box_and_point = detect_one(save_path, retinaface_model, retinaface_args)
+                add_one_to_database(img=img_file, model=arcface_model, name=name, database_path=database_path,
+                                    cpu_or_cuda=cpu_or_cuda)
+                success_names.append(name)
+            else:
+                format_wrong.append(name)
+                continue
+        shutil.rmtree(file_temp_path)
+        # 如果有错误情况
+        if format_wrong or alreadyExist:
+            status = 0
+        else:
+            status = 1
+        fail_names['formatWrong'] = format_wrong
+        fail_names['alreadyExist'] = alreadyExist
+
+        return create_response(status=status,addfile_names=success_names,fail_names=fail_names,database_name=database_name,msg="新增人脸操作执行完成")
+    except RequestEntityTooLarge:
+        return create_response(0,msg='image size should be less than 10M')
+
+
+# 数据库删除人脸，可实现在现有数据库中删除’单/多‘张人脸
+@app.route('/databaseDelete', methods=['POST'])
+def DB_delete_face():
+    try:
+        delete_names = request.form.getlist("delete_names")
+        database_name = request.form.get("database_name")
+        database_path = "./Database/" + database_name + ".npy"
+        if not os.path.exists(database_path):
+            msg = "数据库不存在"
+            return create_response(0,msg=msg)
+        if not delete_names:
+            msg = "delete_names参数为空"
+            return create_response(0,msg=msg)
+        k_v = load_npy(database_path)
+        print(k_v.keys())
+        success_list = []
+        fail_list = []
+        for name in delete_names:
+            if name in k_v.keys():
+                del k_v[name]
+                success_list.append(name)
+            else:
+                fail_list.append(name)
+                continue
+        np.save(database_path, k_v)
+        status = 1
+        if fail_list:
+            status = 0
+        return create_response(status=status,delete_names=success_list,not_exist_names=fail_list,database_name=database_name,
+                               msg="删除人脸操作完成")
+    except RequestEntityTooLarge:
+        return create_response(0,'image size should be less than 10M')
+
+
+# 以图搜图接口：
+# 上传图片压缩包建图片库
+@app.route('/uploadZip', methods=['POST'])
+def upload_Zip():
+    try:
+        zip = request.files['zip_name']
+        dst_dir = './img/search/'
+        if unzip(zip, dst_dir):
+            return create_response('upload zip success')
+        else:
+            return create_response('upload zip file please')
+    except RequestEntityTooLarge:
+        return create_response('image size should be less than 10M')
+
+
+# 以图搜图
+@app.route('/imgSearchImg', methods=['POST'])
+def img_search_img():
+    searchfile = './img/search/face'
+    try:
+        file = request.files['img_name']
+        if file and check_file_format(file.filename, ALLOWED_IMG):
+            img_path = './img/search/' + secure_filename(file.filename)
+            file.save(img_path)
+            check_img_size(img_path)
+            img, box_and_point = detect_one(img_path, retinaface_model,
+                                            retinaface_args)
+            if len(img) == 1:
+                Onename = []
+                num = 0
+                for filenames in os.listdir(searchfile):
+                    imgpath = os.path.join(searchfile, filenames)
+                    imgdata, box_and_point = detect_one(imgpath, retinaface_model, retinaface_args)
+                    result = face_verification(img, imgdata, arcface_model, cpu_or_cuda)
+                    isOne, distance = result.split(' ', -1)[0], result.split(' ', -1)[1]
+                    if isOne == 'same':
+                        Onename.append(filenames)
+                        num += 1
+                return create_response('success', name=Onename, num=num)
+            else:
+                return create_response('image contains no face or more than 1 face')
+        else:
+            return create_response('png jpg jpeg bmp are allowed')
+    except RequestEntityTooLarge:
+        return create_response('image size should be less than 10M')
+
+
+# 人脸聚类接口
+@app.route('/cluster', methods=['POST'])
+def zip_cluster():
+    try:
+        f = request.files['file_name']
+        if f and check_file_format(f.filename, ALLOWED_FILE):
+            zip_name = secure_filename(f.filename)
+            f.save('./img/cluster_tmp_file/' + zip_name)
+            un_zip('./img/cluster_tmp_file/' + zip_name, './img/cluster_tmp_file/')
+            emb_list, name_list = get_claster_tmp_file_embedding("./img/cluster_tmp_file/" + zip_name.rsplit('.')[0],
+                                                                 retinaface_model,
+                                                                 retinaface_args, arcface_model, cpu_or_cuda)
+            return create_cluster_response("success", cluster(emb_list, name_list))
+        else:
+            return create_response('zip are allowed')
+    except RequestEntityTooLarge:
+        return create_response('file size should be less than 100M')
+
+
+# 视频识别接口
+@app.route('/videorecognition', methods=['POST'])
+def video_recognition():
+    try:
+        f = request.files['file_name']
+        if f and check_file_format(f.filename, ALLOWED_VIDEO):
+            video_name = secure_filename(f.filename)
+            f.save('./video/' + video_name)
+            detect_video('./video/' + video_name, './videoout/' + video_name, retinaface_model, arcface_model, k_v,
+                         retinaface_args)
+            return create_response("success")
+        else:
+            return create_response('mp4 are allowed')
+    except RequestEntityTooLarge:
+        return create_response('file size should be less than 100M')
+
+
+@app.route('/download/<string:filename>', methods=['GET'])
+def download(filename):
+    if os.path.isfile(os.path.join('./videoout/', filename)):
+        return send_from_directory('./videoout/', filename, as_attachment=True)
+    else:
+        return create_response("Download failed")
+
+
+if __name__ == '__main__':
+    k_v = load_npy("./Database/student.npy")
+    database_name_list = list(k_v.keys())
+    vector_list = np.array(list(k_v.values()))
+    print(vector_list.shape)
+    #print(database_name_list)
+    nlist = 50
+    quantizer = faiss.IndexFlatL2(512)  # the other index
+    index = faiss.IndexIVFFlat(quantizer, 512, nlist, faiss.METRIC_L2)
+    index.train(vector_list)
+    # index = faiss.IndexFlatL2(512)
+    index.add(vector_list)
+    index.nprobe = 50
+    app.run(host="0.0.0.0", port=5000)
diff --git a/backbones/__init__.py b/backbones/__init__.py
new file mode 100644
index 0000000..4a0bcbf
--- /dev/null
+++ b/backbones/__init__.py
@@ -0,0 +1 @@
+from .iresnet import iresnet18, iresnet34, iresnet50, iresnet100, iresnet200
diff --git a/backbones/__pycache__/__init__.cpython-38.pyc b/backbones/__pycache__/__init__.cpython-38.pyc
new file mode 100644
index 0000000..f50ba25
Binary files /dev/null and b/backbones/__pycache__/__init__.cpython-38.pyc differ
diff --git a/backbones/__pycache__/iresnet.cpython-38.pyc b/backbones/__pycache__/iresnet.cpython-38.pyc
new file mode 100644
index 0000000..42ab5cb
Binary files /dev/null and b/backbones/__pycache__/iresnet.cpython-38.pyc differ
diff --git a/backbones/iresnet.py b/backbones/iresnet.py
new file mode 100644
index 0000000..c6d3b9c
--- /dev/null
+++ b/backbones/iresnet.py
@@ -0,0 +1,187 @@
+import torch
+from torch import nn
+
+__all__ = ['iresnet18', 'iresnet34', 'iresnet50', 'iresnet100', 'iresnet200']
+
+
+def conv3x3(in_planes, out_planes, stride=1, groups=1, dilation=1):
+    """3x3 convolution with padding"""
+    return nn.Conv2d(in_planes,
+                     out_planes,
+                     kernel_size=3,
+                     stride=stride,
+                     padding=dilation,
+                     groups=groups,
+                     bias=False,
+                     dilation=dilation)
+
+
+def conv1x1(in_planes, out_planes, stride=1):
+    """1x1 convolution"""
+    return nn.Conv2d(in_planes,
+                     out_planes,
+                     kernel_size=1,
+                     stride=stride,
+                     bias=False)
+
+
+class IBasicBlock(nn.Module):
+    expansion = 1
+    def __init__(self, inplanes, planes, stride=1, downsample=None,
+                 groups=1, base_width=64, dilation=1):
+        super(IBasicBlock, self).__init__()
+        if groups != 1 or base_width != 64:
+            raise ValueError('BasicBlock only supports groups=1 and base_width=64')
+        if dilation > 1:
+            raise NotImplementedError("Dilation > 1 not supported in BasicBlock")
+        self.bn1 = nn.BatchNorm2d(inplanes, eps=1e-05,)
+        self.conv1 = conv3x3(inplanes, planes)
+        self.bn2 = nn.BatchNorm2d(planes, eps=1e-05,)
+        self.prelu = nn.PReLU(planes)
+        self.conv2 = conv3x3(planes, planes, stride)
+        self.bn3 = nn.BatchNorm2d(planes, eps=1e-05,)
+        self.downsample = downsample
+        self.stride = stride
+
+    def forward(self, x):
+        identity = x
+        out = self.bn1(x)
+        out = self.conv1(out)
+        out = self.bn2(out)
+        out = self.prelu(out)
+        out = self.conv2(out)
+        out = self.bn3(out)
+        if self.downsample is not None:
+            identity = self.downsample(x)
+        out += identity
+        return out
+
+
+class IResNet(nn.Module):
+    fc_scale = 7 * 7
+    def __init__(self,
+                 block, layers, dropout=0, num_features=512, zero_init_residual=False,
+                 groups=1, width_per_group=64, replace_stride_with_dilation=None, fp16=False):
+        super(IResNet, self).__init__()
+        self.fp16 = fp16
+        self.inplanes = 64
+        self.dilation = 1
+        if replace_stride_with_dilation is None:
+            replace_stride_with_dilation = [False, False, False]
+        if len(replace_stride_with_dilation) != 3:
+            raise ValueError("replace_stride_with_dilation should be None "
+                             "or a 3-element tuple, got {}".format(replace_stride_with_dilation))
+        self.groups = groups
+        self.base_width = width_per_group
+        self.conv1 = nn.Conv2d(3, self.inplanes, kernel_size=3, stride=1, padding=1, bias=False)
+        self.bn1 = nn.BatchNorm2d(self.inplanes, eps=1e-05)
+        self.prelu = nn.PReLU(self.inplanes)
+        self.layer1 = self._make_layer(block, 64, layers[0], stride=2)
+        self.layer2 = self._make_layer(block,
+                                       128,
+                                       layers[1],
+                                       stride=2,
+                                       dilate=replace_stride_with_dilation[0])
+        self.layer3 = self._make_layer(block,
+                                       256,
+                                       layers[2],
+                                       stride=2,
+                                       dilate=replace_stride_with_dilation[1])
+        self.layer4 = self._make_layer(block,
+                                       512,
+                                       layers[3],
+                                       stride=2,
+                                       dilate=replace_stride_with_dilation[2])
+        self.bn2 = nn.BatchNorm2d(512 * block.expansion, eps=1e-05,)
+        self.dropout = nn.Dropout(p=dropout, inplace=True)
+        self.fc = nn.Linear(512 * block.expansion * self.fc_scale, num_features)
+        self.features = nn.BatchNorm1d(num_features, eps=1e-05)
+        nn.init.constant_(self.features.weight, 1.0)
+        self.features.weight.requires_grad = False
+
+        for m in self.modules():
+            if isinstance(m, nn.Conv2d):
+                nn.init.normal_(m.weight, 0, 0.1)
+            elif isinstance(m, (nn.BatchNorm2d, nn.GroupNorm)):
+                nn.init.constant_(m.weight, 1)
+                nn.init.constant_(m.bias, 0)
+
+        if zero_init_residual:
+            for m in self.modules():
+                if isinstance(m, IBasicBlock):
+                    nn.init.constant_(m.bn2.weight, 0)
+
+    def _make_layer(self, block, planes, blocks, stride=1, dilate=False):
+        downsample = None
+        previous_dilation = self.dilation
+        if dilate:
+            self.dilation *= stride
+            stride = 1
+        if stride != 1 or self.inplanes != planes * block.expansion:
+            downsample = nn.Sequential(
+                conv1x1(self.inplanes, planes * block.expansion, stride),
+                nn.BatchNorm2d(planes * block.expansion, eps=1e-05, ),
+            )
+        layers = []
+        layers.append(
+            block(self.inplanes, planes, stride, downsample, self.groups,
+                  self.base_width, previous_dilation))
+        self.inplanes = planes * block.expansion
+        for _ in range(1, blocks):
+            layers.append(
+                block(self.inplanes,
+                      planes,
+                      groups=self.groups,
+                      base_width=self.base_width,
+                      dilation=self.dilation))
+
+        return nn.Sequential(*layers)
+
+    def forward(self, x):
+        with torch.cuda.amp.autocast(self.fp16):
+            x = self.conv1(x)
+            x = self.bn1(x)
+            x = self.prelu(x)
+            x = self.layer1(x)
+            x = self.layer2(x)
+            x = self.layer3(x)
+            x = self.layer4(x)
+            x = self.bn2(x)
+            x = torch.flatten(x, 1)
+            x = self.dropout(x)
+        x = self.fc(x.float() if self.fp16 else x)
+        x = self.features(x)
+        return x
+
+
+def _iresnet(arch, block, layers, pretrained, progress, **kwargs):
+    model = IResNet(block, layers, **kwargs)
+    if pretrained:
+        raise ValueError()
+    return model
+
+
+def iresnet18(pretrained=False, progress=True, **kwargs):
+    return _iresnet('iresnet18', IBasicBlock, [2, 2, 2, 2], pretrained,
+                    progress, **kwargs)
+
+
+def iresnet34(pretrained=False, progress=True, **kwargs):
+    return _iresnet('iresnet34', IBasicBlock, [3, 4, 6, 3], pretrained,
+                    progress, **kwargs)
+
+
+def iresnet50(pretrained=False, progress=True, **kwargs):
+    return _iresnet('iresnet50', IBasicBlock, [3, 4, 14, 3], pretrained,
+                    progress, **kwargs)
+
+
+def iresnet100(pretrained=False, progress=True, **kwargs):
+    return _iresnet('iresnet100', IBasicBlock, [3, 13, 30, 3], pretrained,
+                    progress, **kwargs)
+
+
+def iresnet200(pretrained=False, progress=True, **kwargs):
+    return _iresnet('iresnet200', IBasicBlock, [6, 26, 60, 6], pretrained,
+                    progress, **kwargs)
+
diff --git a/centerface.py b/centerface.py
new file mode 100644
index 0000000..0b03dbd
--- /dev/null
+++ b/centerface.py
@@ -0,0 +1,135 @@
+import time
+
+import numpy as np
+import cv2
+import datetime
+
+
+class CenterFace(object):
+    def __init__(self, landmarks=True):
+        self.landmarks = landmarks
+        if self.landmarks:
+            self.net = cv2.dnn.readNetFromONNX('./model/onnx/centerface.onnx')
+        else:
+            self.net = cv2.dnn.readNetFromONNX('./model/onnx/cface.1k.onnx')
+        self.img_h_new, self.img_w_new, self.scale_h, self.scale_w = 0, 0, 0, 0
+
+    def __call__(self, img, height, width, threshold=0.5):
+        self.img_h_new, self.img_w_new, self.scale_h, self.scale_w = self.transform(height, width)
+        return self.inference_opencv(img, threshold)
+
+    def inference_opencv(self, img, threshold):
+        blob = cv2.dnn.blobFromImage(img, scalefactor=1.0, size=(self.img_w_new, self.img_h_new), mean=(0, 0, 0), swapRB=True, crop=False)
+        self.net.setInput(blob)
+        begin = datetime.datetime.now()
+        start_time = time.time()
+
+
+        if self.landmarks:
+            heatmap, scale, offset, lms = self.net.forward(["537", "538", "539", '540'])
+        else:
+            heatmap, scale, offset = self.net.forward(["535", "536", "537"])
+        end = datetime.datetime.now()
+        end_time = time.time()
+        # print("cpuOne time: " + str(end_time - start_time))
+        # print("cpu times = ", end - begin)
+        return self.postprocess(heatmap, lms, offset, scale, threshold)
+
+    def transform(self, h, w):
+        img_h_new, img_w_new = int(np.ceil(h / 32) * 32), int(np.ceil(w / 32) * 32)
+        scale_h, scale_w = img_h_new / h, img_w_new / w
+        return img_h_new, img_w_new, scale_h, scale_w
+
+    def postprocess(self, heatmap, lms, offset, scale, threshold):
+        if self.landmarks:
+            dets, lms = self.decode(heatmap, scale, offset, lms, (self.img_h_new, self.img_w_new), threshold=threshold)
+        else:
+            dets = self.decode(heatmap, scale, offset, None, (self.img_h_new, self.img_w_new), threshold=threshold)
+        if len(dets) > 0:
+            dets[:, 0:4:2], dets[:, 1:4:2] = dets[:, 0:4:2] / self.scale_w, dets[:, 1:4:2] / self.scale_h
+            if self.landmarks:
+                lms[:, 0:10:2], lms[:, 1:10:2] = lms[:, 0:10:2] / self.scale_w, lms[:, 1:10:2] / self.scale_h
+        else:
+            dets = np.empty(shape=[0, 5], dtype=np.float32)
+            if self.landmarks:
+                lms = np.empty(shape=[0, 10], dtype=np.float32)
+        if self.landmarks:
+            return dets, lms
+        else:
+            return dets
+
+    def decode(self, heatmap, scale, offset, landmark, size, threshold=0.1):
+        heatmap = np.squeeze(heatmap)
+        scale0, scale1 = scale[0, 0, :, :], scale[0, 1, :, :]
+        offset0, offset1 = offset[0, 0, :, :], offset[0, 1, :, :]
+        c0, c1 = np.where(heatmap > threshold)
+        if self.landmarks:
+            boxes, lms = [], []
+        else:
+            boxes = []
+        if len(c0) > 0:
+            for i in range(len(c0)):
+                s0, s1 = np.exp(scale0[c0[i], c1[i]]) * 4, np.exp(scale1[c0[i], c1[i]]) * 4
+                o0, o1 = offset0[c0[i], c1[i]], offset1[c0[i], c1[i]]
+                s = heatmap[c0[i], c1[i]]
+                x1, y1 = max(0, (c1[i] + o1 + 0.5) * 4 - s1 / 2), max(0, (c0[i] + o0 + 0.5) * 4 - s0 / 2)
+                x1, y1 = min(x1, size[1]), min(y1, size[0])
+                boxes.append([x1, y1, min(x1 + s1, size[1]), min(y1 + s0, size[0]), s])
+                if self.landmarks:
+                    lm = []
+                    for j in range(5):
+                        lm.append(landmark[0, j * 2 + 1, c0[i], c1[i]] * s1 + x1)
+                        lm.append(landmark[0, j * 2, c0[i], c1[i]] * s0 + y1)
+                    lms.append(lm)
+            boxes = np.asarray(boxes, dtype=np.float32)
+            keep = self.nms(boxes[:, :4], boxes[:, 4], 0.3)
+            boxes = boxes[keep, :]
+            if self.landmarks:
+                lms = np.asarray(lms, dtype=np.float32)
+                lms = lms[keep, :]
+        if self.landmarks:
+            return boxes, lms
+        else:
+            return boxes
+
+    def nms(self, boxes, scores, nms_thresh):
+        x1 = boxes[:, 0]
+        y1 = boxes[:, 1]
+        x2 = boxes[:, 2]
+        y2 = boxes[:, 3]
+        areas = (x2 - x1 + 1) * (y2 - y1 + 1)
+        order = np.argsort(scores)[::-1]
+        num_detections = boxes.shape[0]
+        suppressed = np.zeros((num_detections,), dtype=np.bool)
+
+        keep = []
+        for _i in range(num_detections):
+            i = order[_i]
+            if suppressed[i]:
+                continue
+            keep.append(i)
+
+            ix1 = x1[i]
+            iy1 = y1[i]
+            ix2 = x2[i]
+            iy2 = y2[i]
+            iarea = areas[i]
+
+            for _j in range(_i + 1, num_detections):
+                j = order[_j]
+                if suppressed[j]:
+                    continue
+
+                xx1 = max(ix1, x1[j])
+                yy1 = max(iy1, y1[j])
+                xx2 = min(ix2, x2[j])
+                yy2 = min(iy2, y2[j])
+                w = max(0, xx2 - xx1 + 1)
+                h = max(0, yy2 - yy1 + 1)
+
+                inter = w * h
+                ovr = inter / (iarea + areas[j] - inter)
+                if ovr >= nms_thresh:
+                    suppressed[j] = True
+
+        return keep
diff --git a/config.py b/config.py
new file mode 100644
index 0000000..5fc70e7
--- /dev/null
+++ b/config.py
@@ -0,0 +1,67 @@
+from easydict import EasyDict as edict
+
+config = edict()
+config.dataset = "ms1m-retinaface-t2"
+config.embedding_size = 512
+config.sample_rate = 1
+config.fp16 = False
+config.momentum = 0.9
+config.weight_decay = 5e-4
+config.batch_size = 64
+config.lr = 0.1  # batch size is 512
+config.output = "ms1mv3_arcface_r50"
+
+if config.dataset == "emore":
+    config.rec = "/train_tmp/faces_emore"
+    config.num_classes = 85742
+    config.num_image = 5822653
+    config.num_epoch = 16
+    config.warmup_epoch = -1
+    config.val_targets = ["lfw", ]
+
+    def lr_step_func(epoch):
+        return ((epoch + 1) / (4 + 1)) ** 2 if epoch < -1 else 0.1 ** len(
+            [m for m in [8, 14] if m - 1 <= epoch])
+    config.lr_func = lr_step_func
+
+elif config.dataset == "ms1m-retinaface-t2":
+    config.rec = "/train_tmp/ms1m-retinaface-t2"
+    config.num_classes = 91180
+    config.num_epoch = 25
+    config.warmup_epoch = -1
+    config.val_targets = ["lfw", "cfp_fp", "agedb_30"]
+
+    def lr_step_func(epoch):
+        return ((epoch + 1) / (4 + 1)) ** 2 if epoch < -1 else 0.1 ** len(
+            [m for m in [11, 17, 22] if m - 1 <= epoch])
+    config.lr_func = lr_step_func
+
+elif config.dataset == "glint360k":
+    # make training faster
+    # our RAM is 256G
+    # mount -t tmpfs -o size=140G  tmpfs /train_tmp
+    config.rec = "/train_tmp/glint360k"
+    config.num_classes = 360232
+    config.num_image = 17091657
+    config.num_epoch = 20
+    config.warmup_epoch = -1
+    config.val_targets = ["lfw", "cfp_fp", "agedb_30"]
+
+    def lr_step_func(epoch):
+        return ((epoch + 1) / (4 + 1)) ** 2 if epoch < config.warmup_epoch else 0.1 ** len(
+            [m for m in [8, 12, 15, 18] if m - 1 <= epoch])
+    config.lr_func = lr_step_func
+
+elif config.dataset == "webface":
+    config.rec = "/train_tmp/faces_webface_112x112"
+    config.num_classes = 10572
+    config.num_image = "forget"
+    config.num_epoch = 34
+    config.warmup_epoch = -1
+    config.val_targets = ["lfw", "cfp_fp", "agedb_30"]
+
+    def lr_step_func(epoch):
+        return ((epoch + 1) / (4 + 1)) ** 2 if epoch < config.warmup_epoch else 0.1 ** len(
+            [m for m in [20, 28, 32] if m - 1 <= epoch])
+    config.lr_func = lr_step_func
+
diff --git a/create_database.py b/create_database.py
new file mode 100644
index 0000000..0a4104f
--- /dev/null
+++ b/create_database.py
@@ -0,0 +1,168 @@
+import os
+import time
+import re
+import torch
+import cv2
+import numpy as np
+from backbones import iresnet50,iresnet18,iresnet100
+
+def load_image(img_path):
+    #img = cv2.imread(img_path)
+    img = cv2.imdecode(np.fromfile(img_path,dtype=np.uint8),cv2.IMREAD_COLOR)
+    img = img.transpose((2, 0, 1))
+    img = img[np.newaxis, :, :, :]
+    img = np.array(img, dtype=np.float32)
+    img -= 127.5
+    img /= 127.5
+    return img
+
+def findEuclideanDistance(source_representation, test_representation):
+    euclidean_distance = source_representation - test_representation
+    euclidean_distance = np.sum(np.multiply(euclidean_distance, euclidean_distance))
+    euclidean_distance = np.sqrt(euclidean_distance)
+    return euclidean_distance
+
+def findCosineDistance(source_representation, test_representation):
+    a = np.matmul(np.transpose(source_representation), test_representation)
+    b = np.sum(np.multiply(source_representation, source_representation))
+    c = np.sum(np.multiply(test_representation, test_representation))
+    return 1 - (a / (np.sqrt(b) * np.sqrt(c)))
+
+def l2_normalize(x):
+    return x / np.sqrt(np.sum(np.multiply(x, x)))
+
+def cosin_metric(x1, x2):
+    return np.dot(x1, x2) / (np.linalg.norm(x1) * np.linalg.norm(x2))
+
+def load_npy(path):
+    data = np.load(path,allow_pickle=True)
+    data = data.item()
+    return data
+
+def create_database(path,model,database_path):
+    name_list = os.listdir(path)
+    k_v = {}
+    if os.path.exists(database_path):
+        k_v = np.load(database_path, allow_pickle=True)
+        k_v = k_v.item()
+    for name in name_list:
+        img_path = os.listdir(os.path.join(path,name))
+        for img_name in img_path[:1]:
+            img = load_image(os.path.join(path,name,img_name))
+            img = torch.from_numpy(img)
+            with torch.no_grad():
+                pred = model(img)
+                pred = pred.numpy()
+                k_v[name] = l2_normalize(pred)
+    np.save(database_path, k_v)
+
+def create_database_batch(path,model,database_path):
+    name_list = os.listdir(path)
+    k_v = {}
+    if os.path.exists(database_path):
+        k_v = np.load(database_path, allow_pickle=True)
+        k_v = k_v.item()
+    batch = 256
+    order_name = []
+    order_path = []
+    emb_list = []
+    for name in name_list:
+        img_path = os.listdir(os.path.join(path,name))
+        for img_name in img_path[:1]:
+            order_name.append(name)
+            order_path.append(os.path.join(path,name,img_name))
+    order_img = np.zeros((len(order_path), 3, 112, 112), dtype=np.float32)
+    for index, img_path in enumerate(order_path):
+        order_img[index] = load_image(img_path)
+    print(order_img.shape)
+    order_img = torch.from_numpy(order_img)
+    now = 0
+    number = len(order_img)
+    with torch.no_grad():
+        while now < number:
+            if now + batch < number:
+                emb = model(order_img[now:now+batch])
+            else:
+                emb = model(order_img[now:])
+            now = now + batch
+            for em in emb:
+                emb_list.append(em)
+            print("batch"+str(now))
+
+    for i, emb in enumerate(emb_list):
+        k_v[order_name[i]] = l2_normalize(emb.numpy())
+    np.save(database_path, k_v)
+
+def add_one(img,model,name,database_path):
+    img = torch.from_numpy(img)
+    with torch.no_grad():
+        pred = model(img)
+        pred = pred.numpy()
+        k_v = {}
+        if os.path.exists(database_path):
+            k_v = np.load(database_path, allow_pickle=True)
+            k_v = k_v.item()
+        k_v[name] = l2_normalize(pred)
+        np.save(database_path, k_v)
+
+def findmindistance(pred,threshold,k_v):
+    distance = 10
+    most_like = ""
+    for name in k_v.keys():
+        tmp = findEuclideanDistance(k_v[name],pred)
+        if distance > tmp:
+            distance = tmp
+            most_like = name
+    if distance < threshold:
+        return most_like
+    else:
+        return -1
+
+def findOne(img,model,k_v):
+
+    with torch.no_grad():
+        start_time = time.time()
+        pred = model(img)
+        end_time = time.time()
+        #print("predOne time: " + str(end_time - start_time))
+        pred = pred.numpy()
+        name = findmindistance(l2_normalize(pred),threshold=1.20,k_v=k_v)
+        if name != -1:
+            return name
+        else:
+            return "unknown"
+def findAll(imglist,model,k_v):
+    with torch.no_grad():
+        name_list = []
+        pred = model(imglist)
+        pred = pred.numpy()
+        for pr in pred:
+            name = findmindistance(l2_normalize(pr),threshold=1.20,k_v=k_v)
+            if name != -1:
+                name_list.append(name)
+            else:
+                name_list.append("unknown")
+        return name_list
+
+if __name__=='__main__':
+    model = iresnet100()
+    model.load_state_dict(torch.load("./model/backbone100.pth", map_location="cpu"))
+    model.eval()
+    #img = load_image(r"D:\Download\out\facedatabase\man.jpg")
+    #img = load_image(r"D:\Download\out\facedatabase\man6.jpg")
+    # img = load_image(r"D:\Download\out\alig_students\student.jpg")
+    # print(img.shape)
+    #
+    # k_v = load_npy("./Database/student.npy")
+    # start_time = time.time()
+    # img = torch.from_numpy(img)
+    # name = findOne(img,model,k_v)
+    # mo = r'[\u4e00-\u9fa5]*'
+    # name = re.match(mo,name)
+    # print(name.group(0))
+    # end_time = time.time()
+    # print("findOne time: " + str(end_time - start_time))
+
+    #create_database_batch(r"D:\Download\out\alig_students",model,"./Database/student.npy")
+    create_database_batch(r"D:\Download\out\cfp_database", model, "cfp.npy")
+    #add_one(img,model,"Arminio_Fraga","centerface_lfw.npy")
\ No newline at end of file
diff --git a/data/FDDB/img_list.txt b/data/FDDB/img_list.txt
new file mode 100644
index 0000000..5cf3d31
--- /dev/null
+++ b/data/FDDB/img_list.txt
@@ -0,0 +1,2845 @@
+2002/08/11/big/img_591
+2002/08/26/big/img_265
+2002/07/19/big/img_423
+2002/08/24/big/img_490
+2002/08/31/big/img_17676
+2002/07/31/big/img_228
+2002/07/24/big/img_402
+2002/08/04/big/img_769
+2002/07/19/big/img_581
+2002/08/13/big/img_723
+2002/08/12/big/img_821
+2003/01/17/big/img_610
+2002/08/13/big/img_1116
+2002/08/28/big/img_19238
+2002/08/21/big/img_660
+2002/08/14/big/img_607
+2002/08/05/big/img_3708
+2002/08/19/big/img_511
+2002/08/07/big/img_1316
+2002/07/25/big/img_1047
+2002/07/23/big/img_474
+2002/07/27/big/img_970
+2002/09/02/big/img_15752
+2002/09/01/big/img_16378
+2002/09/01/big/img_16189
+2002/08/26/big/img_276
+2002/07/24/big/img_518
+2002/08/14/big/img_1027
+2002/08/24/big/img_733
+2002/08/15/big/img_249
+2003/01/15/big/img_1371
+2002/08/07/big/img_1348
+2003/01/01/big/img_331
+2002/08/23/big/img_536
+2002/07/30/big/img_224
+2002/08/10/big/img_763
+2002/08/21/big/img_293
+2002/08/15/big/img_1211
+2002/08/15/big/img_1194
+2003/01/15/big/img_390
+2002/08/06/big/img_2893
+2002/08/17/big/img_691
+2002/08/07/big/img_1695
+2002/08/16/big/img_829
+2002/07/25/big/img_201
+2002/08/23/big/img_36
+2003/01/15/big/img_763
+2003/01/15/big/img_637
+2002/08/22/big/img_592
+2002/07/25/big/img_817
+2003/01/15/big/img_1219
+2002/08/05/big/img_3508
+2002/08/15/big/img_1108
+2002/07/19/big/img_488
+2003/01/16/big/img_704
+2003/01/13/big/img_1087
+2002/08/10/big/img_670
+2002/07/24/big/img_104
+2002/08/27/big/img_19823
+2002/09/01/big/img_16229
+2003/01/13/big/img_846
+2002/08/04/big/img_412
+2002/07/22/big/img_554
+2002/08/12/big/img_331
+2002/08/02/big/img_533
+2002/08/12/big/img_259
+2002/08/18/big/img_328
+2003/01/14/big/img_630
+2002/08/05/big/img_3541
+2002/08/06/big/img_2390
+2002/08/20/big/img_150
+2002/08/02/big/img_1231
+2002/08/16/big/img_710
+2002/08/19/big/img_591
+2002/07/22/big/img_725
+2002/07/24/big/img_820
+2003/01/13/big/img_568
+2002/08/22/big/img_853
+2002/08/09/big/img_648
+2002/08/23/big/img_528
+2003/01/14/big/img_888
+2002/08/30/big/img_18201
+2002/08/13/big/img_965
+2003/01/14/big/img_660
+2002/07/19/big/img_517
+2003/01/14/big/img_406
+2002/08/30/big/img_18433
+2002/08/07/big/img_1630
+2002/08/06/big/img_2717
+2002/08/21/big/img_470
+2002/07/23/big/img_633
+2002/08/20/big/img_915
+2002/08/16/big/img_893
+2002/07/29/big/img_644
+2002/08/15/big/img_529
+2002/08/16/big/img_668
+2002/08/07/big/img_1871
+2002/07/25/big/img_192
+2002/07/31/big/img_961
+2002/08/19/big/img_738
+2002/07/31/big/img_382
+2002/08/19/big/img_298
+2003/01/17/big/img_608
+2002/08/21/big/img_514
+2002/07/23/big/img_183
+2003/01/17/big/img_536
+2002/07/24/big/img_478
+2002/08/06/big/img_2997
+2002/09/02/big/img_15380
+2002/08/07/big/img_1153
+2002/07/31/big/img_967
+2002/07/31/big/img_711
+2002/08/26/big/img_664
+2003/01/01/big/img_326
+2002/08/24/big/img_775
+2002/08/08/big/img_961
+2002/08/16/big/img_77
+2002/08/12/big/img_296
+2002/07/22/big/img_905
+2003/01/13/big/img_284
+2002/08/13/big/img_887
+2002/08/24/big/img_849
+2002/07/30/big/img_345
+2002/08/18/big/img_419
+2002/08/01/big/img_1347
+2002/08/05/big/img_3670
+2002/07/21/big/img_479
+2002/08/08/big/img_913
+2002/09/02/big/img_15828
+2002/08/30/big/img_18194
+2002/08/08/big/img_471
+2002/08/22/big/img_734
+2002/08/09/big/img_586
+2002/08/09/big/img_454
+2002/07/29/big/img_47
+2002/07/19/big/img_381
+2002/07/29/big/img_733
+2002/08/20/big/img_327
+2002/07/21/big/img_96
+2002/08/06/big/img_2680
+2002/07/25/big/img_919
+2002/07/21/big/img_158
+2002/07/22/big/img_801
+2002/07/22/big/img_567
+2002/07/24/big/img_804
+2002/07/24/big/img_690
+2003/01/15/big/img_576
+2002/08/14/big/img_335
+2003/01/13/big/img_390
+2002/08/11/big/img_258
+2002/07/23/big/img_917
+2002/08/15/big/img_525
+2003/01/15/big/img_505
+2002/07/30/big/img_886
+2003/01/16/big/img_640
+2003/01/14/big/img_642
+2003/01/17/big/img_844
+2002/08/04/big/img_571
+2002/08/29/big/img_18702
+2003/01/15/big/img_240
+2002/07/29/big/img_553
+2002/08/10/big/img_354
+2002/08/18/big/img_17
+2003/01/15/big/img_782
+2002/07/27/big/img_382
+2002/08/14/big/img_970
+2003/01/16/big/img_70
+2003/01/16/big/img_625
+2002/08/18/big/img_341
+2002/08/26/big/img_188
+2002/08/09/big/img_405
+2002/08/02/big/img_37
+2002/08/13/big/img_748
+2002/07/22/big/img_399
+2002/07/25/big/img_844
+2002/08/12/big/img_340
+2003/01/13/big/img_815
+2002/08/26/big/img_5
+2002/08/10/big/img_158
+2002/08/18/big/img_95
+2002/07/29/big/img_1297
+2003/01/13/big/img_508
+2002/09/01/big/img_16680
+2003/01/16/big/img_338
+2002/08/13/big/img_517
+2002/07/22/big/img_626
+2002/08/06/big/img_3024
+2002/07/26/big/img_499
+2003/01/13/big/img_387
+2002/08/31/big/img_18025
+2002/08/13/big/img_520
+2003/01/16/big/img_576
+2002/07/26/big/img_121
+2002/08/25/big/img_703
+2002/08/26/big/img_615
+2002/08/17/big/img_434
+2002/08/02/big/img_677
+2002/08/18/big/img_276
+2002/08/05/big/img_3672
+2002/07/26/big/img_700
+2002/07/31/big/img_277
+2003/01/14/big/img_220
+2002/08/23/big/img_232
+2002/08/31/big/img_17422
+2002/07/22/big/img_508
+2002/08/13/big/img_681
+2003/01/15/big/img_638
+2002/08/30/big/img_18408
+2003/01/14/big/img_533
+2003/01/17/big/img_12
+2002/08/28/big/img_19388
+2002/08/08/big/img_133
+2002/07/26/big/img_885
+2002/08/19/big/img_387
+2002/08/27/big/img_19976
+2002/08/26/big/img_118
+2002/08/28/big/img_19146
+2002/08/05/big/img_3259
+2002/08/15/big/img_536
+2002/07/22/big/img_279
+2002/07/22/big/img_9
+2002/08/13/big/img_301
+2002/08/15/big/img_974
+2002/08/06/big/img_2355
+2002/08/01/big/img_1526
+2002/08/03/big/img_417
+2002/08/04/big/img_407
+2002/08/15/big/img_1029
+2002/07/29/big/img_700
+2002/08/01/big/img_1463
+2002/08/31/big/img_17365
+2002/07/28/big/img_223
+2002/07/19/big/img_827
+2002/07/27/big/img_531
+2002/07/19/big/img_845
+2002/08/20/big/img_382
+2002/07/31/big/img_268
+2002/08/27/big/img_19705
+2002/08/02/big/img_830
+2002/08/23/big/img_250
+2002/07/20/big/img_777
+2002/08/21/big/img_879
+2002/08/26/big/img_20146
+2002/08/23/big/img_789
+2002/08/06/big/img_2683
+2002/08/25/big/img_576
+2002/08/09/big/img_498
+2002/08/08/big/img_384
+2002/08/26/big/img_592
+2002/07/29/big/img_1470
+2002/08/21/big/img_452
+2002/08/30/big/img_18395
+2002/08/15/big/img_215
+2002/07/21/big/img_643
+2002/07/22/big/img_209
+2003/01/17/big/img_346
+2002/08/25/big/img_658
+2002/08/21/big/img_221
+2002/08/14/big/img_60
+2003/01/17/big/img_885
+2003/01/16/big/img_482
+2002/08/19/big/img_593
+2002/08/08/big/img_233
+2002/07/30/big/img_458
+2002/07/23/big/img_384
+2003/01/15/big/img_670
+2003/01/15/big/img_267
+2002/08/26/big/img_540
+2002/07/29/big/img_552
+2002/07/30/big/img_997
+2003/01/17/big/img_377
+2002/08/21/big/img_265
+2002/08/09/big/img_561
+2002/07/31/big/img_945
+2002/09/02/big/img_15252
+2002/08/11/big/img_276
+2002/07/22/big/img_491
+2002/07/26/big/img_517
+2002/08/14/big/img_726
+2002/08/08/big/img_46
+2002/08/28/big/img_19458
+2002/08/06/big/img_2935
+2002/07/29/big/img_1392
+2002/08/13/big/img_776
+2002/08/24/big/img_616
+2002/08/14/big/img_1065
+2002/07/29/big/img_889
+2002/08/18/big/img_188
+2002/08/07/big/img_1453
+2002/08/02/big/img_760
+2002/07/28/big/img_416
+2002/08/07/big/img_1393
+2002/08/26/big/img_292
+2002/08/26/big/img_301
+2003/01/13/big/img_195
+2002/07/26/big/img_532
+2002/08/20/big/img_550
+2002/08/05/big/img_3658
+2002/08/26/big/img_738
+2002/09/02/big/img_15750
+2003/01/17/big/img_451
+2002/07/23/big/img_339
+2002/08/16/big/img_637
+2002/08/14/big/img_748
+2002/08/06/big/img_2739
+2002/07/25/big/img_482
+2002/08/19/big/img_191
+2002/08/26/big/img_537
+2003/01/15/big/img_716
+2003/01/15/big/img_767
+2002/08/02/big/img_452
+2002/08/08/big/img_1011
+2002/08/10/big/img_144
+2003/01/14/big/img_122
+2002/07/24/big/img_586
+2002/07/24/big/img_762
+2002/08/20/big/img_369
+2002/07/30/big/img_146
+2002/08/23/big/img_396
+2003/01/15/big/img_200
+2002/08/15/big/img_1183
+2003/01/14/big/img_698
+2002/08/09/big/img_792
+2002/08/06/big/img_2347
+2002/07/31/big/img_911
+2002/08/26/big/img_722
+2002/08/23/big/img_621
+2002/08/05/big/img_3790
+2003/01/13/big/img_633
+2002/08/09/big/img_224
+2002/07/24/big/img_454
+2002/07/21/big/img_202
+2002/08/02/big/img_630
+2002/08/30/big/img_18315
+2002/07/19/big/img_491
+2002/09/01/big/img_16456
+2002/08/09/big/img_242
+2002/07/25/big/img_595
+2002/07/22/big/img_522
+2002/08/01/big/img_1593
+2002/07/29/big/img_336
+2002/08/15/big/img_448
+2002/08/28/big/img_19281
+2002/07/29/big/img_342
+2002/08/12/big/img_78
+2003/01/14/big/img_525
+2002/07/28/big/img_147
+2002/08/11/big/img_353
+2002/08/22/big/img_513
+2002/08/04/big/img_721
+2002/08/17/big/img_247
+2003/01/14/big/img_891
+2002/08/20/big/img_853
+2002/07/19/big/img_414
+2002/08/01/big/img_1530
+2003/01/14/big/img_924
+2002/08/22/big/img_468
+2002/08/18/big/img_354
+2002/08/30/big/img_18193
+2002/08/23/big/img_492
+2002/08/15/big/img_871
+2002/08/12/big/img_494
+2002/08/06/big/img_2470
+2002/07/23/big/img_923
+2002/08/26/big/img_155
+2002/08/08/big/img_669
+2002/07/23/big/img_404
+2002/08/28/big/img_19421
+2002/08/29/big/img_18993
+2002/08/25/big/img_416
+2003/01/17/big/img_434
+2002/07/29/big/img_1370
+2002/07/28/big/img_483
+2002/08/11/big/img_50
+2002/08/10/big/img_404
+2002/09/02/big/img_15057
+2003/01/14/big/img_911
+2002/09/01/big/img_16697
+2003/01/16/big/img_665
+2002/09/01/big/img_16708
+2002/08/22/big/img_612
+2002/08/28/big/img_19471
+2002/08/02/big/img_198
+2003/01/16/big/img_527
+2002/08/22/big/img_209
+2002/08/30/big/img_18205
+2003/01/14/big/img_114
+2003/01/14/big/img_1028
+2003/01/16/big/img_894
+2003/01/14/big/img_837
+2002/07/30/big/img_9
+2002/08/06/big/img_2821
+2002/08/04/big/img_85
+2003/01/13/big/img_884
+2002/07/22/big/img_570
+2002/08/07/big/img_1773
+2002/07/26/big/img_208
+2003/01/17/big/img_946
+2002/07/19/big/img_930
+2003/01/01/big/img_698
+2003/01/17/big/img_612
+2002/07/19/big/img_372
+2002/07/30/big/img_721
+2003/01/14/big/img_649
+2002/08/19/big/img_4
+2002/07/25/big/img_1024
+2003/01/15/big/img_601
+2002/08/30/big/img_18470
+2002/07/22/big/img_29
+2002/08/07/big/img_1686
+2002/07/20/big/img_294
+2002/08/14/big/img_800
+2002/08/19/big/img_353
+2002/08/19/big/img_350
+2002/08/05/big/img_3392
+2002/08/09/big/img_622
+2003/01/15/big/img_236
+2002/08/11/big/img_643
+2002/08/05/big/img_3458
+2002/08/12/big/img_413
+2002/08/22/big/img_415
+2002/08/13/big/img_635
+2002/08/07/big/img_1198
+2002/08/04/big/img_873
+2002/08/12/big/img_407
+2003/01/15/big/img_346
+2002/08/02/big/img_275
+2002/08/17/big/img_997
+2002/08/21/big/img_958
+2002/08/20/big/img_579
+2002/07/29/big/img_142
+2003/01/14/big/img_1115
+2002/08/16/big/img_365
+2002/07/29/big/img_1414
+2002/08/17/big/img_489
+2002/08/13/big/img_1010
+2002/07/31/big/img_276
+2002/07/25/big/img_1000
+2002/08/23/big/img_524
+2002/08/28/big/img_19147
+2003/01/13/big/img_433
+2002/08/20/big/img_205
+2003/01/01/big/img_458
+2002/07/29/big/img_1449
+2003/01/16/big/img_696
+2002/08/28/big/img_19296
+2002/08/29/big/img_18688
+2002/08/21/big/img_767
+2002/08/20/big/img_532
+2002/08/26/big/img_187
+2002/07/26/big/img_183
+2002/07/27/big/img_890
+2003/01/13/big/img_576
+2002/07/30/big/img_15
+2002/07/31/big/img_889
+2002/08/31/big/img_17759
+2003/01/14/big/img_1114
+2002/07/19/big/img_445
+2002/08/03/big/img_593
+2002/07/24/big/img_750
+2002/07/30/big/img_133
+2002/08/25/big/img_671
+2002/07/20/big/img_351
+2002/08/31/big/img_17276
+2002/08/05/big/img_3231
+2002/09/02/big/img_15882
+2002/08/14/big/img_115
+2002/08/02/big/img_1148
+2002/07/25/big/img_936
+2002/07/31/big/img_639
+2002/08/04/big/img_427
+2002/08/22/big/img_843
+2003/01/17/big/img_17
+2003/01/13/big/img_690
+2002/08/13/big/img_472
+2002/08/09/big/img_425
+2002/08/05/big/img_3450
+2003/01/17/big/img_439
+2002/08/13/big/img_539
+2002/07/28/big/img_35
+2002/08/16/big/img_241
+2002/08/06/big/img_2898
+2003/01/16/big/img_429
+2002/08/05/big/img_3817
+2002/08/27/big/img_19919
+2002/07/19/big/img_422
+2002/08/15/big/img_560
+2002/07/23/big/img_750
+2002/07/30/big/img_353
+2002/08/05/big/img_43
+2002/08/23/big/img_305
+2002/08/01/big/img_2137
+2002/08/30/big/img_18097
+2002/08/01/big/img_1389
+2002/08/02/big/img_308
+2003/01/14/big/img_652
+2002/08/01/big/img_1798
+2003/01/14/big/img_732
+2003/01/16/big/img_294
+2002/08/26/big/img_213
+2002/07/24/big/img_842
+2003/01/13/big/img_630
+2003/01/13/big/img_634
+2002/08/06/big/img_2285
+2002/08/01/big/img_2162
+2002/08/30/big/img_18134
+2002/08/02/big/img_1045
+2002/08/01/big/img_2143
+2002/07/25/big/img_135
+2002/07/20/big/img_645
+2002/08/05/big/img_3666
+2002/08/14/big/img_523
+2002/08/04/big/img_425
+2003/01/14/big/img_137
+2003/01/01/big/img_176
+2002/08/15/big/img_505
+2002/08/24/big/img_386
+2002/08/05/big/img_3187
+2002/08/15/big/img_419
+2003/01/13/big/img_520
+2002/08/04/big/img_444
+2002/08/26/big/img_483
+2002/08/05/big/img_3449
+2002/08/30/big/img_18409
+2002/08/28/big/img_19455
+2002/08/27/big/img_20090
+2002/07/23/big/img_625
+2002/08/24/big/img_205
+2002/08/08/big/img_938
+2003/01/13/big/img_527
+2002/08/07/big/img_1712
+2002/07/24/big/img_801
+2002/08/09/big/img_579
+2003/01/14/big/img_41
+2003/01/15/big/img_1130
+2002/07/21/big/img_672
+2002/08/07/big/img_1590
+2003/01/01/big/img_532
+2002/08/02/big/img_529
+2002/08/05/big/img_3591
+2002/08/23/big/img_5
+2003/01/14/big/img_882
+2002/08/28/big/img_19234
+2002/07/24/big/img_398
+2003/01/14/big/img_592
+2002/08/22/big/img_548
+2002/08/12/big/img_761
+2003/01/16/big/img_497
+2002/08/18/big/img_133
+2002/08/08/big/img_874
+2002/07/19/big/img_247
+2002/08/15/big/img_170
+2002/08/27/big/img_19679
+2002/08/20/big/img_246
+2002/08/24/big/img_358
+2002/07/29/big/img_599
+2002/08/01/big/img_1555
+2002/07/30/big/img_491
+2002/07/30/big/img_371
+2003/01/16/big/img_682
+2002/07/25/big/img_619
+2003/01/15/big/img_587
+2002/08/02/big/img_1212
+2002/08/01/big/img_2152
+2002/07/25/big/img_668
+2003/01/16/big/img_574
+2002/08/28/big/img_19464
+2002/08/11/big/img_536
+2002/07/24/big/img_201
+2002/08/05/big/img_3488
+2002/07/25/big/img_887
+2002/07/22/big/img_789
+2002/07/30/big/img_432
+2002/08/16/big/img_166
+2002/09/01/big/img_16333
+2002/07/26/big/img_1010
+2002/07/21/big/img_793
+2002/07/22/big/img_720
+2002/07/31/big/img_337
+2002/07/27/big/img_185
+2002/08/23/big/img_440
+2002/07/31/big/img_801
+2002/07/25/big/img_478
+2003/01/14/big/img_171
+2002/08/07/big/img_1054
+2002/09/02/big/img_15659
+2002/07/29/big/img_1348
+2002/08/09/big/img_337
+2002/08/26/big/img_684
+2002/07/31/big/img_537
+2002/08/15/big/img_808
+2003/01/13/big/img_740
+2002/08/07/big/img_1667
+2002/08/03/big/img_404
+2002/08/06/big/img_2520
+2002/07/19/big/img_230
+2002/07/19/big/img_356
+2003/01/16/big/img_627
+2002/08/04/big/img_474
+2002/07/29/big/img_833
+2002/07/25/big/img_176
+2002/08/01/big/img_1684
+2002/08/21/big/img_643
+2002/08/27/big/img_19673
+2002/08/02/big/img_838
+2002/08/06/big/img_2378
+2003/01/15/big/img_48
+2002/07/30/big/img_470
+2002/08/15/big/img_963
+2002/08/24/big/img_444
+2002/08/16/big/img_662
+2002/08/15/big/img_1209
+2002/07/24/big/img_25
+2002/08/06/big/img_2740
+2002/07/29/big/img_996
+2002/08/31/big/img_18074
+2002/08/04/big/img_343
+2003/01/17/big/img_509
+2003/01/13/big/img_726
+2002/08/07/big/img_1466
+2002/07/26/big/img_307
+2002/08/10/big/img_598
+2002/08/13/big/img_890
+2002/08/14/big/img_997
+2002/07/19/big/img_392
+2002/08/02/big/img_475
+2002/08/29/big/img_19038
+2002/07/29/big/img_538
+2002/07/29/big/img_502
+2002/08/02/big/img_364
+2002/08/31/big/img_17353
+2002/08/08/big/img_539
+2002/08/01/big/img_1449
+2002/07/22/big/img_363
+2002/08/02/big/img_90
+2002/09/01/big/img_16867
+2002/08/05/big/img_3371
+2002/07/30/big/img_342
+2002/08/07/big/img_1363
+2002/08/22/big/img_790
+2003/01/15/big/img_404
+2002/08/05/big/img_3447
+2002/09/01/big/img_16167
+2003/01/13/big/img_840
+2002/08/22/big/img_1001
+2002/08/09/big/img_431
+2002/07/27/big/img_618
+2002/07/31/big/img_741
+2002/07/30/big/img_964
+2002/07/25/big/img_86
+2002/07/29/big/img_275
+2002/08/21/big/img_921
+2002/07/26/big/img_892
+2002/08/21/big/img_663
+2003/01/13/big/img_567
+2003/01/14/big/img_719
+2002/07/28/big/img_251
+2003/01/15/big/img_1123
+2002/07/29/big/img_260
+2002/08/24/big/img_337
+2002/08/01/big/img_1914
+2002/08/13/big/img_373
+2003/01/15/big/img_589
+2002/08/13/big/img_906
+2002/07/26/big/img_270
+2002/08/26/big/img_313
+2002/08/25/big/img_694
+2003/01/01/big/img_327
+2002/07/23/big/img_261
+2002/08/26/big/img_642
+2002/07/29/big/img_918
+2002/07/23/big/img_455
+2002/07/24/big/img_612
+2002/07/23/big/img_534
+2002/07/19/big/img_534
+2002/07/19/big/img_726
+2002/08/01/big/img_2146
+2002/08/02/big/img_543
+2003/01/16/big/img_777
+2002/07/30/big/img_484
+2002/08/13/big/img_1161
+2002/07/21/big/img_390
+2002/08/06/big/img_2288
+2002/08/21/big/img_677
+2002/08/13/big/img_747
+2002/08/15/big/img_1248
+2002/07/31/big/img_416
+2002/09/02/big/img_15259
+2002/08/16/big/img_781
+2002/08/24/big/img_754
+2002/07/24/big/img_803
+2002/08/20/big/img_609
+2002/08/28/big/img_19571
+2002/09/01/big/img_16140
+2002/08/26/big/img_769
+2002/07/20/big/img_588
+2002/08/02/big/img_898
+2002/07/21/big/img_466
+2002/08/14/big/img_1046
+2002/07/25/big/img_212
+2002/08/26/big/img_353
+2002/08/19/big/img_810
+2002/08/31/big/img_17824
+2002/08/12/big/img_631
+2002/07/19/big/img_828
+2002/07/24/big/img_130
+2002/08/25/big/img_580
+2002/07/31/big/img_699
+2002/07/23/big/img_808
+2002/07/31/big/img_377
+2003/01/16/big/img_570
+2002/09/01/big/img_16254
+2002/07/21/big/img_471
+2002/08/01/big/img_1548
+2002/08/18/big/img_252
+2002/08/19/big/img_576
+2002/08/20/big/img_464
+2002/07/27/big/img_735
+2002/08/21/big/img_589
+2003/01/15/big/img_1192
+2002/08/09/big/img_302
+2002/07/31/big/img_594
+2002/08/23/big/img_19
+2002/08/29/big/img_18819
+2002/08/19/big/img_293
+2002/07/30/big/img_331
+2002/08/23/big/img_607
+2002/07/30/big/img_363
+2002/08/16/big/img_766
+2003/01/13/big/img_481
+2002/08/06/big/img_2515
+2002/09/02/big/img_15913
+2002/09/02/big/img_15827
+2002/09/02/big/img_15053
+2002/08/07/big/img_1576
+2002/07/23/big/img_268
+2002/08/21/big/img_152
+2003/01/15/big/img_578
+2002/07/21/big/img_589
+2002/07/20/big/img_548
+2002/08/27/big/img_19693
+2002/08/31/big/img_17252
+2002/07/31/big/img_138
+2002/07/23/big/img_372
+2002/08/16/big/img_695
+2002/07/27/big/img_287
+2002/08/15/big/img_315
+2002/08/10/big/img_361
+2002/07/29/big/img_899
+2002/08/13/big/img_771
+2002/08/21/big/img_92
+2003/01/15/big/img_425
+2003/01/16/big/img_450
+2002/09/01/big/img_16942
+2002/08/02/big/img_51
+2002/09/02/big/img_15379
+2002/08/24/big/img_147
+2002/08/30/big/img_18122
+2002/07/26/big/img_950
+2002/08/07/big/img_1400
+2002/08/17/big/img_468
+2002/08/15/big/img_470
+2002/07/30/big/img_318
+2002/07/22/big/img_644
+2002/08/27/big/img_19732
+2002/07/23/big/img_601
+2002/08/26/big/img_398
+2002/08/21/big/img_428
+2002/08/06/big/img_2119
+2002/08/29/big/img_19103
+2003/01/14/big/img_933
+2002/08/11/big/img_674
+2002/08/28/big/img_19420
+2002/08/03/big/img_418
+2002/08/17/big/img_312
+2002/07/25/big/img_1044
+2003/01/17/big/img_671
+2002/08/30/big/img_18297
+2002/07/25/big/img_755
+2002/07/23/big/img_471
+2002/08/21/big/img_39
+2002/07/26/big/img_699
+2003/01/14/big/img_33
+2002/07/31/big/img_411
+2002/08/16/big/img_645
+2003/01/17/big/img_116
+2002/09/02/big/img_15903
+2002/08/20/big/img_120
+2002/08/22/big/img_176
+2002/07/29/big/img_1316
+2002/08/27/big/img_19914
+2002/07/22/big/img_719
+2002/08/28/big/img_19239
+2003/01/13/big/img_385
+2002/08/08/big/img_525
+2002/07/19/big/img_782
+2002/08/13/big/img_843
+2002/07/30/big/img_107
+2002/08/11/big/img_752
+2002/07/29/big/img_383
+2002/08/26/big/img_249
+2002/08/29/big/img_18860
+2002/07/30/big/img_70
+2002/07/26/big/img_194
+2002/08/15/big/img_530
+2002/08/08/big/img_816
+2002/07/31/big/img_286
+2003/01/13/big/img_294
+2002/07/31/big/img_251
+2002/07/24/big/img_13
+2002/08/31/big/img_17938
+2002/07/22/big/img_642
+2003/01/14/big/img_728
+2002/08/18/big/img_47
+2002/08/22/big/img_306
+2002/08/20/big/img_348
+2002/08/15/big/img_764
+2002/08/08/big/img_163
+2002/07/23/big/img_531
+2002/07/23/big/img_467
+2003/01/16/big/img_743
+2003/01/13/big/img_535
+2002/08/02/big/img_523
+2002/08/22/big/img_120
+2002/08/11/big/img_496
+2002/08/29/big/img_19075
+2002/08/08/big/img_465
+2002/08/09/big/img_790
+2002/08/19/big/img_588
+2002/08/23/big/img_407
+2003/01/17/big/img_435
+2002/08/24/big/img_398
+2002/08/27/big/img_19899
+2003/01/15/big/img_335
+2002/08/13/big/img_493
+2002/09/02/big/img_15460
+2002/07/31/big/img_470
+2002/08/05/big/img_3550
+2002/07/28/big/img_123
+2002/08/01/big/img_1498
+2002/08/04/big/img_504
+2003/01/17/big/img_427
+2002/08/27/big/img_19708
+2002/07/27/big/img_861
+2002/07/25/big/img_685
+2002/07/31/big/img_207
+2003/01/14/big/img_745
+2002/08/31/big/img_17756
+2002/08/24/big/img_288
+2002/08/18/big/img_181
+2002/08/10/big/img_520
+2002/08/25/big/img_705
+2002/08/23/big/img_226
+2002/08/04/big/img_727
+2002/07/24/big/img_625
+2002/08/28/big/img_19157
+2002/08/23/big/img_586
+2002/07/31/big/img_232
+2003/01/13/big/img_240
+2003/01/14/big/img_321
+2003/01/15/big/img_533
+2002/07/23/big/img_480
+2002/07/24/big/img_371
+2002/08/21/big/img_702
+2002/08/31/big/img_17075
+2002/09/02/big/img_15278
+2002/07/29/big/img_246
+2003/01/15/big/img_829
+2003/01/15/big/img_1213
+2003/01/16/big/img_441
+2002/08/14/big/img_921
+2002/07/23/big/img_425
+2002/08/15/big/img_296
+2002/07/19/big/img_135
+2002/07/26/big/img_402
+2003/01/17/big/img_88
+2002/08/20/big/img_872
+2002/08/13/big/img_1110
+2003/01/16/big/img_1040
+2002/07/23/big/img_9
+2002/08/13/big/img_700
+2002/08/16/big/img_371
+2002/08/27/big/img_19966
+2003/01/17/big/img_391
+2002/08/18/big/img_426
+2002/08/01/big/img_1618
+2002/07/21/big/img_754
+2003/01/14/big/img_1101
+2003/01/16/big/img_1022
+2002/07/22/big/img_275
+2002/08/24/big/img_86
+2002/08/17/big/img_582
+2003/01/15/big/img_765
+2003/01/17/big/img_449
+2002/07/28/big/img_265
+2003/01/13/big/img_552
+2002/07/28/big/img_115
+2003/01/16/big/img_56
+2002/08/02/big/img_1232
+2003/01/17/big/img_925
+2002/07/22/big/img_445
+2002/07/25/big/img_957
+2002/07/20/big/img_589
+2002/08/31/big/img_17107
+2002/07/29/big/img_483
+2002/08/14/big/img_1063
+2002/08/07/big/img_1545
+2002/08/14/big/img_680
+2002/09/01/big/img_16694
+2002/08/14/big/img_257
+2002/08/11/big/img_726
+2002/07/26/big/img_681
+2002/07/25/big/img_481
+2003/01/14/big/img_737
+2002/08/28/big/img_19480
+2003/01/16/big/img_362
+2002/08/27/big/img_19865
+2003/01/01/big/img_547
+2002/09/02/big/img_15074
+2002/08/01/big/img_1453
+2002/08/22/big/img_594
+2002/08/28/big/img_19263
+2002/08/13/big/img_478
+2002/07/29/big/img_1358
+2003/01/14/big/img_1022
+2002/08/16/big/img_450
+2002/08/02/big/img_159
+2002/07/26/big/img_781
+2003/01/13/big/img_601
+2002/08/20/big/img_407
+2002/08/15/big/img_468
+2002/08/31/big/img_17902
+2002/08/16/big/img_81
+2002/07/25/big/img_987
+2002/07/25/big/img_500
+2002/08/02/big/img_31
+2002/08/18/big/img_538
+2002/08/08/big/img_54
+2002/07/23/big/img_686
+2002/07/24/big/img_836
+2003/01/17/big/img_734
+2002/08/16/big/img_1055
+2003/01/16/big/img_521
+2002/07/25/big/img_612
+2002/08/22/big/img_778
+2002/08/03/big/img_251
+2002/08/12/big/img_436
+2002/08/23/big/img_705
+2002/07/28/big/img_243
+2002/07/25/big/img_1029
+2002/08/20/big/img_287
+2002/08/29/big/img_18739
+2002/08/05/big/img_3272
+2002/07/27/big/img_214
+2003/01/14/big/img_5
+2002/08/01/big/img_1380
+2002/08/29/big/img_19097
+2002/07/30/big/img_486
+2002/08/29/big/img_18707
+2002/08/10/big/img_559
+2002/08/15/big/img_365
+2002/08/09/big/img_525
+2002/08/10/big/img_689
+2002/07/25/big/img_502
+2002/08/03/big/img_667
+2002/08/10/big/img_855
+2002/08/10/big/img_706
+2002/08/18/big/img_603
+2003/01/16/big/img_1055
+2002/08/31/big/img_17890
+2002/08/15/big/img_761
+2003/01/15/big/img_489
+2002/08/26/big/img_351
+2002/08/01/big/img_1772
+2002/08/31/big/img_17729
+2002/07/25/big/img_609
+2003/01/13/big/img_539
+2002/07/27/big/img_686
+2002/07/31/big/img_311
+2002/08/22/big/img_799
+2003/01/16/big/img_936
+2002/08/31/big/img_17813
+2002/08/04/big/img_862
+2002/08/09/big/img_332
+2002/07/20/big/img_148
+2002/08/12/big/img_426
+2002/07/24/big/img_69
+2002/07/27/big/img_685
+2002/08/02/big/img_480
+2002/08/26/big/img_154
+2002/07/24/big/img_598
+2002/08/01/big/img_1881
+2002/08/20/big/img_667
+2003/01/14/big/img_495
+2002/07/21/big/img_744
+2002/07/30/big/img_150
+2002/07/23/big/img_924
+2002/08/08/big/img_272
+2002/07/23/big/img_310
+2002/07/25/big/img_1011
+2002/09/02/big/img_15725
+2002/07/19/big/img_814
+2002/08/20/big/img_936
+2002/07/25/big/img_85
+2002/08/24/big/img_662
+2002/08/09/big/img_495
+2003/01/15/big/img_196
+2002/08/16/big/img_707
+2002/08/28/big/img_19370
+2002/08/06/big/img_2366
+2002/08/06/big/img_3012
+2002/08/01/big/img_1452
+2002/07/31/big/img_742
+2002/07/27/big/img_914
+2003/01/13/big/img_290
+2002/07/31/big/img_288
+2002/08/02/big/img_171
+2002/08/22/big/img_191
+2002/07/27/big/img_1066
+2002/08/12/big/img_383
+2003/01/17/big/img_1018
+2002/08/01/big/img_1785
+2002/08/11/big/img_390
+2002/08/27/big/img_20037
+2002/08/12/big/img_38
+2003/01/15/big/img_103
+2002/08/26/big/img_31
+2002/08/18/big/img_660
+2002/07/22/big/img_694
+2002/08/15/big/img_24
+2002/07/27/big/img_1077
+2002/08/01/big/img_1943
+2002/07/22/big/img_292
+2002/09/01/big/img_16857
+2002/07/22/big/img_892
+2003/01/14/big/img_46
+2002/08/09/big/img_469
+2002/08/09/big/img_414
+2003/01/16/big/img_40
+2002/08/28/big/img_19231
+2002/07/27/big/img_978
+2002/07/23/big/img_475
+2002/07/25/big/img_92
+2002/08/09/big/img_799
+2002/07/25/big/img_491
+2002/08/03/big/img_654
+2003/01/15/big/img_687
+2002/08/11/big/img_478
+2002/08/07/big/img_1664
+2002/08/20/big/img_362
+2002/08/01/big/img_1298
+2003/01/13/big/img_500
+2002/08/06/big/img_2896
+2002/08/30/big/img_18529
+2002/08/16/big/img_1020
+2002/07/29/big/img_892
+2002/08/29/big/img_18726
+2002/07/21/big/img_453
+2002/08/17/big/img_437
+2002/07/19/big/img_665
+2002/07/22/big/img_440
+2002/07/19/big/img_582
+2002/07/21/big/img_233
+2003/01/01/big/img_82
+2002/07/25/big/img_341
+2002/07/29/big/img_864
+2002/08/02/big/img_276
+2002/08/29/big/img_18654
+2002/07/27/big/img_1024
+2002/08/19/big/img_373
+2003/01/15/big/img_241
+2002/07/25/big/img_84
+2002/08/13/big/img_834
+2002/08/10/big/img_511
+2002/08/01/big/img_1627
+2002/08/08/big/img_607
+2002/08/06/big/img_2083
+2002/08/01/big/img_1486
+2002/08/08/big/img_700
+2002/08/01/big/img_1954
+2002/08/21/big/img_54
+2002/07/30/big/img_847
+2002/08/28/big/img_19169
+2002/07/21/big/img_549
+2002/08/03/big/img_693
+2002/07/31/big/img_1002
+2003/01/14/big/img_1035
+2003/01/16/big/img_622
+2002/07/30/big/img_1201
+2002/08/10/big/img_444
+2002/07/31/big/img_374
+2002/08/21/big/img_301
+2002/08/13/big/img_1095
+2003/01/13/big/img_288
+2002/07/25/big/img_232
+2003/01/13/big/img_967
+2002/08/26/big/img_360
+2002/08/05/big/img_67
+2002/08/29/big/img_18969
+2002/07/28/big/img_16
+2002/08/16/big/img_515
+2002/07/20/big/img_708
+2002/08/18/big/img_178
+2003/01/15/big/img_509
+2002/07/25/big/img_430
+2002/08/21/big/img_738
+2002/08/16/big/img_886
+2002/09/02/big/img_15605
+2002/09/01/big/img_16242
+2002/08/24/big/img_711
+2002/07/25/big/img_90
+2002/08/09/big/img_491
+2002/07/30/big/img_534
+2003/01/13/big/img_474
+2002/08/25/big/img_510
+2002/08/15/big/img_555
+2002/08/02/big/img_775
+2002/07/23/big/img_975
+2002/08/19/big/img_229
+2003/01/17/big/img_860
+2003/01/02/big/img_10
+2002/07/23/big/img_542
+2002/08/06/big/img_2535
+2002/07/22/big/img_37
+2002/08/06/big/img_2342
+2002/08/25/big/img_515
+2002/08/25/big/img_336
+2002/08/18/big/img_837
+2002/08/21/big/img_616
+2003/01/17/big/img_24
+2002/07/26/big/img_936
+2002/08/14/big/img_896
+2002/07/29/big/img_465
+2002/07/31/big/img_543
+2002/08/01/big/img_1411
+2002/08/02/big/img_423
+2002/08/21/big/img_44
+2002/07/31/big/img_11
+2003/01/15/big/img_628
+2003/01/15/big/img_605
+2002/07/30/big/img_571
+2002/07/23/big/img_428
+2002/08/15/big/img_942
+2002/07/26/big/img_531
+2003/01/16/big/img_59
+2002/08/02/big/img_410
+2002/07/31/big/img_230
+2002/08/19/big/img_806
+2003/01/14/big/img_462
+2002/08/16/big/img_370
+2002/08/13/big/img_380
+2002/08/16/big/img_932
+2002/07/19/big/img_393
+2002/08/20/big/img_764
+2002/08/15/big/img_616
+2002/07/26/big/img_267
+2002/07/27/big/img_1069
+2002/08/14/big/img_1041
+2003/01/13/big/img_594
+2002/09/01/big/img_16845
+2002/08/09/big/img_229
+2003/01/16/big/img_639
+2002/08/19/big/img_398
+2002/08/18/big/img_978
+2002/08/24/big/img_296
+2002/07/29/big/img_415
+2002/07/30/big/img_923
+2002/08/18/big/img_575
+2002/08/22/big/img_182
+2002/07/25/big/img_806
+2002/07/22/big/img_49
+2002/07/29/big/img_989
+2003/01/17/big/img_789
+2003/01/15/big/img_503
+2002/09/01/big/img_16062
+2003/01/17/big/img_794
+2002/08/15/big/img_564
+2003/01/15/big/img_222
+2002/08/01/big/img_1656
+2003/01/13/big/img_432
+2002/07/19/big/img_426
+2002/08/17/big/img_244
+2002/08/13/big/img_805
+2002/09/02/big/img_15067
+2002/08/11/big/img_58
+2002/08/22/big/img_636
+2002/07/22/big/img_416
+2002/08/13/big/img_836
+2002/08/26/big/img_363
+2002/07/30/big/img_917
+2003/01/14/big/img_206
+2002/08/12/big/img_311
+2002/08/31/big/img_17623
+2002/07/29/big/img_661
+2003/01/13/big/img_417
+2002/08/02/big/img_463
+2002/08/02/big/img_669
+2002/08/26/big/img_670
+2002/08/02/big/img_375
+2002/07/19/big/img_209
+2002/08/08/big/img_115
+2002/08/21/big/img_399
+2002/08/20/big/img_911
+2002/08/07/big/img_1212
+2002/08/20/big/img_578
+2002/08/22/big/img_554
+2002/08/21/big/img_484
+2002/07/25/big/img_450
+2002/08/03/big/img_542
+2002/08/15/big/img_561
+2002/07/23/big/img_360
+2002/08/30/big/img_18137
+2002/07/25/big/img_250
+2002/08/03/big/img_647
+2002/08/20/big/img_375
+2002/08/14/big/img_387
+2002/09/01/big/img_16990
+2002/08/28/big/img_19341
+2003/01/15/big/img_239
+2002/08/20/big/img_528
+2002/08/12/big/img_130
+2002/09/02/big/img_15108
+2003/01/15/big/img_372
+2002/08/16/big/img_678
+2002/08/04/big/img_623
+2002/07/23/big/img_477
+2002/08/28/big/img_19590
+2003/01/17/big/img_978
+2002/09/01/big/img_16692
+2002/07/20/big/img_109
+2002/08/06/big/img_2660
+2003/01/14/big/img_464
+2002/08/09/big/img_618
+2002/07/22/big/img_722
+2002/08/25/big/img_419
+2002/08/03/big/img_314
+2002/08/25/big/img_40
+2002/07/27/big/img_430
+2002/08/10/big/img_569
+2002/08/23/big/img_398
+2002/07/23/big/img_893
+2002/08/16/big/img_261
+2002/08/06/big/img_2668
+2002/07/22/big/img_835
+2002/09/02/big/img_15093
+2003/01/16/big/img_65
+2002/08/21/big/img_448
+2003/01/14/big/img_351
+2003/01/17/big/img_133
+2002/07/28/big/img_493
+2003/01/15/big/img_640
+2002/09/01/big/img_16880
+2002/08/15/big/img_350
+2002/08/20/big/img_624
+2002/08/25/big/img_604
+2002/08/06/big/img_2200
+2002/08/23/big/img_290
+2002/08/13/big/img_1152
+2003/01/14/big/img_251
+2002/08/02/big/img_538
+2002/08/22/big/img_613
+2003/01/13/big/img_351
+2002/08/18/big/img_368
+2002/07/23/big/img_392
+2002/07/25/big/img_198
+2002/07/25/big/img_418
+2002/08/26/big/img_614
+2002/07/23/big/img_405
+2003/01/14/big/img_445
+2002/07/25/big/img_326
+2002/08/10/big/img_734
+2003/01/14/big/img_530
+2002/08/08/big/img_561
+2002/08/29/big/img_18990
+2002/08/10/big/img_576
+2002/07/29/big/img_1494
+2002/07/19/big/img_198
+2002/08/10/big/img_562
+2002/07/22/big/img_901
+2003/01/14/big/img_37
+2002/09/02/big/img_15629
+2003/01/14/big/img_58
+2002/08/01/big/img_1364
+2002/07/27/big/img_636
+2003/01/13/big/img_241
+2002/09/01/big/img_16988
+2003/01/13/big/img_560
+2002/08/09/big/img_533
+2002/07/31/big/img_249
+2003/01/17/big/img_1007
+2002/07/21/big/img_64
+2003/01/13/big/img_537
+2003/01/15/big/img_606
+2002/08/18/big/img_651
+2002/08/24/big/img_405
+2002/07/26/big/img_837
+2002/08/09/big/img_562
+2002/08/01/big/img_1983
+2002/08/03/big/img_514
+2002/07/29/big/img_314
+2002/08/12/big/img_493
+2003/01/14/big/img_121
+2003/01/14/big/img_479
+2002/08/04/big/img_410
+2002/07/22/big/img_607
+2003/01/17/big/img_417
+2002/07/20/big/img_547
+2002/08/13/big/img_396
+2002/08/31/big/img_17538
+2002/08/13/big/img_187
+2002/08/12/big/img_328
+2003/01/14/big/img_569
+2002/07/27/big/img_1081
+2002/08/14/big/img_504
+2002/08/23/big/img_785
+2002/07/26/big/img_339
+2002/08/07/big/img_1156
+2002/08/07/big/img_1456
+2002/08/23/big/img_378
+2002/08/27/big/img_19719
+2002/07/31/big/img_39
+2002/07/31/big/img_883
+2003/01/14/big/img_676
+2002/07/29/big/img_214
+2002/07/26/big/img_669
+2002/07/25/big/img_202
+2002/08/08/big/img_259
+2003/01/17/big/img_943
+2003/01/15/big/img_512
+2002/08/05/big/img_3295
+2002/08/27/big/img_19685
+2002/08/08/big/img_277
+2002/08/30/big/img_18154
+2002/07/22/big/img_663
+2002/08/29/big/img_18914
+2002/07/31/big/img_908
+2002/08/27/big/img_19926
+2003/01/13/big/img_791
+2003/01/15/big/img_827
+2002/08/18/big/img_878
+2002/08/14/big/img_670
+2002/07/20/big/img_182
+2002/08/15/big/img_291
+2002/08/06/big/img_2600
+2002/07/23/big/img_587
+2002/08/14/big/img_577
+2003/01/15/big/img_585
+2002/07/30/big/img_310
+2002/08/03/big/img_658
+2002/08/10/big/img_157
+2002/08/19/big/img_811
+2002/07/29/big/img_1318
+2002/08/04/big/img_104
+2002/07/30/big/img_332
+2002/07/24/big/img_789
+2002/07/29/big/img_516
+2002/07/23/big/img_843
+2002/08/01/big/img_1528
+2002/08/13/big/img_798
+2002/08/07/big/img_1729
+2002/08/28/big/img_19448
+2003/01/16/big/img_95
+2002/08/12/big/img_473
+2002/07/27/big/img_269
+2003/01/16/big/img_621
+2002/07/29/big/img_772
+2002/07/24/big/img_171
+2002/07/19/big/img_429
+2002/08/07/big/img_1933
+2002/08/27/big/img_19629
+2002/08/05/big/img_3688
+2002/08/07/big/img_1691
+2002/07/23/big/img_600
+2002/07/29/big/img_666
+2002/08/25/big/img_566
+2002/08/06/big/img_2659
+2002/08/29/big/img_18929
+2002/08/16/big/img_407
+2002/08/18/big/img_774
+2002/08/19/big/img_249
+2002/08/06/big/img_2427
+2002/08/29/big/img_18899
+2002/08/01/big/img_1818
+2002/07/31/big/img_108
+2002/07/29/big/img_500
+2002/08/11/big/img_115
+2002/07/19/big/img_521
+2002/08/02/big/img_1163
+2002/07/22/big/img_62
+2002/08/13/big/img_466
+2002/08/21/big/img_956
+2002/08/23/big/img_602
+2002/08/20/big/img_858
+2002/07/25/big/img_690
+2002/07/19/big/img_130
+2002/08/04/big/img_874
+2002/07/26/big/img_489
+2002/07/22/big/img_548
+2002/08/10/big/img_191
+2002/07/25/big/img_1051
+2002/08/18/big/img_473
+2002/08/12/big/img_755
+2002/08/18/big/img_413
+2002/08/08/big/img_1044
+2002/08/17/big/img_680
+2002/08/26/big/img_235
+2002/08/20/big/img_330
+2002/08/22/big/img_344
+2002/08/09/big/img_593
+2002/07/31/big/img_1006
+2002/08/14/big/img_337
+2002/08/16/big/img_728
+2002/07/24/big/img_834
+2002/08/04/big/img_552
+2002/09/02/big/img_15213
+2002/07/25/big/img_725
+2002/08/30/big/img_18290
+2003/01/01/big/img_475
+2002/07/27/big/img_1083
+2002/08/29/big/img_18955
+2002/08/31/big/img_17232
+2002/08/08/big/img_480
+2002/08/01/big/img_1311
+2002/07/30/big/img_745
+2002/08/03/big/img_649
+2002/08/12/big/img_193
+2002/07/29/big/img_228
+2002/07/25/big/img_836
+2002/08/20/big/img_400
+2002/07/30/big/img_507
+2002/09/02/big/img_15072
+2002/07/26/big/img_658
+2002/07/28/big/img_503
+2002/08/05/big/img_3814
+2002/08/24/big/img_745
+2003/01/13/big/img_817
+2002/08/08/big/img_579
+2002/07/22/big/img_251
+2003/01/13/big/img_689
+2002/07/25/big/img_407
+2002/08/13/big/img_1050
+2002/08/14/big/img_733
+2002/07/24/big/img_82
+2003/01/17/big/img_288
+2003/01/15/big/img_475
+2002/08/14/big/img_620
+2002/08/21/big/img_167
+2002/07/19/big/img_300
+2002/07/26/big/img_219
+2002/08/01/big/img_1468
+2002/07/23/big/img_260
+2002/08/09/big/img_555
+2002/07/19/big/img_160
+2002/08/02/big/img_1060
+2003/01/14/big/img_149
+2002/08/15/big/img_346
+2002/08/24/big/img_597
+2002/08/22/big/img_502
+2002/08/30/big/img_18228
+2002/07/21/big/img_766
+2003/01/15/big/img_841
+2002/07/24/big/img_516
+2002/08/02/big/img_265
+2002/08/15/big/img_1243
+2003/01/15/big/img_223
+2002/08/04/big/img_236
+2002/07/22/big/img_309
+2002/07/20/big/img_656
+2002/07/31/big/img_412
+2002/09/01/big/img_16462
+2003/01/16/big/img_431
+2002/07/22/big/img_793
+2002/08/15/big/img_877
+2002/07/26/big/img_282
+2002/07/25/big/img_529
+2002/08/24/big/img_613
+2003/01/17/big/img_700
+2002/08/06/big/img_2526
+2002/08/24/big/img_394
+2002/08/21/big/img_521
+2002/08/25/big/img_560
+2002/07/29/big/img_966
+2002/07/25/big/img_448
+2003/01/13/big/img_782
+2002/08/21/big/img_296
+2002/09/01/big/img_16755
+2002/08/05/big/img_3552
+2002/09/02/big/img_15823
+2003/01/14/big/img_193
+2002/07/21/big/img_159
+2002/08/02/big/img_564
+2002/08/16/big/img_300
+2002/07/19/big/img_269
+2002/08/13/big/img_676
+2002/07/28/big/img_57
+2002/08/05/big/img_3318
+2002/07/31/big/img_218
+2002/08/21/big/img_898
+2002/07/29/big/img_109
+2002/07/19/big/img_854
+2002/08/23/big/img_311
+2002/08/14/big/img_318
+2002/07/25/big/img_523
+2002/07/21/big/img_678
+2003/01/17/big/img_690
+2002/08/28/big/img_19503
+2002/08/18/big/img_251
+2002/08/22/big/img_672
+2002/08/20/big/img_663
+2002/08/02/big/img_148
+2002/09/02/big/img_15580
+2002/07/25/big/img_778
+2002/08/14/big/img_565
+2002/08/12/big/img_374
+2002/08/13/big/img_1018
+2002/08/20/big/img_474
+2002/08/25/big/img_33
+2002/08/02/big/img_1190
+2002/08/08/big/img_864
+2002/08/14/big/img_1071
+2002/08/30/big/img_18103
+2002/08/18/big/img_533
+2003/01/16/big/img_650
+2002/07/25/big/img_108
+2002/07/26/big/img_81
+2002/07/27/big/img_543
+2002/07/29/big/img_521
+2003/01/13/big/img_434
+2002/08/26/big/img_674
+2002/08/06/big/img_2932
+2002/08/07/big/img_1262
+2003/01/15/big/img_201
+2003/01/16/big/img_673
+2002/09/02/big/img_15988
+2002/07/29/big/img_1306
+2003/01/14/big/img_1072
+2002/08/30/big/img_18232
+2002/08/05/big/img_3711
+2002/07/23/big/img_775
+2002/08/01/big/img_16
+2003/01/16/big/img_630
+2002/08/22/big/img_695
+2002/08/14/big/img_51
+2002/08/14/big/img_782
+2002/08/24/big/img_742
+2003/01/14/big/img_512
+2003/01/15/big/img_1183
+2003/01/15/big/img_714
+2002/08/01/big/img_2078
+2002/07/31/big/img_682
+2002/09/02/big/img_15687
+2002/07/26/big/img_518
+2002/08/27/big/img_19676
+2002/09/02/big/img_15969
+2002/08/02/big/img_931
+2002/08/25/big/img_508
+2002/08/29/big/img_18616
+2002/07/22/big/img_839
+2002/07/28/big/img_313
+2003/01/14/big/img_155
+2002/08/02/big/img_1105
+2002/08/09/big/img_53
+2002/08/16/big/img_469
+2002/08/15/big/img_502
+2002/08/20/big/img_575
+2002/07/25/big/img_138
+2003/01/16/big/img_579
+2002/07/19/big/img_352
+2003/01/14/big/img_762
+2003/01/01/big/img_588
+2002/08/02/big/img_981
+2002/08/21/big/img_447
+2002/09/01/big/img_16151
+2003/01/14/big/img_769
+2002/08/23/big/img_461
+2002/08/17/big/img_240
+2002/09/02/big/img_15220
+2002/07/19/big/img_408
+2002/09/02/big/img_15496
+2002/07/29/big/img_758
+2002/08/28/big/img_19392
+2002/08/06/big/img_2723
+2002/08/31/big/img_17752
+2002/08/23/big/img_469
+2002/08/13/big/img_515
+2002/09/02/big/img_15551
+2002/08/03/big/img_462
+2002/07/24/big/img_613
+2002/07/22/big/img_61
+2002/08/08/big/img_171
+2002/08/21/big/img_177
+2003/01/14/big/img_105
+2002/08/02/big/img_1017
+2002/08/22/big/img_106
+2002/07/27/big/img_542
+2002/07/21/big/img_665
+2002/07/23/big/img_595
+2002/08/04/big/img_657
+2002/08/29/big/img_19002
+2003/01/15/big/img_550
+2002/08/14/big/img_662
+2002/07/20/big/img_425
+2002/08/30/big/img_18528
+2002/07/26/big/img_611
+2002/07/22/big/img_849
+2002/08/07/big/img_1655
+2002/08/21/big/img_638
+2003/01/17/big/img_732
+2003/01/01/big/img_496
+2002/08/18/big/img_713
+2002/08/08/big/img_109
+2002/07/27/big/img_1008
+2002/07/20/big/img_559
+2002/08/16/big/img_699
+2002/08/31/big/img_17702
+2002/07/31/big/img_1013
+2002/08/01/big/img_2027
+2002/08/02/big/img_1001
+2002/08/03/big/img_210
+2002/08/01/big/img_2087
+2003/01/14/big/img_199
+2002/07/29/big/img_48
+2002/07/19/big/img_727
+2002/08/09/big/img_249
+2002/08/04/big/img_632
+2002/08/22/big/img_620
+2003/01/01/big/img_457
+2002/08/05/big/img_3223
+2002/07/27/big/img_240
+2002/07/25/big/img_797
+2002/08/13/big/img_430
+2002/07/25/big/img_615
+2002/08/12/big/img_28
+2002/07/30/big/img_220
+2002/07/24/big/img_89
+2002/08/21/big/img_357
+2002/08/09/big/img_590
+2003/01/13/big/img_525
+2002/08/17/big/img_818
+2003/01/02/big/img_7
+2002/07/26/big/img_636
+2003/01/13/big/img_1122
+2002/07/23/big/img_810
+2002/08/20/big/img_888
+2002/07/27/big/img_3
+2002/08/15/big/img_451
+2002/09/02/big/img_15787
+2002/07/31/big/img_281
+2002/08/05/big/img_3274
+2002/08/07/big/img_1254
+2002/07/31/big/img_27
+2002/08/01/big/img_1366
+2002/07/30/big/img_182
+2002/08/27/big/img_19690
+2002/07/29/big/img_68
+2002/08/23/big/img_754
+2002/07/30/big/img_540
+2002/08/27/big/img_20063
+2002/08/14/big/img_471
+2002/08/02/big/img_615
+2002/07/30/big/img_186
+2002/08/25/big/img_150
+2002/07/27/big/img_626
+2002/07/20/big/img_225
+2003/01/15/big/img_1252
+2002/07/19/big/img_367
+2003/01/15/big/img_582
+2002/08/09/big/img_572
+2002/08/08/big/img_428
+2003/01/15/big/img_639
+2002/08/28/big/img_19245
+2002/07/24/big/img_321
+2002/08/02/big/img_662
+2002/08/08/big/img_1033
+2003/01/17/big/img_867
+2002/07/22/big/img_652
+2003/01/14/big/img_224
+2002/08/18/big/img_49
+2002/07/26/big/img_46
+2002/08/31/big/img_18021
+2002/07/25/big/img_151
+2002/08/23/big/img_540
+2002/08/25/big/img_693
+2002/07/23/big/img_340
+2002/07/28/big/img_117
+2002/09/02/big/img_15768
+2002/08/26/big/img_562
+2002/07/24/big/img_480
+2003/01/15/big/img_341
+2002/08/10/big/img_783
+2002/08/20/big/img_132
+2003/01/14/big/img_370
+2002/07/20/big/img_720
+2002/08/03/big/img_144
+2002/08/20/big/img_538
+2002/08/01/big/img_1745
+2002/08/11/big/img_683
+2002/08/03/big/img_328
+2002/08/10/big/img_793
+2002/08/14/big/img_689
+2002/08/02/big/img_162
+2003/01/17/big/img_411
+2002/07/31/big/img_361
+2002/08/15/big/img_289
+2002/08/08/big/img_254
+2002/08/15/big/img_996
+2002/08/20/big/img_785
+2002/07/24/big/img_511
+2002/08/06/big/img_2614
+2002/08/29/big/img_18733
+2002/08/17/big/img_78
+2002/07/30/big/img_378
+2002/08/31/big/img_17947
+2002/08/26/big/img_88
+2002/07/30/big/img_558
+2002/08/02/big/img_67
+2003/01/14/big/img_325
+2002/07/29/big/img_1357
+2002/07/19/big/img_391
+2002/07/30/big/img_307
+2003/01/13/big/img_219
+2002/07/24/big/img_807
+2002/08/23/big/img_543
+2002/08/29/big/img_18620
+2002/07/22/big/img_769
+2002/08/26/big/img_503
+2002/07/30/big/img_78
+2002/08/14/big/img_1036
+2002/08/09/big/img_58
+2002/07/24/big/img_616
+2002/08/02/big/img_464
+2002/07/26/big/img_576
+2002/07/22/big/img_273
+2003/01/16/big/img_470
+2002/07/29/big/img_329
+2002/07/30/big/img_1086
+2002/07/31/big/img_353
+2002/09/02/big/img_15275
+2003/01/17/big/img_555
+2002/08/26/big/img_212
+2002/08/01/big/img_1692
+2003/01/15/big/img_600
+2002/07/29/big/img_825
+2002/08/08/big/img_68
+2002/08/10/big/img_719
+2002/07/31/big/img_636
+2002/07/29/big/img_325
+2002/07/21/big/img_515
+2002/07/22/big/img_705
+2003/01/13/big/img_818
+2002/08/09/big/img_486
+2002/08/22/big/img_141
+2002/07/22/big/img_303
+2002/08/09/big/img_393
+2002/07/29/big/img_963
+2002/08/02/big/img_1215
+2002/08/19/big/img_674
+2002/08/12/big/img_690
+2002/08/21/big/img_637
+2002/08/21/big/img_841
+2002/08/24/big/img_71
+2002/07/25/big/img_596
+2002/07/24/big/img_864
+2002/08/18/big/img_293
+2003/01/14/big/img_657
+2002/08/15/big/img_411
+2002/08/16/big/img_348
+2002/08/05/big/img_3157
+2002/07/20/big/img_663
+2003/01/13/big/img_654
+2003/01/16/big/img_433
+2002/08/30/big/img_18200
+2002/08/12/big/img_226
+2003/01/16/big/img_491
+2002/08/08/big/img_666
+2002/07/19/big/img_576
+2003/01/15/big/img_776
+2003/01/16/big/img_899
+2002/07/19/big/img_397
+2002/08/14/big/img_44
+2003/01/15/big/img_762
+2002/08/02/big/img_982
+2002/09/02/big/img_15234
+2002/08/17/big/img_556
+2002/08/21/big/img_410
+2002/08/21/big/img_386
+2002/07/19/big/img_690
+2002/08/05/big/img_3052
+2002/08/14/big/img_219
+2002/08/16/big/img_273
+2003/01/15/big/img_752
+2002/08/08/big/img_184
+2002/07/31/big/img_743
+2002/08/23/big/img_338
+2003/01/14/big/img_1055
+2002/08/05/big/img_3405
+2003/01/15/big/img_17
+2002/08/03/big/img_141
+2002/08/14/big/img_549
+2002/07/27/big/img_1034
+2002/07/31/big/img_932
+2002/08/30/big/img_18487
+2002/09/02/big/img_15814
+2002/08/01/big/img_2086
+2002/09/01/big/img_16535
+2002/07/22/big/img_500
+2003/01/13/big/img_400
+2002/08/25/big/img_607
+2002/08/30/big/img_18384
+2003/01/14/big/img_951
+2002/08/13/big/img_1150
+2002/08/08/big/img_1022
+2002/08/10/big/img_428
+2002/08/28/big/img_19242
+2002/08/05/big/img_3098
+2002/07/23/big/img_400
+2002/08/26/big/img_365
+2002/07/20/big/img_318
+2002/08/13/big/img_740
+2003/01/16/big/img_37
+2002/08/26/big/img_274
+2002/08/02/big/img_205
+2002/08/21/big/img_695
+2002/08/06/big/img_2289
+2002/08/20/big/img_794
+2002/08/18/big/img_438
+2002/08/07/big/img_1380
+2002/08/02/big/img_737
+2002/08/07/big/img_1651
+2002/08/15/big/img_1238
+2002/08/01/big/img_1681
+2002/08/06/big/img_3017
+2002/07/23/big/img_706
+2002/07/31/big/img_392
+2002/08/09/big/img_539
+2002/07/29/big/img_835
+2002/08/26/big/img_723
+2002/08/28/big/img_19235
+2003/01/16/big/img_353
+2002/08/10/big/img_150
+2002/08/29/big/img_19025
+2002/08/21/big/img_310
+2002/08/10/big/img_823
+2002/07/26/big/img_981
+2002/08/11/big/img_288
+2002/08/19/big/img_534
+2002/08/21/big/img_300
+2002/07/31/big/img_49
+2002/07/30/big/img_469
+2002/08/28/big/img_19197
+2002/08/25/big/img_205
+2002/08/10/big/img_390
+2002/08/23/big/img_291
+2002/08/26/big/img_230
+2002/08/18/big/img_76
+2002/07/23/big/img_409
+2002/08/14/big/img_1053
+2003/01/14/big/img_291
+2002/08/10/big/img_503
+2002/08/27/big/img_19928
+2002/08/03/big/img_563
+2002/08/17/big/img_250
+2002/08/06/big/img_2381
+2002/08/17/big/img_948
+2002/08/06/big/img_2710
+2002/07/22/big/img_696
+2002/07/31/big/img_670
+2002/08/12/big/img_594
+2002/07/29/big/img_624
+2003/01/17/big/img_934
+2002/08/03/big/img_584
+2002/08/22/big/img_1003
+2002/08/05/big/img_3396
+2003/01/13/big/img_570
+2002/08/02/big/img_219
+2002/09/02/big/img_15774
+2002/08/16/big/img_818
+2002/08/23/big/img_402
+2003/01/14/big/img_552
+2002/07/29/big/img_71
+2002/08/05/big/img_3592
+2002/08/16/big/img_80
+2002/07/27/big/img_672
+2003/01/13/big/img_470
+2003/01/16/big/img_702
+2002/09/01/big/img_16130
+2002/08/08/big/img_240
+2002/09/01/big/img_16338
+2002/07/26/big/img_312
+2003/01/14/big/img_538
+2002/07/20/big/img_695
+2002/08/30/big/img_18098
+2002/08/25/big/img_259
+2002/08/16/big/img_1042
+2002/08/09/big/img_837
+2002/08/31/big/img_17760
+2002/07/31/big/img_14
+2002/08/09/big/img_361
+2003/01/16/big/img_107
+2002/08/14/big/img_124
+2002/07/19/big/img_463
+2003/01/15/big/img_275
+2002/07/25/big/img_1151
+2002/07/29/big/img_1501
+2002/08/27/big/img_19889
+2002/08/29/big/img_18603
+2003/01/17/big/img_601
+2002/08/25/big/img_355
+2002/08/08/big/img_297
+2002/08/20/big/img_290
+2002/07/31/big/img_195
+2003/01/01/big/img_336
+2002/08/18/big/img_369
+2002/07/25/big/img_621
+2002/08/11/big/img_508
+2003/01/14/big/img_458
+2003/01/15/big/img_795
+2002/08/12/big/img_498
+2002/08/01/big/img_1734
+2002/08/02/big/img_246
+2002/08/16/big/img_565
+2002/08/11/big/img_475
+2002/08/22/big/img_408
+2002/07/28/big/img_78
+2002/07/21/big/img_81
+2003/01/14/big/img_697
+2002/08/14/big/img_661
+2002/08/15/big/img_507
+2002/08/19/big/img_55
+2002/07/22/big/img_152
+2003/01/14/big/img_470
+2002/08/03/big/img_379
+2002/08/22/big/img_506
+2003/01/16/big/img_966
+2002/08/18/big/img_698
+2002/08/24/big/img_528
+2002/08/23/big/img_10
+2002/08/01/big/img_1655
+2002/08/22/big/img_953
+2002/07/19/big/img_630
+2002/07/22/big/img_889
+2002/08/16/big/img_351
+2003/01/16/big/img_83
+2002/07/19/big/img_805
+2002/08/14/big/img_704
+2002/07/19/big/img_389
+2002/08/31/big/img_17765
+2002/07/29/big/img_606
+2003/01/17/big/img_939
+2002/09/02/big/img_15081
+2002/08/21/big/img_181
+2002/07/29/big/img_1321
+2002/07/21/big/img_497
+2002/07/20/big/img_539
+2002/08/24/big/img_119
+2002/08/01/big/img_1281
+2002/07/26/big/img_207
+2002/07/26/big/img_432
+2002/07/27/big/img_1006
+2002/08/05/big/img_3087
+2002/08/14/big/img_252
+2002/08/14/big/img_798
+2002/07/24/big/img_538
+2002/09/02/big/img_15507
+2002/08/08/big/img_901
+2003/01/14/big/img_557
+2002/08/07/big/img_1819
+2002/08/04/big/img_470
+2002/08/01/big/img_1504
+2002/08/16/big/img_1070
+2002/08/16/big/img_372
+2002/08/23/big/img_416
+2002/08/30/big/img_18208
+2002/08/01/big/img_2043
+2002/07/22/big/img_385
+2002/08/22/big/img_466
+2002/08/21/big/img_869
+2002/08/28/big/img_19429
+2002/08/02/big/img_770
+2002/07/23/big/img_433
+2003/01/14/big/img_13
+2002/07/27/big/img_953
+2002/09/02/big/img_15728
+2002/08/01/big/img_1361
+2002/08/29/big/img_18897
+2002/08/26/big/img_534
+2002/08/11/big/img_121
+2002/08/26/big/img_20130
+2002/07/31/big/img_363
+2002/08/13/big/img_978
+2002/07/25/big/img_835
+2002/08/02/big/img_906
+2003/01/14/big/img_548
+2002/07/30/big/img_80
+2002/07/26/big/img_982
+2003/01/16/big/img_99
+2002/08/19/big/img_362
+2002/08/24/big/img_376
+2002/08/07/big/img_1264
+2002/07/27/big/img_938
+2003/01/17/big/img_535
+2002/07/26/big/img_457
+2002/08/08/big/img_848
+2003/01/15/big/img_859
+2003/01/15/big/img_622
+2002/07/30/big/img_403
+2002/07/29/big/img_217
+2002/07/26/big/img_891
+2002/07/24/big/img_70
+2002/08/25/big/img_619
+2002/08/05/big/img_3375
+2002/08/01/big/img_2160
+2002/08/06/big/img_2227
+2003/01/14/big/img_117
+2002/08/14/big/img_227
+2002/08/13/big/img_565
+2002/08/19/big/img_625
+2002/08/03/big/img_812
+2002/07/24/big/img_41
+2002/08/16/big/img_235
+2002/07/29/big/img_759
+2002/07/21/big/img_433
+2002/07/29/big/img_190
+2003/01/16/big/img_435
+2003/01/13/big/img_708
+2002/07/30/big/img_57
+2002/08/22/big/img_162
+2003/01/01/big/img_558
+2003/01/15/big/img_604
+2002/08/16/big/img_935
+2002/08/20/big/img_394
+2002/07/28/big/img_465
+2002/09/02/big/img_15534
+2002/08/16/big/img_87
+2002/07/22/big/img_469
+2002/08/12/big/img_245
+2003/01/13/big/img_236
+2002/08/06/big/img_2736
+2002/08/03/big/img_348
+2003/01/14/big/img_218
+2002/07/26/big/img_232
+2003/01/15/big/img_244
+2002/07/25/big/img_1121
+2002/08/01/big/img_1484
+2002/07/26/big/img_541
+2002/08/07/big/img_1244
+2002/07/31/big/img_3
+2002/08/30/big/img_18437
+2002/08/29/big/img_19094
+2002/08/01/big/img_1355
+2002/08/19/big/img_338
+2002/07/19/big/img_255
+2002/07/21/big/img_76
+2002/08/25/big/img_199
+2002/08/12/big/img_740
+2002/07/30/big/img_852
+2002/08/15/big/img_599
+2002/08/23/big/img_254
+2002/08/19/big/img_125
+2002/07/24/big/img_2
+2002/08/04/big/img_145
+2002/08/05/big/img_3137
+2002/07/28/big/img_463
+2003/01/14/big/img_801
+2002/07/23/big/img_366
+2002/08/26/big/img_600
+2002/08/26/big/img_649
+2002/09/02/big/img_15849
+2002/07/26/big/img_248
+2003/01/13/big/img_200
+2002/08/07/big/img_1794
+2002/08/31/big/img_17270
+2002/08/23/big/img_608
+2003/01/13/big/img_837
+2002/08/23/big/img_581
+2002/08/20/big/img_754
+2002/08/18/big/img_183
+2002/08/20/big/img_328
+2002/07/22/big/img_494
+2002/07/29/big/img_399
+2002/08/28/big/img_19284
+2002/08/08/big/img_566
+2002/07/25/big/img_376
+2002/07/23/big/img_138
+2002/07/25/big/img_435
+2002/08/17/big/img_685
+2002/07/19/big/img_90
+2002/07/20/big/img_716
+2002/08/31/big/img_17458
+2002/08/26/big/img_461
+2002/07/25/big/img_355
+2002/08/06/big/img_2152
+2002/07/27/big/img_932
+2002/07/23/big/img_232
+2002/08/08/big/img_1020
+2002/07/31/big/img_366
+2002/08/06/big/img_2667
+2002/08/21/big/img_465
+2002/08/15/big/img_305
+2002/08/02/big/img_247
+2002/07/28/big/img_46
+2002/08/27/big/img_19922
+2002/08/23/big/img_643
+2003/01/13/big/img_624
+2002/08/23/big/img_625
+2002/08/05/big/img_3787
+2003/01/13/big/img_627
+2002/09/01/big/img_16381
+2002/08/05/big/img_3668
+2002/07/21/big/img_535
+2002/08/27/big/img_19680
+2002/07/22/big/img_413
+2002/07/29/big/img_481
+2003/01/15/big/img_496
+2002/07/23/big/img_701
+2002/08/29/big/img_18670
+2002/07/28/big/img_319
+2003/01/14/big/img_517
+2002/07/26/big/img_256
+2003/01/16/big/img_593
+2002/07/30/big/img_956
+2002/07/30/big/img_667
+2002/07/25/big/img_100
+2002/08/11/big/img_570
+2002/07/26/big/img_745
+2002/08/04/big/img_834
+2002/08/25/big/img_521
+2002/08/01/big/img_2148
+2002/09/02/big/img_15183
+2002/08/22/big/img_514
+2002/08/23/big/img_477
+2002/07/23/big/img_336
+2002/07/26/big/img_481
+2002/08/20/big/img_409
+2002/07/23/big/img_918
+2002/08/09/big/img_474
+2002/08/02/big/img_929
+2002/08/31/big/img_17932
+2002/08/19/big/img_161
+2002/08/09/big/img_667
+2002/07/31/big/img_805
+2002/09/02/big/img_15678
+2002/08/31/big/img_17509
+2002/08/29/big/img_18998
+2002/07/23/big/img_301
+2002/08/07/big/img_1612
+2002/08/06/big/img_2472
+2002/07/23/big/img_466
+2002/08/27/big/img_19634
+2003/01/16/big/img_16
+2002/08/14/big/img_193
+2002/08/21/big/img_340
+2002/08/27/big/img_19799
+2002/08/01/big/img_1345
+2002/08/07/big/img_1448
+2002/08/11/big/img_324
+2003/01/16/big/img_754
+2002/08/13/big/img_418
+2003/01/16/big/img_544
+2002/08/19/big/img_135
+2002/08/10/big/img_455
+2002/08/10/big/img_693
+2002/08/31/big/img_17967
+2002/08/28/big/img_19229
+2002/08/04/big/img_811
+2002/09/01/big/img_16225
+2003/01/16/big/img_428
+2002/09/02/big/img_15295
+2002/07/26/big/img_108
+2002/07/21/big/img_477
+2002/08/07/big/img_1354
+2002/08/23/big/img_246
+2002/08/16/big/img_652
+2002/07/27/big/img_553
+2002/07/31/big/img_346
+2002/08/04/big/img_537
+2002/08/08/big/img_498
+2002/08/29/big/img_18956
+2003/01/13/big/img_922
+2002/08/31/big/img_17425
+2002/07/26/big/img_438
+2002/08/19/big/img_185
+2003/01/16/big/img_33
+2002/08/10/big/img_252
+2002/07/29/big/img_598
+2002/08/27/big/img_19820
+2002/08/06/big/img_2664
+2002/08/20/big/img_705
+2003/01/14/big/img_816
+2002/08/03/big/img_552
+2002/07/25/big/img_561
+2002/07/25/big/img_934
+2002/08/01/big/img_1893
+2003/01/14/big/img_746
+2003/01/16/big/img_519
+2002/08/03/big/img_681
+2002/07/24/big/img_808
+2002/08/14/big/img_803
+2002/08/25/big/img_155
+2002/07/30/big/img_1107
+2002/08/29/big/img_18882
+2003/01/15/big/img_598
+2002/08/19/big/img_122
+2002/07/30/big/img_428
+2002/07/24/big/img_684
+2002/08/22/big/img_192
+2002/08/22/big/img_543
+2002/08/07/big/img_1318
+2002/08/18/big/img_25
+2002/07/26/big/img_583
+2002/07/20/big/img_464
+2002/08/19/big/img_664
+2002/08/24/big/img_861
+2002/09/01/big/img_16136
+2002/08/22/big/img_400
+2002/08/12/big/img_445
+2003/01/14/big/img_174
+2002/08/27/big/img_19677
+2002/08/31/big/img_17214
+2002/08/30/big/img_18175
+2003/01/17/big/img_402
+2002/08/06/big/img_2396
+2002/08/18/big/img_448
+2002/08/21/big/img_165
+2002/08/31/big/img_17609
+2003/01/01/big/img_151
+2002/08/26/big/img_372
+2002/09/02/big/img_15994
+2002/07/26/big/img_660
+2002/09/02/big/img_15197
+2002/07/29/big/img_258
+2002/08/30/big/img_18525
+2003/01/13/big/img_368
+2002/07/29/big/img_1538
+2002/07/21/big/img_787
+2002/08/18/big/img_152
+2002/08/06/big/img_2379
+2003/01/17/big/img_864
+2002/08/27/big/img_19998
+2002/08/01/big/img_1634
+2002/07/25/big/img_414
+2002/08/22/big/img_627
+2002/08/07/big/img_1669
+2002/08/16/big/img_1052
+2002/08/31/big/img_17796
+2002/08/18/big/img_199
+2002/09/02/big/img_15147
+2002/08/09/big/img_460
+2002/08/14/big/img_581
+2002/08/30/big/img_18286
+2002/07/26/big/img_337
+2002/08/18/big/img_589
+2003/01/14/big/img_866
+2002/07/20/big/img_624
+2002/08/01/big/img_1801
+2002/07/24/big/img_683
+2002/08/09/big/img_725
+2003/01/14/big/img_34
+2002/07/30/big/img_144
+2002/07/30/big/img_706
+2002/08/08/big/img_394
+2002/08/19/big/img_619
+2002/08/06/big/img_2703
+2002/08/29/big/img_19034
+2002/07/24/big/img_67
+2002/08/27/big/img_19841
+2002/08/19/big/img_427
+2003/01/14/big/img_333
+2002/09/01/big/img_16406
+2002/07/19/big/img_882
+2002/08/17/big/img_238
+2003/01/14/big/img_739
+2002/07/22/big/img_151
+2002/08/21/big/img_743
+2002/07/25/big/img_1048
+2002/07/30/big/img_395
+2003/01/13/big/img_584
+2002/08/13/big/img_742
+2002/08/13/big/img_1168
+2003/01/14/big/img_147
+2002/07/26/big/img_803
+2002/08/05/big/img_3298
+2002/08/07/big/img_1451
+2002/08/16/big/img_424
+2002/07/29/big/img_1069
+2002/09/01/big/img_16735
+2002/07/21/big/img_637
+2003/01/14/big/img_585
+2002/08/02/big/img_358
+2003/01/13/big/img_358
+2002/08/14/big/img_198
+2002/08/17/big/img_935
+2002/08/04/big/img_42
+2002/08/30/big/img_18245
+2002/07/25/big/img_158
+2002/08/22/big/img_744
+2002/08/06/big/img_2291
+2002/08/05/big/img_3044
+2002/07/30/big/img_272
+2002/08/23/big/img_641
+2002/07/24/big/img_797
+2002/07/30/big/img_392
+2003/01/14/big/img_447
+2002/07/31/big/img_898
+2002/08/06/big/img_2812
+2002/08/13/big/img_564
+2002/07/22/big/img_43
+2002/07/26/big/img_634
+2002/07/19/big/img_843
+2002/08/26/big/img_58
+2002/07/21/big/img_375
+2002/08/25/big/img_729
+2002/07/19/big/img_561
+2003/01/15/big/img_884
+2002/07/25/big/img_891
+2002/08/09/big/img_558
+2002/08/26/big/img_587
+2002/08/13/big/img_1146
+2002/09/02/big/img_15153
+2002/07/26/big/img_316
+2002/08/01/big/img_1940
+2002/08/26/big/img_90
+2003/01/13/big/img_347
+2002/07/25/big/img_520
+2002/08/29/big/img_18718
+2002/08/28/big/img_19219
+2002/08/13/big/img_375
+2002/07/20/big/img_719
+2002/08/31/big/img_17431
+2002/07/28/big/img_192
+2002/08/26/big/img_259
+2002/08/18/big/img_484
+2002/07/29/big/img_580
+2002/07/26/big/img_84
+2002/08/02/big/img_302
+2002/08/31/big/img_17007
+2003/01/15/big/img_543
+2002/09/01/big/img_16488
+2002/08/22/big/img_798
+2002/07/30/big/img_383
+2002/08/04/big/img_668
+2002/08/13/big/img_156
+2002/08/07/big/img_1353
+2002/07/25/big/img_281
+2003/01/14/big/img_587
+2003/01/15/big/img_524
+2002/08/19/big/img_726
+2002/08/21/big/img_709
+2002/08/26/big/img_465
+2002/07/31/big/img_658
+2002/08/28/big/img_19148
+2002/07/23/big/img_423
+2002/08/16/big/img_758
+2002/08/22/big/img_523
+2002/08/16/big/img_591
+2002/08/23/big/img_845
+2002/07/26/big/img_678
+2002/08/09/big/img_806
+2002/08/06/big/img_2369
+2002/07/29/big/img_457
+2002/07/19/big/img_278
+2002/08/30/big/img_18107
+2002/07/26/big/img_444
+2002/08/20/big/img_278
+2002/08/26/big/img_92
+2002/08/26/big/img_257
+2002/07/25/big/img_266
+2002/08/05/big/img_3829
+2002/07/26/big/img_757
+2002/07/29/big/img_1536
+2002/08/09/big/img_472
+2003/01/17/big/img_480
+2002/08/28/big/img_19355
+2002/07/26/big/img_97
+2002/08/06/big/img_2503
+2002/07/19/big/img_254
+2002/08/01/big/img_1470
+2002/08/21/big/img_42
+2002/08/20/big/img_217
+2002/08/06/big/img_2459
+2002/07/19/big/img_552
+2002/08/13/big/img_717
+2002/08/12/big/img_586
+2002/08/20/big/img_411
+2003/01/13/big/img_768
+2002/08/07/big/img_1747
+2002/08/15/big/img_385
+2002/08/01/big/img_1648
+2002/08/15/big/img_311
+2002/08/21/big/img_95
+2002/08/09/big/img_108
+2002/08/21/big/img_398
+2002/08/17/big/img_340
+2002/08/14/big/img_474
+2002/08/13/big/img_294
+2002/08/24/big/img_840
+2002/08/09/big/img_808
+2002/08/23/big/img_491
+2002/07/28/big/img_33
+2003/01/13/big/img_664
+2002/08/02/big/img_261
+2002/08/09/big/img_591
+2002/07/26/big/img_309
+2003/01/14/big/img_372
+2002/08/19/big/img_581
+2002/08/19/big/img_168
+2002/08/26/big/img_422
+2002/07/24/big/img_106
+2002/08/01/big/img_1936
+2002/08/05/big/img_3764
+2002/08/21/big/img_266
+2002/08/31/big/img_17968
+2002/08/01/big/img_1941
+2002/08/15/big/img_550
+2002/08/14/big/img_13
+2002/07/30/big/img_171
+2003/01/13/big/img_490
+2002/07/25/big/img_427
+2002/07/19/big/img_770
+2002/08/12/big/img_759
+2003/01/15/big/img_1360
+2002/08/05/big/img_3692
+2003/01/16/big/img_30
+2002/07/25/big/img_1026
+2002/07/22/big/img_288
+2002/08/29/big/img_18801
+2002/07/24/big/img_793
+2002/08/13/big/img_178
+2002/08/06/big/img_2322
+2003/01/14/big/img_560
+2002/08/18/big/img_408
+2003/01/16/big/img_915
+2003/01/16/big/img_679
+2002/08/07/big/img_1552
+2002/08/29/big/img_19050
+2002/08/01/big/img_2172
+2002/07/31/big/img_30
+2002/07/30/big/img_1019
+2002/07/30/big/img_587
+2003/01/13/big/img_773
+2002/07/30/big/img_410
+2002/07/28/big/img_65
+2002/08/05/big/img_3138
+2002/07/23/big/img_541
+2002/08/22/big/img_963
+2002/07/27/big/img_657
+2002/07/30/big/img_1051
+2003/01/16/big/img_150
+2002/07/31/big/img_519
+2002/08/01/big/img_1961
+2002/08/05/big/img_3752
+2002/07/23/big/img_631
+2003/01/14/big/img_237
+2002/07/28/big/img_21
+2002/07/22/big/img_813
+2002/08/05/big/img_3563
+2003/01/17/big/img_620
+2002/07/19/big/img_523
+2002/07/30/big/img_904
+2002/08/29/big/img_18642
+2002/08/11/big/img_492
+2002/08/01/big/img_2130
+2002/07/25/big/img_618
+2002/08/17/big/img_305
+2003/01/16/big/img_520
+2002/07/26/big/img_495
+2002/08/17/big/img_164
+2002/08/03/big/img_440
+2002/07/24/big/img_441
+2002/08/06/big/img_2146
+2002/08/11/big/img_558
+2002/08/02/big/img_545
+2002/08/31/big/img_18090
+2003/01/01/big/img_136
+2002/07/25/big/img_1099
+2003/01/13/big/img_728
+2003/01/16/big/img_197
+2002/07/26/big/img_651
+2002/08/11/big/img_676
+2003/01/15/big/img_10
+2002/08/21/big/img_250
+2002/08/14/big/img_325
+2002/08/04/big/img_390
+2002/07/24/big/img_554
+2003/01/16/big/img_333
+2002/07/31/big/img_922
+2002/09/02/big/img_15586
+2003/01/16/big/img_184
+2002/07/22/big/img_766
+2002/07/21/big/img_608
+2002/08/07/big/img_1578
+2002/08/17/big/img_961
+2002/07/27/big/img_324
+2002/08/05/big/img_3765
+2002/08/23/big/img_462
+2003/01/16/big/img_382
+2002/08/27/big/img_19838
+2002/08/01/big/img_1505
+2002/08/21/big/img_662
+2002/08/14/big/img_605
+2002/08/19/big/img_816
+2002/07/29/big/img_136
+2002/08/20/big/img_719
+2002/08/06/big/img_2826
+2002/08/10/big/img_630
+2003/01/17/big/img_973
+2002/08/14/big/img_116
+2002/08/02/big/img_666
+2002/08/21/big/img_710
+2002/08/05/big/img_55
+2002/07/31/big/img_229
+2002/08/01/big/img_1549
+2002/07/23/big/img_432
+2002/07/21/big/img_430
+2002/08/21/big/img_549
+2002/08/08/big/img_985
+2002/07/20/big/img_610
+2002/07/23/big/img_978
+2002/08/23/big/img_219
+2002/07/25/big/img_175
+2003/01/15/big/img_230
+2002/08/23/big/img_385
+2002/07/31/big/img_879
+2002/08/12/big/img_495
+2002/08/22/big/img_499
+2002/08/30/big/img_18322
+2002/08/15/big/img_795
+2002/08/13/big/img_835
+2003/01/17/big/img_930
+2002/07/30/big/img_873
+2002/08/11/big/img_257
+2002/07/31/big/img_593
+2002/08/21/big/img_916
+2003/01/13/big/img_814
+2002/07/25/big/img_722
+2002/08/16/big/img_379
+2002/07/31/big/img_497
+2002/07/22/big/img_602
+2002/08/21/big/img_642
+2002/08/21/big/img_614
+2002/08/23/big/img_482
+2002/07/29/big/img_603
+2002/08/13/big/img_705
+2002/07/23/big/img_833
+2003/01/14/big/img_511
+2002/07/24/big/img_376
+2002/08/17/big/img_1030
+2002/08/05/big/img_3576
+2002/08/16/big/img_540
+2002/07/22/big/img_630
+2002/08/10/big/img_180
+2002/08/14/big/img_905
+2002/08/29/big/img_18777
+2002/08/22/big/img_693
+2003/01/16/big/img_933
+2002/08/20/big/img_555
+2002/08/15/big/img_549
+2003/01/14/big/img_830
+2003/01/16/big/img_64
+2002/08/27/big/img_19670
+2002/08/22/big/img_729
+2002/07/27/big/img_981
+2002/08/09/big/img_458
+2003/01/17/big/img_884
+2002/07/25/big/img_639
+2002/08/31/big/img_18008
+2002/08/22/big/img_249
+2002/08/17/big/img_971
+2002/08/04/big/img_308
+2002/07/28/big/img_362
+2002/08/12/big/img_142
+2002/08/26/big/img_61
+2002/08/14/big/img_422
+2002/07/19/big/img_607
+2003/01/15/big/img_717
+2002/08/01/big/img_1475
+2002/08/29/big/img_19061
+2003/01/01/big/img_346
+2002/07/20/big/img_315
+2003/01/15/big/img_756
+2002/08/15/big/img_879
+2002/08/08/big/img_615
+2003/01/13/big/img_431
+2002/08/05/big/img_3233
+2002/08/24/big/img_526
+2003/01/13/big/img_717
+2002/09/01/big/img_16408
+2002/07/22/big/img_217
+2002/07/31/big/img_960
+2002/08/21/big/img_610
+2002/08/05/big/img_3753
+2002/08/03/big/img_151
+2002/08/21/big/img_267
+2002/08/01/big/img_2175
+2002/08/04/big/img_556
+2002/08/21/big/img_527
+2002/09/02/big/img_15800
+2002/07/27/big/img_156
+2002/07/20/big/img_590
+2002/08/15/big/img_700
+2002/08/08/big/img_444
+2002/07/25/big/img_94
+2002/07/24/big/img_778
+2002/08/14/big/img_694
+2002/07/20/big/img_666
+2002/08/02/big/img_200
+2002/08/02/big/img_578
+2003/01/17/big/img_332
+2002/09/01/big/img_16352
+2002/08/27/big/img_19668
+2002/07/23/big/img_823
+2002/08/13/big/img_431
+2003/01/16/big/img_463
+2002/08/27/big/img_19711
+2002/08/23/big/img_154
+2002/07/31/big/img_360
+2002/08/23/big/img_555
+2002/08/10/big/img_561
+2003/01/14/big/img_550
+2002/08/07/big/img_1370
+2002/07/30/big/img_1184
+2002/08/01/big/img_1445
+2002/08/23/big/img_22
+2002/07/30/big/img_606
+2003/01/17/big/img_271
+2002/08/31/big/img_17316
+2002/08/16/big/img_973
+2002/07/26/big/img_77
+2002/07/20/big/img_788
+2002/08/06/big/img_2426
+2002/08/07/big/img_1498
+2002/08/16/big/img_358
+2002/08/06/big/img_2851
+2002/08/12/big/img_359
+2002/08/01/big/img_1521
+2002/08/02/big/img_709
+2002/08/20/big/img_935
+2002/08/12/big/img_188
+2002/08/24/big/img_411
+2002/08/22/big/img_680
+2002/08/06/big/img_2480
+2002/07/20/big/img_627
+2002/07/30/big/img_214
+2002/07/25/big/img_354
+2002/08/02/big/img_636
+2003/01/15/big/img_661
+2002/08/07/big/img_1327
+2002/08/01/big/img_2108
+2002/08/31/big/img_17919
+2002/08/29/big/img_18768
+2002/08/05/big/img_3840
+2002/07/26/big/img_242
+2003/01/14/big/img_451
+2002/08/20/big/img_923
+2002/08/27/big/img_19908
+2002/08/16/big/img_282
+2002/08/19/big/img_440
+2003/01/01/big/img_230
+2002/08/08/big/img_212
+2002/07/20/big/img_443
+2002/08/25/big/img_635
+2003/01/13/big/img_1169
+2002/07/26/big/img_998
+2002/08/15/big/img_995
+2002/08/06/big/img_3002
+2002/07/29/big/img_460
+2003/01/14/big/img_925
+2002/07/23/big/img_539
+2002/08/16/big/img_694
+2003/01/13/big/img_459
+2002/07/23/big/img_249
+2002/08/20/big/img_539
+2002/08/04/big/img_186
+2002/08/26/big/img_264
+2002/07/22/big/img_704
+2002/08/25/big/img_277
+2002/08/22/big/img_988
+2002/07/29/big/img_504
+2002/08/05/big/img_3600
+2002/08/30/big/img_18380
+2003/01/14/big/img_937
+2002/08/21/big/img_254
+2002/08/10/big/img_130
+2002/08/20/big/img_339
+2003/01/14/big/img_428
+2002/08/20/big/img_889
+2002/08/31/big/img_17637
+2002/07/26/big/img_644
+2002/09/01/big/img_16776
+2002/08/06/big/img_2239
+2002/08/06/big/img_2646
+2003/01/13/big/img_491
+2002/08/10/big/img_579
+2002/08/21/big/img_713
+2002/08/22/big/img_482
+2002/07/22/big/img_167
+2002/07/24/big/img_539
+2002/08/14/big/img_721
+2002/07/25/big/img_389
+2002/09/01/big/img_16591
+2002/08/13/big/img_543
+2003/01/14/big/img_432
+2002/08/09/big/img_287
+2002/07/26/big/img_126
+2002/08/23/big/img_412
+2002/08/15/big/img_1034
+2002/08/28/big/img_19485
+2002/07/31/big/img_236
+2002/07/30/big/img_523
+2002/07/19/big/img_141
+2003/01/17/big/img_957
+2002/08/04/big/img_81
+2002/07/25/big/img_206
+2002/08/15/big/img_716
+2002/08/13/big/img_403
+2002/08/15/big/img_685
+2002/07/26/big/img_884
+2002/07/19/big/img_499
+2002/07/23/big/img_772
+2002/07/27/big/img_752
+2003/01/14/big/img_493
+2002/08/25/big/img_664
+2002/07/31/big/img_334
+2002/08/26/big/img_678
+2002/09/01/big/img_16541
+2003/01/14/big/img_347
+2002/07/23/big/img_187
+2002/07/30/big/img_1163
+2002/08/05/big/img_35
+2002/08/22/big/img_944
+2002/08/07/big/img_1239
+2002/07/29/big/img_1215
+2002/08/03/big/img_312
+2002/08/05/big/img_3523
+2002/07/29/big/img_218
+2002/08/13/big/img_672
+2002/08/16/big/img_205
+2002/08/17/big/img_594
+2002/07/29/big/img_1411
+2002/07/30/big/img_942
+2003/01/16/big/img_312
+2002/08/08/big/img_312
+2002/07/25/big/img_15
+2002/08/09/big/img_839
+2002/08/01/big/img_2069
+2002/08/31/big/img_17512
+2002/08/01/big/img_3
+2002/07/31/big/img_320
+2003/01/15/big/img_1265
+2002/08/14/big/img_563
+2002/07/31/big/img_167
+2002/08/20/big/img_374
+2002/08/13/big/img_406
+2002/08/08/big/img_625
+2002/08/02/big/img_314
+2002/08/27/big/img_19964
+2002/09/01/big/img_16670
+2002/07/31/big/img_599
+2002/08/29/big/img_18906
+2002/07/24/big/img_373
+2002/07/26/big/img_513
+2002/09/02/big/img_15497
+2002/08/19/big/img_117
+2003/01/01/big/img_158
+2002/08/24/big/img_178
+2003/01/13/big/img_935
+2002/08/13/big/img_609
+2002/08/30/big/img_18341
+2002/08/25/big/img_674
+2003/01/13/big/img_209
+2002/08/13/big/img_258
+2002/08/05/big/img_3543
+2002/08/07/big/img_1970
+2002/08/06/big/img_3004
+2003/01/17/big/img_487
+2002/08/24/big/img_873
+2002/08/29/big/img_18730
+2002/08/09/big/img_375
+2003/01/16/big/img_751
+2002/08/02/big/img_603
+2002/08/19/big/img_325
+2002/09/01/big/img_16420
+2002/08/05/big/img_3633
+2002/08/21/big/img_516
+2002/07/19/big/img_501
+2002/07/26/big/img_688
+2002/07/24/big/img_256
+2002/07/25/big/img_438
+2002/07/31/big/img_1017
+2002/08/22/big/img_512
+2002/07/21/big/img_543
+2002/08/08/big/img_223
+2002/08/19/big/img_189
+2002/08/12/big/img_630
+2002/07/30/big/img_958
+2002/07/28/big/img_208
+2002/08/31/big/img_17691
+2002/07/22/big/img_542
+2002/07/19/big/img_741
+2002/07/19/big/img_158
+2002/08/15/big/img_399
+2002/08/01/big/img_2159
+2002/08/14/big/img_455
+2002/08/17/big/img_1011
+2002/08/26/big/img_744
+2002/08/12/big/img_624
+2003/01/17/big/img_821
+2002/08/16/big/img_980
+2002/07/28/big/img_281
+2002/07/25/big/img_171
+2002/08/03/big/img_116
+2002/07/22/big/img_467
+2002/07/31/big/img_750
+2002/07/26/big/img_435
+2002/07/19/big/img_822
+2002/08/13/big/img_626
+2002/08/11/big/img_344
+2002/08/02/big/img_473
+2002/09/01/big/img_16817
+2002/08/01/big/img_1275
+2002/08/28/big/img_19270
+2002/07/23/big/img_607
+2002/08/09/big/img_316
+2002/07/29/big/img_626
+2002/07/24/big/img_824
+2002/07/22/big/img_342
+2002/08/08/big/img_794
+2002/08/07/big/img_1209
+2002/07/19/big/img_18
+2002/08/25/big/img_634
+2002/07/24/big/img_730
+2003/01/17/big/img_356
+2002/07/23/big/img_305
+2002/07/30/big/img_453
+2003/01/13/big/img_972
+2002/08/06/big/img_2610
+2002/08/29/big/img_18920
+2002/07/31/big/img_123
+2002/07/26/big/img_979
+2002/08/24/big/img_635
+2002/08/05/big/img_3704
+2002/08/07/big/img_1358
+2002/07/22/big/img_306
+2002/08/13/big/img_619
+2002/08/02/big/img_366
diff --git a/data/__init__.py b/data/__init__.py
new file mode 100644
index 0000000..ea50eba
--- /dev/null
+++ b/data/__init__.py
@@ -0,0 +1,3 @@
+from .wider_face import WiderFaceDetection, detection_collate
+from .data_augment import *
+from .config import *
diff --git a/data/__pycache__/__init__.cpython-38.pyc b/data/__pycache__/__init__.cpython-38.pyc
new file mode 100644
index 0000000..33be196
Binary files /dev/null and b/data/__pycache__/__init__.cpython-38.pyc differ
diff --git a/data/__pycache__/config.cpython-38.pyc b/data/__pycache__/config.cpython-38.pyc
new file mode 100644
index 0000000..282307b
Binary files /dev/null and b/data/__pycache__/config.cpython-38.pyc differ
diff --git a/data/__pycache__/data_augment.cpython-38.pyc b/data/__pycache__/data_augment.cpython-38.pyc
new file mode 100644
index 0000000..f27fec0
Binary files /dev/null and b/data/__pycache__/data_augment.cpython-38.pyc differ
diff --git a/data/__pycache__/wider_face.cpython-38.pyc b/data/__pycache__/wider_face.cpython-38.pyc
new file mode 100644
index 0000000..79903c1
Binary files /dev/null and b/data/__pycache__/wider_face.cpython-38.pyc differ
diff --git a/data/config.py b/data/config.py
new file mode 100644
index 0000000..591f349
--- /dev/null
+++ b/data/config.py
@@ -0,0 +1,42 @@
+# config.py
+
+cfg_mnet = {
+    'name': 'mobilenet0.25',
+    'min_sizes': [[16, 32], [64, 128], [256, 512]],
+    'steps': [8, 16, 32],
+    'variance': [0.1, 0.2],
+    'clip': False,
+    'loc_weight': 2.0,
+    'gpu_train': True,
+    'batch_size': 32,
+    'ngpu': 1,
+    'epoch': 250,
+    'decay1': 190,
+    'decay2': 220,
+    'image_size': 640,
+    'pretrain': True,
+    'return_layers': {'stage1': 1, 'stage2': 2, 'stage3': 3},
+    'in_channel': 32,
+    'out_channel': 64
+}
+
+cfg_re50 = {
+    'name': 'Resnet50',
+    'min_sizes': [[16, 32], [64, 128], [256, 512]],
+    'steps': [8, 16, 32],
+    'variance': [0.1, 0.2],
+    'clip': False,
+    'loc_weight': 2.0,
+    'gpu_train': True,
+    'batch_size': 24,
+    'ngpu': 4,
+    'epoch': 100,
+    'decay1': 70,
+    'decay2': 90,
+    'image_size': 840,
+    'pretrain': True,
+    'return_layers': {'layer2': 1, 'layer3': 2, 'layer4': 3},
+    'in_channel': 256,
+    'out_channel': 256
+}
+
diff --git a/data/data_augment.py b/data/data_augment.py
new file mode 100644
index 0000000..c1b52ae
--- /dev/null
+++ b/data/data_augment.py
@@ -0,0 +1,237 @@
+import cv2
+import numpy as np
+import random
+from utils.box_utils import matrix_iof
+
+
+def _crop(image, boxes, labels, landm, img_dim):
+    height, width, _ = image.shape
+    pad_image_flag = True
+
+    for _ in range(250):
+        """
+        if random.uniform(0, 1) <= 0.2:
+            scale = 1.0
+        else:
+            scale = random.uniform(0.3, 1.0)
+        """
+        PRE_SCALES = [0.3, 0.45, 0.6, 0.8, 1.0]
+        scale = random.choice(PRE_SCALES)
+        short_side = min(width, height)
+        w = int(scale * short_side)
+        h = w
+
+        if width == w:
+            l = 0
+        else:
+            l = random.randrange(width - w)
+        if height == h:
+            t = 0
+        else:
+            t = random.randrange(height - h)
+        roi = np.array((l, t, l + w, t + h))
+
+        value = matrix_iof(boxes, roi[np.newaxis])
+        flag = (value >= 1)
+        if not flag.any():
+            continue
+
+        centers = (boxes[:, :2] + boxes[:, 2:]) / 2
+        mask_a = np.logical_and(roi[:2] < centers, centers < roi[2:]).all(axis=1)
+        boxes_t = boxes[mask_a].copy()
+        labels_t = labels[mask_a].copy()
+        landms_t = landm[mask_a].copy()
+        landms_t = landms_t.reshape([-1, 5, 2])
+
+        if boxes_t.shape[0] == 0:
+            continue
+
+        image_t = image[roi[1]:roi[3], roi[0]:roi[2]]
+
+        boxes_t[:, :2] = np.maximum(boxes_t[:, :2], roi[:2])
+        boxes_t[:, :2] -= roi[:2]
+        boxes_t[:, 2:] = np.minimum(boxes_t[:, 2:], roi[2:])
+        boxes_t[:, 2:] -= roi[:2]
+
+        # landm
+        landms_t[:, :, :2] = landms_t[:, :, :2] - roi[:2]
+        landms_t[:, :, :2] = np.maximum(landms_t[:, :, :2], np.array([0, 0]))
+        landms_t[:, :, :2] = np.minimum(landms_t[:, :, :2], roi[2:] - roi[:2])
+        landms_t = landms_t.reshape([-1, 10])
+
+
+	# make sure that the cropped image contains at least one face > 16 pixel at training image scale
+        b_w_t = (boxes_t[:, 2] - boxes_t[:, 0] + 1) / w * img_dim
+        b_h_t = (boxes_t[:, 3] - boxes_t[:, 1] + 1) / h * img_dim
+        mask_b = np.minimum(b_w_t, b_h_t) > 0.0
+        boxes_t = boxes_t[mask_b]
+        labels_t = labels_t[mask_b]
+        landms_t = landms_t[mask_b]
+
+        if boxes_t.shape[0] == 0:
+            continue
+
+        pad_image_flag = False
+
+        return image_t, boxes_t, labels_t, landms_t, pad_image_flag
+    return image, boxes, labels, landm, pad_image_flag
+
+
+def _distort(image):
+
+    def _convert(image, alpha=1, beta=0):
+        tmp = image.astype(float) * alpha + beta
+        tmp[tmp < 0] = 0
+        tmp[tmp > 255] = 255
+        image[:] = tmp
+
+    image = image.copy()
+
+    if random.randrange(2):
+
+        #brightness distortion
+        if random.randrange(2):
+            _convert(image, beta=random.uniform(-32, 32))
+
+        #contrast distortion
+        if random.randrange(2):
+            _convert(image, alpha=random.uniform(0.5, 1.5))
+
+        image = cv2.cvtColor(image, cv2.COLOR_BGR2HSV)
+
+        #saturation distortion
+        if random.randrange(2):
+            _convert(image[:, :, 1], alpha=random.uniform(0.5, 1.5))
+
+        #hue distortion
+        if random.randrange(2):
+            tmp = image[:, :, 0].astype(int) + random.randint(-18, 18)
+            tmp %= 180
+            image[:, :, 0] = tmp
+
+        image = cv2.cvtColor(image, cv2.COLOR_HSV2BGR)
+
+    else:
+
+        #brightness distortion
+        if random.randrange(2):
+            _convert(image, beta=random.uniform(-32, 32))
+
+        image = cv2.cvtColor(image, cv2.COLOR_BGR2HSV)
+
+        #saturation distortion
+        if random.randrange(2):
+            _convert(image[:, :, 1], alpha=random.uniform(0.5, 1.5))
+
+        #hue distortion
+        if random.randrange(2):
+            tmp = image[:, :, 0].astype(int) + random.randint(-18, 18)
+            tmp %= 180
+            image[:, :, 0] = tmp
+
+        image = cv2.cvtColor(image, cv2.COLOR_HSV2BGR)
+
+        #contrast distortion
+        if random.randrange(2):
+            _convert(image, alpha=random.uniform(0.5, 1.5))
+
+    return image
+
+
+def _expand(image, boxes, fill, p):
+    if random.randrange(2):
+        return image, boxes
+
+    height, width, depth = image.shape
+
+    scale = random.uniform(1, p)
+    w = int(scale * width)
+    h = int(scale * height)
+
+    left = random.randint(0, w - width)
+    top = random.randint(0, h - height)
+
+    boxes_t = boxes.copy()
+    boxes_t[:, :2] += (left, top)
+    boxes_t[:, 2:] += (left, top)
+    expand_image = np.empty(
+        (h, w, depth),
+        dtype=image.dtype)
+    expand_image[:, :] = fill
+    expand_image[top:top + height, left:left + width] = image
+    image = expand_image
+
+    return image, boxes_t
+
+
+def _mirror(image, boxes, landms):
+    _, width, _ = image.shape
+    if random.randrange(2):
+        image = image[:, ::-1]
+        boxes = boxes.copy()
+        boxes[:, 0::2] = width - boxes[:, 2::-2]
+
+        # landm
+        landms = landms.copy()
+        landms = landms.reshape([-1, 5, 2])
+        landms[:, :, 0] = width - landms[:, :, 0]
+        tmp = landms[:, 1, :].copy()
+        landms[:, 1, :] = landms[:, 0, :]
+        landms[:, 0, :] = tmp
+        tmp1 = landms[:, 4, :].copy()
+        landms[:, 4, :] = landms[:, 3, :]
+        landms[:, 3, :] = tmp1
+        landms = landms.reshape([-1, 10])
+
+    return image, boxes, landms
+
+
+def _pad_to_square(image, rgb_mean, pad_image_flag):
+    if not pad_image_flag:
+        return image
+    height, width, _ = image.shape
+    long_side = max(width, height)
+    image_t = np.empty((long_side, long_side, 3), dtype=image.dtype)
+    image_t[:, :] = rgb_mean
+    image_t[0:0 + height, 0:0 + width] = image
+    return image_t
+
+
+def _resize_subtract_mean(image, insize, rgb_mean):
+    interp_methods = [cv2.INTER_LINEAR, cv2.INTER_CUBIC, cv2.INTER_AREA, cv2.INTER_NEAREST, cv2.INTER_LANCZOS4]
+    interp_method = interp_methods[random.randrange(5)]
+    image = cv2.resize(image, (insize, insize), interpolation=interp_method)
+    image = image.astype(np.float32)
+    image -= rgb_mean
+    return image.transpose(2, 0, 1)
+
+
+class preproc(object):
+
+    def __init__(self, img_dim, rgb_means):
+        self.img_dim = img_dim
+        self.rgb_means = rgb_means
+
+    def __call__(self, image, targets):
+        assert targets.shape[0] > 0, "this image does not have gt"
+
+        boxes = targets[:, :4].copy()
+        labels = targets[:, -1].copy()
+        landm = targets[:, 4:-1].copy()
+
+        image_t, boxes_t, labels_t, landm_t, pad_image_flag = _crop(image, boxes, labels, landm, self.img_dim)
+        image_t = _distort(image_t)
+        image_t = _pad_to_square(image_t,self.rgb_means, pad_image_flag)
+        image_t, boxes_t, landm_t = _mirror(image_t, boxes_t, landm_t)
+        height, width, _ = image_t.shape
+        image_t = _resize_subtract_mean(image_t, self.img_dim, self.rgb_means)
+        boxes_t[:, 0::2] /= width
+        boxes_t[:, 1::2] /= height
+
+        landm_t[:, 0::2] /= width
+        landm_t[:, 1::2] /= height
+
+        labels_t = np.expand_dims(labels_t, 1)
+        targets_t = np.hstack((boxes_t, landm_t, labels_t))
+
+        return image_t, targets_t
diff --git a/data/realtime_detect.py b/data/realtime_detect.py
new file mode 100644
index 0000000..d1b0ad5
--- /dev/null
+++ b/data/realtime_detect.py
@@ -0,0 +1,258 @@
+import subprocess
+import time
+import cv2
+import torch
+import numpy as np
+from skimage import transform as trans
+from PIL import Image, ImageDraw, ImageFont
+from data import cfg_mnet, cfg_re50
+from face_api import load_arcface_model, load_npy
+from layers.functions.prior_box import PriorBox
+from retinaface_detect import set_retinaface_conf, load_retinaface_model, findAll
+from utils.nms.py_cpu_nms import py_cpu_nms
+from utils.box_utils import decode, decode_landm
+import faiss
+
+ppi = 1280
+ppi2 = 640
+step = 3
+
+def detect_rtsp(rtsp, out_rtsp, net, arcface_model, k_v, args):
+    tic_total = time.time()
+    cfg = None
+    if args.network == "mobile0.25":
+        cfg = cfg_mnet
+    elif args.network == "resnet50":
+        cfg = cfg_re50
+    device = torch.device("cpu" if args.cpu else "cuda")
+    resize = 1
+
+    # testing begin
+    cap = cv2.VideoCapture(rtsp)
+    ret, frame = cap.read()
+    h, w = frame.shape[:2]
+
+    factor = 0
+    if (w > ppi):
+        factor = h / w
+        frame = cv2.resize(frame, (ppi, int(ppi * factor)))
+        h, w = frame.shape[:2]
+    arf = 1
+    detect_h, detect_w = frame.shape[:2]
+    frame_detect = frame
+    factor2 = 0
+    if (w > ppi2):
+        factor2 = h / w
+        frame_detect = cv2.resize(frame, (ppi2, int(ppi2 * factor2)))
+        detect_h, detect_w = frame_detect.shape[:2]
+        arf = w/detect_w
+    print(w,h)
+    print(detect_w,detect_h)
+
+    #fps = cap.get(cv2.CAP_PROP_FPS)
+    #print(fps)
+    size = (w, h)
+    sizeStr = str(size[0]) + 'x' + str(size[1])
+    if(out_rtsp.startswith("rtsp")):
+        command = ['ffmpeg',
+                   '-y', '-an',
+                   '-f', 'rawvideo',
+                   '-vcodec', 'rawvideo',
+                   '-pix_fmt', 'bgr24',
+                   '-s', sizeStr,
+                   '-r', "25",
+                   '-i', '-',
+                   '-c:v', 'libx265',
+                   '-b:v', '3000k',
+                   '-pix_fmt', 'yuv420p',
+                   '-preset', 'ultrafast',
+                   '-f', 'rtsp',
+                   out_rtsp]
+    pipe = subprocess.Popen(command, shell=False, stdin=subprocess.PIPE)
+    number = step
+    dets = []
+    name_list = []
+    font = ImageFont.truetype("font.ttf", 22)
+    priorbox = PriorBox(cfg, image_size=(detect_h, detect_w))
+    priors = priorbox.forward()
+    priors = priors.to(device)
+    prior_data = priors.data
+
+    scale = torch.Tensor([detect_w, detect_h, detect_w, detect_h])
+    scale = scale.to(device)
+    scale1 = torch.Tensor([detect_w, detect_h, detect_w, detect_h,
+                           detect_w, detect_h, detect_w, detect_h,
+                           detect_w, detect_h])
+    scale1 = scale1.to(device)
+
+    src1 = np.array([
+        [38.3814, 51.6963],
+        [73.6186, 51.5014],
+        [56.1120, 71.7366],
+        [41.6361, 92.3655],
+        [70.8167, 92.2041]], dtype=np.float32)
+    tform = trans.SimilarityTransform()
+
+    while ret:
+        tic_all = time.time()
+        if number == step:
+            tic = time.time()
+            img = np.float32(frame_detect)
+            img -= (104, 117, 123)
+            img = img.transpose(2, 0, 1)
+            img = torch.from_numpy(img).unsqueeze(0)
+            img = img.to(device)
+
+            loc, conf, landms = net(img)  # forward pass
+
+            boxes = decode(loc.data.squeeze(0), prior_data, cfg['variance'])
+            boxes = boxes * scale / resize
+            boxes = boxes.cpu().numpy()
+            scores = conf.squeeze(0).data.cpu().numpy()[:, 1]
+            landms = decode_landm(landms.data.squeeze(0), prior_data, cfg['variance'])
+
+            landms = landms * scale1 / resize
+            landms = landms.cpu().numpy()
+
+            # ignore low scores
+            inds = np.where(scores > args.confidence_threshold)[0]
+            boxes = boxes[inds]
+            landms = landms[inds]
+            scores = scores[inds]
+
+            # keep top-K before NMS
+            order = scores.argsort()[::-1][:args.top_k]
+            boxes = boxes[order]
+            landms = landms[order]
+            scores = scores[order]
+
+            # do NMS
+            dets = np.hstack((boxes, scores[:, np.newaxis])).astype(np.float32, copy=False)
+            keep = py_cpu_nms(dets, args.nms_threshold)
+            # keep = nms(dets, args.nms_threshold,force_cpu=args.cpu)
+            dets = dets[keep, :]
+            landms = landms[keep]
+
+            # keep top-K faster NMS
+            dets = dets[:args.keep_top_k, :]
+            landms = landms[:args.keep_top_k, :]
+
+            dets = np.concatenate((dets, landms), axis=1)
+            face_list = []
+            name_list = []
+            print('net forward time: {:.4f}'.format(time.time() - tic))
+            start_time_findall = time.time()
+            for i, det in enumerate(dets[:1]):
+                if det[4] < args.vis_thres:
+                    continue
+                #boxes, score = det[:4], det[4]
+                dst = np.reshape(landms[i], (5, 2))
+                dst = dst * arf
+
+                tform.estimate(dst, src1)
+                M = tform.params[0:2, :]
+                frame2 = cv2.warpAffine(frame, M, (w, h), borderValue=0.0)
+                img112 = frame2[0:112, 0:112, :]
+                face_list.append(img112)
+
+            if len(face_list) != 0:
+                face_list = np.array(face_list)
+                face_list = face_list.transpose((0, 3, 1, 2))
+                face_list = np.array(face_list, dtype=np.float32)
+                face_list -= 127.5
+                face_list /= 127.5
+                print(face_list.shape)
+                print("warpALL time: " + str(time.time() - start_time_findall ))
+                #start_time = time.time()
+                name_list = findAll(face_list, arcface_model, k_v, "cpu" if args.cpu else "cuda")
+                #print(name_list)
+
+            #print("findOneframe time: " + str(time.time() - start_time_findall))
+            #start_time = time.time()
+            # if (len(dets) != 0):
+            #     for i, det in enumerate(dets[:]):
+            #         if det[4] < args.vis_thres:
+            #             continue
+            #         boxes, score = det[:4], det[4]
+            #         boxes = boxes * arf
+            #         name = name_list[i]
+            #         cv2.rectangle(frame, (int(boxes[0]), int(boxes[1])), (int(boxes[2]), int(boxes[3])), (255, 0, 0), 2)
+            #         cv2.putText(frame, name, (int(boxes[0]), int(boxes[1])), cv2.FONT_HERSHEY_COMPLEX, 0.4,(0, 225, 255), 1)
+            start_time = time.time()
+            if(len(dets) != 0):
+                img_PIL = Image.fromarray(cv2.cvtColor(frame, cv2.COLOR_BGR2RGB))
+                draw = ImageDraw.Draw(img_PIL)
+                for i, det in enumerate(dets[:1]):
+                    if det[4] < args.vis_thres:
+                        continue
+                    boxes, score = det[:4], det[4]
+                    boxes = boxes * arf
+                    name = name_list[i]
+                    if not isinstance(name, np.unicode):
+                        name = name.decode('utf8')
+                    draw.text((int(boxes[0]), int(boxes[1])), name, fill=(255, 0, 0), font=font)
+                    draw.rectangle((int(boxes[0]), int(boxes[1]), int(boxes[2]), int(boxes[3])), outline="green", width=3)
+                frame = cv2.cvtColor(np.asarray(img_PIL), cv2.COLOR_RGB2BGR)
+            pipe.stdin.write(frame.tostring())
+            print("drawOneframe time: " + str(time.time() - start_time))
+            #start_time = time.time()
+            ret, frame = cap.read()
+            frame_detect = frame
+            number = step
+            if (ret != 0 and factor != 0):
+                frame = cv2.resize(frame, (ppi, int(ppi * factor)))
+            if (ret != 0 and factor2 != 0):
+                frame_detect = cv2.resize(frame, (ppi2, int(ppi2 * factor2)))
+            #print("readframe time: " + str(time.time() - start_time))
+        else:
+            number += 1
+            if (len(dets) != 0):
+                for i, det in enumerate(dets[:4]):
+                    if det[4] < args.vis_thres:
+                        continue
+                    boxes, score = det[:4], det[4]
+                    cv2.rectangle(frame, (int(boxes[0]), int(boxes[1])), (int(boxes[2]), int(boxes[3])), (2, 255, 0), 1)
+            # if (len(dets) != 0):
+            #     img_PIL = Image.fromarray(cv2.cvtColor(frame, cv2.COLOR_BGR2RGB))
+            #     draw = ImageDraw.Draw(img_PIL)
+            #     for i, det in enumerate(dets[:4]):
+            #         if det[4] < args.vis_thres:
+            #             continue
+            #         boxes, score = det[:4], det[4]
+            #         name = name_list[i]
+            #         if not isinstance(name, np.unicode):
+            #             name = name.decode('utf8')
+            #         draw.text((int(boxes[0]), int(boxes[1])), name, fill=(255, 0, 0), font=font)
+            #         draw.rectangle((int(boxes[0]), int(boxes[1]), int(boxes[2]), int(boxes[3])), outline="green",
+            #                        width=3)
+            #     frame = cv2.cvtColor(np.asarray(img_PIL), cv2.COLOR_RGB2BGR)
+            start_time = time.time()
+            pipe.stdin.write(frame.tostring())
+            print("writeframe time: " + str(time.time() - start_time))
+            start_time = time.time()
+            ret, frame = cap.read()
+            if (ret != 0 and factor != 0):
+                frame = cv2.resize(frame, (ppi, int(ppi * factor)))
+            print("readframe time: " + str(time.time() - start_time))
+        print('all time: {:.4f}'.format(time.time() - tic_all))
+    cap.release()
+    pipe.terminate()
+    print('total time: {:.4f}'.format(time.time() - tic_total))
+
+if __name__ == "__main__":
+    cpu_or_cuda = "cuda" if torch.cuda.is_available() else "cpu"
+    # 加载人脸识别模型
+    arcface_model = load_arcface_model("./model/backbone100.pth", cpu_or_cuda=cpu_or_cuda)
+    # 加载人脸检测模型
+    retinaface_args = set_retinaface_conf(cpu_or_cuda=cpu_or_cuda)
+    retinaface_model = load_retinaface_model(retinaface_args)
+    k_v = load_npy("./Database/student.npy")
+    #print(list(k_v.keys()))
+    database_name_list = list(k_v.keys())
+    vector_list = np.array(list(k_v.values()))
+    print(vector_list.shape)
+    index = faiss.IndexFlatL2(512)
+    index.add(vector_list)
+
+    #detect_rtsp("software.mp4", 'rtsp://localhost/test2', retinaface_model, arcface_model, index ,database_name_list, retinaface_args)
+    detect_rtsp("cut.mp4", 'rtsp://localhost:5001/test2', retinaface_model, arcface_model, k_v, retinaface_args)
\ No newline at end of file
diff --git a/data/wider_face.py b/data/wider_face.py
new file mode 100644
index 0000000..22f56ef
--- /dev/null
+++ b/data/wider_face.py
@@ -0,0 +1,101 @@
+import os
+import os.path
+import sys
+import torch
+import torch.utils.data as data
+import cv2
+import numpy as np
+
+class WiderFaceDetection(data.Dataset):
+    def __init__(self, txt_path, preproc=None):
+        self.preproc = preproc
+        self.imgs_path = []
+        self.words = []
+        f = open(txt_path,'r')
+        lines = f.readlines()
+        isFirst = True
+        labels = []
+        for line in lines:
+            line = line.rstrip()
+            if line.startswith('#'):
+                if isFirst is True:
+                    isFirst = False
+                else:
+                    labels_copy = labels.copy()
+                    self.words.append(labels_copy)
+                    labels.clear()
+                path = line[2:]
+                path = txt_path.replace('label.txt','images/') + path
+                self.imgs_path.append(path)
+            else:
+                line = line.split(' ')
+                label = [float(x) for x in line]
+                labels.append(label)
+
+        self.words.append(labels)
+
+    def __len__(self):
+        return len(self.imgs_path)
+
+    def __getitem__(self, index):
+        img = cv2.imread(self.imgs_path[index])
+        height, width, _ = img.shape
+
+        labels = self.words[index]
+        annotations = np.zeros((0, 15))
+        if len(labels) == 0:
+            return annotations
+        for idx, label in enumerate(labels):
+            annotation = np.zeros((1, 15))
+            # bbox
+            annotation[0, 0] = label[0]  # x1
+            annotation[0, 1] = label[1]  # y1
+            annotation[0, 2] = label[0] + label[2]  # x2
+            annotation[0, 3] = label[1] + label[3]  # y2
+
+            # landmarks
+            annotation[0, 4] = label[4]    # l0_x
+            annotation[0, 5] = label[5]    # l0_y
+            annotation[0, 6] = label[7]    # l1_x
+            annotation[0, 7] = label[8]    # l1_y
+            annotation[0, 8] = label[10]   # l2_x
+            annotation[0, 9] = label[11]   # l2_y
+            annotation[0, 10] = label[13]  # l3_x
+            annotation[0, 11] = label[14]  # l3_y
+            annotation[0, 12] = label[16]  # l4_x
+            annotation[0, 13] = label[17]  # l4_y
+            if (annotation[0, 4]<0):
+                annotation[0, 14] = -1
+            else:
+                annotation[0, 14] = 1
+
+            annotations = np.append(annotations, annotation, axis=0)
+        target = np.array(annotations)
+        if self.preproc is not None:
+            img, target = self.preproc(img, target)
+
+        return torch.from_numpy(img), target
+
+def detection_collate(batch):
+    """Custom collate fn for dealing with batches of images that have a different
+    number of associated object annotations (bounding boxes).
+
+    Arguments:
+        batch: (tuple) A tuple of tensor images and lists of annotations
+
+    Return:
+        A tuple containing:
+            1) (tensor) batch of images stacked on their 0 dim
+            2) (list of tensors) annotations for a given image are stacked on 0 dim
+    """
+    targets = []
+    imgs = []
+    for _, sample in enumerate(batch):
+        for _, tup in enumerate(sample):
+            if torch.is_tensor(tup):
+                imgs.append(tup)
+            elif isinstance(tup, type(np.empty(0))):
+                annos = torch.from_numpy(tup).float()
+                targets.append(annos)
+
+    return (torch.stack(imgs, 0), targets)
diff --git a/dataset.py b/dataset.py
new file mode 100644
index 0000000..db65236
--- /dev/null
+++ b/dataset.py
@@ -0,0 +1,107 @@
+import numbers
+import os
+import queue as Queue
+import threading
+
+import mxnet as mx
+import numpy as np
+import torch
+from torch.utils.data import DataLoader, Dataset
+from torchvision import transforms
+
+
+class BackgroundGenerator(threading.Thread):
+    def __init__(self, generator, local_rank, max_prefetch=6):
+        super(BackgroundGenerator, self).__init__()
+        self.queue = Queue.Queue(max_prefetch)
+        self.generator = generator
+        self.local_rank = local_rank
+        self.daemon = True
+        self.start()
+
+    def run(self):
+        torch.cuda.set_device(self.local_rank)
+        for item in self.generator:
+            self.queue.put(item)
+        self.queue.put(None)
+
+    def next(self):
+        next_item = self.queue.get()
+        if next_item is None:
+            raise StopIteration
+        return next_item
+
+    def __next__(self):
+        return self.next()
+
+    def __iter__(self):
+        return self
+
+
+class DataLoaderX(DataLoader):
+    def __init__(self, local_rank, **kwargs):
+        super(DataLoaderX, self).__init__(**kwargs)
+        self.stream = torch.cuda.Stream(local_rank)
+        self.local_rank = local_rank
+
+    def __iter__(self):
+        self.iter = super(DataLoaderX, self).__iter__()
+        self.iter = BackgroundGenerator(self.iter, self.local_rank)
+        self.preload()
+        return self
+
+    def preload(self):
+        self.batch = next(self.iter, None)
+        if self.batch is None:
+            return None
+        with torch.cuda.stream(self.stream):
+            for k in range(len(self.batch)):
+                self.batch[k] = self.batch[k].to(device=self.local_rank,
+                                                 non_blocking=True)
+
+    def __next__(self):
+        torch.cuda.current_stream().wait_stream(self.stream)
+        batch = self.batch
+        if batch is None:
+            raise StopIteration
+        self.preload()
+        return batch
+
+
+class MXFaceDataset(Dataset):
+    def __init__(self, root_dir, local_rank):
+        super(MXFaceDataset, self).__init__()
+        self.transform = transforms.Compose(
+            [transforms.ToPILImage(),
+             transforms.RandomHorizontalFlip(),
+             transforms.ToTensor(),
+             transforms.Normalize(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5]),
+             ])
+        self.root_dir = root_dir
+        self.local_rank = local_rank
+        path_imgrec = os.path.join(root_dir, 'train.rec')
+        path_imgidx = os.path.join(root_dir, 'train.idx')
+        self.imgrec = mx.recordio.MXIndexedRecordIO(path_imgidx, path_imgrec, 'r')
+        s = self.imgrec.read_idx(0)
+        header, _ = mx.recordio.unpack(s)
+        if header.flag > 0:
+            self.header0 = (int(header.label[0]), int(header.label[1]))
+            self.imgidx = np.array(range(1, int(header.label[0])))
+        else:
+            self.imgidx = np.array(list(self.imgrec.keys))
+
+    def __getitem__(self, index):
+        idx = self.imgidx[index]
+        s = self.imgrec.read_idx(idx)
+        header, img = mx.recordio.unpack(s)
+        label = header.label
+        if not isinstance(label, numbers.Number):
+            label = label[0]
+        label = torch.tensor(label, dtype=torch.long)
+        sample = mx.image.imdecode(img).asnumpy()
+        if self.transform is not None:
+            sample = self.transform(sample)
+        return sample, label
+
+    def __len__(self):
+        return len(self.imgidx)
diff --git a/eval/__init__.py b/eval/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/eval/verification.py b/eval/verification.py
new file mode 100644
index 0000000..edacf8d
--- /dev/null
+++ b/eval/verification.py
@@ -0,0 +1,409 @@
+"""Helper for evaluation on the Labeled Faces in the Wild dataset 
+"""
+
+# MIT License
+#
+# Copyright (c) 2016 David Sandberg
+#
+# Permission is hereby granted, free of charge, to any person obtaining a copy
+# of this software and associated documentation files (the "Software"), to deal
+# in the Software without restriction, including without limitation the rights
+# to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+# copies of the Software, and to permit persons to whom the Software is
+# furnished to do so, subject to the following conditions:
+#
+# The above copyright notice and this permission notice shall be included in all
+# copies or substantial portions of the Software.
+#
+# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+# AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+# OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+# SOFTWARE.
+
+
+import datetime
+import os
+import pickle
+
+import mxnet as mx
+import numpy as np
+import sklearn
+import torch
+from mxnet import ndarray as nd
+from scipy import interpolate
+from sklearn.decomposition import PCA
+from sklearn.model_selection import KFold
+
+
+class LFold:
+    def __init__(self, n_splits=2, shuffle=False):
+        self.n_splits = n_splits
+        if self.n_splits > 1:
+            self.k_fold = KFold(n_splits=n_splits, shuffle=shuffle)
+
+    def split(self, indices):
+        if self.n_splits > 1:
+            return self.k_fold.split(indices)
+        else:
+            return [(indices, indices)]
+
+
+def calculate_roc(thresholds,
+                  embeddings1,
+                  embeddings2,
+                  actual_issame,
+                  nrof_folds=10,
+                  pca=0):
+    assert (embeddings1.shape[0] == embeddings2.shape[0])
+    assert (embeddings1.shape[1] == embeddings2.shape[1])
+    nrof_pairs = min(len(actual_issame), embeddings1.shape[0])
+    nrof_thresholds = len(thresholds)
+    k_fold = LFold(n_splits=nrof_folds, shuffle=False)
+
+    tprs = np.zeros((nrof_folds, nrof_thresholds))
+    fprs = np.zeros((nrof_folds, nrof_thresholds))
+    accuracy = np.zeros((nrof_folds))
+    indices = np.arange(nrof_pairs)
+
+    if pca == 0:
+        diff = np.subtract(embeddings1, embeddings2)
+        dist = np.sum(np.square(diff), 1)
+
+    for fold_idx, (train_set, test_set) in enumerate(k_fold.split(indices)):
+        if pca > 0:
+            print('doing pca on', fold_idx)
+            embed1_train = embeddings1[train_set]
+            embed2_train = embeddings2[train_set]
+            _embed_train = np.concatenate((embed1_train, embed2_train), axis=0)
+            pca_model = PCA(n_components=pca)
+            pca_model.fit(_embed_train)
+            embed1 = pca_model.transform(embeddings1)
+            embed2 = pca_model.transform(embeddings2)
+            embed1 = sklearn.preprocessing.normalize(embed1)
+            embed2 = sklearn.preprocessing.normalize(embed2)
+            diff = np.subtract(embed1, embed2)
+            dist = np.sum(np.square(diff), 1)
+
+        # Find the best threshold for the fold
+        acc_train = np.zeros((nrof_thresholds))
+        for threshold_idx, threshold in enumerate(thresholds):
+            _, _, acc_train[threshold_idx] = calculate_accuracy(
+                threshold, dist[train_set], actual_issame[train_set])
+        best_threshold_index = np.argmax(acc_train)
+        for threshold_idx, threshold in enumerate(thresholds):
+            tprs[fold_idx, threshold_idx], fprs[fold_idx, threshold_idx], _ = calculate_accuracy(
+                threshold, dist[test_set],
+                actual_issame[test_set])
+        _, _, accuracy[fold_idx] = calculate_accuracy(
+            thresholds[best_threshold_index], dist[test_set],
+            actual_issame[test_set])
+
+    tpr = np.mean(tprs, 0)
+    fpr = np.mean(fprs, 0)
+    return tpr, fpr, accuracy
+
+
+def calculate_accuracy(threshold, dist, actual_issame):
+    predict_issame = np.less(dist, threshold)
+    tp = np.sum(np.logical_and(predict_issame, actual_issame))
+    fp = np.sum(np.logical_and(predict_issame, np.logical_not(actual_issame)))
+    tn = np.sum(
+        np.logical_and(np.logical_not(predict_issame),
+                       np.logical_not(actual_issame)))
+    fn = np.sum(np.logical_and(np.logical_not(predict_issame), actual_issame))
+
+    tpr = 0 if (tp + fn == 0) else float(tp) / float(tp + fn)
+    fpr = 0 if (fp + tn == 0) else float(fp) / float(fp + tn)
+    acc = float(tp + tn) / dist.size
+    return tpr, fpr, acc
+
+
+def calculate_val(thresholds,
+                  embeddings1,
+                  embeddings2,
+                  actual_issame,
+                  far_target,
+                  nrof_folds=10):
+    assert (embeddings1.shape[0] == embeddings2.shape[0])
+    assert (embeddings1.shape[1] == embeddings2.shape[1])
+    nrof_pairs = min(len(actual_issame), embeddings1.shape[0])
+    nrof_thresholds = len(thresholds)
+    k_fold = LFold(n_splits=nrof_folds, shuffle=False)
+
+    val = np.zeros(nrof_folds)
+    far = np.zeros(nrof_folds)
+
+    diff = np.subtract(embeddings1, embeddings2)
+    dist = np.sum(np.square(diff), 1)
+    indices = np.arange(nrof_pairs)
+
+    for fold_idx, (train_set, test_set) in enumerate(k_fold.split(indices)):
+
+        # Find the threshold that gives FAR = far_target
+        far_train = np.zeros(nrof_thresholds)
+        for threshold_idx, threshold in enumerate(thresholds):
+            _, far_train[threshold_idx] = calculate_val_far(
+                threshold, dist[train_set], actual_issame[train_set])
+        if np.max(far_train) >= far_target:
+            f = interpolate.interp1d(far_train, thresholds, kind='slinear')
+            threshold = f(far_target)
+        else:
+            threshold = 0.0
+
+        val[fold_idx], far[fold_idx] = calculate_val_far(
+            threshold, dist[test_set], actual_issame[test_set])
+
+    val_mean = np.mean(val)
+    far_mean = np.mean(far)
+    val_std = np.std(val)
+    return val_mean, val_std, far_mean
+
+
+def calculate_val_far(threshold, dist, actual_issame):
+    predict_issame = np.less(dist, threshold)
+    true_accept = np.sum(np.logical_and(predict_issame, actual_issame))
+    false_accept = np.sum(
+        np.logical_and(predict_issame, np.logical_not(actual_issame)))
+    n_same = np.sum(actual_issame)
+    n_diff = np.sum(np.logical_not(actual_issame))
+    # print(true_accept, false_accept)
+    # print(n_same, n_diff)
+    val = float(true_accept) / float(n_same)
+    far = float(false_accept) / float(n_diff)
+    return val, far
+
+
+def evaluate(embeddings, actual_issame, nrof_folds=10, pca=0):
+    # Calculate evaluation metrics
+    thresholds = np.arange(0, 4, 0.01)
+    embeddings1 = embeddings[0::2]
+    embeddings2 = embeddings[1::2]
+    tpr, fpr, accuracy = calculate_roc(thresholds,
+                                       embeddings1,
+                                       embeddings2,
+                                       np.asarray(actual_issame),
+                                       nrof_folds=nrof_folds,
+                                       pca=pca)
+    thresholds = np.arange(0, 4, 0.001)
+    val, val_std, far = calculate_val(thresholds,
+                                      embeddings1,
+                                      embeddings2,
+                                      np.asarray(actual_issame),
+                                      1e-3,
+                                      nrof_folds=nrof_folds)
+    return tpr, fpr, accuracy, val, val_std, far
+
+@torch.no_grad()
+def load_bin(path, image_size):
+    try:
+        with open(path, 'rb') as f:
+            bins, issame_list = pickle.load(f)  # py2
+    except UnicodeDecodeError as e:
+        with open(path, 'rb') as f:
+            bins, issame_list = pickle.load(f, encoding='bytes')  # py3
+    data_list = []
+    for flip in [0, 1]:
+        data = torch.empty((len(issame_list) * 2, 3, image_size[0], image_size[1]))
+        data_list.append(data)
+    for idx in range(len(issame_list) * 2):
+        _bin = bins[idx]
+        img = mx.image.imdecode(_bin)
+        if img.shape[1] != image_size[0]:
+            img = mx.image.resize_short(img, image_size[0])
+        img = nd.transpose(img, axes=(2, 0, 1))
+        for flip in [0, 1]:
+            if flip == 1:
+                img = mx.ndarray.flip(data=img, axis=2)
+            data_list[flip][idx][:] = torch.from_numpy(img.asnumpy())
+        if idx % 1000 == 0:
+            print('loading bin', idx)
+    print(data_list[0].shape)
+    return data_list, issame_list
+
+@torch.no_grad()
+def test(data_set, backbone, batch_size, nfolds=10):
+    print('testing verification..')
+    data_list = data_set[0]
+    issame_list = data_set[1]
+    embeddings_list = []
+    time_consumed = 0.0
+    for i in range(len(data_list)):
+        data = data_list[i]
+        embeddings = None
+        ba = 0
+        while ba < data.shape[0]:
+            bb = min(ba + batch_size, data.shape[0])
+            count = bb - ba
+            _data = data[bb - batch_size: bb]
+            time0 = datetime.datetime.now()
+            img = ((_data / 255) - 0.5) / 0.5
+            net_out: torch.Tensor = backbone(img)
+            _embeddings = net_out.detach().cpu().numpy()
+            time_now = datetime.datetime.now()
+            diff = time_now - time0
+            time_consumed += diff.total_seconds()
+            if embeddings is None:
+                embeddings = np.zeros((data.shape[0], _embeddings.shape[1]))
+            embeddings[ba:bb, :] = _embeddings[(batch_size - count):, :]
+            ba = bb
+        embeddings_list.append(embeddings)
+
+    _xnorm = 0.0
+    _xnorm_cnt = 0
+    for embed in embeddings_list:
+        for i in range(embed.shape[0]):
+            _em = embed[i]
+            _norm = np.linalg.norm(_em)
+            _xnorm += _norm
+            _xnorm_cnt += 1
+    _xnorm /= _xnorm_cnt
+
+    embeddings = embeddings_list[0].copy()
+    embeddings = sklearn.preprocessing.normalize(embeddings)
+    acc1 = 0.0
+    std1 = 0.0
+    embeddings = embeddings_list[0] + embeddings_list[1]
+    embeddings = sklearn.preprocessing.normalize(embeddings)
+    print(embeddings.shape)
+    print('infer time', time_consumed)
+    _, _, accuracy, val, val_std, far = evaluate(embeddings, issame_list, nrof_folds=nfolds)
+    acc2, std2 = np.mean(accuracy), np.std(accuracy)
+    return acc1, std1, acc2, std2, _xnorm, embeddings_list
+
+
+def dumpR(data_set,
+          backbone,
+          batch_size,
+          name='',
+          data_extra=None,
+          label_shape=None):
+    print('dump verification embedding..')
+    data_list = data_set[0]
+    issame_list = data_set[1]
+    embeddings_list = []
+    time_consumed = 0.0
+    for i in range(len(data_list)):
+        data = data_list[i]
+        embeddings = None
+        ba = 0
+        while ba < data.shape[0]:
+            bb = min(ba + batch_size, data.shape[0])
+            count = bb - ba
+
+            _data = nd.slice_axis(data, axis=0, begin=bb - batch_size, end=bb)
+            time0 = datetime.datetime.now()
+            if data_extra is None:
+                db = mx.io.DataBatch(data=(_data,), label=(_label,))
+            else:
+                db = mx.io.DataBatch(data=(_data, _data_extra),
+                                     label=(_label,))
+            model.forward(db, is_train=False)
+            net_out = model.get_outputs()
+            _embeddings = net_out[0].asnumpy()
+            time_now = datetime.datetime.now()
+            diff = time_now - time0
+            time_consumed += diff.total_seconds()
+            if embeddings is None:
+                embeddings = np.zeros((data.shape[0], _embeddings.shape[1]))
+            embeddings[ba:bb, :] = _embeddings[(batch_size - count):, :]
+            ba = bb
+        embeddings_list.append(embeddings)
+    embeddings = embeddings_list[0] + embeddings_list[1]
+    embeddings = sklearn.preprocessing.normalize(embeddings)
+    actual_issame = np.asarray(issame_list)
+    outname = os.path.join('temp.bin')
+    with open(outname, 'wb') as f:
+        pickle.dump((embeddings, issame_list),
+                    f,
+                    protocol=pickle.HIGHEST_PROTOCOL)
+
+
+# if __name__ == '__main__':
+#
+#     parser = argparse.ArgumentParser(description='do verification')
+#     # general
+#     parser.add_argument('--data-dir', default='', help='')
+#     parser.add_argument('--model',
+#                         default='../model/softmax,50',
+#                         help='path to load model.')
+#     parser.add_argument('--target',
+#                         default='lfw,cfp_ff,cfp_fp,agedb_30',
+#                         help='test targets.')
+#     parser.add_argument('--gpu', default=0, type=int, help='gpu id')
+#     parser.add_argument('--batch-size', default=32, type=int, help='')
+#     parser.add_argument('--max', default='', type=str, help='')
+#     parser.add_argument('--mode', default=0, type=int, help='')
+#     parser.add_argument('--nfolds', default=10, type=int, help='')
+#     args = parser.parse_args()
+#     image_size = [112, 112]
+#     print('image_size', image_size)
+#     ctx = mx.gpu(args.gpu)
+#     nets = []
+#     vec = args.model.split(',')
+#     prefix = args.model.split(',')[0]
+#     epochs = []
+#     if len(vec) == 1:
+#         pdir = os.path.dirname(prefix)
+#         for fname in os.listdir(pdir):
+#             if not fname.endswith('.params'):
+#                 continue
+#             _file = os.path.join(pdir, fname)
+#             if _file.startswith(prefix):
+#                 epoch = int(fname.split('.')[0].split('-')[1])
+#                 epochs.append(epoch)
+#         epochs = sorted(epochs, reverse=True)
+#         if len(args.max) > 0:
+#             _max = [int(x) for x in args.max.split(',')]
+#             assert len(_max) == 2
+#             if len(epochs) > _max[1]:
+#                 epochs = epochs[_max[0]:_max[1]]
+#
+#     else:
+#         epochs = [int(x) for x in vec[1].split('|')]
+#     print('model number', len(epochs))
+#     time0 = datetime.datetime.now()
+#     for epoch in epochs:
+#         print('loading', prefix, epoch)
+#         sym, arg_params, aux_params = mx.model.load_checkpoint(prefix, epoch)
+#         # arg_params, aux_params = ch_dev(arg_params, aux_params, ctx)
+#         all_layers = sym.get_internals()
+#         sym = all_layers['fc1_output']
+#         model = mx.mod.Module(symbol=sym, context=ctx, label_names=None)
+#         # model.bind(data_shapes=[('data', (args.batch_size, 3, image_size[0], image_size[1]))], label_shapes=[('softmax_label', (args.batch_size,))])
+#         model.bind(data_shapes=[('data', (args.batch_size, 3, image_size[0],
+#                                           image_size[1]))])
+#         model.set_params(arg_params, aux_params)
+#         nets.append(model)
+#     time_now = datetime.datetime.now()
+#     diff = time_now - time0
+#     print('model loading time', diff.total_seconds())
+#
+#     ver_list = []
+#     ver_name_list = []
+#     for name in args.target.split(','):
+#         path = os.path.join(args.data_dir, name + ".bin")
+#         if os.path.exists(path):
+#             print('loading.. ', name)
+#             data_set = load_bin(path, image_size)
+#             ver_list.append(data_set)
+#             ver_name_list.append(name)
+#
+#     if args.mode == 0:
+#         for i in range(len(ver_list)):
+#             results = []
+#             for model in nets:
+#                 acc1, std1, acc2, std2, xnorm, embeddings_list = test(
+#                     ver_list[i], model, args.batch_size, args.nfolds)
+#                 print('[%s]XNorm: %f' % (ver_name_list[i], xnorm))
+#                 print('[%s]Accuracy: %1.5f+-%1.5f' % (ver_name_list[i], acc1, std1))
+#                 print('[%s]Accuracy-Flip: %1.5f+-%1.5f' % (ver_name_list[i], acc2, std2))
+#                 results.append(acc2)
+#             print('Max of [%s] is %1.5f' % (ver_name_list[i], np.max(results)))
+#     elif args.mode == 1:
+#         raise ValueError
+#     else:
+#         model = nets[0]
+#         dumpR(ver_list[0], model, args.batch_size, args.target)
diff --git a/eval_ijbc.py b/eval_ijbc.py
new file mode 100644
index 0000000..c144e4e
--- /dev/null
+++ b/eval_ijbc.py
@@ -0,0 +1,483 @@
+# coding: utf-8
+
+import os
+import pickle
+
+import matplotlib
+import pandas as pd
+
+matplotlib.use('Agg')
+import matplotlib.pyplot as plt
+import timeit
+import sklearn
+import argparse
+from sklearn.metrics import roc_curve, auc
+
+from menpo.visualize.viewmatplotlib import sample_colours_from_colourmap
+from prettytable import PrettyTable
+from pathlib import Path
+import sys
+import warnings
+
+sys.path.insert(0, "../")
+warnings.filterwarnings("ignore")
+
+parser = argparse.ArgumentParser(description='do ijb test')
+# general
+parser.add_argument('--model-prefix', default='', help='path to load model.')
+parser.add_argument('--image-path', default='', type=str, help='')
+parser.add_argument('--result-dir', default='.', type=str, help='')
+parser.add_argument('--batch-size', default=128, type=int, help='')
+parser.add_argument('--network', default='iresnet50', type=str, help='')
+parser.add_argument('--job', default='insightface', type=str, help='job name')
+parser.add_argument('--target', default='IJBC', type=str, help='target, set to IJBC or IJBB')
+args = parser.parse_args()
+
+target = args.target
+model_path = args.model_prefix
+image_path = args.image_path
+result_dir = args.result_dir
+gpu_id = None
+use_norm_score = True  # if Ture, TestMode(N1)
+use_detector_score = True  # if Ture, TestMode(D1)
+use_flip_test = True  # if Ture, TestMode(F1)
+job = args.job
+batch_size = args.batch_size
+
+import cv2
+import numpy as np
+import torch
+from skimage import transform as trans
+import backbones
+
+
+class Embedding(object):
+    def __init__(self, prefix, data_shape, batch_size=1):
+        image_size = (112, 112)
+        self.image_size = image_size
+        weight = torch.load(prefix)
+        resnet = eval("backbones.{}".format(args.network))(False).cuda()
+        resnet.load_state_dict(weight)
+        model = torch.nn.DataParallel(resnet)
+        self.model = model
+        self.model.eval()
+        src = np.array([
+            [30.2946, 51.6963],
+            [65.5318, 51.5014],
+            [48.0252, 71.7366],
+            [33.5493, 92.3655],
+            [62.7299, 92.2041]], dtype=np.float32)
+        src[:, 0] += 8.0
+        self.src = src
+        self.batch_size = batch_size
+        self.data_shape = data_shape
+
+    def get(self, rimg, landmark):
+
+        assert landmark.shape[0] == 68 or landmark.shape[0] == 5
+        assert landmark.shape[1] == 2
+        if landmark.shape[0] == 68:
+            landmark5 = np.zeros((5, 2), dtype=np.float32)
+            landmark5[0] = (landmark[36] + landmark[39]) / 2
+            landmark5[1] = (landmark[42] + landmark[45]) / 2
+            landmark5[2] = landmark[30]
+            landmark5[3] = landmark[48]
+            landmark5[4] = landmark[54]
+        else:
+            landmark5 = landmark
+        tform = trans.SimilarityTransform()
+        tform.estimate(landmark5, self.src)
+        M = tform.params[0:2, :]
+        img = cv2.warpAffine(rimg,
+                             M, (self.image_size[1], self.image_size[0]),
+                             borderValue=0.0)
+        img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
+        img_flip = np.fliplr(img)
+        img = np.transpose(img, (2, 0, 1))  # 3*112*112, RGB
+        img_flip = np.transpose(img_flip, (2, 0, 1))
+        input_blob = np.zeros((2, 3, self.image_size[1], self.image_size[0]), dtype=np.uint8)
+        input_blob[0] = img
+        input_blob[1] = img_flip
+        return input_blob
+
+    @torch.no_grad()
+    def forward_db(self, batch_data):
+        imgs = torch.Tensor(batch_data).cuda()
+        imgs.div_(255).sub_(0.5).div_(0.5)
+        feat = self.model(imgs)
+        feat = feat.reshape([self.batch_size, 2 * feat.shape[1]])
+        return feat.cpu().numpy()
+
+
+# 将一个list尽量均分成n份，限制len(list)==n，份数大于原list内元素个数则分配空list[]
+def divideIntoNstrand(listTemp, n):
+    twoList = [[] for i in range(n)]
+    for i, e in enumerate(listTemp):
+        twoList[i % n].append(e)
+    return twoList
+
+
+def read_template_media_list(path):
+    # ijb_meta = np.loadtxt(path, dtype=str)
+    ijb_meta = pd.read_csv(path, sep=' ', header=None).values
+    templates = ijb_meta[:, 1].astype(np.int)
+    medias = ijb_meta[:, 2].astype(np.int)
+    return templates, medias
+
+
+# In[ ]:
+
+
+def read_template_pair_list(path):
+    # pairs = np.loadtxt(path, dtype=str)
+    pairs = pd.read_csv(path, sep=' ', header=None).values
+    # print(pairs.shape)
+    # print(pairs[:, 0].astype(np.int))
+    t1 = pairs[:, 0].astype(np.int)
+    t2 = pairs[:, 1].astype(np.int)
+    label = pairs[:, 2].astype(np.int)
+    return t1, t2, label
+
+
+# In[ ]:
+
+
+def read_image_feature(path):
+    with open(path, 'rb') as fid:
+        img_feats = pickle.load(fid)
+    return img_feats
+
+
+# In[ ]:
+
+
+def get_image_feature(img_path, files_list, model_path, epoch, gpu_id):
+    batch_size = args.batch_size
+    data_shape = (3, 112, 112)
+
+    files = files_list
+    print('files:', len(files))
+    rare_size = len(files) % batch_size
+    faceness_scores = []
+    batch = 0
+    img_feats = np.empty((len(files), 1024), dtype=np.float32)
+
+    batch_data = np.empty((2 * batch_size, 3, 112, 112))
+    embedding = Embedding(model_path, data_shape, batch_size)
+    for img_index, each_line in enumerate(files[:len(files) - rare_size]):
+        name_lmk_score = each_line.strip().split(' ')
+        img_name = os.path.join(img_path, name_lmk_score[0])
+        img = cv2.imread(img_name)
+        lmk = np.array([float(x) for x in name_lmk_score[1:-1]],
+                       dtype=np.float32)
+        lmk = lmk.reshape((5, 2))
+        input_blob = embedding.get(img, lmk)
+
+        batch_data[2 * (img_index - batch * batch_size)][:] = input_blob[0]
+        batch_data[2 * (img_index - batch * batch_size) + 1][:] = input_blob[1]
+        if (img_index + 1) % batch_size == 0:
+            print('batch', batch)
+            img_feats[batch * batch_size:batch * batch_size +
+                                         batch_size][:] = embedding.forward_db(batch_data)
+            batch += 1
+        faceness_scores.append(name_lmk_score[-1])
+
+    batch_data = np.empty((2 * rare_size, 3, 112, 112))
+    embedding = Embedding(model_path, data_shape, rare_size)
+    for img_index, each_line in enumerate(files[len(files) - rare_size:]):
+        name_lmk_score = each_line.strip().split(' ')
+        img_name = os.path.join(img_path, name_lmk_score[0])
+        img = cv2.imread(img_name)
+        lmk = np.array([float(x) for x in name_lmk_score[1:-1]],
+                       dtype=np.float32)
+        lmk = lmk.reshape((5, 2))
+        input_blob = embedding.get(img, lmk)
+        batch_data[2 * img_index][:] = input_blob[0]
+        batch_data[2 * img_index + 1][:] = input_blob[1]
+        if (img_index + 1) % rare_size == 0:
+            print('batch', batch)
+            img_feats[len(files) -
+                      rare_size:][:] = embedding.forward_db(batch_data)
+            batch += 1
+        faceness_scores.append(name_lmk_score[-1])
+    faceness_scores = np.array(faceness_scores).astype(np.float32)
+    # img_feats = np.ones( (len(files), 1024), dtype=np.float32) * 0.01
+    # faceness_scores = np.ones( (len(files), ), dtype=np.float32 )
+    return img_feats, faceness_scores
+
+
+# In[ ]:
+
+
+def image2template_feature(img_feats=None, templates=None, medias=None):
+    # ==========================================================
+    # 1. face image feature l2 normalization. img_feats:[number_image x feats_dim]
+    # 2. compute media feature.
+    # 3. compute template feature.
+    # ==========================================================
+    unique_templates = np.unique(templates)
+    template_feats = np.zeros((len(unique_templates), img_feats.shape[1]))
+
+    for count_template, uqt in enumerate(unique_templates):
+
+        (ind_t,) = np.where(templates == uqt)
+        face_norm_feats = img_feats[ind_t]
+        face_medias = medias[ind_t]
+        unique_medias, unique_media_counts = np.unique(face_medias,
+                                                       return_counts=True)
+        media_norm_feats = []
+        for u, ct in zip(unique_medias, unique_media_counts):
+            (ind_m,) = np.where(face_medias == u)
+            if ct == 1:
+                media_norm_feats += [face_norm_feats[ind_m]]
+            else:  # image features from the same video will be aggregated into one feature
+                media_norm_feats += [
+                    np.mean(face_norm_feats[ind_m], axis=0, keepdims=True)
+                ]
+        media_norm_feats = np.array(media_norm_feats)
+        # media_norm_feats = media_norm_feats / np.sqrt(np.sum(media_norm_feats ** 2, -1, keepdims=True))
+        template_feats[count_template] = np.sum(media_norm_feats, axis=0)
+        if count_template % 2000 == 0:
+            print('Finish Calculating {} template features.'.format(
+                count_template))
+    # template_norm_feats = template_feats / np.sqrt(np.sum(template_feats ** 2, -1, keepdims=True))
+    template_norm_feats = sklearn.preprocessing.normalize(template_feats)
+    # print(template_norm_feats.shape)
+    return template_norm_feats, unique_templates
+
+
+# In[ ]:
+
+
+def verification(template_norm_feats=None,
+                 unique_templates=None,
+                 p1=None,
+                 p2=None):
+    # ==========================================================
+    #         Compute set-to-set Similarity Score.
+    # ==========================================================
+    template2id = np.zeros((max(unique_templates) + 1, 1), dtype=int)
+    for count_template, uqt in enumerate(unique_templates):
+        template2id[uqt] = count_template
+
+    score = np.zeros((len(p1),))  # save cosine distance between pairs
+
+    total_pairs = np.array(range(len(p1)))
+    batchsize = 100000  # small batchsize instead of all pairs in one batch due to the memory limiation
+    sublists = [
+        total_pairs[i:i + batchsize] for i in range(0, len(p1), batchsize)
+    ]
+    total_sublists = len(sublists)
+    for c, s in enumerate(sublists):
+        feat1 = template_norm_feats[template2id[p1[s]]]
+        feat2 = template_norm_feats[template2id[p2[s]]]
+        similarity_score = np.sum(feat1 * feat2, -1)
+        score[s] = similarity_score.flatten()
+        if c % 10 == 0:
+            print('Finish {}/{} pairs.'.format(c, total_sublists))
+    return score
+
+
+# In[ ]:
+def verification2(template_norm_feats=None,
+                  unique_templates=None,
+                  p1=None,
+                  p2=None):
+    template2id = np.zeros((max(unique_templates) + 1, 1), dtype=int)
+    for count_template, uqt in enumerate(unique_templates):
+        template2id[uqt] = count_template
+    score = np.zeros((len(p1),))  # save cosine distance between pairs
+    total_pairs = np.array(range(len(p1)))
+    batchsize = 100000  # small batchsize instead of all pairs in one batch due to the memory limiation
+    sublists = [
+        total_pairs[i:i + batchsize] for i in range(0, len(p1), batchsize)
+    ]
+    total_sublists = len(sublists)
+    for c, s in enumerate(sublists):
+        feat1 = template_norm_feats[template2id[p1[s]]]
+        feat2 = template_norm_feats[template2id[p2[s]]]
+        similarity_score = np.sum(feat1 * feat2, -1)
+        score[s] = similarity_score.flatten()
+        if c % 10 == 0:
+            print('Finish {}/{} pairs.'.format(c, total_sublists))
+    return score
+
+
+def read_score(path):
+    with open(path, 'rb') as fid:
+        img_feats = pickle.load(fid)
+    return img_feats
+
+
+# # Step1: Load Meta Data
+
+# In[ ]:
+
+assert target == 'IJBC' or target == 'IJBB'
+
+# =============================================================
+# load image and template relationships for template feature embedding
+# tid --> template id,  mid --> media id
+# format:
+#           image_name tid mid
+# =============================================================
+start = timeit.default_timer()
+templates, medias = read_template_media_list(
+    os.path.join('%s/meta' % image_path,
+                 '%s_face_tid_mid.txt' % target.lower()))
+stop = timeit.default_timer()
+print('Time: %.2f s. ' % (stop - start))
+
+# In[ ]:
+
+# =============================================================
+# load template pairs for template-to-template verification
+# tid : template id,  label : 1/0
+# format:
+#           tid_1 tid_2 label
+# =============================================================
+start = timeit.default_timer()
+p1, p2, label = read_template_pair_list(
+    os.path.join('%s/meta' % image_path,
+                 '%s_template_pair_label.txt' % target.lower()))
+stop = timeit.default_timer()
+print('Time: %.2f s. ' % (stop - start))
+
+# # Step 2: Get Image Features
+
+# In[ ]:
+
+# =============================================================
+# load image features
+# format:
+#           img_feats: [image_num x feats_dim] (227630, 512)
+# =============================================================
+start = timeit.default_timer()
+img_path = '%s/loose_crop' % image_path
+img_list_path = '%s/meta/%s_name_5pts_score.txt' % (image_path, target.lower())
+img_list = open(img_list_path)
+files = img_list.readlines()
+# files_list = divideIntoNstrand(files, rank_size)
+files_list = files
+
+# img_feats
+# for i in range(rank_size):
+img_feats, faceness_scores = get_image_feature(img_path, files_list,
+                                               model_path, 0, gpu_id)
+stop = timeit.default_timer()
+print('Time: %.2f s. ' % (stop - start))
+print('Feature Shape: ({} , {}) .'.format(img_feats.shape[0],
+                                          img_feats.shape[1]))
+
+# # Step3: Get Template Features
+
+# In[ ]:
+
+# =============================================================
+# compute template features from image features.
+# =============================================================
+start = timeit.default_timer()
+# ==========================================================
+# Norm feature before aggregation into template feature?
+# Feature norm from embedding network and faceness score are able to decrease weights for noise samples (not face).
+# ==========================================================
+# 1. FaceScore （Feature Norm）
+# 2. FaceScore （Detector）
+
+if use_flip_test:
+    # concat --- F1
+    # img_input_feats = img_feats
+    # add --- F2
+    img_input_feats = img_feats[:, 0:img_feats.shape[1] //
+                                     2] + img_feats[:, img_feats.shape[1] // 2:]
+else:
+    img_input_feats = img_feats[:, 0:img_feats.shape[1] // 2]
+
+if use_norm_score:
+    img_input_feats = img_input_feats
+else:
+    # normalise features to remove norm information
+    img_input_feats = img_input_feats / np.sqrt(
+        np.sum(img_input_feats ** 2, -1, keepdims=True))
+
+if use_detector_score:
+    print(img_input_feats.shape, faceness_scores.shape)
+    img_input_feats = img_input_feats * faceness_scores[:, np.newaxis]
+else:
+    img_input_feats = img_input_feats
+
+template_norm_feats, unique_templates = image2template_feature(
+    img_input_feats, templates, medias)
+stop = timeit.default_timer()
+print('Time: %.2f s. ' % (stop - start))
+
+# # Step 4: Get Template Similarity Scores
+
+# In[ ]:
+
+# =============================================================
+# compute verification scores between template pairs.
+# =============================================================
+start = timeit.default_timer()
+score = verification(template_norm_feats, unique_templates, p1, p2)
+stop = timeit.default_timer()
+print('Time: %.2f s. ' % (stop - start))
+
+# In[ ]:
+save_path = os.path.join(result_dir, args.job)
+# save_path = result_dir + '/%s_result' % target
+
+if not os.path.exists(save_path):
+    os.makedirs(save_path)
+
+score_save_file = os.path.join(save_path, "%s.npy" % target.lower())
+np.save(score_save_file, score)
+
+# # Step 5: Get ROC Curves and TPR@FPR Table
+
+# In[ ]:
+
+files = [score_save_file]
+methods = []
+scores = []
+for file in files:
+    methods.append(Path(file).stem)
+    scores.append(np.load(file))
+
+methods = np.array(methods)
+scores = dict(zip(methods, scores))
+colours = dict(
+    zip(methods, sample_colours_from_colourmap(methods.shape[0], 'Set2')))
+x_labels = [10 ** -6, 10 ** -5, 10 ** -4, 10 ** -3, 10 ** -2, 10 ** -1]
+tpr_fpr_table = PrettyTable(['Methods'] + [str(x) for x in x_labels])
+fig = plt.figure()
+for method in methods:
+    fpr, tpr, _ = roc_curve(label, scores[method])
+    roc_auc = auc(fpr, tpr)
+    fpr = np.flipud(fpr)
+    tpr = np.flipud(tpr)  # select largest tpr at same fpr
+    plt.plot(fpr,
+             tpr,
+             color=colours[method],
+             lw=1,
+             label=('[%s (AUC = %0.4f %%)]' %
+                    (method.split('-')[-1], roc_auc * 100)))
+    tpr_fpr_row = []
+    tpr_fpr_row.append("%s-%s" % (method, target))
+    for fpr_iter in np.arange(len(x_labels)):
+        _, min_index = min(
+            list(zip(abs(fpr - x_labels[fpr_iter]), range(len(fpr)))))
+        tpr_fpr_row.append('%.2f' % (tpr[min_index] * 100))
+    tpr_fpr_table.add_row(tpr_fpr_row)
+plt.xlim([10 ** -6, 0.1])
+plt.ylim([0.3, 1.0])
+plt.grid(linestyle='--', linewidth=1)
+plt.xticks(x_labels)
+plt.yticks(np.linspace(0.3, 1.0, 8, endpoint=True))
+plt.xscale('log')
+plt.xlabel('False Positive Rate')
+plt.ylabel('True Positive Rate')
+plt.title('ROC on IJB')
+plt.legend(loc="lower right")
+fig.savefig(os.path.join(save_path, '%s.pdf' % target.lower()))
+print(tpr_fpr_table)
diff --git a/face_api.py b/face_api.py
new file mode 100644
index 0000000..f4c3197
--- /dev/null
+++ b/face_api.py
@@ -0,0 +1,377 @@
+import os
+import time
+import re
+import torch
+import cv2
+import numpy as np
+
+from anti import anti_spoofing, load_anti_model
+from backbones import iresnet50, iresnet18, iresnet100
+from retinaface_detect import load_retinaface_model, detect_one, detect_video, set_retinaface_conf
+from torch2trt import torch2trt, TRTModule
+
+threshold = 0.7
+
+
+# 读取112x112的本地图片并变换通道位置归一化
+def load_image(img_path):
+    img = cv2.imdecode(np.fromfile(img_path, dtype=np.uint8), cv2.IMREAD_COLOR)
+    img = img.transpose((2, 0, 1))
+    img = img[np.newaxis, :, :, :]
+    img = np.array(img, dtype=np.float32)
+    img -= 127.5
+    img /= 127.5
+    return img
+
+
+# 计算两个特征向量的欧式距离
+def findEuclideanDistance(source_representation, test_representation):
+    euclidean_distance = source_representation - test_representation
+    euclidean_distance = np.sum(np.multiply(euclidean_distance, euclidean_distance))
+    euclidean_distance = np.sqrt(euclidean_distance)
+    return euclidean_distance
+
+
+# 计算两个特征向量的余弦距离
+def findCosineDistance(source_representation, test_representation):
+    a = np.matmul(np.transpose(source_representation), test_representation)
+    b = np.sum(np.multiply(source_representation, source_representation))
+    c = np.sum(np.multiply(test_representation, test_representation))
+    return 1 - (a / (np.sqrt(b) * np.sqrt(c)))
+
+
+# 归一化欧氏距离
+def l2_normalize(x):
+    return x / np.sqrt(np.sum(np.multiply(x, x)))
+
+
+# 归一化余弦距离
+def cosin_metric(x1, x2):
+    return np.dot(x1, x2) / (np.linalg.norm(x1) * np.linalg.norm(x2))
+
+
+# 加载保存的姓名、人脸特征向量的人脸库
+def load_npy(path):
+    data = np.load(path, allow_pickle=True)
+    data = data.item()
+    return data
+
+
+# 批量化生成人脸特征向量并保存到人脸库
+def create_database_batch(path, model, database_path):
+    name_list = os.listdir(path)
+    k_v = {}
+    if os.path.exists(database_path):
+        k_v = np.load(database_path, allow_pickle=True)
+        k_v = k_v.item()
+    batch = 256
+    order_name = []
+    order_path = []
+    emb_list = []
+    for name in name_list[:]:
+        img_path = os.path.join(path, name)
+        # for img_name in img_path[:1]:
+        order_name.append(name[:-4])
+        order_path.append(img_path)
+    order_img = np.zeros((len(order_path), 3, 112, 112), dtype=np.float32)
+    for index, img_path in enumerate(order_path):
+        order_img[index] = load_image(img_path)
+    print(order_img.shape)
+    order_img = torch.from_numpy(order_img)
+    order_img = order_img.to(torch.device("cuda" if cpu_or_cuda == "cuda" else "cpu"))
+    now = 0
+    number = len(order_img)
+    with torch.no_grad():
+        while now < number:
+            if now + batch < number:
+                emb = model(order_img[now:now + batch])
+            else:
+                emb = model(order_img[now:])
+            now = now + batch
+            emb = emb.cpu().numpy()
+            for em in emb:
+                emb_list.append(em)
+            print("batch" + str(now))
+
+    for i, emb in enumerate(emb_list):
+        k_v[order_name[i]] = l2_normalize(emb)
+    np.save(database_path, k_v)
+
+def create_database_from_img(order_name, order_img, model, database_path, cpu_or_cuda):
+    k_v = {}
+    if os.path.exists(database_path):
+        k_v = np.load(database_path, allow_pickle=True)
+        k_v = k_v.item()
+    batch = 256
+    emb_list = []
+
+    print(order_img.shape)
+    order_img = torch.from_numpy(order_img)
+    order_img = order_img.to(torch.device("cuda" if cpu_or_cuda == "cuda" else "cpu"))
+    now = 0
+    number = len(order_img)
+    with torch.no_grad():
+        while now < number:
+            if now + batch < number:
+                emb = model(order_img[now:now + batch])
+            else:
+                emb = model(order_img[now:])
+            now = now + batch
+            emb = emb.cpu().numpy()
+            for em in emb:
+                emb_list.append(em)
+            print("batch" + str(now))
+    for i, emb in enumerate(emb_list):
+        k_v[order_name[i]] = l2_normalize(emb)
+    np.save(database_path, k_v)
+
+# 向人脸库中新增一个人的姓名和人脸特征向量，若人脸库不存在则创建
+def add_one_to_database(img, model, name, database_path, cpu_or_cuda):
+    img = torch.from_numpy(img)
+    img = img.to(torch.device("cuda" if cpu_or_cuda == "cuda" else "cpu"))
+    with torch.no_grad():
+        pred = model(img)
+        pred = pred.cpu().numpy()
+        k_v = {}
+        if os.path.exists(database_path):
+            k_v = np.load(database_path, allow_pickle=True)
+            k_v = k_v.item()
+        k_v[name] = l2_normalize(pred)
+        np.save(database_path, k_v)
+
+
+# 计算此特征向量与人脸库中的哪个人脸特征向量距离最近
+def findmindistance(pred, threshold, k_v):
+    distance = 10
+    most_like = ""
+    for name in k_v.keys():
+        tmp = findEuclideanDistance(k_v[name], pred)
+        if distance > tmp:
+            distance = tmp
+            most_like = name
+    if distance < threshold:
+        return most_like, distance
+    else:
+        return -1, distance
+
+
+def faiss_find_face(pred, index, database_name_list):
+    name_list = []
+    start_time = time.time()
+    D, I = index.search(pred, 1)
+    end_time = time.time()
+    # print("faiss cost %fs" % (end_time - start_time))
+    # print(D, I)
+    if len(pred) == 1:
+        if D[0][0] < threshold:
+            # print(database_name_list[I[0][0]])
+            return database_name_list[I[0][0]], D[0][0]
+        else:
+            return "unknown", D[0][0]
+    else:
+        for i,index in enumerate(I):
+            if D[i][0] < threshold:
+                #print(database_name_list[I[0][0]])
+                name_list.append(database_name_list[index[0]]+str(D[i][0]))
+            else:
+                name_list.append("unknown"+str(D[i][0]))
+        return name_list
+
+
+# 从人脸库中找到单个人脸
+def findOne(img, model, index, database_name_list, cpu_or_cuda):
+    img = torch.from_numpy(img)
+    img = img.to(torch.device("cuda" if cpu_or_cuda == "cuda" else "cpu"))
+    with torch.no_grad():
+        start_time = time.time()
+        pred = model(img)
+        end_time = time.time()
+        print("predOne time: " + str(end_time - start_time))
+        pred = pred.cpu().numpy()
+        # start_time = time.time()
+        # name, distance = findmindistance(l2_normalize(pred), threshold=threshold, k_v=k_v)
+        # end_time = time.time()
+        # print("baoli time: " + str(end_time - start_time))
+        name, distance = faiss_find_face(l2_normalize(pred), index, database_name_list)
+        print(pred.shape)
+        if name != -1:
+            mo = r'[\u4e00-\u9fa5_a-zA-Z0-9]*'
+            name = re.match(mo, name)
+            return name.group(0), distance
+        else:
+            return "unknown", distance
+
+
+# 从人脸库中找到传入的人脸列表中的所有人脸
+def findAll(imglist, model, index ,database_name_list, cpu_or_cuda):
+    imglist = torch.from_numpy(imglist)
+    imglist = imglist.to(torch.device("cuda" if cpu_or_cuda == "cuda" else "cpu"))
+    with torch.no_grad():
+        name_list =[]
+        start_time = time.time()
+        pred = model(imglist)
+        end_time = time.time()
+        print("predOne time: " + str(end_time - start_time))
+        pred = pred.cpu().numpy()
+        start_time = time.time()
+        #name_list = faiss_find_face(l2_normalize(pred), index, database_name_list)
+        for pr in pred:
+            pr = np.expand_dims(l2_normalize(pr), 0)
+            #     #print(pr.shape)
+            name, distance = faiss_find_face(l2_normalize(pr), index, database_name_list)
+            #name_list.append(name+" "+str(distance))
+            name_list.append(name)
+        # for pr in pred:
+        #     name, distance = findmindistance(l2_normalize(pr), threshold=threshold, k_v=k_v)
+        #     if name != -1:
+        #         mo = r'[\u4e00-\u9fa5_a-zA-Z]*'
+        #         name = re.match(mo, name)
+        #         name_list.append(name.group(0) + str(distance))
+        #     else:
+        #         name_list.append("unknown" + str(distance))
+        end_time = time.time()
+        print("searchALL time: " + str(end_time - start_time))
+        return name_list
+
+
+# 提取为512维特征向量
+def embedding(order_img, model, cpu_or_cuda):
+    number = len(order_img)
+    order_img = torch.from_numpy(order_img)
+    order_img = order_img.to(torch.device("cuda" if cpu_or_cuda == "cuda" else "cpu"))
+    batch = 64
+    emb_list = []
+    now = 0
+    with torch.no_grad():
+        while now < number:
+            if now + batch < number:
+                emb = model(order_img[now:now + batch])
+            else:
+                emb = model(order_img[now:])
+            now = now + batch
+            emb = emb.cpu().numpy()
+            for em in emb:
+                emb_list.append(l2_normalize(em))
+            # print("batch" + str(now))
+        emb_list = np.array(emb_list)
+        return emb_list
+
+
+# 处理聚类人脸文件夹，返回特征向量列表，文件名列表
+def get_claster_tmp_file_embedding(file_path, retinaface_model, retinaface_args, arcface_model, cpu_or_cuda):
+    img_name = os.listdir(file_path)
+    img_list = []
+    for name in img_name:
+        all_face, box_and_point = detect_one(os.path.join(file_path, name), retinaface_model, retinaface_args)
+        img_list.append(all_face[0])
+    img_list = np.array(img_list)
+    # print(img_list.shape)
+    emb_list = embedding(img_list, arcface_model, cpu_or_cuda)
+    return emb_list, img_name
+
+
+# 同一个人聚为一类
+def cluster(emb_list, name_list):
+    all_claster = []
+    cla = []
+    in_claster_name = []
+    img_number = len(emb_list)
+    for index, emb in enumerate(emb_list):
+        if name_list[index] in in_claster_name:
+            continue
+        for j in range(img_number - index - 1):
+            if findEuclideanDistance(emb, emb_list[index + 1 + j]) < threshold:
+                if name_list[index + 1 + j] not in in_claster_name:
+                    cla.append(name_list[index + 1 + j])
+                    in_claster_name.append(name_list[index + 1 + j])
+        cla.append(name_list[index])
+        in_claster_name.append(name_list[index])
+        all_claster.append(cla)
+        cla = []
+    return all_claster
+
+
+# 加载人脸识别模型
+def load_arcface_model(model_path, cpu_or_cuda):
+    if cpu_or_cuda == "trt":
+        model = TRTModule()
+        model.load_state_dict(torch.load('./model/arcface_trt.pth'))
+    elif cpu_or_cuda == "trt_new":
+        model = iresnet100()
+        model.load_state_dict(torch.load(model_path, map_location="cuda"))
+        model = model.eval()
+        model.to(torch.device("cuda"))
+        x = torch.ones((1, 3, 112, 112)).to(torch.device("cuda"))
+        model = torch2trt(model, [x], max_batch_size=4)
+        torch.save(model.state_dict(), './model/arcface_trt.pth')
+    else:
+        model = iresnet100()
+        model.load_state_dict(torch.load(model_path, map_location=cpu_or_cuda))
+        model = model.eval()
+        model.to(torch.device("cuda" if cpu_or_cuda == "cuda" else "cpu"))
+    return model
+
+
+# 对比两张人脸是否相同
+def face_verification(img1, img2, model, cpu_or_cuda):
+    img_list = np.concatenate((img1, img2), axis=0)
+    img_list = torch.from_numpy(img_list)
+    img_list = img_list.to(torch.device("cuda" if cpu_or_cuda == "cuda" else "cpu"))
+    with torch.no_grad():
+        pred = model(img_list)
+        pred = pred.cpu().numpy()
+        distance = findEuclideanDistance(l2_normalize(pred[0]), l2_normalize(pred[1]))
+        # print("EuclideanDistance is :" + str(distance))
+        if distance < threshold:
+            return 'same ',distance
+        else:
+            return 'different ', distance
+
+
+if __name__ == '__main__':
+    cpu_or_cuda = "cuda" if torch.cuda.is_available() else "cpu"
+    arcface_model = load_arcface_model("./model/backbone100.pth", cpu_or_cuda=cpu_or_cuda)
+    # retinaface_args = set_retinaface_conf(cpu_or_cuda=cpu_or_cuda)
+    # retinaface_model = load_retinaface_model(retinaface_args)
+    #
+    # anti_spoofing_model_path = "model/anti_spoof_models"
+    # anti_model = load_anti_model(anti_spoofing_model_path, 0)
+    #
+    # k_v = load_npy("./Database/student.npy")
+    # 对比两张人脸
+    # img1, box_and_point = detect_one("D:\Download\lfw\lfw\Aaron_Peirsol\Aaron_Peirsol_0001.jpg", retinaface_model, retinaface_args)
+    # img2, box_and_point = detect_one("D:\Download\lfw\lfw\Aaron_Peirsol\Aaron_Peirsol_0002.jpg", retinaface_model, retinaface_args)
+    # print(face_verification(img1, img2, arcface_model))
+
+    # img3 = load_image(r"D:\Download\out\alig_students\student.jpg")
+    # img3 = torch.from_numpy(img3)
+    # 单张人脸活体检测
+    # img3, b_p = detect_one(r"C:\Users\ASUS\Desktop\face\IMG_20210525_113950.jpg", retinaface_model, retinaface_args)
+    # b = b_p[0]
+    # w = b[2] - b[0]
+    # h = b[3] - b[1]
+    # b[2] = w
+    # b[3] = h
+    # label, value = anti_spoofing("./img/recognition/000_0.bmp", "model/anti_spoof_models", 0, np.array(b[:4], int), anti_model)
+    # print(label,value)
+    # name = findOne(img3, arcface_model, k_v, cpu_or_cuda)
+    # print(name)
+
+    # 人脸聚类
+    # emb_list, name_list = get_claster_tmp_file_embedding("./img/cluster_tmp_file/face", retinaface_model,
+    #                                                      retinaface_args, arcface_model, cpu_or_cuda)
+    # print(cluster(emb_list, name_list))
+
+    # img3, box_and_point = detect_one("D:\Download\out\students\student.jpg", retinaface_model, retinaface_args)
+    # print(embedding(img3,arcface_model).shape)
+
+    # 人脸库中增加一张人脸
+    # add_one_to_database(img1,arcface_model,"Aaron_Peirsol","./Database/student.npy")
+    # name = findOne(img1, arcface_model, k_v)
+    # print(name)
+
+    # 人脸库中批量增加人脸
+    create_database_batch(r"D:\Download\out\alig_students_all", arcface_model, "./Database/sfz.npy")
+
+    # 识别视频中的人脸
+    # detect_video("software.mp4","out.avi",retinaface_model,arcface_model,k_v,retinaface_args)
diff --git a/font.ttf b/font.ttf
new file mode 100644
index 0000000..6f92880
Binary files /dev/null and b/font.ttf differ
diff --git a/gender_age.py b/gender_age.py
new file mode 100644
index 0000000..d13b1ec
--- /dev/null
+++ b/gender_age.py
@@ -0,0 +1,98 @@
+import datetime
+import mxnet as mx
+import numpy as np
+from retinaface_detect import detect_one, load_retinaface_model, set_retinaface_conf
+
+
+# 年龄性别配置
+class ConfGenderModel(object):
+    def __init__(self, image_size, image, model, gpu, det):
+        self.image_size = image_size
+        self.image = image
+        self.gpu = gpu
+        self.model = model
+        self.det = det
+
+
+# 实例化一个配置
+def set_gender_conf():
+    args = ConfGenderModel(image_size='112,112',
+                           image=r'C:\Users\ASUS\Desktop\man.png',
+                           gpu=-1,
+                           model='model/model,0',
+                           det=0)
+    return args
+
+
+# 加载性别年龄模型
+def load_gender_model(args, layer):
+    if args.gpu >= 0:
+        ctx = mx.gpu(args.gpu)
+    else:
+        ctx = mx.cpu()
+    _vec = args.image_size.split(',')
+    assert len(_vec) == 2
+    image_size = (int(_vec[0]), int(_vec[1]))
+
+    _vec = args.model.split(',')
+    assert len(_vec) == 2
+    prefix = _vec[0]
+    epoch = int(_vec[1])
+    print('loading', prefix, epoch)
+    sym, arg_params, aux_params = mx.model.load_checkpoint(prefix, epoch)
+    all_layers = sym.get_internals()
+    sym = all_layers[layer + '_output']
+    model = mx.mod.Module(symbol=sym, context=ctx, label_names=None)
+    model.bind(data_shapes=[('data', (1, 3, image_size[0], image_size[1]))])
+    model.set_params(arg_params, aux_params)
+    return model
+
+
+# 前向推理
+def get_ga(model, img):
+    # print(data)
+    model.forward(img, is_train=False)
+    ret = model.get_outputs()[0].asnumpy()
+    g = ret[:, 0:2].flatten()
+    gender = np.argmax(g)
+    a = ret[:, 2:202].reshape((100, 2))
+    a = np.argmax(a, axis=1)
+    age = int(sum(a))
+    return gender, age
+
+
+# 预测人脸列表中每个人的性别年龄
+def gender_age(img_list, gender_model):
+    gender_list = []
+    age_list = []
+    if len(img_list) == 0:
+        print("find no face")
+    else:
+        time_now = datetime.datetime.now()
+        img_list *= 127.5
+        img_list += 127.5
+
+        for img in img_list:
+            img = np.expand_dims(img, axis=0)
+            img = mx.nd.array(img)
+            img = mx.io.DataBatch(data=(img,))
+            gender, age = get_ga(gender_model, img)
+            if gender == 1:
+                gender_list.append("man")
+            else:
+                gender_list.append('woman')
+            age_list.append(age)
+        time_now2 = datetime.datetime.now()
+        diff = time_now2 - time_now
+        print('time cost', diff.total_seconds())
+    return gender_list,age_list
+
+
+if __name__ == "__main__":
+    args = set_gender_conf()
+    retinaface_args = set_retinaface_conf()
+    gender_model = load_gender_model(args, 'fc1')
+    retinaface_model = load_retinaface_model(retinaface_args)
+    img_list, box_and_point = detect_one(args.image, retinaface_model,retinaface_args)
+    gender_list, age_list = gender_age(img_list, gender_model)
+    print(gender_list)
diff --git a/gender_model.py b/gender_model.py
new file mode 100644
index 0000000..421af8a
--- /dev/null
+++ b/gender_model.py
@@ -0,0 +1,49 @@
+import numpy as np
+import mxnet as mx
+
+
+# 加载性别年龄模型
+def get_model(ctx, image_size, model_str, layer):
+    _vec = model_str.split(',')
+    assert len(_vec) == 2
+    prefix = _vec[0]
+    epoch = int(_vec[1])
+    print('loading', prefix, epoch)
+    sym, arg_params, aux_params = mx.model.load_checkpoint(prefix, epoch)
+    all_layers = sym.get_internals()
+    sym = all_layers[layer + '_output']
+    model = mx.mod.Module(symbol=sym, context=ctx, label_names=None)
+    model.bind(data_shapes=[('data', (1, 3, image_size[0], image_size[1]))])
+    model.set_params(arg_params, aux_params)
+    return model
+
+
+class GenderModel:
+    def __init__(self, args):
+        self.args = args
+        if args.gpu >= 0:
+            ctx = mx.gpu(args.gpu)
+        else:
+            ctx = mx.cpu()
+        _vec = args.image_size.split(',')
+        assert len(_vec) == 2
+        image_size = (int(_vec[0]), int(_vec[1]))
+        self.model = None
+        if len(args.model) > 0:
+            self.model = get_model(ctx, image_size, args.model, 'fc1')
+
+        self.det_minsize = 50
+        self.det_threshold = [0.6, 0.7, 0.8]
+        # self.det_factor = 0.9
+        self.image_size = image_size
+
+    def get_ga(self, data):
+        # print(data)
+        self.model.forward(data, is_train=False)
+        ret = self.model.get_outputs()[0].asnumpy()
+        g = ret[:, 0:2].flatten()
+        gender = np.argmax(g)
+        a = ret[:, 2:202].reshape((100, 2))
+        a = np.argmax(a, axis=1)
+        age = int(sum(a))
+        return gender, age
diff --git a/img/search/000_1.bmp b/img/search/000_1.bmp
new file mode 100644
index 0000000..d4c77a5
Binary files /dev/null and b/img/search/000_1.bmp differ
diff --git a/img/search/002_1.bmp b/img/search/002_1.bmp
new file mode 100644
index 0000000..1e5ad7b
Binary files /dev/null and b/img/search/002_1.bmp differ
diff --git a/img/search/377_3.bmp b/img/search/377_3.bmp
new file mode 100644
index 0000000..3d314b7
Binary files /dev/null and b/img/search/377_3.bmp differ
diff --git a/img/search/face/000_0.bmp b/img/search/face/000_0.bmp
new file mode 100644
index 0000000..75d84b6
Binary files /dev/null and b/img/search/face/000_0.bmp differ
diff --git a/img/search/face/000_1.bmp b/img/search/face/000_1.bmp
new file mode 100644
index 0000000..d4c77a5
Binary files /dev/null and b/img/search/face/000_1.bmp differ
diff --git a/img/search/face/000_2.bmp b/img/search/face/000_2.bmp
new file mode 100644
index 0000000..ecefec5
Binary files /dev/null and b/img/search/face/000_2.bmp differ
diff --git a/img/search/face/000_3.bmp b/img/search/face/000_3.bmp
new file mode 100644
index 0000000..78fd0e2
Binary files /dev/null and b/img/search/face/000_3.bmp differ
diff --git a/img/search/face/000_4.bmp b/img/search/face/000_4.bmp
new file mode 100644
index 0000000..2c5d50d
Binary files /dev/null and b/img/search/face/000_4.bmp differ
diff --git a/img/search/face/001_0.bmp b/img/search/face/001_0.bmp
new file mode 100644
index 0000000..d201e49
Binary files /dev/null and b/img/search/face/001_0.bmp differ
diff --git a/img/search/face/001_1.bmp b/img/search/face/001_1.bmp
new file mode 100644
index 0000000..c13a006
Binary files /dev/null and b/img/search/face/001_1.bmp differ
diff --git a/img/search/face/001_2.bmp b/img/search/face/001_2.bmp
new file mode 100644
index 0000000..df7ca39
Binary files /dev/null and b/img/search/face/001_2.bmp differ
diff --git a/img/search/face/001_3.bmp b/img/search/face/001_3.bmp
new file mode 100644
index 0000000..c6f3ecd
Binary files /dev/null and b/img/search/face/001_3.bmp differ
diff --git a/img/search/face/001_4.bmp b/img/search/face/001_4.bmp
new file mode 100644
index 0000000..16d1b39
Binary files /dev/null and b/img/search/face/001_4.bmp differ
diff --git a/img/search/face/002_0.bmp b/img/search/face/002_0.bmp
new file mode 100644
index 0000000..667917a
Binary files /dev/null and b/img/search/face/002_0.bmp differ
diff --git a/img/search/face/002_1.bmp b/img/search/face/002_1.bmp
new file mode 100644
index 0000000..1e5ad7b
Binary files /dev/null and b/img/search/face/002_1.bmp differ
diff --git a/img/search/face/002_2.bmp b/img/search/face/002_2.bmp
new file mode 100644
index 0000000..48359a6
Binary files /dev/null and b/img/search/face/002_2.bmp differ
diff --git a/img/search/face/002_3.bmp b/img/search/face/002_3.bmp
new file mode 100644
index 0000000..3bb2807
Binary files /dev/null and b/img/search/face/002_3.bmp differ
diff --git a/img/search/face/002_4.bmp b/img/search/face/002_4.bmp
new file mode 100644
index 0000000..555944e
Binary files /dev/null and b/img/search/face/002_4.bmp differ
diff --git a/img/search/face/003_0.bmp b/img/search/face/003_0.bmp
new file mode 100644
index 0000000..17ba734
Binary files /dev/null and b/img/search/face/003_0.bmp differ
diff --git a/img/search/face/003_1.bmp b/img/search/face/003_1.bmp
new file mode 100644
index 0000000..6955c10
Binary files /dev/null and b/img/search/face/003_1.bmp differ
diff --git a/img/search/face/003_2.bmp b/img/search/face/003_2.bmp
new file mode 100644
index 0000000..fc945c9
Binary files /dev/null and b/img/search/face/003_2.bmp differ
diff --git a/img/search/face/003_3.bmp b/img/search/face/003_3.bmp
new file mode 100644
index 0000000..340d156
Binary files /dev/null and b/img/search/face/003_3.bmp differ
diff --git a/img/search/face/003_4.bmp b/img/search/face/003_4.bmp
new file mode 100644
index 0000000..284b89c
Binary files /dev/null and b/img/search/face/003_4.bmp differ
diff --git a/img/search/face/004_0.bmp b/img/search/face/004_0.bmp
new file mode 100644
index 0000000..676125f
Binary files /dev/null and b/img/search/face/004_0.bmp differ
diff --git a/img/search/face/004_1.bmp b/img/search/face/004_1.bmp
new file mode 100644
index 0000000..abefa6e
Binary files /dev/null and b/img/search/face/004_1.bmp differ
diff --git a/img/search/face/004_2.bmp b/img/search/face/004_2.bmp
new file mode 100644
index 0000000..94bdac5
Binary files /dev/null and b/img/search/face/004_2.bmp differ
diff --git a/img/search/face/004_3.bmp b/img/search/face/004_3.bmp
new file mode 100644
index 0000000..df046a7
Binary files /dev/null and b/img/search/face/004_3.bmp differ
diff --git a/img/search/face/004_4.bmp b/img/search/face/004_4.bmp
new file mode 100644
index 0000000..ffcd27e
Binary files /dev/null and b/img/search/face/004_4.bmp differ
diff --git a/layers/__init__.py b/layers/__init__.py
new file mode 100644
index 0000000..53a3f4b
--- /dev/null
+++ b/layers/__init__.py
@@ -0,0 +1,2 @@
+from .functions import *
+from .modules import *
diff --git a/layers/__pycache__/__init__.cpython-38.pyc b/layers/__pycache__/__init__.cpython-38.pyc
new file mode 100644
index 0000000..e1f3c64
Binary files /dev/null and b/layers/__pycache__/__init__.cpython-38.pyc differ
diff --git a/layers/functions/__pycache__/prior_box.cpython-38.pyc b/layers/functions/__pycache__/prior_box.cpython-38.pyc
new file mode 100644
index 0000000..f7049a4
Binary files /dev/null and b/layers/functions/__pycache__/prior_box.cpython-38.pyc differ
diff --git a/layers/functions/prior_box.py b/layers/functions/prior_box.py
new file mode 100644
index 0000000..80c7f85
--- /dev/null
+++ b/layers/functions/prior_box.py
@@ -0,0 +1,34 @@
+import torch
+from itertools import product as product
+import numpy as np
+from math import ceil
+
+
+class PriorBox(object):
+    def __init__(self, cfg, image_size=None, phase='train'):
+        super(PriorBox, self).__init__()
+        self.min_sizes = cfg['min_sizes']
+        self.steps = cfg['steps']
+        self.clip = cfg['clip']
+        self.image_size = image_size
+        self.feature_maps = [[ceil(self.image_size[0]/step), ceil(self.image_size[1]/step)] for step in self.steps]
+        self.name = "s"
+
+    def forward(self):
+        anchors = []
+        for k, f in enumerate(self.feature_maps):
+            min_sizes = self.min_sizes[k]
+            for i, j in product(range(f[0]), range(f[1])):
+                for min_size in min_sizes:
+                    s_kx = min_size / self.image_size[1]
+                    s_ky = min_size / self.image_size[0]
+                    dense_cx = [x * self.steps[k] / self.image_size[1] for x in [j + 0.5]]
+                    dense_cy = [y * self.steps[k] / self.image_size[0] for y in [i + 0.5]]
+                    for cy, cx in product(dense_cy, dense_cx):
+                        anchors += [cx, cy, s_kx, s_ky]
+
+        # back to torch land
+        output = torch.Tensor(anchors).view(-1, 4)
+        if self.clip:
+            output.clamp_(max=1, min=0)
+        return output
diff --git a/layers/modules/__init__.py b/layers/modules/__init__.py
new file mode 100644
index 0000000..cf24bdd
--- /dev/null
+++ b/layers/modules/__init__.py
@@ -0,0 +1,3 @@
+from .multibox_loss import MultiBoxLoss
+
+__all__ = ['MultiBoxLoss']
diff --git a/layers/modules/__pycache__/__init__.cpython-38.pyc b/layers/modules/__pycache__/__init__.cpython-38.pyc
new file mode 100644
index 0000000..5644757
Binary files /dev/null and b/layers/modules/__pycache__/__init__.cpython-38.pyc differ
diff --git a/layers/modules/__pycache__/multibox_loss.cpython-38.pyc b/layers/modules/__pycache__/multibox_loss.cpython-38.pyc
new file mode 100644
index 0000000..a011c52
Binary files /dev/null and b/layers/modules/__pycache__/multibox_loss.cpython-38.pyc differ
diff --git a/layers/modules/multibox_loss.py b/layers/modules/multibox_loss.py
new file mode 100644
index 0000000..0966204
--- /dev/null
+++ b/layers/modules/multibox_loss.py
@@ -0,0 +1,125 @@
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+from torch.autograd import Variable
+from utils.box_utils import match, log_sum_exp
+from data import cfg_mnet
+GPU = cfg_mnet['gpu_train']
+
+class MultiBoxLoss(nn.Module):
+    """SSD Weighted Loss Function
+    Compute Targets:
+        1) Produce Confidence Target Indices by matching  ground truth boxes
+           with (default) 'priorboxes' that have jaccard index > threshold parameter
+           (default threshold: 0.5).
+        2) Produce localization target by 'encoding' variance into offsets of ground
+           truth boxes and their matched  'priorboxes'.
+        3) Hard negative mining to filter the excessive number of negative examples
+           that comes with using a large number of default bounding boxes.
+           (default negative:positive ratio 3:1)
+    Objective Loss:
+        L(x,c,l,g) = (Lconf(x, c) + αLloc(x,l,g)) / N
+        Where, Lconf is the CrossEntropy Loss and Lloc is the SmoothL1 Loss
+        weighted by α which is set to 1 by cross val.
+        Args:
+            c: class confidences,
+            l: predicted boxes,
+            g: ground truth boxes
+            N: number of matched default boxes
+        See: https://arxiv.org/pdf/1512.02325.pdf for more details.
+    """
+
+    def __init__(self, num_classes, overlap_thresh, prior_for_matching, bkg_label, neg_mining, neg_pos, neg_overlap, encode_target):
+        super(MultiBoxLoss, self).__init__()
+        self.num_classes = num_classes
+        self.threshold = overlap_thresh
+        self.background_label = bkg_label
+        self.encode_target = encode_target
+        self.use_prior_for_matching = prior_for_matching
+        self.do_neg_mining = neg_mining
+        self.negpos_ratio = neg_pos
+        self.neg_overlap = neg_overlap
+        self.variance = [0.1, 0.2]
+
+    def forward(self, predictions, priors, targets):
+        """Multibox Loss
+        Args:
+            predictions (tuple): A tuple containing loc preds, conf preds,
+            and prior boxes from SSD net.
+                conf shape: torch.size(batch_size,num_priors,num_classes)
+                loc shape: torch.size(batch_size,num_priors,4)
+                priors shape: torch.size(num_priors,4)
+
+            ground_truth (tensor): Ground truth boxes and labels for a batch,
+                shape: [batch_size,num_objs,5] (last idx is the label).
+        """
+
+        loc_data, conf_data, landm_data = predictions
+        priors = priors
+        num = loc_data.size(0)
+        num_priors = (priors.size(0))
+
+        # match priors (default boxes) and ground truth boxes
+        loc_t = torch.Tensor(num, num_priors, 4)
+        landm_t = torch.Tensor(num, num_priors, 10)
+        conf_t = torch.LongTensor(num, num_priors)
+        for idx in range(num):
+            truths = targets[idx][:, :4].data
+            labels = targets[idx][:, -1].data
+            landms = targets[idx][:, 4:14].data
+            defaults = priors.data
+            match(self.threshold, truths, defaults, self.variance, labels, landms, loc_t, conf_t, landm_t, idx)
+        if GPU:
+            loc_t = loc_t.cuda()
+            conf_t = conf_t.cuda()
+            landm_t = landm_t.cuda()
+
+        zeros = torch.tensor(0).cuda()
+        # landm Loss (Smooth L1)
+        # Shape: [batch,num_priors,10]
+        pos1 = conf_t > zeros
+        num_pos_landm = pos1.long().sum(1, keepdim=True)
+        N1 = max(num_pos_landm.data.sum().float(), 1)
+        pos_idx1 = pos1.unsqueeze(pos1.dim()).expand_as(landm_data)
+        landm_p = landm_data[pos_idx1].view(-1, 10)
+        landm_t = landm_t[pos_idx1].view(-1, 10)
+        loss_landm = F.smooth_l1_loss(landm_p, landm_t, reduction='sum')
+
+
+        pos = conf_t != zeros
+        conf_t[pos] = 1
+
+        # Localization Loss (Smooth L1)
+        # Shape: [batch,num_priors,4]
+        pos_idx = pos.unsqueeze(pos.dim()).expand_as(loc_data)
+        loc_p = loc_data[pos_idx].view(-1, 4)
+        loc_t = loc_t[pos_idx].view(-1, 4)
+        loss_l = F.smooth_l1_loss(loc_p, loc_t, reduction='sum')
+
+        # Compute max conf across batch for hard negative mining
+        batch_conf = conf_data.view(-1, self.num_classes)
+        loss_c = log_sum_exp(batch_conf) - batch_conf.gather(1, conf_t.view(-1, 1))
+
+        # Hard Negative Mining
+        loss_c[pos.view(-1, 1)] = 0 # filter out pos boxes for now
+        loss_c = loss_c.view(num, -1)
+        _, loss_idx = loss_c.sort(1, descending=True)
+        _, idx_rank = loss_idx.sort(1)
+        num_pos = pos.long().sum(1, keepdim=True)
+        num_neg = torch.clamp(self.negpos_ratio*num_pos, max=pos.size(1)-1)
+        neg = idx_rank < num_neg.expand_as(idx_rank)
+
+        # Confidence Loss Including Positive and Negative Examples
+        pos_idx = pos.unsqueeze(2).expand_as(conf_data)
+        neg_idx = neg.unsqueeze(2).expand_as(conf_data)
+        conf_p = conf_data[(pos_idx+neg_idx).gt(0)].view(-1,self.num_classes)
+        targets_weighted = conf_t[(pos+neg).gt(0)]
+        loss_c = F.cross_entropy(conf_p, targets_weighted, reduction='sum')
+
+        # Sum of losses: L(x,c,l,g) = (Lconf(x, c) + αLloc(x,l,g)) / N
+        N = max(num_pos.data.sum().float(), 1)
+        loss_l /= N
+        loss_c /= N
+        loss_landm /= N1
+
+        return loss_l, loss_c, loss_landm
diff --git a/losses.py b/losses.py
new file mode 100644
index 0000000..b96ed65
--- /dev/null
+++ b/losses.py
@@ -0,0 +1,33 @@
+import torch
+from torch import nn
+
+
+class CosFace(nn.Module):
+    def __init__(self, s=64.0, m=0.40):
+        super(CosFace, self).__init__()
+        self.s = s
+        self.m = m
+
+    def forward(self, cosine, label):
+        index = torch.where(label != -1)[0]
+        m_hot = torch.zeros(index.size()[0], cosine.size()[1], device=cosine.device)
+        m_hot.scatter_(1, label[index, None], self.m)
+        cosine[index] -= m_hot
+        ret = cosine * self.s
+        return ret
+
+
+class ArcFace(nn.Module):
+    def __init__(self, s=64.0, m=0.5):
+        super(ArcFace, self).__init__()
+        self.s = s
+        self.m = m
+
+    def forward(self, cosine: torch.Tensor, label):
+        index = torch.where(label != -1)[0]
+        m_hot = torch.zeros(index.size()[0], cosine.size()[1], device=cosine.device)
+        m_hot.scatter_(1, label[index, None], self.m)
+        cosine.acos_()
+        cosine[index] += m_hot
+        cosine.cos_().mul_(self.s)
+        return cosine
diff --git a/model/anti_spoof_models/2.7_80x80_MiniFASNetV2.pth b/model/anti_spoof_models/2.7_80x80_MiniFASNetV2.pth
new file mode 100644
index 0000000..47c4af2
Binary files /dev/null and b/model/anti_spoof_models/2.7_80x80_MiniFASNetV2.pth differ
diff --git a/model/anti_spoof_models/4_0_0_80x80_MiniFASNetV1SE.pth b/model/anti_spoof_models/4_0_0_80x80_MiniFASNetV1SE.pth
new file mode 100644
index 0000000..55a25b3
Binary files /dev/null and b/model/anti_spoof_models/4_0_0_80x80_MiniFASNetV1SE.pth differ
diff --git a/model/backbone100.pth b/model/backbone100.pth
new file mode 100644
index 0000000..81ebb87
Binary files /dev/null and b/model/backbone100.pth differ
diff --git a/model/log b/model/log
new file mode 100644
index 0000000..06cc63b
--- /dev/null
+++ b/model/log
@@ -0,0 +1,7094 @@
+gpu num: 4
+num_layers 50
+image_size [112, 112]
+num_classes 85742
+Called with argument: Namespace(batch_size=512, beta=1000.0, beta_freeze=0, beta_min=5.0, center_alpha=0.5, center_scale=0.003, ckpt=1, coco_scale=8.679543201625835, ctx_num=4, data_dir='/cache/jiaguo/faces_ms1mr_112x112', easy_margin=0, emb_size=512, end_epoch=100000, gamma=0.12, image_channel=3, image_h=112, image_w=112, images_per_identity=0, incay=0.0, loss_type=4, lr=0.1, lr_steps='50000,80000,100000', margin=4, margin_m=0.5, margin_s=64.0, margin_verbose=2000, max_steps=0, mom=0.9, network='r50', num_classes=85742, num_layers=50, patch='0_0_96_112_0', per_batch_size=128, power=1.0, prefix='../model-r50-am-lfw/model', pretrained='', rand_mirror=1, rescale_threshold=0, retrain=False, scale=0.9993, target='cfp_ff,cfp_fp,agedb_30,lfw', triplet_alpha=0.3, triplet_bag_size=3600, triplet_max_ap=0.0, use_deformable=0, use_val=False, verbose=2000, version_input=1, version_output='E', version_se=0, version_unit=3, wd=0.0005)
+init resnet 50
+0 1 E 3
+INFO:root:loading recordio /cache/jiaguo/faces_ms1mr_112x112/train.rec...
+header0 label [ 3850179.  3935921.]
+id2range 85742
+rand_mirror 1
+[11:19:40] src/engine/engine.cc:54: MXNet start using engine: ThreadedEnginePerDevice
+loading bin 1000
+loading bin 2000
+loading bin 3000
+loading bin 4000
+loading bin 5000
+loading bin 6000
+loading bin 7000
+loading bin 8000
+loading bin 9000
+loading bin 10000
+loading bin 11000
+loading bin 12000
+loading bin 13000
+loading bin 14000
+(14000L, 3L, 112L, 112L)
+ver cfp_ff
+loading bin 1000
+loading bin 2000
+loading bin 3000
+loading bin 4000
+loading bin 5000
+loading bin 6000
+loading bin 7000
+loading bin 8000
+loading bin 9000
+loading bin 10000
+loading bin 11000
+loading bin 12000
+loading bin 13000
+loading bin 14000
+(14000L, 3L, 112L, 112L)
+ver cfp_fp
+loading bin 1000
+loading bin 2000
+loading bin 3000
+loading bin 4000
+loading bin 5000
+loading bin 6000
+loading bin 7000
+loading bin 8000
+loading bin 9000
+loading bin 10000
+loading bin 11000
+loading bin 12000
+(12000L, 3L, 112L, 112L)
+ver agedb_30
+loading bin 1000
+loading bin 2000
+loading bin 3000
+loading bin 4000
+loading bin 5000
+loading bin 6000
+loading bin 7000
+loading bin 8000
+loading bin 9000
+loading bin 10000
+loading bin 11000
+loading bin 12000
+(12000L, 3L, 112L, 112L)
+ver lfw
+lr_steps [50000, 80000, 100000]
+/root/codebase/mxnet/python/mxnet/module/base_module.py:466: UserWarning: Optimizer created manually outside Module but rescale_grad is not normalized to 1.0/batch_size/num_workers (0.25 vs. 0.001953125). Is this intended?
+  optimizer_params=optimizer_params)
+call reset()
+[1][MARGIN]0.000434,-0.478536
+INFO:root:Epoch[0] Batch [20]	Speed: 644.79 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [40]	Speed: 634.07 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [60]	Speed: 649.95 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [80]	Speed: 649.72 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [100]	Speed: 650.81 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [120]	Speed: 649.96 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [140]	Speed: 651.57 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [160]	Speed: 650.47 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [180]	Speed: 650.82 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [200]	Speed: 651.51 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [220]	Speed: 652.71 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [240]	Speed: 653.87 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [260]	Speed: 654.40 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [280]	Speed: 640.71 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [300]	Speed: 653.38 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [320]	Speed: 655.12 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [340]	Speed: 652.97 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [360]	Speed: 655.28 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [380]	Speed: 656.21 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [400]	Speed: 656.65 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [420]	Speed: 656.34 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [440]	Speed: 656.27 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [460]	Speed: 657.00 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [480]	Speed: 658.23 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [500]	Speed: 656.69 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [520]	Speed: 655.78 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [540]	Speed: 659.12 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [560]	Speed: 658.79 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [580]	Speed: 658.13 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [600]	Speed: 657.53 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [620]	Speed: 660.71 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [640]	Speed: 663.04 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [660]	Speed: 662.37 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [680]	Speed: 662.88 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [700]	Speed: 659.24 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [720]	Speed: 663.00 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [740]	Speed: 648.51 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [760]	Speed: 664.60 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [780]	Speed: 665.42 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [800]	Speed: 665.97 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [820]	Speed: 668.72 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [840]	Speed: 667.42 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [860]	Speed: 666.18 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [880]	Speed: 666.68 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [900]	Speed: 666.40 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [920]	Speed: 667.43 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [940]	Speed: 669.26 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [960]	Speed: 670.06 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [980]	Speed: 668.92 samples/sec	acc=0.000000
+lr-batch-epoch: 0.1 999 0
+INFO:root:Epoch[0] Batch [1000]	Speed: 669.35 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [1020]	Speed: 669.08 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [1040]	Speed: 669.27 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [1060]	Speed: 671.89 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [1080]	Speed: 668.86 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [1100]	Speed: 670.39 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [1120]	Speed: 670.10 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [1140]	Speed: 671.93 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [1160]	Speed: 671.13 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [1180]	Speed: 670.72 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [1200]	Speed: 669.89 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [1220]	Speed: 672.17 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [1240]	Speed: 669.87 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [1260]	Speed: 673.92 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [1280]	Speed: 671.42 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [1300]	Speed: 669.78 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [1320]	Speed: 659.88 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [1340]	Speed: 671.13 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [1360]	Speed: 673.34 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [1380]	Speed: 670.05 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [1400]	Speed: 669.00 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [1420]	Speed: 672.09 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [1440]	Speed: 670.99 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [1460]	Speed: 671.36 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [1480]	Speed: 670.33 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [1500]	Speed: 670.73 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [1520]	Speed: 670.46 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [1540]	Speed: 672.16 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [1560]	Speed: 674.16 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [1580]	Speed: 672.88 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [1600]	Speed: 672.23 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [1620]	Speed: 673.29 samples/sec	acc=0.000098
+INFO:root:Epoch[0] Batch [1640]	Speed: 673.69 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [1660]	Speed: 673.66 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [1680]	Speed: 674.84 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [1700]	Speed: 674.14 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [1720]	Speed: 673.76 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [1740]	Speed: 674.11 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [1760]	Speed: 672.24 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [1780]	Speed: 673.66 samples/sec	acc=0.000195
+INFO:root:Epoch[0] Batch [1800]	Speed: 673.61 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [1820]	Speed: 674.43 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [1840]	Speed: 673.21 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [1860]	Speed: 674.09 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [1880]	Speed: 675.00 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [1900]	Speed: 675.02 samples/sec	acc=0.000293
+INFO:root:Epoch[0] Batch [1920]	Speed: 674.10 samples/sec	acc=0.000000
+INFO:root:Epoch[0] Batch [1940]	Speed: 675.82 samples/sec	acc=0.000586
+INFO:root:Epoch[0] Batch [1960]	Speed: 674.90 samples/sec	acc=0.000293
+INFO:root:Epoch[0] Batch [1980]	Speed: 662.58 samples/sec	acc=0.000293
+[2000][MARGIN]0.318108,-0.171479
+lr-batch-epoch: 0.1 1999 0
+testing verification..
+(14000, 512)
+infer time 22.275142
+[cfp_ff][2000]XNorm: 21.556408
+[cfp_ff][2000]Accuracy-Flip: 0.89814+-0.01236
+testing verification..
+(14000, 512)
+infer time 17.385799
+[cfp_fp][2000]XNorm: 20.292884
+[cfp_fp][2000]Accuracy-Flip: 0.69000+-0.01651
+testing verification..
+(12000, 512)
+infer time 14.193146
+[agedb_30][2000]XNorm: 21.596348
+[agedb_30][2000]Accuracy-Flip: 0.71433+-0.01589
+testing verification..
+(12000, 512)
+infer time 13.689882
+[lfw][2000]XNorm: 22.501038
+[lfw][2000]Accuracy-Flip: 0.92283+-0.01750
+[2000]Accuracy-Highest: 0.92283
+INFO:root:Epoch[0] Batch [2000]	Speed: 98.00 samples/sec	acc=0.000293
+INFO:root:Epoch[0] Batch [2020]	Speed: 673.67 samples/sec	acc=0.000293
+INFO:root:Epoch[0] Batch [2040]	Speed: 674.79 samples/sec	acc=0.000195
+INFO:root:Epoch[0] Batch [2060]	Speed: 673.67 samples/sec	acc=0.000195
+INFO:root:Epoch[0] Batch [2080]	Speed: 676.43 samples/sec	acc=0.000391
+INFO:root:Epoch[0] Batch [2100]	Speed: 674.95 samples/sec	acc=0.000293
+INFO:root:Epoch[0] Batch [2120]	Speed: 674.58 samples/sec	acc=0.000098
+INFO:root:Epoch[0] Batch [2140]	Speed: 675.61 samples/sec	acc=0.000586
+INFO:root:Epoch[0] Batch [2160]	Speed: 675.09 samples/sec	acc=0.000586
+INFO:root:Epoch[0] Batch [2180]	Speed: 676.43 samples/sec	acc=0.000781
+INFO:root:Epoch[0] Batch [2200]	Speed: 675.79 samples/sec	acc=0.000195
+INFO:root:Epoch[0] Batch [2220]	Speed: 675.75 samples/sec	acc=0.000488
+INFO:root:Epoch[0] Batch [2240]	Speed: 675.89 samples/sec	acc=0.001270
+INFO:root:Epoch[0] Batch [2260]	Speed: 676.21 samples/sec	acc=0.000684
+INFO:root:Epoch[0] Batch [2280]	Speed: 675.04 samples/sec	acc=0.000781
+INFO:root:Epoch[0] Batch [2300]	Speed: 676.03 samples/sec	acc=0.001172
+INFO:root:Epoch[0] Batch [2320]	Speed: 676.00 samples/sec	acc=0.001074
+INFO:root:Epoch[0] Batch [2340]	Speed: 675.01 samples/sec	acc=0.001465
+INFO:root:Epoch[0] Batch [2360]	Speed: 675.63 samples/sec	acc=0.001367
+INFO:root:Epoch[0] Batch [2380]	Speed: 674.71 samples/sec	acc=0.000879
+INFO:root:Epoch[0] Batch [2400]	Speed: 675.81 samples/sec	acc=0.001758
+INFO:root:Epoch[0] Batch [2420]	Speed: 675.55 samples/sec	acc=0.001367
+INFO:root:Epoch[0] Batch [2440]	Speed: 675.67 samples/sec	acc=0.001660
+INFO:root:Epoch[0] Batch [2460]	Speed: 675.68 samples/sec	acc=0.001465
+INFO:root:Epoch[0] Batch [2480]	Speed: 674.96 samples/sec	acc=0.003125
+INFO:root:Epoch[0] Batch [2500]	Speed: 675.34 samples/sec	acc=0.002344
+INFO:root:Epoch[0] Batch [2520]	Speed: 675.26 samples/sec	acc=0.002441
+INFO:root:Epoch[0] Batch [2540]	Speed: 674.46 samples/sec	acc=0.002539
+INFO:root:Epoch[0] Batch [2560]	Speed: 676.14 samples/sec	acc=0.003223
+INFO:root:Epoch[0] Batch [2580]	Speed: 675.05 samples/sec	acc=0.002539
+INFO:root:Epoch[0] Batch [2600]	Speed: 676.11 samples/sec	acc=0.003223
+INFO:root:Epoch[0] Batch [2620]	Speed: 675.14 samples/sec	acc=0.004004
+INFO:root:Epoch[0] Batch [2640]	Speed: 666.70 samples/sec	acc=0.003613
+INFO:root:Epoch[0] Batch [2660]	Speed: 675.77 samples/sec	acc=0.004395
+INFO:root:Epoch[0] Batch [2680]	Speed: 676.31 samples/sec	acc=0.004199
+INFO:root:Epoch[0] Batch [2700]	Speed: 675.34 samples/sec	acc=0.005957
+INFO:root:Epoch[0] Batch [2720]	Speed: 675.18 samples/sec	acc=0.003809
+INFO:root:Epoch[0] Batch [2740]	Speed: 675.65 samples/sec	acc=0.004492
+INFO:root:Epoch[0] Batch [2760]	Speed: 675.55 samples/sec	acc=0.006641
+INFO:root:Epoch[0] Batch [2780]	Speed: 675.93 samples/sec	acc=0.005664
+INFO:root:Epoch[0] Batch [2800]	Speed: 676.44 samples/sec	acc=0.006836
+INFO:root:Epoch[0] Batch [2820]	Speed: 676.61 samples/sec	acc=0.006348
+INFO:root:Epoch[0] Batch [2840]	Speed: 677.42 samples/sec	acc=0.008105
+INFO:root:Epoch[0] Batch [2860]	Speed: 675.81 samples/sec	acc=0.007715
+INFO:root:Epoch[0] Batch [2880]	Speed: 676.39 samples/sec	acc=0.008789
+INFO:root:Epoch[0] Batch [2900]	Speed: 675.64 samples/sec	acc=0.006934
+INFO:root:Epoch[0] Batch [2920]	Speed: 675.39 samples/sec	acc=0.007227
+INFO:root:Epoch[0] Batch [2940]	Speed: 675.92 samples/sec	acc=0.008398
+INFO:root:Epoch[0] Batch [2960]	Speed: 677.12 samples/sec	acc=0.010645
+INFO:root:Epoch[0] Batch [2980]	Speed: 675.77 samples/sec	acc=0.009668
+lr-batch-epoch: 0.1 2999 0
+INFO:root:Epoch[0] Batch [3000]	Speed: 677.35 samples/sec	acc=0.008984
+INFO:root:Epoch[0] Batch [3020]	Speed: 676.25 samples/sec	acc=0.009375
+INFO:root:Epoch[0] Batch [3040]	Speed: 676.78 samples/sec	acc=0.011914
+INFO:root:Epoch[0] Batch [3060]	Speed: 677.34 samples/sec	acc=0.011328
+INFO:root:Epoch[0] Batch [3080]	Speed: 676.81 samples/sec	acc=0.012402
+INFO:root:Epoch[0] Batch [3100]	Speed: 676.99 samples/sec	acc=0.012598
+INFO:root:Epoch[0] Batch [3120]	Speed: 676.19 samples/sec	acc=0.012598
+INFO:root:Epoch[0] Batch [3140]	Speed: 677.12 samples/sec	acc=0.014063
+INFO:root:Epoch[0] Batch [3160]	Speed: 676.20 samples/sec	acc=0.013184
+INFO:root:Epoch[0] Batch [3180]	Speed: 676.86 samples/sec	acc=0.014258
+INFO:root:Epoch[0] Batch [3200]	Speed: 676.69 samples/sec	acc=0.012988
+INFO:root:Epoch[0] Batch [3220]	Speed: 676.08 samples/sec	acc=0.015625
+INFO:root:Epoch[0] Batch [3240]	Speed: 676.61 samples/sec	acc=0.012598
+INFO:root:Epoch[0] Batch [3260]	Speed: 677.18 samples/sec	acc=0.016016
+INFO:root:Epoch[0] Batch [3280]	Speed: 676.36 samples/sec	acc=0.014844
+INFO:root:Epoch[0] Batch [3300]	Speed: 668.66 samples/sec	acc=0.014648
+INFO:root:Epoch[0] Batch [3320]	Speed: 676.78 samples/sec	acc=0.016797
+INFO:root:Epoch[0] Batch [3340]	Speed: 677.91 samples/sec	acc=0.017285
+INFO:root:Epoch[0] Batch [3360]	Speed: 676.52 samples/sec	acc=0.019824
+INFO:root:Epoch[0] Batch [3380]	Speed: 676.83 samples/sec	acc=0.018750
+INFO:root:Epoch[0] Batch [3400]	Speed: 675.56 samples/sec	acc=0.015625
+INFO:root:Epoch[0] Batch [3420]	Speed: 677.13 samples/sec	acc=0.016992
+INFO:root:Epoch[0] Batch [3440]	Speed: 676.76 samples/sec	acc=0.021094
+INFO:root:Epoch[0] Batch [3460]	Speed: 675.81 samples/sec	acc=0.021484
+INFO:root:Epoch[0] Batch [3480]	Speed: 676.76 samples/sec	acc=0.019336
+INFO:root:Epoch[0] Batch [3500]	Speed: 675.75 samples/sec	acc=0.018457
+INFO:root:Epoch[0] Batch [3520]	Speed: 675.31 samples/sec	acc=0.019434
+INFO:root:Epoch[0] Batch [3540]	Speed: 676.22 samples/sec	acc=0.018359
+INFO:root:Epoch[0] Batch [3560]	Speed: 675.88 samples/sec	acc=0.024219
+INFO:root:Epoch[0] Batch [3580]	Speed: 676.76 samples/sec	acc=0.017285
+INFO:root:Epoch[0] Batch [3600]	Speed: 678.39 samples/sec	acc=0.018457
+INFO:root:Epoch[0] Batch [3620]	Speed: 677.13 samples/sec	acc=0.022949
+INFO:root:Epoch[0] Batch [3640]	Speed: 677.37 samples/sec	acc=0.023730
+INFO:root:Epoch[0] Batch [3660]	Speed: 677.03 samples/sec	acc=0.021484
+INFO:root:Epoch[0] Batch [3680]	Speed: 677.22 samples/sec	acc=0.024414
+INFO:root:Epoch[0] Batch [3700]	Speed: 677.14 samples/sec	acc=0.025586
+INFO:root:Epoch[0] Batch [3720]	Speed: 677.20 samples/sec	acc=0.022559
+INFO:root:Epoch[0] Batch [3740]	Speed: 677.84 samples/sec	acc=0.021973
+INFO:root:Epoch[0] Batch [3760]	Speed: 677.04 samples/sec	acc=0.023438
+INFO:root:Epoch[0] Batch [3780]	Speed: 676.00 samples/sec	acc=0.023730
+INFO:root:Epoch[0] Batch [3800]	Speed: 677.31 samples/sec	acc=0.026953
+INFO:root:Epoch[0] Batch [3820]	Speed: 676.48 samples/sec	acc=0.025195
+INFO:root:Epoch[0] Batch [3840]	Speed: 675.81 samples/sec	acc=0.027148
+INFO:root:Epoch[0] Batch [3860]	Speed: 676.56 samples/sec	acc=0.027441
+INFO:root:Epoch[0] Batch [3880]	Speed: 676.52 samples/sec	acc=0.026367
+INFO:root:Epoch[0] Batch [3900]	Speed: 677.22 samples/sec	acc=0.026270
+INFO:root:Epoch[0] Batch [3920]	Speed: 675.63 samples/sec	acc=0.025391
+INFO:root:Epoch[0] Batch [3940]	Speed: 677.07 samples/sec	acc=0.026660
+INFO:root:Epoch[0] Batch [3960]	Speed: 676.32 samples/sec	acc=0.030078
+INFO:root:Epoch[0] Batch [3980]	Speed: 675.38 samples/sec	acc=0.027734
+[4000][MARGIN]0.635072,0.194337
+lr-batch-epoch: 0.1 3999 0
+testing verification..
+(14000, 512)
+infer time 18.636469
+[cfp_ff][4000]XNorm: 19.935919
+[cfp_ff][4000]Accuracy-Flip: 0.97486+-0.00576
+testing verification..
+(14000, 512)
+infer time 17.165106
+[cfp_fp][4000]XNorm: 16.943145
+[cfp_fp][4000]Accuracy-Flip: 0.77571+-0.01934
+testing verification..
+(12000, 512)
+infer time 13.460029
+[agedb_30][4000]XNorm: 19.915963
+[agedb_30][4000]Accuracy-Flip: 0.87317+-0.02042
+testing verification..
+(12000, 512)
+infer time 14.412261
+[lfw][4000]XNorm: 20.575942
+[lfw][4000]Accuracy-Flip: 0.97883+-0.00628
+[4000]Accuracy-Highest: 0.97883
+INFO:root:Epoch[0] Batch [4000]	Speed: 100.99 samples/sec	acc=0.031152
+INFO:root:Epoch[0] Batch [4020]	Speed: 670.46 samples/sec	acc=0.029102
+INFO:root:Epoch[0] Batch [4040]	Speed: 676.63 samples/sec	acc=0.027930
+INFO:root:Epoch[0] Batch [4060]	Speed: 677.21 samples/sec	acc=0.029102
+INFO:root:Epoch[0] Batch [4080]	Speed: 677.37 samples/sec	acc=0.027832
+INFO:root:Epoch[0] Batch [4100]	Speed: 675.95 samples/sec	acc=0.030469
+INFO:root:Epoch[0] Batch [4120]	Speed: 679.02 samples/sec	acc=0.030762
+INFO:root:Epoch[0] Batch [4140]	Speed: 678.00 samples/sec	acc=0.029297
+INFO:root:Epoch[0] Batch [4160]	Speed: 677.56 samples/sec	acc=0.033301
+INFO:root:Epoch[0] Batch [4180]	Speed: 677.75 samples/sec	acc=0.028320
+INFO:root:Epoch[0] Batch [4200]	Speed: 677.43 samples/sec	acc=0.033984
+INFO:root:Epoch[0] Batch [4220]	Speed: 678.17 samples/sec	acc=0.029687
+INFO:root:Epoch[0] Batch [4240]	Speed: 676.25 samples/sec	acc=0.031250
+INFO:root:Epoch[0] Batch [4260]	Speed: 676.69 samples/sec	acc=0.030176
+INFO:root:Epoch[0] Batch [4280]	Speed: 675.17 samples/sec	acc=0.035937
+INFO:root:Epoch[0] Batch [4300]	Speed: 676.69 samples/sec	acc=0.037695
+INFO:root:Epoch[0] Batch [4320]	Speed: 676.37 samples/sec	acc=0.033203
+INFO:root:Epoch[0] Batch [4340]	Speed: 677.01 samples/sec	acc=0.032617
+INFO:root:Epoch[0] Batch [4360]	Speed: 678.14 samples/sec	acc=0.034961
+INFO:root:Epoch[0] Batch [4380]	Speed: 678.45 samples/sec	acc=0.034180
+INFO:root:Epoch[0] Batch [4400]	Speed: 678.02 samples/sec	acc=0.033691
+INFO:root:Epoch[0] Batch [4420]	Speed: 677.47 samples/sec	acc=0.034082
+INFO:root:Epoch[0] Batch [4440]	Speed: 678.48 samples/sec	acc=0.037793
+INFO:root:Epoch[0] Batch [4460]	Speed: 678.17 samples/sec	acc=0.034375
+INFO:root:Epoch[0] Batch [4480]	Speed: 677.33 samples/sec	acc=0.035449
+INFO:root:Epoch[0] Batch [4500]	Speed: 677.29 samples/sec	acc=0.035742
+INFO:root:Epoch[0] Batch [4520]	Speed: 676.83 samples/sec	acc=0.034766
+INFO:root:Epoch[0] Batch [4540]	Speed: 678.16 samples/sec	acc=0.038965
+INFO:root:Epoch[0] Batch [4560]	Speed: 678.53 samples/sec	acc=0.034277
+INFO:root:Epoch[0] Batch [4580]	Speed: 677.40 samples/sec	acc=0.037500
+INFO:root:Epoch[0] Batch [4600]	Speed: 675.07 samples/sec	acc=0.035840
+INFO:root:Epoch[0] Batch [4620]	Speed: 677.33 samples/sec	acc=0.036621
+INFO:root:Epoch[0] Batch [4640]	Speed: 676.37 samples/sec	acc=0.039453
+INFO:root:Epoch[0] Batch [4660]	Speed: 676.62 samples/sec	acc=0.034473
+INFO:root:Epoch[0] Batch [4680]	Speed: 677.97 samples/sec	acc=0.038379
+INFO:root:Epoch[0] Batch [4700]	Speed: 677.84 samples/sec	acc=0.037598
+INFO:root:Epoch[0] Batch [4720]	Speed: 677.82 samples/sec	acc=0.037012
+INFO:root:Epoch[0] Batch [4740]	Speed: 667.47 samples/sec	acc=0.043164
+INFO:root:Epoch[0] Batch [4760]	Speed: 677.44 samples/sec	acc=0.041797
+INFO:root:Epoch[0] Batch [4780]	Speed: 679.78 samples/sec	acc=0.045215
+INFO:root:Epoch[0] Batch [4800]	Speed: 679.15 samples/sec	acc=0.041504
+INFO:root:Epoch[0] Batch [4820]	Speed: 661.68 samples/sec	acc=0.039355
+INFO:root:Epoch[0] Batch [4840]	Speed: 661.76 samples/sec	acc=0.041797
+INFO:root:Epoch[0] Batch [4860]	Speed: 661.89 samples/sec	acc=0.040234
+INFO:root:Epoch[0] Batch [4880]	Speed: 662.45 samples/sec	acc=0.041406
+INFO:root:Epoch[0] Batch [4900]	Speed: 660.34 samples/sec	acc=0.043066
+INFO:root:Epoch[0] Batch [4920]	Speed: 661.82 samples/sec	acc=0.039941
+INFO:root:Epoch[0] Batch [4940]	Speed: 661.75 samples/sec	acc=0.039453
+INFO:root:Epoch[0] Batch [4960]	Speed: 661.17 samples/sec	acc=0.043359
+INFO:root:Epoch[0] Batch [4980]	Speed: 661.79 samples/sec	acc=0.040625
+lr-batch-epoch: 0.1 4999 0
+INFO:root:Epoch[0] Batch [5000]	Speed: 661.08 samples/sec	acc=0.043652
+INFO:root:Epoch[0] Batch [5020]	Speed: 662.01 samples/sec	acc=0.042188
+INFO:root:Epoch[0] Batch [5040]	Speed: 661.66 samples/sec	acc=0.042090
+INFO:root:Epoch[0] Batch [5060]	Speed: 662.41 samples/sec	acc=0.044824
+INFO:root:Epoch[0] Batch [5080]	Speed: 661.84 samples/sec	acc=0.043848
+INFO:root:Epoch[0] Batch [5100]	Speed: 661.70 samples/sec	acc=0.044824
+INFO:root:Epoch[0] Batch [5120]	Speed: 662.14 samples/sec	acc=0.046777
+INFO:root:Epoch[0] Batch [5140]	Speed: 660.41 samples/sec	acc=0.048535
+INFO:root:Epoch[0] Batch [5160]	Speed: 662.28 samples/sec	acc=0.043066
+INFO:root:Epoch[0] Batch [5180]	Speed: 662.27 samples/sec	acc=0.049316
+INFO:root:Epoch[0] Batch [5200]	Speed: 662.37 samples/sec	acc=0.043457
+INFO:root:Epoch[0] Batch [5220]	Speed: 660.97 samples/sec	acc=0.044238
+INFO:root:Epoch[0] Batch [5240]	Speed: 660.86 samples/sec	acc=0.045801
+INFO:root:Epoch[0] Batch [5260]	Speed: 662.95 samples/sec	acc=0.049609
+INFO:root:Epoch[0] Batch [5280]	Speed: 661.55 samples/sec	acc=0.043359
+INFO:root:Epoch[0] Batch [5300]	Speed: 661.68 samples/sec	acc=0.046777
+INFO:root:Epoch[0] Batch [5320]	Speed: 662.14 samples/sec	acc=0.047461
+INFO:root:Epoch[0] Batch [5340]	Speed: 660.87 samples/sec	acc=0.047266
+INFO:root:Epoch[0] Batch [5360]	Speed: 661.63 samples/sec	acc=0.051562
+INFO:root:Epoch[0] Batch [5380]	Speed: 662.28 samples/sec	acc=0.046777
+INFO:root:Epoch[0] Batch [5400]	Speed: 662.36 samples/sec	acc=0.049609
+INFO:root:Epoch[0] Batch [5420]	Speed: 661.43 samples/sec	acc=0.046191
+INFO:root:Epoch[0] Batch [5440]	Speed: 662.07 samples/sec	acc=0.049805
+INFO:root:Epoch[0] Batch [5460]	Speed: 662.13 samples/sec	acc=0.046875
+INFO:root:Epoch[0] Batch [5480]	Speed: 662.57 samples/sec	acc=0.049512
+INFO:root:Epoch[0] Batch [5500]	Speed: 677.52 samples/sec	acc=0.047363
+INFO:root:Epoch[0] Batch [5520]	Speed: 679.41 samples/sec	acc=0.048242
+INFO:root:Epoch[0] Batch [5540]	Speed: 679.69 samples/sec	acc=0.050781
+INFO:root:Epoch[0] Batch [5560]	Speed: 678.48 samples/sec	acc=0.049512
+INFO:root:Epoch[0] Batch [5580]	Speed: 676.77 samples/sec	acc=0.050977
+INFO:root:Epoch[0] Batch [5600]	Speed: 678.05 samples/sec	acc=0.051953
+INFO:root:Epoch[0] Batch [5620]	Speed: 679.09 samples/sec	acc=0.050586
+INFO:root:Epoch[0] Batch [5640]	Speed: 678.14 samples/sec	acc=0.046680
+INFO:root:Epoch[0] Batch [5660]	Speed: 678.66 samples/sec	acc=0.052344
+INFO:root:Epoch[0] Batch [5680]	Speed: 677.87 samples/sec	acc=0.052441
+INFO:root:Epoch[0] Batch [5700]	Speed: 677.25 samples/sec	acc=0.052051
+INFO:root:Epoch[0] Batch [5720]	Speed: 678.25 samples/sec	acc=0.053516
+INFO:root:Epoch[0] Batch [5740]	Speed: 675.71 samples/sec	acc=0.056152
+INFO:root:Epoch[0] Batch [5760]	Speed: 679.53 samples/sec	acc=0.051953
+INFO:root:Epoch[0] Batch [5780]	Speed: 678.92 samples/sec	acc=0.052930
+INFO:root:Epoch[0] Batch [5800]	Speed: 677.03 samples/sec	acc=0.052148
+INFO:root:Epoch[0] Batch [5820]	Speed: 676.98 samples/sec	acc=0.053027
+INFO:root:Epoch[0] Batch [5840]	Speed: 678.34 samples/sec	acc=0.052344
+INFO:root:Epoch[0] Batch [5860]	Speed: 678.96 samples/sec	acc=0.052930
+INFO:root:Epoch[0] Batch [5880]	Speed: 677.13 samples/sec	acc=0.055957
+INFO:root:Epoch[0] Batch [5900]	Speed: 677.95 samples/sec	acc=0.056152
+INFO:root:Epoch[0] Batch [5920]	Speed: 677.03 samples/sec	acc=0.051367
+INFO:root:Epoch[0] Batch [5940]	Speed: 677.81 samples/sec	acc=0.056250
+INFO:root:Epoch[0] Batch [5960]	Speed: 677.87 samples/sec	acc=0.052832
+INFO:root:Epoch[0] Batch [5980]	Speed: 678.27 samples/sec	acc=0.054102
+[6000][MARGIN]0.638322,0.196974
+lr-batch-epoch: 0.1 5999 0
+testing verification..
+(14000, 512)
+infer time 20.587736
+[cfp_ff][6000]XNorm: 22.175584
+[cfp_ff][6000]Accuracy-Flip: 0.98514+-0.00583
+testing verification..
+(14000, 512)
+infer time 21.94076
+[cfp_fp][6000]XNorm: 18.532788
+[cfp_fp][6000]Accuracy-Flip: 0.81614+-0.01800
+testing verification..
+(12000, 512)
+infer time 17.659599
+[agedb_30][6000]XNorm: 22.109134
+[agedb_30][6000]Accuracy-Flip: 0.91667+-0.02018
+testing verification..
+(12000, 512)
+infer time 18.542575
+[lfw][6000]XNorm: 22.595654
+[lfw][6000]Accuracy-Flip: 0.98567+-0.00642
+[6000]Accuracy-Highest: 0.98567
+INFO:root:Epoch[0] Batch [6000]	Speed: 88.05 samples/sec	acc=0.055957
+INFO:root:Epoch[0] Batch [6020]	Speed: 679.23 samples/sec	acc=0.053906
+INFO:root:Epoch[0] Batch [6040]	Speed: 678.16 samples/sec	acc=0.058105
+INFO:root:Epoch[0] Batch [6060]	Speed: 676.01 samples/sec	acc=0.058496
+INFO:root:Epoch[0] Batch [6080]	Speed: 676.47 samples/sec	acc=0.055273
+INFO:root:Epoch[0] Batch [6100]	Speed: 677.98 samples/sec	acc=0.056445
+INFO:root:Epoch[0] Batch [6120]	Speed: 676.84 samples/sec	acc=0.053906
+INFO:root:Epoch[0] Batch [6140]	Speed: 678.95 samples/sec	acc=0.059766
+INFO:root:Epoch[0] Batch [6160]	Speed: 669.69 samples/sec	acc=0.050977
+INFO:root:Epoch[0] Batch [6180]	Speed: 677.19 samples/sec	acc=0.056152
+INFO:root:Epoch[0] Batch [6200]	Speed: 675.56 samples/sec	acc=0.055859
+INFO:root:Epoch[0] Batch [6220]	Speed: 676.91 samples/sec	acc=0.058496
+INFO:root:Epoch[0] Batch [6240]	Speed: 677.78 samples/sec	acc=0.056738
+INFO:root:Epoch[0] Batch [6260]	Speed: 678.77 samples/sec	acc=0.054297
+INFO:root:Epoch[0] Batch [6280]	Speed: 680.08 samples/sec	acc=0.055078
+INFO:root:Epoch[0] Batch [6300]	Speed: 676.31 samples/sec	acc=0.059570
+INFO:root:Epoch[0] Batch [6320]	Speed: 679.00 samples/sec	acc=0.063281
+INFO:root:Epoch[0] Batch [6340]	Speed: 677.99 samples/sec	acc=0.059473
+INFO:root:Epoch[0] Batch [6360]	Speed: 680.58 samples/sec	acc=0.055469
+INFO:root:Epoch[0] Batch [6380]	Speed: 662.02 samples/sec	acc=0.061035
+INFO:root:Epoch[0] Batch [6400]	Speed: 662.17 samples/sec	acc=0.061230
+INFO:root:Epoch[0] Batch [6420]	Speed: 662.59 samples/sec	acc=0.058887
+INFO:root:Epoch[0] Batch [6440]	Speed: 661.62 samples/sec	acc=0.061426
+INFO:root:Epoch[0] Batch [6460]	Speed: 662.99 samples/sec	acc=0.060742
+INFO:root:Epoch[0] Batch [6480]	Speed: 662.39 samples/sec	acc=0.058789
+INFO:root:Epoch[0] Batch [6500]	Speed: 662.39 samples/sec	acc=0.059180
+INFO:root:Epoch[0] Batch [6520]	Speed: 661.72 samples/sec	acc=0.058203
+INFO:root:Epoch[0] Batch [6540]	Speed: 662.39 samples/sec	acc=0.058789
+INFO:root:Epoch[0] Batch [6560]	Speed: 662.52 samples/sec	acc=0.062012
+INFO:root:Epoch[0] Batch [6580]	Speed: 662.48 samples/sec	acc=0.057129
+INFO:root:Epoch[0] Batch [6600]	Speed: 662.49 samples/sec	acc=0.060742
+INFO:root:Epoch[0] Batch [6620]	Speed: 661.21 samples/sec	acc=0.062012
+INFO:root:Epoch[0] Batch [6640]	Speed: 661.41 samples/sec	acc=0.063770
+INFO:root:Epoch[0] Batch [6660]	Speed: 661.61 samples/sec	acc=0.059277
+INFO:root:Epoch[0] Batch [6680]	Speed: 661.67 samples/sec	acc=0.061035
+INFO:root:Epoch[0] Batch [6700]	Speed: 662.48 samples/sec	acc=0.062988
+INFO:root:Epoch[0] Batch [6720]	Speed: 662.67 samples/sec	acc=0.063965
+INFO:root:Epoch[0] Batch [6740]	Speed: 661.64 samples/sec	acc=0.064648
+INFO:root:Epoch[0] Batch [6760]	Speed: 662.10 samples/sec	acc=0.065137
+INFO:root:Epoch[0] Batch [6780]	Speed: 660.91 samples/sec	acc=0.059668
+INFO:root:Epoch[0] Batch [6800]	Speed: 661.81 samples/sec	acc=0.062402
+INFO:root:Epoch[0] Batch [6820]	Speed: 662.05 samples/sec	acc=0.062891
+INFO:root:Epoch[0] Batch [6840]	Speed: 662.09 samples/sec	acc=0.063379
+INFO:root:Epoch[0] Batch [6860]	Speed: 662.57 samples/sec	acc=0.067773
+INFO:root:Epoch[0] Batch [6880]	Speed: 655.70 samples/sec	acc=0.064355
+INFO:root:Epoch[0] Batch [6900]	Speed: 677.25 samples/sec	acc=0.062695
+INFO:root:Epoch[0] Batch [6920]	Speed: 679.46 samples/sec	acc=0.066602
+INFO:root:Epoch[0] Batch [6940]	Speed: 678.32 samples/sec	acc=0.063086
+INFO:root:Epoch[0] Batch [6960]	Speed: 677.74 samples/sec	acc=0.065820
+INFO:root:Epoch[0] Batch [6980]	Speed: 680.36 samples/sec	acc=0.066309
+lr-batch-epoch: 0.1 6999 0
+INFO:root:Epoch[0] Batch [7000]	Speed: 677.42 samples/sec	acc=0.061719
+INFO:root:Epoch[0] Batch [7020]	Speed: 678.84 samples/sec	acc=0.065137
+INFO:root:Epoch[0] Batch [7040]	Speed: 678.47 samples/sec	acc=0.064453
+INFO:root:Epoch[0] Batch [7060]	Speed: 677.29 samples/sec	acc=0.066992
+INFO:root:Epoch[0] Batch [7080]	Speed: 678.12 samples/sec	acc=0.061328
+INFO:root:Epoch[0] Batch [7100]	Speed: 678.82 samples/sec	acc=0.064160
+INFO:root:Epoch[0] Batch [7120]	Speed: 677.84 samples/sec	acc=0.066211
+INFO:root:Epoch[0] Batch [7140]	Speed: 678.83 samples/sec	acc=0.075293
+INFO:root:Epoch[0] Batch [7160]	Speed: 678.00 samples/sec	acc=0.061914
+INFO:root:Epoch[0] Batch [7180]	Speed: 679.14 samples/sec	acc=0.065137
+INFO:root:Epoch[0] Batch [7200]	Speed: 679.30 samples/sec	acc=0.063965
+INFO:root:Epoch[0] Batch [7220]	Speed: 680.12 samples/sec	acc=0.066504
+INFO:root:Epoch[0] Batch [7240]	Speed: 677.71 samples/sec	acc=0.062402
+INFO:root:Epoch[0] Batch [7260]	Speed: 679.25 samples/sec	acc=0.064844
+INFO:root:Epoch[0] Batch [7280]	Speed: 678.19 samples/sec	acc=0.068750
+INFO:root:Epoch[0] Batch [7300]	Speed: 679.31 samples/sec	acc=0.070996
+INFO:root:Epoch[0] Batch [7320]	Speed: 678.34 samples/sec	acc=0.064355
+INFO:root:Epoch[0] Batch [7340]	Speed: 677.93 samples/sec	acc=0.063867
+INFO:root:Epoch[0] Batch [7360]	Speed: 678.70 samples/sec	acc=0.070508
+INFO:root:Epoch[0] Batch [7380]	Speed: 678.44 samples/sec	acc=0.065137
+INFO:root:Epoch[0] Batch [7400]	Speed: 679.43 samples/sec	acc=0.069141
+INFO:root:Epoch[0] Batch [7420]	Speed: 677.16 samples/sec	acc=0.069238
+INFO:root:Epoch[0] Batch [7440]	Speed: 675.98 samples/sec	acc=0.066895
+INFO:root:Epoch[0] Batch [7460]	Speed: 678.25 samples/sec	acc=0.071875
+INFO:root:Epoch[0] Batch [7480]	Speed: 676.22 samples/sec	acc=0.069824
+INFO:root:Epoch[0] Batch [7500]	Speed: 677.61 samples/sec	acc=0.068262
+INFO:root:Epoch[0] Train-acc=0.065213
+INFO:root:Epoch[0] Time cost=6026.197
+call reset()
+INFO:root:Epoch[1] Batch [20]	Speed: 678.56 samples/sec	acc=0.088356
+INFO:root:Epoch[1] Batch [40]	Speed: 670.34 samples/sec	acc=0.085742
+INFO:root:Epoch[1] Batch [60]	Speed: 679.25 samples/sec	acc=0.086230
+INFO:root:Epoch[1] Batch [80]	Speed: 677.77 samples/sec	acc=0.080762
+INFO:root:Epoch[1] Batch [100]	Speed: 679.00 samples/sec	acc=0.084863
+INFO:root:Epoch[1] Batch [120]	Speed: 677.89 samples/sec	acc=0.082520
+INFO:root:Epoch[1] Batch [140]	Speed: 677.44 samples/sec	acc=0.084277
+INFO:root:Epoch[1] Batch [160]	Speed: 678.36 samples/sec	acc=0.082812
+INFO:root:Epoch[1] Batch [180]	Speed: 677.89 samples/sec	acc=0.080859
+INFO:root:Epoch[1] Batch [200]	Speed: 670.57 samples/sec	acc=0.080664
+INFO:root:Epoch[1] Batch [220]	Speed: 662.27 samples/sec	acc=0.079590
+INFO:root:Epoch[1] Batch [240]	Speed: 661.53 samples/sec	acc=0.075391
+INFO:root:Epoch[1] Batch [260]	Speed: 661.56 samples/sec	acc=0.077832
+INFO:root:Epoch[1] Batch [280]	Speed: 661.32 samples/sec	acc=0.074805
+INFO:root:Epoch[1] Batch [300]	Speed: 662.06 samples/sec	acc=0.076367
+INFO:root:Epoch[1] Batch [320]	Speed: 661.87 samples/sec	acc=0.075098
+INFO:root:Epoch[1] Batch [340]	Speed: 661.84 samples/sec	acc=0.074707
+INFO:root:Epoch[1] Batch [360]	Speed: 663.11 samples/sec	acc=0.081445
+INFO:root:Epoch[1] Batch [380]	Speed: 662.29 samples/sec	acc=0.080371
+INFO:root:Epoch[1] Batch [400]	Speed: 662.15 samples/sec	acc=0.074805
+INFO:root:Epoch[1] Batch [420]	Speed: 663.09 samples/sec	acc=0.078613
+INFO:root:Epoch[1] Batch [440]	Speed: 662.54 samples/sec	acc=0.078320
+INFO:root:Epoch[1] Batch [460]	Speed: 662.79 samples/sec	acc=0.073340
+[8000][MARGIN]0.659157,0.225854
+INFO:root:Epoch[1] Batch [480]	Speed: 662.05 samples/sec	acc=0.074023
+lr-batch-epoch: 0.1 480 1
+testing verification..
+(14000, 512)
+infer time 20.835884
+[cfp_ff][8000]XNorm: 19.584579
+[cfp_ff][8000]Accuracy-Flip: 0.98857+-0.00433
+testing verification..
+(14000, 512)
+infer time 21.983813
+[cfp_fp][8000]XNorm: 16.207827
+[cfp_fp][8000]Accuracy-Flip: 0.81771+-0.01746
+testing verification..
+(12000, 512)
+infer time 18.553163
+[agedb_30][8000]XNorm: 19.670710
+[agedb_30][8000]Accuracy-Flip: 0.92450+-0.01976
+testing verification..
+(12000, 512)
+infer time 17.92847
+[lfw][8000]XNorm: 19.712476
+[lfw][8000]Accuracy-Flip: 0.99033+-0.00386
+[8000]Accuracy-Highest: 0.99033
+INFO:root:Epoch[1] Batch [500]	Speed: 88.19 samples/sec	acc=0.075098
+INFO:root:Epoch[1] Batch [520]	Speed: 677.44 samples/sec	acc=0.077051
+INFO:root:Epoch[1] Batch [540]	Speed: 677.87 samples/sec	acc=0.079883
+INFO:root:Epoch[1] Batch [560]	Speed: 678.35 samples/sec	acc=0.079297
+INFO:root:Epoch[1] Batch [580]	Speed: 677.45 samples/sec	acc=0.076758
+INFO:root:Epoch[1] Batch [600]	Speed: 678.32 samples/sec	acc=0.077637
+INFO:root:Epoch[1] Batch [620]	Speed: 677.88 samples/sec	acc=0.079199
+INFO:root:Epoch[1] Batch [640]	Speed: 663.81 samples/sec	acc=0.078711
+INFO:root:Epoch[1] Batch [660]	Speed: 663.58 samples/sec	acc=0.077734
+INFO:root:Epoch[1] Batch [680]	Speed: 664.08 samples/sec	acc=0.073340
+INFO:root:Epoch[1] Batch [700]	Speed: 665.07 samples/sec	acc=0.078320
+INFO:root:Epoch[1] Batch [720]	Speed: 664.18 samples/sec	acc=0.069824
+INFO:root:Epoch[1] Batch [740]	Speed: 664.43 samples/sec	acc=0.073926
+INFO:root:Epoch[1] Batch [760]	Speed: 664.01 samples/sec	acc=0.075781
+INFO:root:Epoch[1] Batch [780]	Speed: 664.59 samples/sec	acc=0.076855
+INFO:root:Epoch[1] Batch [800]	Speed: 664.93 samples/sec	acc=0.078223
+INFO:root:Epoch[1] Batch [820]	Speed: 664.85 samples/sec	acc=0.076660
+INFO:root:Epoch[1] Batch [840]	Speed: 667.14 samples/sec	acc=0.075684
+INFO:root:Epoch[1] Batch [860]	Speed: 664.40 samples/sec	acc=0.073926
+INFO:root:Epoch[1] Batch [880]	Speed: 664.33 samples/sec	acc=0.078516
+INFO:root:Epoch[1] Batch [900]	Speed: 664.83 samples/sec	acc=0.077441
+INFO:root:Epoch[1] Batch [920]	Speed: 664.61 samples/sec	acc=0.078125
+INFO:root:Epoch[1] Batch [940]	Speed: 665.42 samples/sec	acc=0.076074
+INFO:root:Epoch[1] Batch [960]	Speed: 664.69 samples/sec	acc=0.074707
+INFO:root:Epoch[1] Batch [980]	Speed: 664.73 samples/sec	acc=0.074512
+INFO:root:Epoch[1] Batch [1000]	Speed: 665.19 samples/sec	acc=0.077539
+INFO:root:Epoch[1] Batch [1020]	Speed: 664.25 samples/sec	acc=0.071191
+INFO:root:Epoch[1] Batch [1040]	Speed: 664.30 samples/sec	acc=0.073828
+INFO:root:Epoch[1] Batch [1060]	Speed: 664.79 samples/sec	acc=0.072949
+INFO:root:Epoch[1] Batch [1080]	Speed: 664.35 samples/sec	acc=0.073535
+INFO:root:Epoch[1] Batch [1100]	Speed: 665.18 samples/sec	acc=0.077148
+INFO:root:Epoch[1] Batch [1120]	Speed: 664.84 samples/sec	acc=0.073047
+INFO:root:Epoch[1] Batch [1140]	Speed: 664.19 samples/sec	acc=0.079297
+INFO:root:Epoch[1] Batch [1160]	Speed: 664.03 samples/sec	acc=0.073145
+INFO:root:Epoch[1] Batch [1180]	Speed: 664.45 samples/sec	acc=0.072363
+INFO:root:Epoch[1] Batch [1200]	Speed: 664.66 samples/sec	acc=0.077441
+INFO:root:Epoch[1] Batch [1220]	Speed: 664.29 samples/sec	acc=0.072363
+INFO:root:Epoch[1] Batch [1240]	Speed: 665.36 samples/sec	acc=0.081348
+INFO:root:Epoch[1] Batch [1260]	Speed: 665.23 samples/sec	acc=0.072754
+INFO:root:Epoch[1] Batch [1280]	Speed: 663.95 samples/sec	acc=0.071387
+INFO:root:Epoch[1] Batch [1300]	Speed: 664.83 samples/sec	acc=0.076758
+INFO:root:Epoch[1] Batch [1320]	Speed: 663.75 samples/sec	acc=0.077148
+INFO:root:Epoch[1] Batch [1340]	Speed: 665.27 samples/sec	acc=0.075586
+INFO:root:Epoch[1] Batch [1360]	Speed: 664.15 samples/sec	acc=0.075879
+INFO:root:Epoch[1] Batch [1380]	Speed: 664.03 samples/sec	acc=0.074512
+INFO:root:Epoch[1] Batch [1400]	Speed: 663.55 samples/sec	acc=0.076172
+INFO:root:Epoch[1] Batch [1420]	Speed: 664.51 samples/sec	acc=0.080859
+INFO:root:Epoch[1] Batch [1440]	Speed: 665.22 samples/sec	acc=0.073828
+INFO:root:Epoch[1] Batch [1460]	Speed: 663.78 samples/sec	acc=0.080566
+INFO:root:Epoch[1] Batch [1480]	Speed: 664.50 samples/sec	acc=0.080566
+lr-batch-epoch: 0.1 1480 1
+INFO:root:Epoch[1] Batch [1500]	Speed: 664.67 samples/sec	acc=0.076660
+INFO:root:Epoch[1] Batch [1520]	Speed: 664.57 samples/sec	acc=0.075586
+INFO:root:Epoch[1] Batch [1540]	Speed: 663.94 samples/sec	acc=0.078125
+INFO:root:Epoch[1] Batch [1560]	Speed: 665.17 samples/sec	acc=0.077051
+INFO:root:Epoch[1] Batch [1580]	Speed: 663.92 samples/sec	acc=0.080371
+INFO:root:Epoch[1] Batch [1600]	Speed: 664.96 samples/sec	acc=0.077930
+INFO:root:Epoch[1] Batch [1620]	Speed: 661.54 samples/sec	acc=0.083203
+INFO:root:Epoch[1] Batch [1640]	Speed: 677.62 samples/sec	acc=0.076172
+INFO:root:Epoch[1] Batch [1660]	Speed: 671.80 samples/sec	acc=0.079004
+INFO:root:Epoch[1] Batch [1680]	Speed: 664.55 samples/sec	acc=0.077441
+INFO:root:Epoch[1] Batch [1700]	Speed: 664.60 samples/sec	acc=0.077637
+INFO:root:Epoch[1] Batch [1720]	Speed: 665.05 samples/sec	acc=0.081641
+INFO:root:Epoch[1] Batch [1740]	Speed: 664.88 samples/sec	acc=0.078418
+INFO:root:Epoch[1] Batch [1760]	Speed: 664.12 samples/sec	acc=0.080078
+INFO:root:Epoch[1] Batch [1780]	Speed: 665.52 samples/sec	acc=0.077441
+INFO:root:Epoch[1] Batch [1800]	Speed: 664.75 samples/sec	acc=0.079687
+INFO:root:Epoch[1] Batch [1820]	Speed: 664.27 samples/sec	acc=0.083301
+INFO:root:Epoch[1] Batch [1840]	Speed: 664.70 samples/sec	acc=0.081250
+INFO:root:Epoch[1] Batch [1860]	Speed: 664.75 samples/sec	acc=0.075684
+INFO:root:Epoch[1] Batch [1880]	Speed: 664.46 samples/sec	acc=0.080859
+INFO:root:Epoch[1] Batch [1900]	Speed: 664.87 samples/sec	acc=0.080762
+INFO:root:Epoch[1] Batch [1920]	Speed: 664.11 samples/sec	acc=0.078906
+INFO:root:Epoch[1] Batch [1940]	Speed: 664.11 samples/sec	acc=0.079004
+INFO:root:Epoch[1] Batch [1960]	Speed: 665.33 samples/sec	acc=0.080176
+INFO:root:Epoch[1] Batch [1980]	Speed: 663.48 samples/sec	acc=0.079883
+INFO:root:Epoch[1] Batch [2000]	Speed: 663.18 samples/sec	acc=0.084473
+INFO:root:Epoch[1] Batch [2020]	Speed: 665.34 samples/sec	acc=0.087207
+INFO:root:Epoch[1] Batch [2040]	Speed: 664.37 samples/sec	acc=0.081055
+INFO:root:Epoch[1] Batch [2060]	Speed: 664.42 samples/sec	acc=0.079590
+INFO:root:Epoch[1] Batch [2080]	Speed: 665.51 samples/sec	acc=0.075977
+INFO:root:Epoch[1] Batch [2100]	Speed: 665.09 samples/sec	acc=0.085645
+INFO:root:Epoch[1] Batch [2120]	Speed: 665.31 samples/sec	acc=0.080078
+INFO:root:Epoch[1] Batch [2140]	Speed: 664.63 samples/sec	acc=0.087500
+INFO:root:Epoch[1] Batch [2160]	Speed: 664.01 samples/sec	acc=0.084375
+INFO:root:Epoch[1] Batch [2180]	Speed: 664.13 samples/sec	acc=0.082227
+INFO:root:Epoch[1] Batch [2200]	Speed: 665.01 samples/sec	acc=0.083105
+INFO:root:Epoch[1] Batch [2220]	Speed: 664.57 samples/sec	acc=0.079785
+INFO:root:Epoch[1] Batch [2240]	Speed: 664.51 samples/sec	acc=0.086523
+INFO:root:Epoch[1] Batch [2260]	Speed: 664.53 samples/sec	acc=0.078223
+INFO:root:Epoch[1] Batch [2280]	Speed: 664.83 samples/sec	acc=0.087988
+INFO:root:Epoch[1] Batch [2300]	Speed: 664.73 samples/sec	acc=0.081543
+INFO:root:Epoch[1] Batch [2320]	Speed: 664.37 samples/sec	acc=0.084668
+INFO:root:Epoch[1] Batch [2340]	Speed: 665.15 samples/sec	acc=0.081738
+INFO:root:Epoch[1] Batch [2360]	Speed: 663.92 samples/sec	acc=0.079004
+INFO:root:Epoch[1] Batch [2380]	Speed: 665.51 samples/sec	acc=0.082031
+INFO:root:Epoch[1] Batch [2400]	Speed: 664.35 samples/sec	acc=0.077051
+INFO:root:Epoch[1] Batch [2420]	Speed: 662.95 samples/sec	acc=0.082227
+INFO:root:Epoch[1] Batch [2440]	Speed: 676.10 samples/sec	acc=0.081836
+INFO:root:Epoch[1] Batch [2460]	Speed: 663.98 samples/sec	acc=0.086719
+[10000][MARGIN]0.664993,0.231848
+INFO:root:Epoch[1] Batch [2480]	Speed: 663.22 samples/sec	acc=0.082617
+lr-batch-epoch: 0.1 2480 1
+testing verification..
+(14000, 512)
+infer time 17.877836
+[cfp_ff][10000]XNorm: 20.343479
+[cfp_ff][10000]Accuracy-Flip: 0.98929+-0.00327
+testing verification..
+(14000, 512)
+infer time 19.623478
+[cfp_fp][10000]XNorm: 17.196823
+[cfp_fp][10000]Accuracy-Flip: 0.82243+-0.01548
+testing verification..
+(12000, 512)
+infer time 15.526678
+[agedb_30][10000]XNorm: 20.347628
+[agedb_30][10000]Accuracy-Flip: 0.92933+-0.01628
+testing verification..
+(12000, 512)
+infer time 16.828989
+[lfw][10000]XNorm: 20.625510
+[lfw][10000]Accuracy-Flip: 0.98800+-0.00482
+[10000]Accuracy-Highest: 0.99033
+INFO:root:Epoch[1] Batch [2500]	Speed: 98.69 samples/sec	acc=0.086621
+INFO:root:Epoch[1] Batch [2520]	Speed: 678.48 samples/sec	acc=0.088672
+INFO:root:Epoch[1] Batch [2540]	Speed: 678.65 samples/sec	acc=0.089551
+INFO:root:Epoch[1] Batch [2560]	Speed: 669.09 samples/sec	acc=0.082324
+INFO:root:Epoch[1] Batch [2580]	Speed: 664.26 samples/sec	acc=0.083105
+INFO:root:Epoch[1] Batch [2600]	Speed: 663.67 samples/sec	acc=0.087207
+INFO:root:Epoch[1] Batch [2620]	Speed: 664.71 samples/sec	acc=0.085352
+INFO:root:Epoch[1] Batch [2640]	Speed: 664.20 samples/sec	acc=0.085840
+INFO:root:Epoch[1] Batch [2660]	Speed: 663.51 samples/sec	acc=0.073438
+INFO:root:Epoch[1] Batch [2680]	Speed: 665.76 samples/sec	acc=0.082129
+INFO:root:Epoch[1] Batch [2700]	Speed: 664.60 samples/sec	acc=0.081934
+INFO:root:Epoch[1] Batch [2720]	Speed: 664.04 samples/sec	acc=0.085352
+INFO:root:Epoch[1] Batch [2740]	Speed: 665.08 samples/sec	acc=0.082617
+INFO:root:Epoch[1] Batch [2760]	Speed: 664.56 samples/sec	acc=0.091699
+INFO:root:Epoch[1] Batch [2780]	Speed: 664.34 samples/sec	acc=0.085352
+INFO:root:Epoch[1] Batch [2800]	Speed: 664.53 samples/sec	acc=0.083496
+INFO:root:Epoch[1] Batch [2820]	Speed: 664.41 samples/sec	acc=0.088477
+INFO:root:Epoch[1] Batch [2840]	Speed: 663.93 samples/sec	acc=0.083008
+INFO:root:Epoch[1] Batch [2860]	Speed: 665.67 samples/sec	acc=0.091309
+INFO:root:Epoch[1] Batch [2880]	Speed: 665.30 samples/sec	acc=0.085254
+INFO:root:Epoch[1] Batch [2900]	Speed: 664.62 samples/sec	acc=0.086816
+INFO:root:Epoch[1] Batch [2920]	Speed: 664.53 samples/sec	acc=0.087695
+INFO:root:Epoch[1] Batch [2940]	Speed: 664.82 samples/sec	acc=0.088574
+INFO:root:Epoch[1] Batch [2960]	Speed: 664.41 samples/sec	acc=0.090039
+INFO:root:Epoch[1] Batch [2980]	Speed: 664.68 samples/sec	acc=0.088672
+INFO:root:Epoch[1] Batch [3000]	Speed: 664.52 samples/sec	acc=0.082422
+INFO:root:Epoch[1] Batch [3020]	Speed: 662.59 samples/sec	acc=0.089453
+INFO:root:Epoch[1] Batch [3040]	Speed: 664.72 samples/sec	acc=0.084570
+INFO:root:Epoch[1] Batch [3060]	Speed: 664.22 samples/sec	acc=0.086230
+INFO:root:Epoch[1] Batch [3080]	Speed: 664.08 samples/sec	acc=0.085645
+INFO:root:Epoch[1] Batch [3100]	Speed: 664.31 samples/sec	acc=0.084863
+INFO:root:Epoch[1] Batch [3120]	Speed: 665.35 samples/sec	acc=0.090723
+INFO:root:Epoch[1] Batch [3140]	Speed: 663.29 samples/sec	acc=0.085156
+INFO:root:Epoch[1] Batch [3160]	Speed: 664.39 samples/sec	acc=0.082031
+INFO:root:Epoch[1] Batch [3180]	Speed: 663.95 samples/sec	acc=0.092871
+INFO:root:Epoch[1] Batch [3200]	Speed: 664.39 samples/sec	acc=0.086523
+INFO:root:Epoch[1] Batch [3220]	Speed: 666.84 samples/sec	acc=0.087109
+INFO:root:Epoch[1] Batch [3240]	Speed: 664.72 samples/sec	acc=0.088867
+INFO:root:Epoch[1] Batch [3260]	Speed: 664.66 samples/sec	acc=0.089355
+INFO:root:Epoch[1] Batch [3280]	Speed: 664.14 samples/sec	acc=0.086328
+INFO:root:Epoch[1] Batch [3300]	Speed: 664.52 samples/sec	acc=0.083691
+INFO:root:Epoch[1] Batch [3320]	Speed: 663.75 samples/sec	acc=0.088477
+INFO:root:Epoch[1] Batch [3340]	Speed: 664.86 samples/sec	acc=0.092871
+INFO:root:Epoch[1] Batch [3360]	Speed: 663.90 samples/sec	acc=0.086035
+INFO:root:Epoch[1] Batch [3380]	Speed: 665.00 samples/sec	acc=0.086914
+INFO:root:Epoch[1] Batch [3400]	Speed: 664.49 samples/sec	acc=0.085547
+INFO:root:Epoch[1] Batch [3420]	Speed: 664.78 samples/sec	acc=0.088867
+INFO:root:Epoch[1] Batch [3440]	Speed: 664.42 samples/sec	acc=0.089551
+INFO:root:Epoch[1] Batch [3460]	Speed: 664.11 samples/sec	acc=0.080762
+INFO:root:Epoch[1] Batch [3480]	Speed: 664.17 samples/sec	acc=0.087500
+lr-batch-epoch: 0.1 3480 1
+INFO:root:Epoch[1] Batch [3500]	Speed: 663.87 samples/sec	acc=0.089844
+INFO:root:Epoch[1] Batch [3520]	Speed: 665.16 samples/sec	acc=0.089551
+INFO:root:Epoch[1] Batch [3540]	Speed: 664.13 samples/sec	acc=0.087500
+INFO:root:Epoch[1] Batch [3560]	Speed: 664.60 samples/sec	acc=0.085547
+INFO:root:Epoch[1] Batch [3580]	Speed: 664.75 samples/sec	acc=0.084570
+INFO:root:Epoch[1] Batch [3600]	Speed: 665.00 samples/sec	acc=0.095020
+INFO:root:Epoch[1] Batch [3620]	Speed: 663.77 samples/sec	acc=0.087793
+INFO:root:Epoch[1] Batch [3640]	Speed: 664.30 samples/sec	acc=0.091895
+INFO:root:Epoch[1] Batch [3660]	Speed: 664.50 samples/sec	acc=0.089355
+INFO:root:Epoch[1] Batch [3680]	Speed: 664.58 samples/sec	acc=0.089941
+INFO:root:Epoch[1] Batch [3700]	Speed: 664.02 samples/sec	acc=0.090039
+INFO:root:Epoch[1] Batch [3720]	Speed: 664.71 samples/sec	acc=0.088965
+INFO:root:Epoch[1] Batch [3740]	Speed: 664.16 samples/sec	acc=0.089355
+INFO:root:Epoch[1] Batch [3760]	Speed: 664.09 samples/sec	acc=0.087012
+INFO:root:Epoch[1] Batch [3780]	Speed: 664.70 samples/sec	acc=0.090723
+INFO:root:Epoch[1] Batch [3800]	Speed: 664.31 samples/sec	acc=0.089355
+INFO:root:Epoch[1] Batch [3820]	Speed: 664.69 samples/sec	acc=0.096973
+INFO:root:Epoch[1] Batch [3840]	Speed: 664.36 samples/sec	acc=0.084961
+INFO:root:Epoch[1] Batch [3860]	Speed: 664.40 samples/sec	acc=0.091992
+INFO:root:Epoch[1] Batch [3880]	Speed: 664.58 samples/sec	acc=0.092480
+INFO:root:Epoch[1] Batch [3900]	Speed: 664.23 samples/sec	acc=0.092578
+INFO:root:Epoch[1] Batch [3920]	Speed: 664.49 samples/sec	acc=0.093066
+INFO:root:Epoch[1] Batch [3940]	Speed: 663.86 samples/sec	acc=0.091309
+INFO:root:Epoch[1] Batch [3960]	Speed: 665.03 samples/sec	acc=0.090527
+INFO:root:Epoch[1] Batch [3980]	Speed: 663.71 samples/sec	acc=0.084863
+INFO:root:Epoch[1] Batch [4000]	Speed: 664.63 samples/sec	acc=0.095703
+INFO:root:Epoch[1] Batch [4020]	Speed: 669.26 samples/sec	acc=0.095703
+INFO:root:Epoch[1] Batch [4040]	Speed: 675.43 samples/sec	acc=0.089648
+INFO:root:Epoch[1] Batch [4060]	Speed: 664.82 samples/sec	acc=0.086328
+INFO:root:Epoch[1] Batch [4080]	Speed: 663.26 samples/sec	acc=0.092480
+INFO:root:Epoch[1] Batch [4100]	Speed: 664.69 samples/sec	acc=0.092188
+INFO:root:Epoch[1] Batch [4120]	Speed: 664.61 samples/sec	acc=0.086914
+INFO:root:Epoch[1] Batch [4140]	Speed: 663.93 samples/sec	acc=0.089258
+INFO:root:Epoch[1] Batch [4160]	Speed: 663.79 samples/sec	acc=0.096094
+INFO:root:Epoch[1] Batch [4180]	Speed: 664.65 samples/sec	acc=0.090820
+INFO:root:Epoch[1] Batch [4200]	Speed: 664.80 samples/sec	acc=0.093359
+INFO:root:Epoch[1] Batch [4220]	Speed: 664.67 samples/sec	acc=0.091699
+INFO:root:Epoch[1] Batch [4240]	Speed: 664.96 samples/sec	acc=0.091602
+INFO:root:Epoch[1] Batch [4260]	Speed: 665.19 samples/sec	acc=0.094824
+INFO:root:Epoch[1] Batch [4280]	Speed: 663.33 samples/sec	acc=0.092676
+INFO:root:Epoch[1] Batch [4300]	Speed: 664.50 samples/sec	acc=0.093555
+INFO:root:Epoch[1] Batch [4320]	Speed: 664.55 samples/sec	acc=0.092969
+INFO:root:Epoch[1] Batch [4340]	Speed: 664.03 samples/sec	acc=0.090918
+INFO:root:Epoch[1] Batch [4360]	Speed: 664.82 samples/sec	acc=0.090527
+INFO:root:Epoch[1] Batch [4380]	Speed: 665.27 samples/sec	acc=0.087793
+INFO:root:Epoch[1] Batch [4400]	Speed: 664.36 samples/sec	acc=0.096582
+INFO:root:Epoch[1] Batch [4420]	Speed: 664.62 samples/sec	acc=0.088477
+INFO:root:Epoch[1] Batch [4440]	Speed: 664.28 samples/sec	acc=0.090625
+INFO:root:Epoch[1] Batch [4460]	Speed: 663.94 samples/sec	acc=0.089258
+[12000][MARGIN]0.663869,0.231542
+INFO:root:Epoch[1] Batch [4480]	Speed: 664.20 samples/sec	acc=0.093359
+lr-batch-epoch: 0.1 4480 1
+testing verification..
+(14000, 512)
+infer time 19.055237
+[cfp_ff][12000]XNorm: 20.237718
+[cfp_ff][12000]Accuracy-Flip: 0.99200+-0.00321
+testing verification..
+(14000, 512)
+infer time 18.214119
+[cfp_fp][12000]XNorm: 16.544081
+[cfp_fp][12000]Accuracy-Flip: 0.82214+-0.01447
+testing verification..
+(12000, 512)
+infer time 19.305643
+[agedb_30][12000]XNorm: 19.872737
+[agedb_30][12000]Accuracy-Flip: 0.93100+-0.01537
+testing verification..
+(12000, 512)
+infer time 17.90811
+[lfw][12000]XNorm: 20.345964
+[lfw][12000]Accuracy-Flip: 0.99017+-0.00540
+[12000]Accuracy-Highest: 0.99033
+INFO:root:Epoch[1] Batch [4500]	Speed: 91.87 samples/sec	acc=0.093457
+INFO:root:Epoch[1] Batch [4520]	Speed: 663.13 samples/sec	acc=0.094629
+INFO:root:Epoch[1] Batch [4540]	Speed: 664.12 samples/sec	acc=0.089648
+INFO:root:Epoch[1] Batch [4560]	Speed: 663.19 samples/sec	acc=0.095703
+INFO:root:Epoch[1] Batch [4580]	Speed: 663.06 samples/sec	acc=0.092285
+INFO:root:Epoch[1] Batch [4600]	Speed: 664.42 samples/sec	acc=0.089746
+INFO:root:Epoch[1] Batch [4620]	Speed: 663.47 samples/sec	acc=0.091992
+INFO:root:Epoch[1] Batch [4640]	Speed: 664.15 samples/sec	acc=0.090430
+INFO:root:Epoch[1] Batch [4660]	Speed: 664.55 samples/sec	acc=0.095215
+INFO:root:Epoch[1] Batch [4680]	Speed: 664.79 samples/sec	acc=0.095508
+INFO:root:Epoch[1] Batch [4700]	Speed: 664.10 samples/sec	acc=0.092480
+INFO:root:Epoch[1] Batch [4720]	Speed: 664.65 samples/sec	acc=0.092578
+INFO:root:Epoch[1] Batch [4740]	Speed: 661.35 samples/sec	acc=0.096484
+INFO:root:Epoch[1] Batch [4760]	Speed: 663.40 samples/sec	acc=0.094238
+INFO:root:Epoch[1] Batch [4780]	Speed: 664.87 samples/sec	acc=0.095508
+INFO:root:Epoch[1] Batch [4800]	Speed: 665.01 samples/sec	acc=0.095117
+INFO:root:Epoch[1] Batch [4820]	Speed: 673.97 samples/sec	acc=0.097266
+INFO:root:Epoch[1] Batch [4840]	Speed: 680.04 samples/sec	acc=0.091016
+INFO:root:Epoch[1] Batch [4860]	Speed: 671.80 samples/sec	acc=0.089355
+INFO:root:Epoch[1] Batch [4880]	Speed: 664.33 samples/sec	acc=0.091406
+INFO:root:Epoch[1] Batch [4900]	Speed: 665.06 samples/sec	acc=0.097461
+INFO:root:Epoch[1] Batch [4920]	Speed: 664.58 samples/sec	acc=0.092188
+INFO:root:Epoch[1] Batch [4940]	Speed: 664.17 samples/sec	acc=0.094336
+INFO:root:Epoch[1] Batch [4960]	Speed: 665.38 samples/sec	acc=0.094238
+INFO:root:Epoch[1] Batch [4980]	Speed: 664.12 samples/sec	acc=0.097070
+INFO:root:Epoch[1] Batch [5000]	Speed: 664.07 samples/sec	acc=0.095898
+INFO:root:Epoch[1] Batch [5020]	Speed: 664.17 samples/sec	acc=0.093750
+INFO:root:Epoch[1] Batch [5040]	Speed: 664.34 samples/sec	acc=0.098828
+INFO:root:Epoch[1] Batch [5060]	Speed: 664.32 samples/sec	acc=0.093555
+INFO:root:Epoch[1] Batch [5080]	Speed: 664.87 samples/sec	acc=0.094531
+INFO:root:Epoch[1] Batch [5100]	Speed: 664.44 samples/sec	acc=0.093652
+INFO:root:Epoch[1] Batch [5120]	Speed: 664.01 samples/sec	acc=0.089941
+INFO:root:Epoch[1] Batch [5140]	Speed: 664.60 samples/sec	acc=0.096387
+INFO:root:Epoch[1] Batch [5160]	Speed: 664.69 samples/sec	acc=0.097070
+INFO:root:Epoch[1] Batch [5180]	Speed: 664.06 samples/sec	acc=0.097852
+INFO:root:Epoch[1] Batch [5200]	Speed: 665.19 samples/sec	acc=0.095898
+INFO:root:Epoch[1] Batch [5220]	Speed: 664.77 samples/sec	acc=0.095801
+INFO:root:Epoch[1] Batch [5240]	Speed: 664.18 samples/sec	acc=0.095996
+INFO:root:Epoch[1] Batch [5260]	Speed: 665.28 samples/sec	acc=0.096582
+INFO:root:Epoch[1] Batch [5280]	Speed: 663.67 samples/sec	acc=0.102148
+INFO:root:Epoch[1] Batch [5300]	Speed: 664.59 samples/sec	acc=0.095801
+INFO:root:Epoch[1] Batch [5320]	Speed: 665.23 samples/sec	acc=0.098633
+INFO:root:Epoch[1] Batch [5340]	Speed: 664.13 samples/sec	acc=0.099414
+INFO:root:Epoch[1] Batch [5360]	Speed: 663.60 samples/sec	acc=0.093262
+INFO:root:Epoch[1] Batch [5380]	Speed: 664.76 samples/sec	acc=0.095020
+INFO:root:Epoch[1] Batch [5400]	Speed: 664.85 samples/sec	acc=0.091602
+INFO:root:Epoch[1] Batch [5420]	Speed: 663.86 samples/sec	acc=0.102734
+INFO:root:Epoch[1] Batch [5440]	Speed: 664.46 samples/sec	acc=0.094434
+INFO:root:Epoch[1] Batch [5460]	Speed: 664.61 samples/sec	acc=0.090820
+INFO:root:Epoch[1] Batch [5480]	Speed: 664.34 samples/sec	acc=0.090723
+lr-batch-epoch: 0.1 5480 1
+INFO:root:Epoch[1] Batch [5500]	Speed: 664.63 samples/sec	acc=0.094434
+INFO:root:Epoch[1] Batch [5520]	Speed: 664.64 samples/sec	acc=0.097559
+INFO:root:Epoch[1] Batch [5540]	Speed: 664.70 samples/sec	acc=0.098730
+INFO:root:Epoch[1] Batch [5560]	Speed: 665.25 samples/sec	acc=0.094238
+INFO:root:Epoch[1] Batch [5580]	Speed: 664.78 samples/sec	acc=0.091992
+INFO:root:Epoch[1] Batch [5600]	Speed: 663.40 samples/sec	acc=0.094922
+INFO:root:Epoch[1] Batch [5620]	Speed: 676.34 samples/sec	acc=0.093750
+INFO:root:Epoch[1] Batch [5640]	Speed: 664.83 samples/sec	acc=0.097754
+INFO:root:Epoch[1] Batch [5660]	Speed: 664.16 samples/sec	acc=0.098340
+INFO:root:Epoch[1] Batch [5680]	Speed: 664.74 samples/sec	acc=0.096875
+INFO:root:Epoch[1] Batch [5700]	Speed: 664.90 samples/sec	acc=0.096191
+INFO:root:Epoch[1] Batch [5720]	Speed: 664.13 samples/sec	acc=0.095410
+INFO:root:Epoch[1] Batch [5740]	Speed: 665.64 samples/sec	acc=0.096973
+INFO:root:Epoch[1] Batch [5760]	Speed: 664.07 samples/sec	acc=0.099512
+INFO:root:Epoch[1] Batch [5780]	Speed: 664.54 samples/sec	acc=0.096777
+INFO:root:Epoch[1] Batch [5800]	Speed: 664.87 samples/sec	acc=0.096191
+INFO:root:Epoch[1] Batch [5820]	Speed: 664.23 samples/sec	acc=0.096973
+INFO:root:Epoch[1] Batch [5840]	Speed: 663.94 samples/sec	acc=0.100098
+INFO:root:Epoch[1] Batch [5860]	Speed: 664.64 samples/sec	acc=0.098535
+INFO:root:Epoch[1] Batch [5880]	Speed: 664.42 samples/sec	acc=0.098926
+INFO:root:Epoch[1] Batch [5900]	Speed: 664.01 samples/sec	acc=0.100488
+INFO:root:Epoch[1] Batch [5920]	Speed: 664.99 samples/sec	acc=0.097656
+INFO:root:Epoch[1] Batch [5940]	Speed: 664.46 samples/sec	acc=0.094238
+INFO:root:Epoch[1] Batch [5960]	Speed: 663.98 samples/sec	acc=0.097363
+INFO:root:Epoch[1] Batch [5980]	Speed: 665.40 samples/sec	acc=0.092188
+INFO:root:Epoch[1] Batch [6000]	Speed: 665.01 samples/sec	acc=0.093457
+INFO:root:Epoch[1] Batch [6020]	Speed: 664.35 samples/sec	acc=0.100586
+INFO:root:Epoch[1] Batch [6040]	Speed: 665.23 samples/sec	acc=0.102344
+INFO:root:Epoch[1] Batch [6060]	Speed: 664.94 samples/sec	acc=0.101562
+INFO:root:Epoch[1] Batch [6080]	Speed: 664.54 samples/sec	acc=0.099512
+INFO:root:Epoch[1] Batch [6100]	Speed: 665.41 samples/sec	acc=0.099316
+INFO:root:Epoch[1] Batch [6120]	Speed: 664.04 samples/sec	acc=0.100781
+INFO:root:Epoch[1] Batch [6140]	Speed: 663.61 samples/sec	acc=0.103125
+INFO:root:Epoch[1] Batch [6160]	Speed: 665.18 samples/sec	acc=0.098633
+INFO:root:Epoch[1] Batch [6180]	Speed: 664.86 samples/sec	acc=0.101074
+INFO:root:Epoch[1] Batch [6200]	Speed: 664.23 samples/sec	acc=0.097852
+INFO:root:Epoch[1] Batch [6220]	Speed: 665.03 samples/sec	acc=0.097949
+INFO:root:Epoch[1] Batch [6240]	Speed: 664.08 samples/sec	acc=0.099219
+INFO:root:Epoch[1] Batch [6260]	Speed: 663.75 samples/sec	acc=0.097754
+INFO:root:Epoch[1] Batch [6280]	Speed: 664.83 samples/sec	acc=0.099902
+INFO:root:Epoch[1] Batch [6300]	Speed: 664.99 samples/sec	acc=0.092969
+INFO:root:Epoch[1] Batch [6320]	Speed: 664.05 samples/sec	acc=0.102246
+INFO:root:Epoch[1] Batch [6340]	Speed: 664.66 samples/sec	acc=0.096387
+INFO:root:Epoch[1] Batch [6360]	Speed: 665.46 samples/sec	acc=0.098535
+INFO:root:Epoch[1] Batch [6380]	Speed: 664.46 samples/sec	acc=0.104395
+INFO:root:Epoch[1] Batch [6400]	Speed: 665.45 samples/sec	acc=0.100977
+INFO:root:Epoch[1] Batch [6420]	Speed: 664.39 samples/sec	acc=0.102051
+INFO:root:Epoch[1] Batch [6440]	Speed: 664.14 samples/sec	acc=0.098633
+INFO:root:Epoch[1] Batch [6460]	Speed: 664.70 samples/sec	acc=0.095508
+[14000][MARGIN]0.663905,0.231038
+INFO:root:Epoch[1] Batch [6480]	Speed: 664.85 samples/sec	acc=0.102734
+lr-batch-epoch: 0.1 6480 1
+testing verification..
+(14000, 512)
+infer time 23.191147
+[cfp_ff][14000]XNorm: 21.783922
+[cfp_ff][14000]Accuracy-Flip: 0.99243+-0.00384
+testing verification..
+(14000, 512)
+infer time 24.450724
+[cfp_fp][14000]XNorm: 18.047383
+[cfp_fp][14000]Accuracy-Flip: 0.83314+-0.01562
+testing verification..
+(12000, 512)
+infer time 24.3648
+[agedb_30][14000]XNorm: 21.715938
+[agedb_30][14000]Accuracy-Flip: 0.93733+-0.01510
+testing verification..
+(12000, 512)
+infer time 21.910281
+[lfw][14000]XNorm: 22.156482
+[lfw][14000]Accuracy-Flip: 0.99100+-0.00389
+saving 7
+INFO:root:Saved checkpoint to "../model-r50-am-lfw/model-0007.params"
+[14000]Accuracy-Highest: 0.99100
+INFO:root:Epoch[1] Batch [6500]	Speed: 76.89 samples/sec	acc=0.098828
+INFO:root:Epoch[1] Batch [6520]	Speed: 663.29 samples/sec	acc=0.096973
+INFO:root:Epoch[1] Batch [6540]	Speed: 663.21 samples/sec	acc=0.094141
+INFO:root:Epoch[1] Batch [6560]	Speed: 663.48 samples/sec	acc=0.098633
+INFO:root:Epoch[1] Batch [6580]	Speed: 665.25 samples/sec	acc=0.094922
+INFO:root:Epoch[1] Batch [6600]	Speed: 662.71 samples/sec	acc=0.094629
+INFO:root:Epoch[1] Batch [6620]	Speed: 664.04 samples/sec	acc=0.104590
+INFO:root:Epoch[1] Batch [6640]	Speed: 664.52 samples/sec	acc=0.095312
+INFO:root:Epoch[1] Batch [6660]	Speed: 664.97 samples/sec	acc=0.103223
+INFO:root:Epoch[1] Batch [6680]	Speed: 664.04 samples/sec	acc=0.098926
+INFO:root:Epoch[1] Batch [6700]	Speed: 665.37 samples/sec	acc=0.101758
+INFO:root:Epoch[1] Batch [6720]	Speed: 665.17 samples/sec	acc=0.101660
+INFO:root:Epoch[1] Batch [6740]	Speed: 664.60 samples/sec	acc=0.103613
+INFO:root:Epoch[1] Batch [6760]	Speed: 664.41 samples/sec	acc=0.103711
+INFO:root:Epoch[1] Batch [6780]	Speed: 665.11 samples/sec	acc=0.102441
+INFO:root:Epoch[1] Batch [6800]	Speed: 664.46 samples/sec	acc=0.100391
+INFO:root:Epoch[1] Batch [6820]	Speed: 665.60 samples/sec	acc=0.102344
+INFO:root:Epoch[1] Batch [6840]	Speed: 664.00 samples/sec	acc=0.103516
+INFO:root:Epoch[1] Batch [6860]	Speed: 664.43 samples/sec	acc=0.099023
+INFO:root:Epoch[1] Batch [6880]	Speed: 664.94 samples/sec	acc=0.101074
+INFO:root:Epoch[1] Batch [6900]	Speed: 672.25 samples/sec	acc=0.104980
+INFO:root:Epoch[1] Batch [6920]	Speed: 678.31 samples/sec	acc=0.100684
+INFO:root:Epoch[1] Batch [6940]	Speed: 679.02 samples/sec	acc=0.097168
+INFO:root:Epoch[1] Batch [6960]	Speed: 678.62 samples/sec	acc=0.100586
+INFO:root:Epoch[1] Batch [6980]	Speed: 678.71 samples/sec	acc=0.100879
+INFO:root:Epoch[1] Batch [7000]	Speed: 679.13 samples/sec	acc=0.102344
+INFO:root:Epoch[1] Batch [7020]	Speed: 671.83 samples/sec	acc=0.098633
+INFO:root:Epoch[1] Batch [7040]	Speed: 664.19 samples/sec	acc=0.103320
+INFO:root:Epoch[1] Batch [7060]	Speed: 664.75 samples/sec	acc=0.097363
+INFO:root:Epoch[1] Batch [7080]	Speed: 664.33 samples/sec	acc=0.106250
+INFO:root:Epoch[1] Batch [7100]	Speed: 664.17 samples/sec	acc=0.103418
+INFO:root:Epoch[1] Batch [7120]	Speed: 667.57 samples/sec	acc=0.107422
+INFO:root:Epoch[1] Batch [7140]	Speed: 678.84 samples/sec	acc=0.098437
+INFO:root:Epoch[1] Batch [7160]	Speed: 673.04 samples/sec	acc=0.099609
+INFO:root:Epoch[1] Batch [7180]	Speed: 664.41 samples/sec	acc=0.103516
+INFO:root:Epoch[1] Batch [7200]	Speed: 664.60 samples/sec	acc=0.104102
+INFO:root:Epoch[1] Batch [7220]	Speed: 663.82 samples/sec	acc=0.097266
+INFO:root:Epoch[1] Batch [7240]	Speed: 665.29 samples/sec	acc=0.101758
+INFO:root:Epoch[1] Batch [7260]	Speed: 664.91 samples/sec	acc=0.104004
+INFO:root:Epoch[1] Batch [7280]	Speed: 664.18 samples/sec	acc=0.104980
+INFO:root:Epoch[1] Batch [7300]	Speed: 664.74 samples/sec	acc=0.104980
+INFO:root:Epoch[1] Batch [7320]	Speed: 664.91 samples/sec	acc=0.099707
+INFO:root:Epoch[1] Batch [7340]	Speed: 664.00 samples/sec	acc=0.103125
+INFO:root:Epoch[1] Batch [7360]	Speed: 665.13 samples/sec	acc=0.103516
+INFO:root:Epoch[1] Batch [7380]	Speed: 664.32 samples/sec	acc=0.100098
+INFO:root:Epoch[1] Batch [7400]	Speed: 664.56 samples/sec	acc=0.099902
+INFO:root:Epoch[1] Batch [7420]	Speed: 664.68 samples/sec	acc=0.098633
+INFO:root:Epoch[1] Batch [7440]	Speed: 664.59 samples/sec	acc=0.102930
+INFO:root:Epoch[1] Batch [7460]	Speed: 663.32 samples/sec	acc=0.100977
+INFO:root:Epoch[1] Batch [7480]	Speed: 665.12 samples/sec	acc=0.100977
+lr-batch-epoch: 0.1 7480 1
+INFO:root:Epoch[1] Batch [7500]	Speed: 664.33 samples/sec	acc=0.096973
+INFO:root:Epoch[1] Train-acc=0.102322
+INFO:root:Epoch[1] Time cost=6187.711
+call reset()
+INFO:root:Epoch[2] Batch [20]	Speed: 662.76 samples/sec	acc=0.123791
+INFO:root:Epoch[2] Batch [40]	Speed: 664.15 samples/sec	acc=0.120020
+INFO:root:Epoch[2] Batch [60]	Speed: 664.97 samples/sec	acc=0.121094
+INFO:root:Epoch[2] Batch [80]	Speed: 665.05 samples/sec	acc=0.122461
+INFO:root:Epoch[2] Batch [100]	Speed: 664.38 samples/sec	acc=0.123242
+INFO:root:Epoch[2] Batch [120]	Speed: 665.15 samples/sec	acc=0.120410
+INFO:root:Epoch[2] Batch [140]	Speed: 665.00 samples/sec	acc=0.112891
+INFO:root:Epoch[2] Batch [160]	Speed: 664.95 samples/sec	acc=0.114355
+INFO:root:Epoch[2] Batch [180]	Speed: 665.33 samples/sec	acc=0.116504
+INFO:root:Epoch[2] Batch [200]	Speed: 665.11 samples/sec	acc=0.114551
+INFO:root:Epoch[2] Batch [220]	Speed: 664.61 samples/sec	acc=0.119141
+INFO:root:Epoch[2] Batch [240]	Speed: 665.19 samples/sec	acc=0.114941
+INFO:root:Epoch[2] Batch [260]	Speed: 663.86 samples/sec	acc=0.110937
+INFO:root:Epoch[2] Batch [280]	Speed: 664.48 samples/sec	acc=0.108594
+INFO:root:Epoch[2] Batch [300]	Speed: 664.79 samples/sec	acc=0.114062
+INFO:root:Epoch[2] Batch [320]	Speed: 665.11 samples/sec	acc=0.112988
+INFO:root:Epoch[2] Batch [340]	Speed: 664.95 samples/sec	acc=0.116406
+INFO:root:Epoch[2] Batch [360]	Speed: 667.06 samples/sec	acc=0.107910
+INFO:root:Epoch[2] Batch [380]	Speed: 665.33 samples/sec	acc=0.107617
+INFO:root:Epoch[2] Batch [400]	Speed: 664.36 samples/sec	acc=0.106934
+INFO:root:Epoch[2] Batch [420]	Speed: 664.53 samples/sec	acc=0.107031
+INFO:root:Epoch[2] Batch [440]	Speed: 664.55 samples/sec	acc=0.106152
+INFO:root:Epoch[2] Batch [460]	Speed: 664.04 samples/sec	acc=0.111914
+INFO:root:Epoch[2] Batch [480]	Speed: 665.19 samples/sec	acc=0.105957
+INFO:root:Epoch[2] Batch [500]	Speed: 665.20 samples/sec	acc=0.106348
+INFO:root:Epoch[2] Batch [520]	Speed: 665.11 samples/sec	acc=0.107715
+INFO:root:Epoch[2] Batch [540]	Speed: 664.99 samples/sec	acc=0.100879
+INFO:root:Epoch[2] Batch [560]	Speed: 664.56 samples/sec	acc=0.103711
+INFO:root:Epoch[2] Batch [580]	Speed: 664.09 samples/sec	acc=0.112793
+INFO:root:Epoch[2] Batch [600]	Speed: 665.09 samples/sec	acc=0.104980
+INFO:root:Epoch[2] Batch [620]	Speed: 664.69 samples/sec	acc=0.102930
+INFO:root:Epoch[2] Batch [640]	Speed: 664.65 samples/sec	acc=0.105762
+INFO:root:Epoch[2] Batch [660]	Speed: 664.24 samples/sec	acc=0.107227
+INFO:root:Epoch[2] Batch [680]	Speed: 664.73 samples/sec	acc=0.108691
+INFO:root:Epoch[2] Batch [700]	Speed: 664.50 samples/sec	acc=0.107813
+INFO:root:Epoch[2] Batch [720]	Speed: 664.69 samples/sec	acc=0.105566
+INFO:root:Epoch[2] Batch [740]	Speed: 665.19 samples/sec	acc=0.112793
+INFO:root:Epoch[2] Batch [760]	Speed: 664.86 samples/sec	acc=0.109277
+INFO:root:Epoch[2] Batch [780]	Speed: 663.88 samples/sec	acc=0.108496
+INFO:root:Epoch[2] Batch [800]	Speed: 664.99 samples/sec	acc=0.100684
+INFO:root:Epoch[2] Batch [820]	Speed: 664.61 samples/sec	acc=0.109473
+INFO:root:Epoch[2] Batch [840]	Speed: 664.20 samples/sec	acc=0.103027
+INFO:root:Epoch[2] Batch [860]	Speed: 665.03 samples/sec	acc=0.105273
+INFO:root:Epoch[2] Batch [880]	Speed: 665.11 samples/sec	acc=0.101660
+INFO:root:Epoch[2] Batch [900]	Speed: 664.63 samples/sec	acc=0.106738
+INFO:root:Epoch[2] Batch [920]	Speed: 663.97 samples/sec	acc=0.107520
+INFO:root:Epoch[2] Batch [940]	Speed: 664.86 samples/sec	acc=0.103125
+INFO:root:Epoch[2] Batch [960]	Speed: 664.08 samples/sec	acc=0.107422
+[16000][MARGIN]0.651529,0.216234
+lr-batch-epoch: 0.1 961 2
+testing verification..
+(14000, 512)
+infer time 22.022497
+[cfp_ff][16000]XNorm: 22.037617
+[cfp_ff][16000]Accuracy-Flip: 0.99143+-0.00461
+testing verification..
+(14000, 512)
+infer time 22.0002
+[cfp_fp][16000]XNorm: 18.299705
+[cfp_fp][16000]Accuracy-Flip: 0.82743+-0.01448
+testing verification..
+(12000, 512)
+infer time 19.077122
+[agedb_30][16000]XNorm: 21.726744
+[agedb_30][16000]Accuracy-Flip: 0.93767+-0.01450
+testing verification..
+(12000, 512)
+infer time 20.571949
+[lfw][16000]XNorm: 22.139647
+[lfw][16000]Accuracy-Flip: 0.99033+-0.00427
+[16000]Accuracy-Highest: 0.99100
+INFO:root:Epoch[2] Batch [980]	Speed: 88.23 samples/sec	acc=0.105664
+INFO:root:Epoch[2] Batch [1000]	Speed: 663.50 samples/sec	acc=0.103809
+INFO:root:Epoch[2] Batch [1020]	Speed: 663.57 samples/sec	acc=0.110156
+INFO:root:Epoch[2] Batch [1040]	Speed: 664.63 samples/sec	acc=0.105078
+INFO:root:Epoch[2] Batch [1060]	Speed: 664.12 samples/sec	acc=0.110645
+INFO:root:Epoch[2] Batch [1080]	Speed: 670.58 samples/sec	acc=0.110547
+INFO:root:Epoch[2] Batch [1100]	Speed: 685.56 samples/sec	acc=0.110059
+INFO:root:Epoch[2] Batch [1120]	Speed: 664.68 samples/sec	acc=0.107324
+INFO:root:Epoch[2] Batch [1140]	Speed: 664.30 samples/sec	acc=0.108594
+INFO:root:Epoch[2] Batch [1160]	Speed: 665.77 samples/sec	acc=0.105957
+INFO:root:Epoch[2] Batch [1180]	Speed: 664.53 samples/sec	acc=0.101758
+INFO:root:Epoch[2] Batch [1200]	Speed: 664.85 samples/sec	acc=0.108594
+INFO:root:Epoch[2] Batch [1220]	Speed: 664.07 samples/sec	acc=0.107520
+INFO:root:Epoch[2] Batch [1240]	Speed: 664.89 samples/sec	acc=0.103516
+INFO:root:Epoch[2] Batch [1260]	Speed: 664.97 samples/sec	acc=0.107520
+INFO:root:Epoch[2] Batch [1280]	Speed: 665.06 samples/sec	acc=0.106543
+INFO:root:Epoch[2] Batch [1300]	Speed: 664.37 samples/sec	acc=0.104590
+INFO:root:Epoch[2] Batch [1320]	Speed: 664.46 samples/sec	acc=0.111230
+INFO:root:Epoch[2] Batch [1340]	Speed: 664.77 samples/sec	acc=0.107324
+INFO:root:Epoch[2] Batch [1360]	Speed: 664.64 samples/sec	acc=0.107813
+INFO:root:Epoch[2] Batch [1380]	Speed: 665.25 samples/sec	acc=0.106250
+INFO:root:Epoch[2] Batch [1400]	Speed: 664.91 samples/sec	acc=0.108398
+INFO:root:Epoch[2] Batch [1420]	Speed: 664.60 samples/sec	acc=0.105469
+INFO:root:Epoch[2] Batch [1440]	Speed: 663.94 samples/sec	acc=0.104492
+INFO:root:Epoch[2] Batch [1460]	Speed: 665.11 samples/sec	acc=0.105273
+INFO:root:Epoch[2] Batch [1480]	Speed: 665.18 samples/sec	acc=0.104688
+INFO:root:Epoch[2] Batch [1500]	Speed: 663.93 samples/sec	acc=0.109570
+INFO:root:Epoch[2] Batch [1520]	Speed: 664.76 samples/sec	acc=0.109375
+INFO:root:Epoch[2] Batch [1540]	Speed: 663.55 samples/sec	acc=0.107813
+INFO:root:Epoch[2] Batch [1560]	Speed: 665.20 samples/sec	acc=0.112793
+INFO:root:Epoch[2] Batch [1580]	Speed: 665.05 samples/sec	acc=0.104004
+INFO:root:Epoch[2] Batch [1600]	Speed: 664.81 samples/sec	acc=0.110449
+INFO:root:Epoch[2] Batch [1620]	Speed: 664.04 samples/sec	acc=0.103711
+INFO:root:Epoch[2] Batch [1640]	Speed: 664.94 samples/sec	acc=0.103613
+INFO:root:Epoch[2] Batch [1660]	Speed: 664.82 samples/sec	acc=0.106543
+INFO:root:Epoch[2] Batch [1680]	Speed: 664.07 samples/sec	acc=0.115625
+INFO:root:Epoch[2] Batch [1700]	Speed: 664.73 samples/sec	acc=0.108594
+INFO:root:Epoch[2] Batch [1720]	Speed: 664.96 samples/sec	acc=0.111621
+INFO:root:Epoch[2] Batch [1740]	Speed: 664.57 samples/sec	acc=0.108203
+INFO:root:Epoch[2] Batch [1760]	Speed: 664.65 samples/sec	acc=0.105273
+INFO:root:Epoch[2] Batch [1780]	Speed: 664.55 samples/sec	acc=0.109570
+INFO:root:Epoch[2] Batch [1800]	Speed: 663.90 samples/sec	acc=0.108887
+INFO:root:Epoch[2] Batch [1820]	Speed: 665.03 samples/sec	acc=0.110937
+INFO:root:Epoch[2] Batch [1840]	Speed: 664.37 samples/sec	acc=0.108789
+INFO:root:Epoch[2] Batch [1860]	Speed: 665.64 samples/sec	acc=0.108691
+INFO:root:Epoch[2] Batch [1880]	Speed: 664.35 samples/sec	acc=0.106738
+INFO:root:Epoch[2] Batch [1900]	Speed: 664.58 samples/sec	acc=0.110742
+INFO:root:Epoch[2] Batch [1920]	Speed: 664.56 samples/sec	acc=0.116797
+INFO:root:Epoch[2] Batch [1940]	Speed: 664.19 samples/sec	acc=0.107129
+INFO:root:Epoch[2] Batch [1960]	Speed: 664.40 samples/sec	acc=0.107227
+lr-batch-epoch: 0.1 1961 2
+INFO:root:Epoch[2] Batch [1980]	Speed: 664.40 samples/sec	acc=0.108008
+INFO:root:Epoch[2] Batch [2000]	Speed: 664.47 samples/sec	acc=0.106055
+INFO:root:Epoch[2] Batch [2020]	Speed: 663.48 samples/sec	acc=0.112207
+INFO:root:Epoch[2] Batch [2040]	Speed: 664.45 samples/sec	acc=0.104199
+INFO:root:Epoch[2] Batch [2060]	Speed: 664.24 samples/sec	acc=0.110937
+INFO:root:Epoch[2] Batch [2080]	Speed: 664.63 samples/sec	acc=0.104590
+INFO:root:Epoch[2] Batch [2100]	Speed: 664.37 samples/sec	acc=0.110254
+INFO:root:Epoch[2] Batch [2120]	Speed: 664.96 samples/sec	acc=0.109668
+INFO:root:Epoch[2] Batch [2140]	Speed: 664.44 samples/sec	acc=0.111523
+INFO:root:Epoch[2] Batch [2160]	Speed: 664.79 samples/sec	acc=0.107520
+INFO:root:Epoch[2] Batch [2180]	Speed: 664.34 samples/sec	acc=0.106445
+INFO:root:Epoch[2] Batch [2200]	Speed: 664.62 samples/sec	acc=0.109961
+INFO:root:Epoch[2] Batch [2220]	Speed: 664.52 samples/sec	acc=0.109375
+INFO:root:Epoch[2] Batch [2240]	Speed: 665.07 samples/sec	acc=0.109570
+INFO:root:Epoch[2] Batch [2260]	Speed: 664.79 samples/sec	acc=0.108496
+INFO:root:Epoch[2] Batch [2280]	Speed: 664.37 samples/sec	acc=0.107422
+INFO:root:Epoch[2] Batch [2300]	Speed: 663.81 samples/sec	acc=0.102930
+INFO:root:Epoch[2] Batch [2320]	Speed: 664.82 samples/sec	acc=0.106738
+INFO:root:Epoch[2] Batch [2340]	Speed: 663.48 samples/sec	acc=0.111230
+INFO:root:Epoch[2] Batch [2360]	Speed: 664.29 samples/sec	acc=0.109180
+INFO:root:Epoch[2] Batch [2380]	Speed: 664.07 samples/sec	acc=0.107617
+INFO:root:Epoch[2] Batch [2400]	Speed: 664.46 samples/sec	acc=0.111621
+INFO:root:Epoch[2] Batch [2420]	Speed: 663.83 samples/sec	acc=0.111426
+INFO:root:Epoch[2] Batch [2440]	Speed: 664.06 samples/sec	acc=0.111523
+INFO:root:Epoch[2] Batch [2460]	Speed: 664.59 samples/sec	acc=0.106934
+INFO:root:Epoch[2] Batch [2480]	Speed: 664.31 samples/sec	acc=0.112109
+INFO:root:Epoch[2] Batch [2500]	Speed: 664.29 samples/sec	acc=0.114258
+INFO:root:Epoch[2] Batch [2520]	Speed: 665.16 samples/sec	acc=0.108594
+INFO:root:Epoch[2] Batch [2540]	Speed: 664.23 samples/sec	acc=0.109180
+INFO:root:Epoch[2] Batch [2560]	Speed: 663.99 samples/sec	acc=0.116113
+INFO:root:Epoch[2] Batch [2580]	Speed: 664.65 samples/sec	acc=0.110059
+INFO:root:Epoch[2] Batch [2600]	Speed: 664.12 samples/sec	acc=0.110059
+INFO:root:Epoch[2] Batch [2620]	Speed: 663.70 samples/sec	acc=0.109180
+INFO:root:Epoch[2] Batch [2640]	Speed: 661.06 samples/sec	acc=0.111230
+INFO:root:Epoch[2] Batch [2660]	Speed: 671.20 samples/sec	acc=0.113770
+INFO:root:Epoch[2] Batch [2680]	Speed: 664.35 samples/sec	acc=0.106934
+INFO:root:Epoch[2] Batch [2700]	Speed: 664.46 samples/sec	acc=0.109570
+INFO:root:Epoch[2] Batch [2720]	Speed: 664.56 samples/sec	acc=0.109863
+INFO:root:Epoch[2] Batch [2740]	Speed: 664.62 samples/sec	acc=0.109473
+INFO:root:Epoch[2] Batch [2760]	Speed: 663.39 samples/sec	acc=0.113281
+INFO:root:Epoch[2] Batch [2780]	Speed: 663.43 samples/sec	acc=0.107031
+INFO:root:Epoch[2] Batch [2800]	Speed: 663.70 samples/sec	acc=0.112207
+INFO:root:Epoch[2] Batch [2820]	Speed: 664.16 samples/sec	acc=0.110937
+INFO:root:Epoch[2] Batch [2840]	Speed: 664.13 samples/sec	acc=0.114551
+INFO:root:Epoch[2] Batch [2860]	Speed: 664.69 samples/sec	acc=0.112500
+INFO:root:Epoch[2] Batch [2880]	Speed: 664.43 samples/sec	acc=0.110156
+INFO:root:Epoch[2] Batch [2900]	Speed: 664.40 samples/sec	acc=0.106445
+INFO:root:Epoch[2] Batch [2920]	Speed: 664.79 samples/sec	acc=0.112598
+INFO:root:Epoch[2] Batch [2940]	Speed: 663.88 samples/sec	acc=0.108496
+INFO:root:Epoch[2] Batch [2960]	Speed: 664.18 samples/sec	acc=0.109766
+[18000][MARGIN]0.643295,0.205146
+lr-batch-epoch: 0.1 2961 2
+testing verification..
+(14000, 512)
+infer time 20.873428
+[cfp_ff][18000]XNorm: 21.546624
+[cfp_ff][18000]Accuracy-Flip: 0.99343+-0.00339
+testing verification..
+(14000, 512)
+infer time 21.734194
+[cfp_fp][18000]XNorm: 17.674893
+[cfp_fp][18000]Accuracy-Flip: 0.83671+-0.01759
+testing verification..
+(12000, 512)
+infer time 21.034575
+[agedb_30][18000]XNorm: 21.225530
+[agedb_30][18000]Accuracy-Flip: 0.94467+-0.01388
+testing verification..
+(12000, 512)
+infer time 19.427078
+[lfw][18000]XNorm: 21.675605
+[lfw][18000]Accuracy-Flip: 0.99167+-0.00489
+saving 9
+INFO:root:Saved checkpoint to "../model-r50-am-lfw/model-0009.params"
+[18000]Accuracy-Highest: 0.99167
+INFO:root:Epoch[2] Batch [2980]	Speed: 84.94 samples/sec	acc=0.110156
+INFO:root:Epoch[2] Batch [3000]	Speed: 663.02 samples/sec	acc=0.112598
+INFO:root:Epoch[2] Batch [3020]	Speed: 663.09 samples/sec	acc=0.109082
+INFO:root:Epoch[2] Batch [3040]	Speed: 663.72 samples/sec	acc=0.113867
+INFO:root:Epoch[2] Batch [3060]	Speed: 663.60 samples/sec	acc=0.114160
+INFO:root:Epoch[2] Batch [3080]	Speed: 664.46 samples/sec	acc=0.107227
+INFO:root:Epoch[2] Batch [3100]	Speed: 664.43 samples/sec	acc=0.110254
+INFO:root:Epoch[2] Batch [3120]	Speed: 664.16 samples/sec	acc=0.117578
+INFO:root:Epoch[2] Batch [3140]	Speed: 663.31 samples/sec	acc=0.111914
+INFO:root:Epoch[2] Batch [3160]	Speed: 663.83 samples/sec	acc=0.110937
+INFO:root:Epoch[2] Batch [3180]	Speed: 664.06 samples/sec	acc=0.108691
+INFO:root:Epoch[2] Batch [3200]	Speed: 664.65 samples/sec	acc=0.107617
+INFO:root:Epoch[2] Batch [3220]	Speed: 664.69 samples/sec	acc=0.112500
+INFO:root:Epoch[2] Batch [3240]	Speed: 664.71 samples/sec	acc=0.110352
+INFO:root:Epoch[2] Batch [3260]	Speed: 664.45 samples/sec	acc=0.110645
+INFO:root:Epoch[2] Batch [3280]	Speed: 665.02 samples/sec	acc=0.104785
+INFO:root:Epoch[2] Batch [3300]	Speed: 664.39 samples/sec	acc=0.112793
+INFO:root:Epoch[2] Batch [3320]	Speed: 663.92 samples/sec	acc=0.107227
+INFO:root:Epoch[2] Batch [3340]	Speed: 664.14 samples/sec	acc=0.115137
+INFO:root:Epoch[2] Batch [3360]	Speed: 668.48 samples/sec	acc=0.111621
+INFO:root:Epoch[2] Batch [3380]	Speed: 683.07 samples/sec	acc=0.108789
+INFO:root:Epoch[2] Batch [3400]	Speed: 675.24 samples/sec	acc=0.112305
+INFO:root:Epoch[2] Batch [3420]	Speed: 664.20 samples/sec	acc=0.109375
+INFO:root:Epoch[2] Batch [3440]	Speed: 664.32 samples/sec	acc=0.111914
+INFO:root:Epoch[2] Batch [3460]	Speed: 663.89 samples/sec	acc=0.114844
+INFO:root:Epoch[2] Batch [3480]	Speed: 664.25 samples/sec	acc=0.111914
+INFO:root:Epoch[2] Batch [3500]	Speed: 663.71 samples/sec	acc=0.114355
+INFO:root:Epoch[2] Batch [3520]	Speed: 664.43 samples/sec	acc=0.111816
+INFO:root:Epoch[2] Batch [3540]	Speed: 663.37 samples/sec	acc=0.112207
+INFO:root:Epoch[2] Batch [3560]	Speed: 664.55 samples/sec	acc=0.111816
+INFO:root:Epoch[2] Batch [3580]	Speed: 664.70 samples/sec	acc=0.108008
+INFO:root:Epoch[2] Batch [3600]	Speed: 664.80 samples/sec	acc=0.110742
+INFO:root:Epoch[2] Batch [3620]	Speed: 664.22 samples/sec	acc=0.109570
+INFO:root:Epoch[2] Batch [3640]	Speed: 663.62 samples/sec	acc=0.111719
+INFO:root:Epoch[2] Batch [3660]	Speed: 664.16 samples/sec	acc=0.114648
+INFO:root:Epoch[2] Batch [3680]	Speed: 664.17 samples/sec	acc=0.111328
+INFO:root:Epoch[2] Batch [3700]	Speed: 664.86 samples/sec	acc=0.112793
+INFO:root:Epoch[2] Batch [3720]	Speed: 664.13 samples/sec	acc=0.107227
+INFO:root:Epoch[2] Batch [3740]	Speed: 664.20 samples/sec	acc=0.116797
+INFO:root:Epoch[2] Batch [3760]	Speed: 664.91 samples/sec	acc=0.115723
+INFO:root:Epoch[2] Batch [3780]	Speed: 663.87 samples/sec	acc=0.117773
+INFO:root:Epoch[2] Batch [3800]	Speed: 664.63 samples/sec	acc=0.108398
+INFO:root:Epoch[2] Batch [3820]	Speed: 664.50 samples/sec	acc=0.114746
+INFO:root:Epoch[2] Batch [3840]	Speed: 664.13 samples/sec	acc=0.106348
+INFO:root:Epoch[2] Batch [3860]	Speed: 663.95 samples/sec	acc=0.119531
+INFO:root:Epoch[2] Batch [3880]	Speed: 664.61 samples/sec	acc=0.110840
+INFO:root:Epoch[2] Batch [3900]	Speed: 664.16 samples/sec	acc=0.118555
+INFO:root:Epoch[2] Batch [3920]	Speed: 664.38 samples/sec	acc=0.111523
+INFO:root:Epoch[2] Batch [3940]	Speed: 664.28 samples/sec	acc=0.112305
+INFO:root:Epoch[2] Batch [3960]	Speed: 664.88 samples/sec	acc=0.116016
+lr-batch-epoch: 0.1 3961 2
+INFO:root:Epoch[2] Batch [3980]	Speed: 664.94 samples/sec	acc=0.113672
+INFO:root:Epoch[2] Batch [4000]	Speed: 664.07 samples/sec	acc=0.108008
+INFO:root:Epoch[2] Batch [4020]	Speed: 664.84 samples/sec	acc=0.117090
+INFO:root:Epoch[2] Batch [4040]	Speed: 664.60 samples/sec	acc=0.113867
+INFO:root:Epoch[2] Batch [4060]	Speed: 664.30 samples/sec	acc=0.116309
+INFO:root:Epoch[2] Batch [4080]	Speed: 663.47 samples/sec	acc=0.118262
+INFO:root:Epoch[2] Batch [4100]	Speed: 664.70 samples/sec	acc=0.114355
+INFO:root:Epoch[2] Batch [4120]	Speed: 663.99 samples/sec	acc=0.112109
+INFO:root:Epoch[2] Batch [4140]	Speed: 669.73 samples/sec	acc=0.109082
+INFO:root:Epoch[2] Batch [4160]	Speed: 680.88 samples/sec	acc=0.111914
+INFO:root:Epoch[2] Batch [4180]	Speed: 669.61 samples/sec	acc=0.110547
+INFO:root:Epoch[2] Batch [4200]	Speed: 665.34 samples/sec	acc=0.113477
+INFO:root:Epoch[2] Batch [4220]	Speed: 664.35 samples/sec	acc=0.117285
+INFO:root:Epoch[2] Batch [4240]	Speed: 664.92 samples/sec	acc=0.114062
+INFO:root:Epoch[2] Batch [4260]	Speed: 664.83 samples/sec	acc=0.113867
+INFO:root:Epoch[2] Batch [4280]	Speed: 663.98 samples/sec	acc=0.115430
+INFO:root:Epoch[2] Batch [4300]	Speed: 664.89 samples/sec	acc=0.114941
+INFO:root:Epoch[2] Batch [4320]	Speed: 664.82 samples/sec	acc=0.112207
+INFO:root:Epoch[2] Batch [4340]	Speed: 664.90 samples/sec	acc=0.114258
+INFO:root:Epoch[2] Batch [4360]	Speed: 664.17 samples/sec	acc=0.110840
+INFO:root:Epoch[2] Batch [4380]	Speed: 664.28 samples/sec	acc=0.118359
+INFO:root:Epoch[2] Batch [4400]	Speed: 663.80 samples/sec	acc=0.115723
+INFO:root:Epoch[2] Batch [4420]	Speed: 664.33 samples/sec	acc=0.114355
+INFO:root:Epoch[2] Batch [4440]	Speed: 665.21 samples/sec	acc=0.111719
+INFO:root:Epoch[2] Batch [4460]	Speed: 664.03 samples/sec	acc=0.116504
+INFO:root:Epoch[2] Batch [4480]	Speed: 664.72 samples/sec	acc=0.111133
+INFO:root:Epoch[2] Batch [4500]	Speed: 663.72 samples/sec	acc=0.119629
+INFO:root:Epoch[2] Batch [4520]	Speed: 664.06 samples/sec	acc=0.108594
+INFO:root:Epoch[2] Batch [4540]	Speed: 664.94 samples/sec	acc=0.111914
+INFO:root:Epoch[2] Batch [4560]	Speed: 663.78 samples/sec	acc=0.110742
+INFO:root:Epoch[2] Batch [4580]	Speed: 665.08 samples/sec	acc=0.117188
+INFO:root:Epoch[2] Batch [4600]	Speed: 664.37 samples/sec	acc=0.110937
+INFO:root:Epoch[2] Batch [4620]	Speed: 664.12 samples/sec	acc=0.110840
+INFO:root:Epoch[2] Batch [4640]	Speed: 664.57 samples/sec	acc=0.117383
+INFO:root:Epoch[2] Batch [4660]	Speed: 663.86 samples/sec	acc=0.113672
+INFO:root:Epoch[2] Batch [4680]	Speed: 664.00 samples/sec	acc=0.116113
+INFO:root:Epoch[2] Batch [4700]	Speed: 664.21 samples/sec	acc=0.121289
+INFO:root:Epoch[2] Batch [4720]	Speed: 664.58 samples/sec	acc=0.114355
+INFO:root:Epoch[2] Batch [4740]	Speed: 664.36 samples/sec	acc=0.114160
+INFO:root:Epoch[2] Batch [4760]	Speed: 664.57 samples/sec	acc=0.117090
+INFO:root:Epoch[2] Batch [4780]	Speed: 664.18 samples/sec	acc=0.116113
+INFO:root:Epoch[2] Batch [4800]	Speed: 664.03 samples/sec	acc=0.118652
+INFO:root:Epoch[2] Batch [4820]	Speed: 664.77 samples/sec	acc=0.112305
+INFO:root:Epoch[2] Batch [4840]	Speed: 663.82 samples/sec	acc=0.121484
+INFO:root:Epoch[2] Batch [4860]	Speed: 664.63 samples/sec	acc=0.115918
+INFO:root:Epoch[2] Batch [4880]	Speed: 664.45 samples/sec	acc=0.116016
+INFO:root:Epoch[2] Batch [4900]	Speed: 664.80 samples/sec	acc=0.120020
+INFO:root:Epoch[2] Batch [4920]	Speed: 663.74 samples/sec	acc=0.119531
+INFO:root:Epoch[2] Batch [4940]	Speed: 669.45 samples/sec	acc=0.110937
+INFO:root:Epoch[2] Batch [4960]	Speed: 665.32 samples/sec	acc=0.109863
+[20000][MARGIN]0.663238,0.230801
+lr-batch-epoch: 0.1 4961 2
+testing verification..
+(14000, 512)
+infer time 21.918577
+[cfp_ff][20000]XNorm: 20.374364
+[cfp_ff][20000]Accuracy-Flip: 0.99314+-0.00325
+testing verification..
+(14000, 512)
+infer time 19.171411
+[cfp_fp][20000]XNorm: 16.566911
+[cfp_fp][20000]Accuracy-Flip: 0.81929+-0.01688
+testing verification..
+(12000, 512)
+infer time 18.896416
+[agedb_30][20000]XNorm: 19.540734
+[agedb_30][20000]Accuracy-Flip: 0.94333+-0.01216
+testing verification..
+(12000, 512)
+infer time 20.152508
+[lfw][20000]XNorm: 20.211320
+[lfw][20000]Accuracy-Flip: 0.99067+-0.00410
+[20000]Accuracy-Highest: 0.99167
+INFO:root:Epoch[2] Batch [4980]	Speed: 91.09 samples/sec	acc=0.118750
+INFO:root:Epoch[2] Batch [5000]	Speed: 672.88 samples/sec	acc=0.119922
+INFO:root:Epoch[2] Batch [5020]	Speed: 663.05 samples/sec	acc=0.116016
+INFO:root:Epoch[2] Batch [5040]	Speed: 663.79 samples/sec	acc=0.118066
+INFO:root:Epoch[2] Batch [5060]	Speed: 663.57 samples/sec	acc=0.115039
+INFO:root:Epoch[2] Batch [5080]	Speed: 663.88 samples/sec	acc=0.112305
+INFO:root:Epoch[2] Batch [5100]	Speed: 664.05 samples/sec	acc=0.119336
+INFO:root:Epoch[2] Batch [5120]	Speed: 664.60 samples/sec	acc=0.118652
+INFO:root:Epoch[2] Batch [5140]	Speed: 663.68 samples/sec	acc=0.117676
+INFO:root:Epoch[2] Batch [5160]	Speed: 665.14 samples/sec	acc=0.116406
+INFO:root:Epoch[2] Batch [5180]	Speed: 663.46 samples/sec	acc=0.110449
+INFO:root:Epoch[2] Batch [5200]	Speed: 663.83 samples/sec	acc=0.112109
+INFO:root:Epoch[2] Batch [5220]	Speed: 664.03 samples/sec	acc=0.116602
+INFO:root:Epoch[2] Batch [5240]	Speed: 664.81 samples/sec	acc=0.116309
+INFO:root:Epoch[2] Batch [5260]	Speed: 663.35 samples/sec	acc=0.117090
+INFO:root:Epoch[2] Batch [5280]	Speed: 664.49 samples/sec	acc=0.115918
+INFO:root:Epoch[2] Batch [5300]	Speed: 663.17 samples/sec	acc=0.119531
+INFO:root:Epoch[2] Batch [5320]	Speed: 664.53 samples/sec	acc=0.114844
+INFO:root:Epoch[2] Batch [5340]	Speed: 662.99 samples/sec	acc=0.116699
+INFO:root:Epoch[2] Batch [5360]	Speed: 663.62 samples/sec	acc=0.114258
+INFO:root:Epoch[2] Batch [5380]	Speed: 663.86 samples/sec	acc=0.111133
+INFO:root:Epoch[2] Batch [5400]	Speed: 664.32 samples/sec	acc=0.115430
+INFO:root:Epoch[2] Batch [5420]	Speed: 664.78 samples/sec	acc=0.111816
+INFO:root:Epoch[2] Batch [5440]	Speed: 664.39 samples/sec	acc=0.111230
+INFO:root:Epoch[2] Batch [5460]	Speed: 664.17 samples/sec	acc=0.117676
+INFO:root:Epoch[2] Batch [5480]	Speed: 664.50 samples/sec	acc=0.115625
+INFO:root:Epoch[2] Batch [5500]	Speed: 664.36 samples/sec	acc=0.118066
+INFO:root:Epoch[2] Batch [5520]	Speed: 664.05 samples/sec	acc=0.120996
+INFO:root:Epoch[2] Batch [5540]	Speed: 664.47 samples/sec	acc=0.118750
+INFO:root:Epoch[2] Batch [5560]	Speed: 664.39 samples/sec	acc=0.123047
+INFO:root:Epoch[2] Batch [5580]	Speed: 663.95 samples/sec	acc=0.117090
+INFO:root:Epoch[2] Batch [5600]	Speed: 664.22 samples/sec	acc=0.121387
+INFO:root:Epoch[2] Batch [5620]	Speed: 664.51 samples/sec	acc=0.119824
+INFO:root:Epoch[2] Batch [5640]	Speed: 664.47 samples/sec	acc=0.118164
+INFO:root:Epoch[2] Batch [5660]	Speed: 666.73 samples/sec	acc=0.115137
+INFO:root:Epoch[2] Batch [5680]	Speed: 664.81 samples/sec	acc=0.116895
+INFO:root:Epoch[2] Batch [5700]	Speed: 663.82 samples/sec	acc=0.122656
+INFO:root:Epoch[2] Batch [5720]	Speed: 664.55 samples/sec	acc=0.114062
+INFO:root:Epoch[2] Batch [5740]	Speed: 664.58 samples/sec	acc=0.114453
+INFO:root:Epoch[2] Batch [5760]	Speed: 664.23 samples/sec	acc=0.118164
+INFO:root:Epoch[2] Batch [5780]	Speed: 664.38 samples/sec	acc=0.116699
+INFO:root:Epoch[2] Batch [5800]	Speed: 664.22 samples/sec	acc=0.117676
+INFO:root:Epoch[2] Batch [5820]	Speed: 663.73 samples/sec	acc=0.119336
+INFO:root:Epoch[2] Batch [5840]	Speed: 664.61 samples/sec	acc=0.116602
+INFO:root:Epoch[2] Batch [5860]	Speed: 664.85 samples/sec	acc=0.114941
+INFO:root:Epoch[2] Batch [5880]	Speed: 664.40 samples/sec	acc=0.118164
+INFO:root:Epoch[2] Batch [5900]	Speed: 663.76 samples/sec	acc=0.119531
+INFO:root:Epoch[2] Batch [5920]	Speed: 664.06 samples/sec	acc=0.119043
+INFO:root:Epoch[2] Batch [5940]	Speed: 663.59 samples/sec	acc=0.121387
+INFO:root:Epoch[2] Batch [5960]	Speed: 664.12 samples/sec	acc=0.112793
+lr-batch-epoch: 0.1 5961 2
+INFO:root:Epoch[2] Batch [5980]	Speed: 664.73 samples/sec	acc=0.115039
+INFO:root:Epoch[2] Batch [6000]	Speed: 663.84 samples/sec	acc=0.111133
+INFO:root:Epoch[2] Batch [6020]	Speed: 664.09 samples/sec	acc=0.118555
+INFO:root:Epoch[2] Batch [6040]	Speed: 664.42 samples/sec	acc=0.117090
+INFO:root:Epoch[2] Batch [6060]	Speed: 664.06 samples/sec	acc=0.117480
+INFO:root:Epoch[2] Batch [6080]	Speed: 664.31 samples/sec	acc=0.119531
+INFO:root:Epoch[2] Batch [6100]	Speed: 663.97 samples/sec	acc=0.119434
+INFO:root:Epoch[2] Batch [6120]	Speed: 663.05 samples/sec	acc=0.123633
+INFO:root:Epoch[2] Batch [6140]	Speed: 664.41 samples/sec	acc=0.117871
+INFO:root:Epoch[2] Batch [6160]	Speed: 663.85 samples/sec	acc=0.110449
+INFO:root:Epoch[2] Batch [6180]	Speed: 663.81 samples/sec	acc=0.113867
+INFO:root:Epoch[2] Batch [6200]	Speed: 664.09 samples/sec	acc=0.114941
+INFO:root:Epoch[2] Batch [6220]	Speed: 664.49 samples/sec	acc=0.119238
+INFO:root:Epoch[2] Batch [6240]	Speed: 664.11 samples/sec	acc=0.115820
+INFO:root:Epoch[2] Batch [6260]	Speed: 663.29 samples/sec	acc=0.115234
+INFO:root:Epoch[2] Batch [6280]	Speed: 664.97 samples/sec	acc=0.114355
+INFO:root:Epoch[2] Batch [6300]	Speed: 663.77 samples/sec	acc=0.117188
+INFO:root:Epoch[2] Batch [6320]	Speed: 663.55 samples/sec	acc=0.123535
+INFO:root:Epoch[2] Batch [6340]	Speed: 664.37 samples/sec	acc=0.122070
+INFO:root:Epoch[2] Batch [6360]	Speed: 664.77 samples/sec	acc=0.125977
+INFO:root:Epoch[2] Batch [6380]	Speed: 663.94 samples/sec	acc=0.114746
+INFO:root:Epoch[2] Batch [6400]	Speed: 664.46 samples/sec	acc=0.124316
+INFO:root:Epoch[2] Batch [6420]	Speed: 663.56 samples/sec	acc=0.118262
+INFO:root:Epoch[2] Batch [6440]	Speed: 669.30 samples/sec	acc=0.118164
+INFO:root:Epoch[2] Batch [6460]	Speed: 681.44 samples/sec	acc=0.121094
+INFO:root:Epoch[2] Batch [6480]	Speed: 680.55 samples/sec	acc=0.124023
+INFO:root:Epoch[2] Batch [6500]	Speed: 680.84 samples/sec	acc=0.119531
+INFO:root:Epoch[2] Batch [6520]	Speed: 680.07 samples/sec	acc=0.124219
+INFO:root:Epoch[2] Batch [6540]	Speed: 681.65 samples/sec	acc=0.122754
+INFO:root:Epoch[2] Batch [6560]	Speed: 678.54 samples/sec	acc=0.115527
+INFO:root:Epoch[2] Batch [6580]	Speed: 663.81 samples/sec	acc=0.116895
+INFO:root:Epoch[2] Batch [6600]	Speed: 664.42 samples/sec	acc=0.114844
+INFO:root:Epoch[2] Batch [6620]	Speed: 663.95 samples/sec	acc=0.119141
+INFO:root:Epoch[2] Batch [6640]	Speed: 662.97 samples/sec	acc=0.119824
+INFO:root:Epoch[2] Batch [6660]	Speed: 664.26 samples/sec	acc=0.119922
+INFO:root:Epoch[2] Batch [6680]	Speed: 664.31 samples/sec	acc=0.117285
+INFO:root:Epoch[2] Batch [6700]	Speed: 663.47 samples/sec	acc=0.114355
+INFO:root:Epoch[2] Batch [6720]	Speed: 664.60 samples/sec	acc=0.123145
+INFO:root:Epoch[2] Batch [6740]	Speed: 663.39 samples/sec	acc=0.121484
+INFO:root:Epoch[2] Batch [6760]	Speed: 663.38 samples/sec	acc=0.117969
+INFO:root:Epoch[2] Batch [6780]	Speed: 664.36 samples/sec	acc=0.125000
+INFO:root:Epoch[2] Batch [6800]	Speed: 664.32 samples/sec	acc=0.116992
+INFO:root:Epoch[2] Batch [6820]	Speed: 664.38 samples/sec	acc=0.116211
+INFO:root:Epoch[2] Batch [6840]	Speed: 663.61 samples/sec	acc=0.114062
+INFO:root:Epoch[2] Batch [6860]	Speed: 664.10 samples/sec	acc=0.119629
+INFO:root:Epoch[2] Batch [6880]	Speed: 664.32 samples/sec	acc=0.121582
+INFO:root:Epoch[2] Batch [6900]	Speed: 664.65 samples/sec	acc=0.118359
+INFO:root:Epoch[2] Batch [6920]	Speed: 663.93 samples/sec	acc=0.115234
+INFO:root:Epoch[2] Batch [6940]	Speed: 663.81 samples/sec	acc=0.119434
+INFO:root:Epoch[2] Batch [6960]	Speed: 663.18 samples/sec	acc=0.113574
+[22000][MARGIN]0.657348,0.223681
+lr-batch-epoch: 0.1 6961 2
+testing verification..
+(14000, 512)
+infer time 21.128909
+[cfp_ff][22000]XNorm: 21.352920
+[cfp_ff][22000]Accuracy-Flip: 0.99257+-0.00167
+testing verification..
+(14000, 512)
+infer time 18.872781
+[cfp_fp][22000]XNorm: 17.430041
+[cfp_fp][22000]Accuracy-Flip: 0.82871+-0.01249
+testing verification..
+(12000, 512)
+infer time 21.459833
+[agedb_30][22000]XNorm: 20.922552
+[agedb_30][22000]Accuracy-Flip: 0.94383+-0.00913
+testing verification..
+(12000, 512)
+infer time 17.803807
+[lfw][22000]XNorm: 21.271461
+[lfw][22000]Accuracy-Flip: 0.99250+-0.00344
+saving 11
+INFO:root:Saved checkpoint to "../model-r50-am-lfw/model-0011.params"
+[22000]Accuracy-Highest: 0.99250
+INFO:root:Epoch[2] Batch [6980]	Speed: 88.76 samples/sec	acc=0.117480
+INFO:root:Epoch[2] Batch [7000]	Speed: 663.49 samples/sec	acc=0.120117
+INFO:root:Epoch[2] Batch [7020]	Speed: 662.53 samples/sec	acc=0.117871
+INFO:root:Epoch[2] Batch [7040]	Speed: 663.60 samples/sec	acc=0.118457
+INFO:root:Epoch[2] Batch [7060]	Speed: 663.42 samples/sec	acc=0.116602
+INFO:root:Epoch[2] Batch [7080]	Speed: 664.05 samples/sec	acc=0.119922
+INFO:root:Epoch[2] Batch [7100]	Speed: 664.54 samples/sec	acc=0.122461
+INFO:root:Epoch[2] Batch [7120]	Speed: 664.20 samples/sec	acc=0.123145
+INFO:root:Epoch[2] Batch [7140]	Speed: 665.31 samples/sec	acc=0.123340
+INFO:root:Epoch[2] Batch [7160]	Speed: 664.74 samples/sec	acc=0.124609
+INFO:root:Epoch[2] Batch [7180]	Speed: 664.66 samples/sec	acc=0.120020
+INFO:root:Epoch[2] Batch [7200]	Speed: 664.34 samples/sec	acc=0.127148
+INFO:root:Epoch[2] Batch [7220]	Speed: 671.26 samples/sec	acc=0.123633
+INFO:root:Epoch[2] Batch [7240]	Speed: 664.98 samples/sec	acc=0.129297
+INFO:root:Epoch[2] Batch [7260]	Speed: 663.81 samples/sec	acc=0.118750
+INFO:root:Epoch[2] Batch [7280]	Speed: 664.59 samples/sec	acc=0.120605
+INFO:root:Epoch[2] Batch [7300]	Speed: 664.42 samples/sec	acc=0.125098
+INFO:root:Epoch[2] Batch [7320]	Speed: 664.16 samples/sec	acc=0.124414
+INFO:root:Epoch[2] Batch [7340]	Speed: 664.47 samples/sec	acc=0.117969
+INFO:root:Epoch[2] Batch [7360]	Speed: 664.39 samples/sec	acc=0.121875
+INFO:root:Epoch[2] Batch [7380]	Speed: 663.71 samples/sec	acc=0.122949
+INFO:root:Epoch[2] Batch [7400]	Speed: 663.93 samples/sec	acc=0.124219
+INFO:root:Epoch[2] Batch [7420]	Speed: 664.72 samples/sec	acc=0.121289
+INFO:root:Epoch[2] Batch [7440]	Speed: 664.18 samples/sec	acc=0.113965
+INFO:root:Epoch[2] Batch [7460]	Speed: 663.63 samples/sec	acc=0.124316
+INFO:root:Epoch[2] Batch [7480]	Speed: 664.10 samples/sec	acc=0.120898
+INFO:root:Epoch[2] Batch [7500]	Speed: 664.24 samples/sec	acc=0.120410
+INFO:root:Epoch[2] Train-acc=0.120551
+INFO:root:Epoch[2] Time cost=6193.579
+call reset()
+INFO:root:Epoch[3] Batch [20]	Speed: 676.59 samples/sec	acc=0.149833
+INFO:root:Epoch[3] Batch [40]	Speed: 672.43 samples/sec	acc=0.142578
+INFO:root:Epoch[3] Batch [60]	Speed: 663.42 samples/sec	acc=0.137402
+INFO:root:Epoch[3] Batch [80]	Speed: 664.03 samples/sec	acc=0.135645
+INFO:root:Epoch[3] Batch [100]	Speed: 664.10 samples/sec	acc=0.132715
+INFO:root:Epoch[3] Batch [120]	Speed: 664.15 samples/sec	acc=0.133203
+INFO:root:Epoch[3] Batch [140]	Speed: 663.85 samples/sec	acc=0.134473
+INFO:root:Epoch[3] Batch [160]	Speed: 664.61 samples/sec	acc=0.131152
+INFO:root:Epoch[3] Batch [180]	Speed: 664.18 samples/sec	acc=0.137793
+INFO:root:Epoch[3] Batch [200]	Speed: 664.27 samples/sec	acc=0.135742
+INFO:root:Epoch[3] Batch [220]	Speed: 665.12 samples/sec	acc=0.131836
+INFO:root:Epoch[3] Batch [240]	Speed: 663.97 samples/sec	acc=0.139746
+INFO:root:Epoch[3] Batch [260]	Speed: 664.24 samples/sec	acc=0.136230
+INFO:root:Epoch[3] Batch [280]	Speed: 665.08 samples/sec	acc=0.133691
+INFO:root:Epoch[3] Batch [300]	Speed: 664.23 samples/sec	acc=0.137012
+INFO:root:Epoch[3] Batch [320]	Speed: 663.24 samples/sec	acc=0.131836
+INFO:root:Epoch[3] Batch [340]	Speed: 665.29 samples/sec	acc=0.129395
+INFO:root:Epoch[3] Batch [360]	Speed: 664.53 samples/sec	acc=0.133594
+INFO:root:Epoch[3] Batch [380]	Speed: 664.21 samples/sec	acc=0.127637
+INFO:root:Epoch[3] Batch [400]	Speed: 663.98 samples/sec	acc=0.125879
+INFO:root:Epoch[3] Batch [420]	Speed: 664.86 samples/sec	acc=0.128906
+INFO:root:Epoch[3] Batch [440]	Speed: 663.66 samples/sec	acc=0.132910
+lr-batch-epoch: 0.1 442 3
+INFO:root:Epoch[3] Batch [460]	Speed: 673.99 samples/sec	acc=0.125000
+INFO:root:Epoch[3] Batch [480]	Speed: 663.55 samples/sec	acc=0.136426
+INFO:root:Epoch[3] Batch [500]	Speed: 664.04 samples/sec	acc=0.123145
+INFO:root:Epoch[3] Batch [520]	Speed: 665.16 samples/sec	acc=0.129004
+INFO:root:Epoch[3] Batch [540]	Speed: 663.61 samples/sec	acc=0.119531
+INFO:root:Epoch[3] Batch [560]	Speed: 664.36 samples/sec	acc=0.123340
+INFO:root:Epoch[3] Batch [580]	Speed: 664.27 samples/sec	acc=0.124707
+INFO:root:Epoch[3] Batch [600]	Speed: 663.76 samples/sec	acc=0.127832
+INFO:root:Epoch[3] Batch [620]	Speed: 664.59 samples/sec	acc=0.125098
+INFO:root:Epoch[3] Batch [640]	Speed: 664.99 samples/sec	acc=0.118945
+INFO:root:Epoch[3] Batch [660]	Speed: 664.18 samples/sec	acc=0.132129
+INFO:root:Epoch[3] Batch [680]	Speed: 664.41 samples/sec	acc=0.120996
+INFO:root:Epoch[3] Batch [700]	Speed: 665.25 samples/sec	acc=0.120996
+INFO:root:Epoch[3] Batch [720]	Speed: 664.00 samples/sec	acc=0.125391
+INFO:root:Epoch[3] Batch [740]	Speed: 664.30 samples/sec	acc=0.121289
+INFO:root:Epoch[3] Batch [760]	Speed: 664.60 samples/sec	acc=0.124023
+INFO:root:Epoch[3] Batch [780]	Speed: 664.30 samples/sec	acc=0.123145
+INFO:root:Epoch[3] Batch [800]	Speed: 664.70 samples/sec	acc=0.122168
+INFO:root:Epoch[3] Batch [820]	Speed: 663.92 samples/sec	acc=0.122949
+INFO:root:Epoch[3] Batch [840]	Speed: 664.20 samples/sec	acc=0.126172
+INFO:root:Epoch[3] Batch [860]	Speed: 664.74 samples/sec	acc=0.127051
+INFO:root:Epoch[3] Batch [880]	Speed: 665.04 samples/sec	acc=0.123047
+INFO:root:Epoch[3] Batch [900]	Speed: 664.94 samples/sec	acc=0.125488
+INFO:root:Epoch[3] Batch [920]	Speed: 662.39 samples/sec	acc=0.130371
+INFO:root:Epoch[3] Batch [940]	Speed: 665.08 samples/sec	acc=0.122266
+INFO:root:Epoch[3] Batch [960]	Speed: 664.76 samples/sec	acc=0.116992
+INFO:root:Epoch[3] Batch [980]	Speed: 664.07 samples/sec	acc=0.122656
+INFO:root:Epoch[3] Batch [1000]	Speed: 664.29 samples/sec	acc=0.124902
+INFO:root:Epoch[3] Batch [1020]	Speed: 665.25 samples/sec	acc=0.124121
+INFO:root:Epoch[3] Batch [1040]	Speed: 663.85 samples/sec	acc=0.123926
+INFO:root:Epoch[3] Batch [1060]	Speed: 664.97 samples/sec	acc=0.123145
+INFO:root:Epoch[3] Batch [1080]	Speed: 663.92 samples/sec	acc=0.121582
+INFO:root:Epoch[3] Batch [1100]	Speed: 663.69 samples/sec	acc=0.127344
+INFO:root:Epoch[3] Batch [1120]	Speed: 665.25 samples/sec	acc=0.124609
+INFO:root:Epoch[3] Batch [1140]	Speed: 664.36 samples/sec	acc=0.121680
+INFO:root:Epoch[3] Batch [1160]	Speed: 662.74 samples/sec	acc=0.122559
+INFO:root:Epoch[3] Batch [1180]	Speed: 664.77 samples/sec	acc=0.122461
+INFO:root:Epoch[3] Batch [1200]	Speed: 665.04 samples/sec	acc=0.123633
+INFO:root:Epoch[3] Batch [1220]	Speed: 664.49 samples/sec	acc=0.123535
+INFO:root:Epoch[3] Batch [1240]	Speed: 661.68 samples/sec	acc=0.125488
+INFO:root:Epoch[3] Batch [1260]	Speed: 670.20 samples/sec	acc=0.121484
+INFO:root:Epoch[3] Batch [1280]	Speed: 664.20 samples/sec	acc=0.119141
+INFO:root:Epoch[3] Batch [1300]	Speed: 664.92 samples/sec	acc=0.124414
+INFO:root:Epoch[3] Batch [1320]	Speed: 664.64 samples/sec	acc=0.119629
+INFO:root:Epoch[3] Batch [1340]	Speed: 663.19 samples/sec	acc=0.122266
+INFO:root:Epoch[3] Batch [1360]	Speed: 665.31 samples/sec	acc=0.119629
+INFO:root:Epoch[3] Batch [1380]	Speed: 664.81 samples/sec	acc=0.119824
+INFO:root:Epoch[3] Batch [1400]	Speed: 664.02 samples/sec	acc=0.125781
+INFO:root:Epoch[3] Batch [1420]	Speed: 664.94 samples/sec	acc=0.121680
+INFO:root:Epoch[3] Batch [1440]	Speed: 664.45 samples/sec	acc=0.124805
+[24000][MARGIN]0.676005,0.245434
+lr-batch-epoch: 0.1 1442 3
+testing verification..
+(14000, 512)
+infer time 18.967896
+[cfp_ff][24000]XNorm: 21.066906
+[cfp_ff][24000]Accuracy-Flip: 0.99229+-0.00475
+testing verification..
+(14000, 512)
+infer time 20.284738
+[cfp_fp][24000]XNorm: 17.000686
+[cfp_fp][24000]Accuracy-Flip: 0.83029+-0.01477
+testing verification..
+(12000, 512)
+infer time 18.408323
+[agedb_30][24000]XNorm: 20.699430
+[agedb_30][24000]Accuracy-Flip: 0.94183+-0.00911
+testing verification..
+(12000, 512)
+infer time 17.257163
+[lfw][24000]XNorm: 20.954269
+[lfw][24000]Accuracy-Flip: 0.99233+-0.00442
+[24000]Accuracy-Highest: 0.99250
+INFO:root:Epoch[3] Batch [1460]	Speed: 95.17 samples/sec	acc=0.122559
+INFO:root:Epoch[3] Batch [1480]	Speed: 684.66 samples/sec	acc=0.128809
+INFO:root:Epoch[3] Batch [1500]	Speed: 664.17 samples/sec	acc=0.120117
+INFO:root:Epoch[3] Batch [1520]	Speed: 663.18 samples/sec	acc=0.122461
+INFO:root:Epoch[3] Batch [1540]	Speed: 663.68 samples/sec	acc=0.128516
+INFO:root:Epoch[3] Batch [1560]	Speed: 664.18 samples/sec	acc=0.121289
+INFO:root:Epoch[3] Batch [1580]	Speed: 663.50 samples/sec	acc=0.123730
+INFO:root:Epoch[3] Batch [1600]	Speed: 664.59 samples/sec	acc=0.119727
+INFO:root:Epoch[3] Batch [1620]	Speed: 664.57 samples/sec	acc=0.120117
+INFO:root:Epoch[3] Batch [1640]	Speed: 664.05 samples/sec	acc=0.122168
+INFO:root:Epoch[3] Batch [1660]	Speed: 664.38 samples/sec	acc=0.124512
+INFO:root:Epoch[3] Batch [1680]	Speed: 663.93 samples/sec	acc=0.126660
+INFO:root:Epoch[3] Batch [1700]	Speed: 664.04 samples/sec	acc=0.128516
+INFO:root:Epoch[3] Batch [1720]	Speed: 664.56 samples/sec	acc=0.120215
+INFO:root:Epoch[3] Batch [1740]	Speed: 663.86 samples/sec	acc=0.120898
+INFO:root:Epoch[3] Batch [1760]	Speed: 662.76 samples/sec	acc=0.123535
+INFO:root:Epoch[3] Batch [1780]	Speed: 664.90 samples/sec	acc=0.126270
+INFO:root:Epoch[3] Batch [1800]	Speed: 663.03 samples/sec	acc=0.121484
+INFO:root:Epoch[3] Batch [1820]	Speed: 663.67 samples/sec	acc=0.122266
+INFO:root:Epoch[3] Batch [1840]	Speed: 664.01 samples/sec	acc=0.124805
+INFO:root:Epoch[3] Batch [1860]	Speed: 664.74 samples/sec	acc=0.122559
+INFO:root:Epoch[3] Batch [1880]	Speed: 662.70 samples/sec	acc=0.121289
+INFO:root:Epoch[3] Batch [1900]	Speed: 664.50 samples/sec	acc=0.128320
+INFO:root:Epoch[3] Batch [1920]	Speed: 664.27 samples/sec	acc=0.125977
+INFO:root:Epoch[3] Batch [1940]	Speed: 663.74 samples/sec	acc=0.124902
+INFO:root:Epoch[3] Batch [1960]	Speed: 664.55 samples/sec	acc=0.122754
+INFO:root:Epoch[3] Batch [1980]	Speed: 663.80 samples/sec	acc=0.126270
+INFO:root:Epoch[3] Batch [2000]	Speed: 664.25 samples/sec	acc=0.124023
+INFO:root:Epoch[3] Batch [2020]	Speed: 664.63 samples/sec	acc=0.129199
+INFO:root:Epoch[3] Batch [2040]	Speed: 663.88 samples/sec	acc=0.126270
+INFO:root:Epoch[3] Batch [2060]	Speed: 664.78 samples/sec	acc=0.119824
+INFO:root:Epoch[3] Batch [2080]	Speed: 663.88 samples/sec	acc=0.122070
+INFO:root:Epoch[3] Batch [2100]	Speed: 663.73 samples/sec	acc=0.118066
+INFO:root:Epoch[3] Batch [2120]	Speed: 664.68 samples/sec	acc=0.119336
+INFO:root:Epoch[3] Batch [2140]	Speed: 664.16 samples/sec	acc=0.121777
+INFO:root:Epoch[3] Batch [2160]	Speed: 663.81 samples/sec	acc=0.122461
+INFO:root:Epoch[3] Batch [2180]	Speed: 663.03 samples/sec	acc=0.123340
+INFO:root:Epoch[3] Batch [2200]	Speed: 664.74 samples/sec	acc=0.119043
+INFO:root:Epoch[3] Batch [2220]	Speed: 663.71 samples/sec	acc=0.124902
+INFO:root:Epoch[3] Batch [2240]	Speed: 663.56 samples/sec	acc=0.126953
+INFO:root:Epoch[3] Batch [2260]	Speed: 665.15 samples/sec	acc=0.121680
+INFO:root:Epoch[3] Batch [2280]	Speed: 664.00 samples/sec	acc=0.119434
+INFO:root:Epoch[3] Batch [2300]	Speed: 663.77 samples/sec	acc=0.128809
+INFO:root:Epoch[3] Batch [2320]	Speed: 664.50 samples/sec	acc=0.122461
+INFO:root:Epoch[3] Batch [2340]	Speed: 663.68 samples/sec	acc=0.120898
+INFO:root:Epoch[3] Batch [2360]	Speed: 663.62 samples/sec	acc=0.124219
+INFO:root:Epoch[3] Batch [2380]	Speed: 664.04 samples/sec	acc=0.123535
+INFO:root:Epoch[3] Batch [2400]	Speed: 664.30 samples/sec	acc=0.125195
+INFO:root:Epoch[3] Batch [2420]	Speed: 663.62 samples/sec	acc=0.117969
+INFO:root:Epoch[3] Batch [2440]	Speed: 663.52 samples/sec	acc=0.125488
+lr-batch-epoch: 0.1 2442 3
+INFO:root:Epoch[3] Batch [2460]	Speed: 663.67 samples/sec	acc=0.124512
+INFO:root:Epoch[3] Batch [2480]	Speed: 664.31 samples/sec	acc=0.126562
+INFO:root:Epoch[3] Batch [2500]	Speed: 664.13 samples/sec	acc=0.122754
+INFO:root:Epoch[3] Batch [2520]	Speed: 664.67 samples/sec	acc=0.128906
+INFO:root:Epoch[3] Batch [2540]	Speed: 662.74 samples/sec	acc=0.127637
+INFO:root:Epoch[3] Batch [2560]	Speed: 664.93 samples/sec	acc=0.121387
+INFO:root:Epoch[3] Batch [2580]	Speed: 664.35 samples/sec	acc=0.123340
+INFO:root:Epoch[3] Batch [2600]	Speed: 663.53 samples/sec	acc=0.122754
+INFO:root:Epoch[3] Batch [2620]	Speed: 664.64 samples/sec	acc=0.119922
+INFO:root:Epoch[3] Batch [2640]	Speed: 664.33 samples/sec	acc=0.122168
+INFO:root:Epoch[3] Batch [2660]	Speed: 662.44 samples/sec	acc=0.126660
+INFO:root:Epoch[3] Batch [2680]	Speed: 664.43 samples/sec	acc=0.121582
+INFO:root:Epoch[3] Batch [2700]	Speed: 664.71 samples/sec	acc=0.120996
+INFO:root:Epoch[3] Batch [2720]	Speed: 663.88 samples/sec	acc=0.124023
+INFO:root:Epoch[3] Batch [2740]	Speed: 664.39 samples/sec	acc=0.126172
+INFO:root:Epoch[3] Batch [2760]	Speed: 664.13 samples/sec	acc=0.126855
+INFO:root:Epoch[3] Batch [2780]	Speed: 663.57 samples/sec	acc=0.128125
+INFO:root:Epoch[3] Batch [2800]	Speed: 663.28 samples/sec	acc=0.126172
+INFO:root:Epoch[3] Batch [2820]	Speed: 663.83 samples/sec	acc=0.122754
+INFO:root:Epoch[3] Batch [2840]	Speed: 658.99 samples/sec	acc=0.129883
+INFO:root:Epoch[3] Batch [2860]	Speed: 680.46 samples/sec	acc=0.124512
+INFO:root:Epoch[3] Batch [2880]	Speed: 663.96 samples/sec	acc=0.123828
+INFO:root:Epoch[3] Batch [2900]	Speed: 663.27 samples/sec	acc=0.121875
+INFO:root:Epoch[3] Batch [2920]	Speed: 665.04 samples/sec	acc=0.122168
+INFO:root:Epoch[3] Batch [2940]	Speed: 664.46 samples/sec	acc=0.127930
+INFO:root:Epoch[3] Batch [2960]	Speed: 663.56 samples/sec	acc=0.129590
+INFO:root:Epoch[3] Batch [2980]	Speed: 663.95 samples/sec	acc=0.132031
+INFO:root:Epoch[3] Batch [3000]	Speed: 664.42 samples/sec	acc=0.124316
+INFO:root:Epoch[3] Batch [3020]	Speed: 663.19 samples/sec	acc=0.125879
+INFO:root:Epoch[3] Batch [3040]	Speed: 664.69 samples/sec	acc=0.128223
+INFO:root:Epoch[3] Batch [3060]	Speed: 664.21 samples/sec	acc=0.122852
+INFO:root:Epoch[3] Batch [3080]	Speed: 663.99 samples/sec	acc=0.125195
+INFO:root:Epoch[3] Batch [3100]	Speed: 664.47 samples/sec	acc=0.120801
+INFO:root:Epoch[3] Batch [3120]	Speed: 663.47 samples/sec	acc=0.122363
+INFO:root:Epoch[3] Batch [3140]	Speed: 663.53 samples/sec	acc=0.134668
+INFO:root:Epoch[3] Batch [3160]	Speed: 664.83 samples/sec	acc=0.125488
+INFO:root:Epoch[3] Batch [3180]	Speed: 663.79 samples/sec	acc=0.131738
+INFO:root:Epoch[3] Batch [3200]	Speed: 663.86 samples/sec	acc=0.128125
+INFO:root:Epoch[3] Batch [3220]	Speed: 664.37 samples/sec	acc=0.126660
+INFO:root:Epoch[3] Batch [3240]	Speed: 664.69 samples/sec	acc=0.123438
+INFO:root:Epoch[3] Batch [3260]	Speed: 663.81 samples/sec	acc=0.131348
+INFO:root:Epoch[3] Batch [3280]	Speed: 664.47 samples/sec	acc=0.128125
+INFO:root:Epoch[3] Batch [3300]	Speed: 664.63 samples/sec	acc=0.124121
+INFO:root:Epoch[3] Batch [3320]	Speed: 664.18 samples/sec	acc=0.123633
+INFO:root:Epoch[3] Batch [3340]	Speed: 664.41 samples/sec	acc=0.130371
+INFO:root:Epoch[3] Batch [3360]	Speed: 664.15 samples/sec	acc=0.121973
+INFO:root:Epoch[3] Batch [3380]	Speed: 663.45 samples/sec	acc=0.123145
+INFO:root:Epoch[3] Batch [3400]	Speed: 664.41 samples/sec	acc=0.127832
+INFO:root:Epoch[3] Batch [3420]	Speed: 664.75 samples/sec	acc=0.124316
+INFO:root:Epoch[3] Batch [3440]	Speed: 663.48 samples/sec	acc=0.128809
+[26000][MARGIN]0.661023,0.226349
+lr-batch-epoch: 0.1 3442 3
+testing verification..
+(14000, 512)
+infer time 21.8588
+[cfp_ff][26000]XNorm: 22.559293
+[cfp_ff][26000]Accuracy-Flip: 0.99257+-0.00325
+testing verification..
+(14000, 512)
+infer time 20.407895
+[cfp_fp][26000]XNorm: 18.426936
+[cfp_fp][26000]Accuracy-Flip: 0.83086+-0.01707
+testing verification..
+(12000, 512)
+infer time 21.819959
+[agedb_30][26000]XNorm: 22.785963
+[agedb_30][26000]Accuracy-Flip: 0.94317+-0.01084
+testing verification..
+(12000, 512)
+infer time 19.544551
+[lfw][26000]XNorm: 22.925718
+[lfw][26000]Accuracy-Flip: 0.99167+-0.00350
+[26000]Accuracy-Highest: 0.99250
+INFO:root:Epoch[3] Batch [3460]	Speed: 87.18 samples/sec	acc=0.121484
+INFO:root:Epoch[3] Batch [3480]	Speed: 664.00 samples/sec	acc=0.127832
+INFO:root:Epoch[3] Batch [3500]	Speed: 663.16 samples/sec	acc=0.128711
+INFO:root:Epoch[3] Batch [3520]	Speed: 664.71 samples/sec	acc=0.128223
+INFO:root:Epoch[3] Batch [3540]	Speed: 663.81 samples/sec	acc=0.123926
+INFO:root:Epoch[3] Batch [3560]	Speed: 662.96 samples/sec	acc=0.127441
+INFO:root:Epoch[3] Batch [3580]	Speed: 663.53 samples/sec	acc=0.126074
+INFO:root:Epoch[3] Batch [3600]	Speed: 663.95 samples/sec	acc=0.129199
+INFO:root:Epoch[3] Batch [3620]	Speed: 663.67 samples/sec	acc=0.128320
+INFO:root:Epoch[3] Batch [3640]	Speed: 667.46 samples/sec	acc=0.125098
+INFO:root:Epoch[3] Batch [3660]	Speed: 671.07 samples/sec	acc=0.126758
+INFO:root:Epoch[3] Batch [3680]	Speed: 663.59 samples/sec	acc=0.129590
+INFO:root:Epoch[3] Batch [3700]	Speed: 665.08 samples/sec	acc=0.127051
+INFO:root:Epoch[3] Batch [3720]	Speed: 663.88 samples/sec	acc=0.121875
+INFO:root:Epoch[3] Batch [3740]	Speed: 663.13 samples/sec	acc=0.124219
+INFO:root:Epoch[3] Batch [3760]	Speed: 664.26 samples/sec	acc=0.129785
+INFO:root:Epoch[3] Batch [3780]	Speed: 664.86 samples/sec	acc=0.125684
+INFO:root:Epoch[3] Batch [3800]	Speed: 662.89 samples/sec	acc=0.129199
+INFO:root:Epoch[3] Batch [3820]	Speed: 664.38 samples/sec	acc=0.122949
+INFO:root:Epoch[3] Batch [3840]	Speed: 664.12 samples/sec	acc=0.128516
+INFO:root:Epoch[3] Batch [3860]	Speed: 663.67 samples/sec	acc=0.126855
+INFO:root:Epoch[3] Batch [3880]	Speed: 664.96 samples/sec	acc=0.127441
+INFO:root:Epoch[3] Batch [3900]	Speed: 664.07 samples/sec	acc=0.125586
+INFO:root:Epoch[3] Batch [3920]	Speed: 662.60 samples/sec	acc=0.131348
+INFO:root:Epoch[3] Batch [3940]	Speed: 663.37 samples/sec	acc=0.126660
+INFO:root:Epoch[3] Batch [3960]	Speed: 664.11 samples/sec	acc=0.132129
+INFO:root:Epoch[3] Batch [3980]	Speed: 664.05 samples/sec	acc=0.130566
+INFO:root:Epoch[3] Batch [4000]	Speed: 664.85 samples/sec	acc=0.131055
+INFO:root:Epoch[3] Batch [4020]	Speed: 664.73 samples/sec	acc=0.125293
+INFO:root:Epoch[3] Batch [4040]	Speed: 653.47 samples/sec	acc=0.126172
+INFO:root:Epoch[3] Batch [4060]	Speed: 664.52 samples/sec	acc=0.122949
+INFO:root:Epoch[3] Batch [4080]	Speed: 664.28 samples/sec	acc=0.123047
+INFO:root:Epoch[3] Batch [4100]	Speed: 663.82 samples/sec	acc=0.124219
+INFO:root:Epoch[3] Batch [4120]	Speed: 664.02 samples/sec	acc=0.128809
+INFO:root:Epoch[3] Batch [4140]	Speed: 664.64 samples/sec	acc=0.123340
+INFO:root:Epoch[3] Batch [4160]	Speed: 664.12 samples/sec	acc=0.122852
+INFO:root:Epoch[3] Batch [4180]	Speed: 665.16 samples/sec	acc=0.124902
+INFO:root:Epoch[3] Batch [4200]	Speed: 663.81 samples/sec	acc=0.121582
+INFO:root:Epoch[3] Batch [4220]	Speed: 663.79 samples/sec	acc=0.117969
+INFO:root:Epoch[3] Batch [4240]	Speed: 664.55 samples/sec	acc=0.125586
+INFO:root:Epoch[3] Batch [4260]	Speed: 664.06 samples/sec	acc=0.128516
+INFO:root:Epoch[3] Batch [4280]	Speed: 663.59 samples/sec	acc=0.123047
+INFO:root:Epoch[3] Batch [4300]	Speed: 665.40 samples/sec	acc=0.130957
+INFO:root:Epoch[3] Batch [4320]	Speed: 664.98 samples/sec	acc=0.123438
+INFO:root:Epoch[3] Batch [4340]	Speed: 663.44 samples/sec	acc=0.130957
+INFO:root:Epoch[3] Batch [4360]	Speed: 663.80 samples/sec	acc=0.123730
+INFO:root:Epoch[3] Batch [4380]	Speed: 663.76 samples/sec	acc=0.133008
+INFO:root:Epoch[3] Batch [4400]	Speed: 663.94 samples/sec	acc=0.125684
+INFO:root:Epoch[3] Batch [4420]	Speed: 665.00 samples/sec	acc=0.128125
+INFO:root:Epoch[3] Batch [4440]	Speed: 671.72 samples/sec	acc=0.129883
+lr-batch-epoch: 0.1 4442 3
+INFO:root:Epoch[3] Batch [4460]	Speed: 664.25 samples/sec	acc=0.128906
+INFO:root:Epoch[3] Batch [4480]	Speed: 663.98 samples/sec	acc=0.130371
+INFO:root:Epoch[3] Batch [4500]	Speed: 663.85 samples/sec	acc=0.127051
+INFO:root:Epoch[3] Batch [4520]	Speed: 663.60 samples/sec	acc=0.128125
+INFO:root:Epoch[3] Batch [4540]	Speed: 664.99 samples/sec	acc=0.127148
+INFO:root:Epoch[3] Batch [4560]	Speed: 664.75 samples/sec	acc=0.126855
+INFO:root:Epoch[3] Batch [4580]	Speed: 663.85 samples/sec	acc=0.131543
+INFO:root:Epoch[3] Batch [4600]	Speed: 664.22 samples/sec	acc=0.129688
+INFO:root:Epoch[3] Batch [4620]	Speed: 663.74 samples/sec	acc=0.128711
+INFO:root:Epoch[3] Batch [4640]	Speed: 664.12 samples/sec	acc=0.132715
+INFO:root:Epoch[3] Batch [4660]	Speed: 664.73 samples/sec	acc=0.130273
+INFO:root:Epoch[3] Batch [4680]	Speed: 664.34 samples/sec	acc=0.128125
+INFO:root:Epoch[3] Batch [4700]	Speed: 663.96 samples/sec	acc=0.126074
+INFO:root:Epoch[3] Batch [4720]	Speed: 664.89 samples/sec	acc=0.127051
+INFO:root:Epoch[3] Batch [4740]	Speed: 663.51 samples/sec	acc=0.129102
+INFO:root:Epoch[3] Batch [4760]	Speed: 663.56 samples/sec	acc=0.124219
+INFO:root:Epoch[3] Batch [4780]	Speed: 664.69 samples/sec	acc=0.124121
+INFO:root:Epoch[3] Batch [4800]	Speed: 663.76 samples/sec	acc=0.127930
+INFO:root:Epoch[3] Batch [4820]	Speed: 663.68 samples/sec	acc=0.125586
+INFO:root:Epoch[3] Batch [4840]	Speed: 664.41 samples/sec	acc=0.125391
+INFO:root:Epoch[3] Batch [4860]	Speed: 663.67 samples/sec	acc=0.121094
+INFO:root:Epoch[3] Batch [4880]	Speed: 663.83 samples/sec	acc=0.129297
+INFO:root:Epoch[3] Batch [4900]	Speed: 664.27 samples/sec	acc=0.123145
+INFO:root:Epoch[3] Batch [4920]	Speed: 663.89 samples/sec	acc=0.130859
+INFO:root:Epoch[3] Batch [4940]	Speed: 662.91 samples/sec	acc=0.127441
+INFO:root:Epoch[3] Batch [4960]	Speed: 665.17 samples/sec	acc=0.130664
+INFO:root:Epoch[3] Batch [4980]	Speed: 664.26 samples/sec	acc=0.127539
+INFO:root:Epoch[3] Batch [5000]	Speed: 663.62 samples/sec	acc=0.134668
+INFO:root:Epoch[3] Batch [5020]	Speed: 663.40 samples/sec	acc=0.125098
+INFO:root:Epoch[3] Batch [5040]	Speed: 663.95 samples/sec	acc=0.124121
+INFO:root:Epoch[3] Batch [5060]	Speed: 663.76 samples/sec	acc=0.128125
+INFO:root:Epoch[3] Batch [5080]	Speed: 664.72 samples/sec	acc=0.124219
+INFO:root:Epoch[3] Batch [5100]	Speed: 663.57 samples/sec	acc=0.127344
+INFO:root:Epoch[3] Batch [5120]	Speed: 664.31 samples/sec	acc=0.129297
+INFO:root:Epoch[3] Batch [5140]	Speed: 663.71 samples/sec	acc=0.134570
+INFO:root:Epoch[3] Batch [5160]	Speed: 664.63 samples/sec	acc=0.125684
+INFO:root:Epoch[3] Batch [5180]	Speed: 663.41 samples/sec	acc=0.130566
+INFO:root:Epoch[3] Batch [5200]	Speed: 664.59 samples/sec	acc=0.130566
+INFO:root:Epoch[3] Batch [5220]	Speed: 658.54 samples/sec	acc=0.127148
+INFO:root:Epoch[3] Batch [5240]	Speed: 675.42 samples/sec	acc=0.125879
+INFO:root:Epoch[3] Batch [5260]	Speed: 677.59 samples/sec	acc=0.131348
+INFO:root:Epoch[3] Batch [5280]	Speed: 679.10 samples/sec	acc=0.129395
+INFO:root:Epoch[3] Batch [5300]	Speed: 663.36 samples/sec	acc=0.123535
+INFO:root:Epoch[3] Batch [5320]	Speed: 664.75 samples/sec	acc=0.122949
+INFO:root:Epoch[3] Batch [5340]	Speed: 663.91 samples/sec	acc=0.131543
+INFO:root:Epoch[3] Batch [5360]	Speed: 663.92 samples/sec	acc=0.132324
+INFO:root:Epoch[3] Batch [5380]	Speed: 664.66 samples/sec	acc=0.130566
+INFO:root:Epoch[3] Batch [5400]	Speed: 663.47 samples/sec	acc=0.125684
+INFO:root:Epoch[3] Batch [5420]	Speed: 662.88 samples/sec	acc=0.136230
+INFO:root:Epoch[3] Batch [5440]	Speed: 663.58 samples/sec	acc=0.126758
+[28000][MARGIN]0.672505,0.241496
+lr-batch-epoch: 0.1 5442 3
+testing verification..
+(14000, 512)
+infer time 20.845058
+[cfp_ff][28000]XNorm: 21.863884
+[cfp_ff][28000]Accuracy-Flip: 0.99343+-0.00321
+testing verification..
+(14000, 512)
+infer time 20.262105
+[cfp_fp][28000]XNorm: 17.841239
+[cfp_fp][28000]Accuracy-Flip: 0.83929+-0.01692
+testing verification..
+(12000, 512)
+infer time 20.014003
+[agedb_30][28000]XNorm: 21.388760
+[agedb_30][28000]Accuracy-Flip: 0.94683+-0.01092
+testing verification..
+(12000, 512)
+infer time 19.268613
+[lfw][28000]XNorm: 21.858025
+[lfw][28000]Accuracy-Flip: 0.99267+-0.00455
+saving 14
+INFO:root:Saved checkpoint to "../model-r50-am-lfw/model-0014.params"
+[28000]Accuracy-Highest: 0.99267
+INFO:root:Epoch[3] Batch [5460]	Speed: 87.04 samples/sec	acc=0.131445
+INFO:root:Epoch[3] Batch [5480]	Speed: 677.46 samples/sec	acc=0.123926
+INFO:root:Epoch[3] Batch [5500]	Speed: 663.41 samples/sec	acc=0.128418
+INFO:root:Epoch[3] Batch [5520]	Speed: 663.99 samples/sec	acc=0.122363
+INFO:root:Epoch[3] Batch [5540]	Speed: 663.13 samples/sec	acc=0.129883
+INFO:root:Epoch[3] Batch [5560]	Speed: 664.72 samples/sec	acc=0.127344
+INFO:root:Epoch[3] Batch [5580]	Speed: 664.61 samples/sec	acc=0.127344
+INFO:root:Epoch[3] Batch [5600]	Speed: 663.83 samples/sec	acc=0.127637
+INFO:root:Epoch[3] Batch [5620]	Speed: 664.57 samples/sec	acc=0.129102
+INFO:root:Epoch[3] Batch [5640]	Speed: 655.02 samples/sec	acc=0.123535
+INFO:root:Epoch[3] Batch [5660]	Speed: 663.64 samples/sec	acc=0.128027
+INFO:root:Epoch[3] Batch [5680]	Speed: 664.13 samples/sec	acc=0.125098
+INFO:root:Epoch[3] Batch [5700]	Speed: 663.81 samples/sec	acc=0.124707
+INFO:root:Epoch[3] Batch [5720]	Speed: 663.50 samples/sec	acc=0.133398
+INFO:root:Epoch[3] Batch [5740]	Speed: 663.96 samples/sec	acc=0.129785
+INFO:root:Epoch[3] Batch [5760]	Speed: 664.77 samples/sec	acc=0.134180
+INFO:root:Epoch[3] Batch [5780]	Speed: 645.82 samples/sec	acc=0.131152
+INFO:root:Epoch[3] Batch [5800]	Speed: 664.52 samples/sec	acc=0.129297
+INFO:root:Epoch[3] Batch [5820]	Speed: 663.54 samples/sec	acc=0.130469
+INFO:root:Epoch[3] Batch [5840]	Speed: 663.43 samples/sec	acc=0.128906
+INFO:root:Epoch[3] Batch [5860]	Speed: 657.87 samples/sec	acc=0.126465
+INFO:root:Epoch[3] Batch [5880]	Speed: 663.86 samples/sec	acc=0.130957
+INFO:root:Epoch[3] Batch [5900]	Speed: 663.04 samples/sec	acc=0.132422
+INFO:root:Epoch[3] Batch [5920]	Speed: 663.75 samples/sec	acc=0.130371
+INFO:root:Epoch[3] Batch [5940]	Speed: 662.68 samples/sec	acc=0.125488
+INFO:root:Epoch[3] Batch [5960]	Speed: 663.85 samples/sec	acc=0.124414
+INFO:root:Epoch[3] Batch [5980]	Speed: 663.71 samples/sec	acc=0.129785
+INFO:root:Epoch[3] Batch [6000]	Speed: 663.49 samples/sec	acc=0.126953
+INFO:root:Epoch[3] Batch [6020]	Speed: 663.35 samples/sec	acc=0.127734
+INFO:root:Epoch[3] Batch [6040]	Speed: 663.89 samples/sec	acc=0.129297
+INFO:root:Epoch[3] Batch [6060]	Speed: 663.14 samples/sec	acc=0.126367
+INFO:root:Epoch[3] Batch [6080]	Speed: 662.93 samples/sec	acc=0.129004
+INFO:root:Epoch[3] Batch [6100]	Speed: 664.81 samples/sec	acc=0.127148
+INFO:root:Epoch[3] Batch [6120]	Speed: 664.00 samples/sec	acc=0.125684
+INFO:root:Epoch[3] Batch [6140]	Speed: 662.86 samples/sec	acc=0.130176
+INFO:root:Epoch[3] Batch [6160]	Speed: 664.56 samples/sec	acc=0.126855
+INFO:root:Epoch[3] Batch [6180]	Speed: 663.60 samples/sec	acc=0.129980
+INFO:root:Epoch[3] Batch [6200]	Speed: 663.43 samples/sec	acc=0.128320
+INFO:root:Epoch[3] Batch [6220]	Speed: 662.30 samples/sec	acc=0.124902
+INFO:root:Epoch[3] Batch [6240]	Speed: 664.36 samples/sec	acc=0.126172
+INFO:root:Epoch[3] Batch [6260]	Speed: 663.59 samples/sec	acc=0.124902
+INFO:root:Epoch[3] Batch [6280]	Speed: 664.64 samples/sec	acc=0.122559
+INFO:root:Epoch[3] Batch [6300]	Speed: 663.70 samples/sec	acc=0.127441
+INFO:root:Epoch[3] Batch [6320]	Speed: 663.98 samples/sec	acc=0.138184
+INFO:root:Epoch[3] Batch [6340]	Speed: 664.47 samples/sec	acc=0.126758
+INFO:root:Epoch[3] Batch [6360]	Speed: 663.74 samples/sec	acc=0.123340
+INFO:root:Epoch[3] Batch [6380]	Speed: 663.78 samples/sec	acc=0.126172
+INFO:root:Epoch[3] Batch [6400]	Speed: 664.66 samples/sec	acc=0.135059
+INFO:root:Epoch[3] Batch [6420]	Speed: 663.86 samples/sec	acc=0.119141
+INFO:root:Epoch[3] Batch [6440]	Speed: 662.84 samples/sec	acc=0.127734
+lr-batch-epoch: 0.1 6442 3
+INFO:root:Epoch[3] Batch [6460]	Speed: 663.55 samples/sec	acc=0.128320
+INFO:root:Epoch[3] Batch [6480]	Speed: 664.26 samples/sec	acc=0.128613
+INFO:root:Epoch[3] Batch [6500]	Speed: 663.40 samples/sec	acc=0.131348
+INFO:root:Epoch[3] Batch [6520]	Speed: 664.24 samples/sec	acc=0.128516
+INFO:root:Epoch[3] Batch [6540]	Speed: 664.42 samples/sec	acc=0.125879
+INFO:root:Epoch[3] Batch [6560]	Speed: 663.78 samples/sec	acc=0.133301
+INFO:root:Epoch[3] Batch [6580]	Speed: 664.83 samples/sec	acc=0.129785
+INFO:root:Epoch[3] Batch [6600]	Speed: 664.32 samples/sec	acc=0.128613
+INFO:root:Epoch[3] Batch [6620]	Speed: 662.97 samples/sec	acc=0.129199
+INFO:root:Epoch[3] Batch [6640]	Speed: 664.96 samples/sec	acc=0.127148
+INFO:root:Epoch[3] Batch [6660]	Speed: 664.42 samples/sec	acc=0.128711
+INFO:root:Epoch[3] Batch [6680]	Speed: 663.26 samples/sec	acc=0.130957
+INFO:root:Epoch[3] Batch [6700]	Speed: 664.43 samples/sec	acc=0.131250
+INFO:root:Epoch[3] Batch [6720]	Speed: 664.17 samples/sec	acc=0.129199
+INFO:root:Epoch[3] Batch [6740]	Speed: 663.50 samples/sec	acc=0.124023
+INFO:root:Epoch[3] Batch [6760]	Speed: 667.68 samples/sec	acc=0.125195
+INFO:root:Epoch[3] Batch [6780]	Speed: 664.16 samples/sec	acc=0.128516
+INFO:root:Epoch[3] Batch [6800]	Speed: 664.15 samples/sec	acc=0.124512
+INFO:root:Epoch[3] Batch [6820]	Speed: 663.64 samples/sec	acc=0.129395
+INFO:root:Epoch[3] Batch [6840]	Speed: 664.13 samples/sec	acc=0.127832
+INFO:root:Epoch[3] Batch [6860]	Speed: 663.07 samples/sec	acc=0.129590
+INFO:root:Epoch[3] Batch [6880]	Speed: 664.00 samples/sec	acc=0.126074
+INFO:root:Epoch[3] Batch [6900]	Speed: 663.31 samples/sec	acc=0.131934
+INFO:root:Epoch[3] Batch [6920]	Speed: 663.62 samples/sec	acc=0.138770
+INFO:root:Epoch[3] Batch [6940]	Speed: 664.47 samples/sec	acc=0.127832
+INFO:root:Epoch[3] Batch [6960]	Speed: 663.95 samples/sec	acc=0.125293
+INFO:root:Epoch[3] Batch [6980]	Speed: 663.62 samples/sec	acc=0.129102
+INFO:root:Epoch[3] Batch [7000]	Speed: 663.89 samples/sec	acc=0.124902
+INFO:root:Epoch[3] Batch [7020]	Speed: 664.24 samples/sec	acc=0.129883
+INFO:root:Epoch[3] Batch [7040]	Speed: 662.99 samples/sec	acc=0.135742
+INFO:root:Epoch[3] Batch [7060]	Speed: 664.31 samples/sec	acc=0.132422
+INFO:root:Epoch[3] Batch [7080]	Speed: 663.11 samples/sec	acc=0.136719
+INFO:root:Epoch[3] Batch [7100]	Speed: 663.35 samples/sec	acc=0.131934
+INFO:root:Epoch[3] Batch [7120]	Speed: 664.66 samples/sec	acc=0.124707
+INFO:root:Epoch[3] Batch [7140]	Speed: 664.30 samples/sec	acc=0.127441
+INFO:root:Epoch[3] Batch [7160]	Speed: 663.58 samples/sec	acc=0.131055
+INFO:root:Epoch[3] Batch [7180]	Speed: 663.99 samples/sec	acc=0.128223
+INFO:root:Epoch[3] Batch [7200]	Speed: 664.17 samples/sec	acc=0.129102
+INFO:root:Epoch[3] Batch [7220]	Speed: 663.18 samples/sec	acc=0.131543
+INFO:root:Epoch[3] Batch [7240]	Speed: 664.60 samples/sec	acc=0.126270
+INFO:root:Epoch[3] Batch [7260]	Speed: 663.31 samples/sec	acc=0.128809
+INFO:root:Epoch[3] Batch [7280]	Speed: 661.80 samples/sec	acc=0.131250
+INFO:root:Epoch[3] Batch [7300]	Speed: 664.73 samples/sec	acc=0.132910
+INFO:root:Epoch[3] Batch [7320]	Speed: 666.62 samples/sec	acc=0.129980
+INFO:root:Epoch[3] Batch [7340]	Speed: 677.70 samples/sec	acc=0.128906
+INFO:root:Epoch[3] Batch [7360]	Speed: 677.18 samples/sec	acc=0.123145
+INFO:root:Epoch[3] Batch [7380]	Speed: 679.76 samples/sec	acc=0.124219
+INFO:root:Epoch[3] Batch [7400]	Speed: 662.22 samples/sec	acc=0.126660
+INFO:root:Epoch[3] Batch [7420]	Speed: 664.41 samples/sec	acc=0.128906
+INFO:root:Epoch[3] Batch [7440]	Speed: 663.52 samples/sec	acc=0.130176
+[30000][MARGIN]0.661733,0.226331
+lr-batch-epoch: 0.1 7442 3
+testing verification..
+(14000, 512)
+infer time 19.90458
+[cfp_ff][30000]XNorm: 23.170623
+[cfp_ff][30000]Accuracy-Flip: 0.99271+-0.00329
+testing verification..
+(14000, 512)
+infer time 22.25676
+[cfp_fp][30000]XNorm: 19.201502
+[cfp_fp][30000]Accuracy-Flip: 0.82100+-0.01378
+testing verification..
+(12000, 512)
+infer time 20.529218
+[agedb_30][30000]XNorm: 22.540779
+[agedb_30][30000]Accuracy-Flip: 0.94417+-0.00932
+testing verification..
+(12000, 512)
+infer time 18.718563
+[lfw][30000]XNorm: 23.312340
+[lfw][30000]Accuracy-Flip: 0.99283+-0.00441
+saving 15
+INFO:root:Saved checkpoint to "../model-r50-am-lfw/model-0015.params"
+[30000]Accuracy-Highest: 0.99283
+INFO:root:Epoch[3] Batch [7460]	Speed: 86.43 samples/sec	acc=0.133203
+INFO:root:Epoch[3] Batch [7480]	Speed: 663.38 samples/sec	acc=0.127441
+INFO:root:Epoch[3] Batch [7500]	Speed: 662.21 samples/sec	acc=0.132617
+INFO:root:Epoch[3] Train-acc=0.129449
+INFO:root:Epoch[3] Time cost=6194.280
+call reset()
+INFO:root:Epoch[4] Batch [20]	Speed: 673.87 samples/sec	acc=0.155041
+INFO:root:Epoch[4] Batch [40]	Speed: 663.79 samples/sec	acc=0.158008
+INFO:root:Epoch[4] Batch [60]	Speed: 662.85 samples/sec	acc=0.154590
+INFO:root:Epoch[4] Batch [80]	Speed: 664.58 samples/sec	acc=0.150977
+INFO:root:Epoch[4] Batch [100]	Speed: 663.91 samples/sec	acc=0.148535
+INFO:root:Epoch[4] Batch [120]	Speed: 663.78 samples/sec	acc=0.144043
+INFO:root:Epoch[4] Batch [140]	Speed: 663.78 samples/sec	acc=0.150684
+INFO:root:Epoch[4] Batch [160]	Speed: 664.75 samples/sec	acc=0.147656
+INFO:root:Epoch[4] Batch [180]	Speed: 664.28 samples/sec	acc=0.144727
+INFO:root:Epoch[4] Batch [200]	Speed: 663.78 samples/sec	acc=0.141211
+INFO:root:Epoch[4] Batch [220]	Speed: 664.62 samples/sec	acc=0.149023
+INFO:root:Epoch[4] Batch [240]	Speed: 663.82 samples/sec	acc=0.140039
+INFO:root:Epoch[4] Batch [260]	Speed: 664.21 samples/sec	acc=0.136719
+INFO:root:Epoch[4] Batch [280]	Speed: 664.80 samples/sec	acc=0.136230
+INFO:root:Epoch[4] Batch [300]	Speed: 664.31 samples/sec	acc=0.144922
+INFO:root:Epoch[4] Batch [320]	Speed: 664.58 samples/sec	acc=0.142871
+INFO:root:Epoch[4] Batch [340]	Speed: 664.43 samples/sec	acc=0.137891
+INFO:root:Epoch[4] Batch [360]	Speed: 664.43 samples/sec	acc=0.138086
+INFO:root:Epoch[4] Batch [380]	Speed: 664.11 samples/sec	acc=0.139648
+INFO:root:Epoch[4] Batch [400]	Speed: 663.53 samples/sec	acc=0.140137
+INFO:root:Epoch[4] Batch [420]	Speed: 664.35 samples/sec	acc=0.139746
+INFO:root:Epoch[4] Batch [440]	Speed: 664.55 samples/sec	acc=0.140234
+INFO:root:Epoch[4] Batch [460]	Speed: 664.32 samples/sec	acc=0.136523
+INFO:root:Epoch[4] Batch [480]	Speed: 664.06 samples/sec	acc=0.138770
+INFO:root:Epoch[4] Batch [500]	Speed: 664.08 samples/sec	acc=0.133008
+INFO:root:Epoch[4] Batch [520]	Speed: 664.79 samples/sec	acc=0.137988
+INFO:root:Epoch[4] Batch [540]	Speed: 664.51 samples/sec	acc=0.134277
+INFO:root:Epoch[4] Batch [560]	Speed: 664.15 samples/sec	acc=0.136035
+INFO:root:Epoch[4] Batch [580]	Speed: 664.48 samples/sec	acc=0.123633
+INFO:root:Epoch[4] Batch [600]	Speed: 664.55 samples/sec	acc=0.136230
+INFO:root:Epoch[4] Batch [620]	Speed: 664.31 samples/sec	acc=0.130957
+INFO:root:Epoch[4] Batch [640]	Speed: 664.24 samples/sec	acc=0.130078
+INFO:root:Epoch[4] Batch [660]	Speed: 664.04 samples/sec	acc=0.136621
+INFO:root:Epoch[4] Batch [680]	Speed: 665.17 samples/sec	acc=0.132812
+INFO:root:Epoch[4] Batch [700]	Speed: 664.30 samples/sec	acc=0.137500
+INFO:root:Epoch[4] Batch [720]	Speed: 664.43 samples/sec	acc=0.137891
+INFO:root:Epoch[4] Batch [740]	Speed: 664.00 samples/sec	acc=0.126758
+INFO:root:Epoch[4] Batch [760]	Speed: 664.27 samples/sec	acc=0.142383
+INFO:root:Epoch[4] Batch [780]	Speed: 664.59 samples/sec	acc=0.134082
+INFO:root:Epoch[4] Batch [800]	Speed: 664.90 samples/sec	acc=0.128027
+INFO:root:Epoch[4] Batch [820]	Speed: 683.19 samples/sec	acc=0.136914
+INFO:root:Epoch[4] Batch [840]	Speed: 663.24 samples/sec	acc=0.137891
+INFO:root:Epoch[4] Batch [860]	Speed: 664.29 samples/sec	acc=0.133594
+INFO:root:Epoch[4] Batch [880]	Speed: 664.50 samples/sec	acc=0.134473
+INFO:root:Epoch[4] Batch [900]	Speed: 664.25 samples/sec	acc=0.128711
+INFO:root:Epoch[4] Batch [920]	Speed: 664.04 samples/sec	acc=0.127930
+lr-batch-epoch: 0.1 923 4
+INFO:root:Epoch[4] Batch [940]	Speed: 664.25 samples/sec	acc=0.130957
+INFO:root:Epoch[4] Batch [960]	Speed: 663.67 samples/sec	acc=0.128711
+INFO:root:Epoch[4] Batch [980]	Speed: 663.79 samples/sec	acc=0.131836
+INFO:root:Epoch[4] Batch [1000]	Speed: 663.87 samples/sec	acc=0.129102
+INFO:root:Epoch[4] Batch [1020]	Speed: 664.87 samples/sec	acc=0.127930
+INFO:root:Epoch[4] Batch [1040]	Speed: 663.82 samples/sec	acc=0.130566
+INFO:root:Epoch[4] Batch [1060]	Speed: 664.12 samples/sec	acc=0.131836
+INFO:root:Epoch[4] Batch [1080]	Speed: 664.72 samples/sec	acc=0.127637
+INFO:root:Epoch[4] Batch [1100]	Speed: 664.79 samples/sec	acc=0.124707
+INFO:root:Epoch[4] Batch [1120]	Speed: 665.24 samples/sec	acc=0.127930
+INFO:root:Epoch[4] Batch [1140]	Speed: 663.19 samples/sec	acc=0.132324
+INFO:root:Epoch[4] Batch [1160]	Speed: 663.95 samples/sec	acc=0.134180
+INFO:root:Epoch[4] Batch [1180]	Speed: 664.73 samples/sec	acc=0.128516
+INFO:root:Epoch[4] Batch [1200]	Speed: 664.67 samples/sec	acc=0.131641
+INFO:root:Epoch[4] Batch [1220]	Speed: 664.35 samples/sec	acc=0.133105
+INFO:root:Epoch[4] Batch [1240]	Speed: 664.22 samples/sec	acc=0.136816
+INFO:root:Epoch[4] Batch [1260]	Speed: 664.81 samples/sec	acc=0.136914
+INFO:root:Epoch[4] Batch [1280]	Speed: 664.22 samples/sec	acc=0.134082
+INFO:root:Epoch[4] Batch [1300]	Speed: 665.18 samples/sec	acc=0.132812
+INFO:root:Epoch[4] Batch [1320]	Speed: 664.18 samples/sec	acc=0.133008
+INFO:root:Epoch[4] Batch [1340]	Speed: 664.17 samples/sec	acc=0.136426
+INFO:root:Epoch[4] Batch [1360]	Speed: 663.57 samples/sec	acc=0.133887
+INFO:root:Epoch[4] Batch [1380]	Speed: 664.20 samples/sec	acc=0.129785
+INFO:root:Epoch[4] Batch [1400]	Speed: 664.27 samples/sec	acc=0.123633
+INFO:root:Epoch[4] Batch [1420]	Speed: 664.61 samples/sec	acc=0.129980
+INFO:root:Epoch[4] Batch [1440]	Speed: 663.41 samples/sec	acc=0.129980
+INFO:root:Epoch[4] Batch [1460]	Speed: 664.07 samples/sec	acc=0.138867
+INFO:root:Epoch[4] Batch [1480]	Speed: 664.34 samples/sec	acc=0.124219
+INFO:root:Epoch[4] Batch [1500]	Speed: 664.73 samples/sec	acc=0.127344
+INFO:root:Epoch[4] Batch [1520]	Speed: 663.56 samples/sec	acc=0.135742
+INFO:root:Epoch[4] Batch [1540]	Speed: 664.24 samples/sec	acc=0.127930
+INFO:root:Epoch[4] Batch [1560]	Speed: 664.63 samples/sec	acc=0.131152
+INFO:root:Epoch[4] Batch [1580]	Speed: 659.18 samples/sec	acc=0.136816
+INFO:root:Epoch[4] Batch [1600]	Speed: 681.27 samples/sec	acc=0.129688
+INFO:root:Epoch[4] Batch [1620]	Speed: 663.35 samples/sec	acc=0.128027
+INFO:root:Epoch[4] Batch [1640]	Speed: 664.00 samples/sec	acc=0.135059
+INFO:root:Epoch[4] Batch [1660]	Speed: 664.82 samples/sec	acc=0.131250
+INFO:root:Epoch[4] Batch [1680]	Speed: 663.46 samples/sec	acc=0.134277
+INFO:root:Epoch[4] Batch [1700]	Speed: 663.90 samples/sec	acc=0.140332
+INFO:root:Epoch[4] Batch [1720]	Speed: 663.26 samples/sec	acc=0.132715
+INFO:root:Epoch[4] Batch [1740]	Speed: 664.70 samples/sec	acc=0.136719
+INFO:root:Epoch[4] Batch [1760]	Speed: 664.55 samples/sec	acc=0.131055
+INFO:root:Epoch[4] Batch [1780]	Speed: 663.91 samples/sec	acc=0.136230
+INFO:root:Epoch[4] Batch [1800]	Speed: 664.39 samples/sec	acc=0.132617
+INFO:root:Epoch[4] Batch [1820]	Speed: 664.18 samples/sec	acc=0.132227
+INFO:root:Epoch[4] Batch [1840]	Speed: 663.53 samples/sec	acc=0.134473
+INFO:root:Epoch[4] Batch [1860]	Speed: 664.66 samples/sec	acc=0.132715
+INFO:root:Epoch[4] Batch [1880]	Speed: 663.26 samples/sec	acc=0.131055
+INFO:root:Epoch[4] Batch [1900]	Speed: 663.04 samples/sec	acc=0.128711
+INFO:root:Epoch[4] Batch [1920]	Speed: 664.32 samples/sec	acc=0.134375
+[32000][MARGIN]0.665787,0.234442
+lr-batch-epoch: 0.1 1923 4
+testing verification..
+(14000, 512)
+infer time 23.744189
+[cfp_ff][32000]XNorm: 21.983676
+[cfp_ff][32000]Accuracy-Flip: 0.99286+-0.00389
+testing verification..
+(14000, 512)
+infer time 22.832868
+[cfp_fp][32000]XNorm: 18.096088
+[cfp_fp][32000]Accuracy-Flip: 0.84700+-0.02116
+testing verification..
+(12000, 512)
+infer time 17.128199
+[agedb_30][32000]XNorm: 21.840082
+[agedb_30][32000]Accuracy-Flip: 0.94300+-0.00963
+testing verification..
+(12000, 512)
+infer time 20.034206
+[lfw][32000]XNorm: 21.846459
+[lfw][32000]Accuracy-Flip: 0.99183+-0.00329
+[32000]Accuracy-Highest: 0.99283
+INFO:root:Epoch[4] Batch [1940]	Speed: 86.30 samples/sec	acc=0.131445
+INFO:root:Epoch[4] Batch [1960]	Speed: 682.94 samples/sec	acc=0.131152
+INFO:root:Epoch[4] Batch [1980]	Speed: 662.51 samples/sec	acc=0.135352
+INFO:root:Epoch[4] Batch [2000]	Speed: 662.63 samples/sec	acc=0.136230
+INFO:root:Epoch[4] Batch [2020]	Speed: 662.96 samples/sec	acc=0.130078
+INFO:root:Epoch[4] Batch [2040]	Speed: 663.31 samples/sec	acc=0.130469
+INFO:root:Epoch[4] Batch [2060]	Speed: 664.32 samples/sec	acc=0.135254
+INFO:root:Epoch[4] Batch [2080]	Speed: 664.79 samples/sec	acc=0.136133
+INFO:root:Epoch[4] Batch [2100]	Speed: 663.51 samples/sec	acc=0.132129
+INFO:root:Epoch[4] Batch [2120]	Speed: 663.86 samples/sec	acc=0.130176
+INFO:root:Epoch[4] Batch [2140]	Speed: 664.23 samples/sec	acc=0.127930
+INFO:root:Epoch[4] Batch [2160]	Speed: 664.24 samples/sec	acc=0.135352
+INFO:root:Epoch[4] Batch [2180]	Speed: 663.77 samples/sec	acc=0.141797
+INFO:root:Epoch[4] Batch [2200]	Speed: 664.26 samples/sec	acc=0.133105
+INFO:root:Epoch[4] Batch [2220]	Speed: 663.63 samples/sec	acc=0.135156
+INFO:root:Epoch[4] Batch [2240]	Speed: 664.33 samples/sec	acc=0.133008
+INFO:root:Epoch[4] Batch [2260]	Speed: 664.29 samples/sec	acc=0.134277
+INFO:root:Epoch[4] Batch [2280]	Speed: 663.78 samples/sec	acc=0.137305
+INFO:root:Epoch[4] Batch [2300]	Speed: 663.89 samples/sec	acc=0.128027
+INFO:root:Epoch[4] Batch [2320]	Speed: 663.85 samples/sec	acc=0.131543
+INFO:root:Epoch[4] Batch [2340]	Speed: 663.91 samples/sec	acc=0.137891
+INFO:root:Epoch[4] Batch [2360]	Speed: 665.93 samples/sec	acc=0.132715
+INFO:root:Epoch[4] Batch [2380]	Speed: 671.25 samples/sec	acc=0.132520
+INFO:root:Epoch[4] Batch [2400]	Speed: 663.57 samples/sec	acc=0.134570
+INFO:root:Epoch[4] Batch [2420]	Speed: 664.41 samples/sec	acc=0.141016
+INFO:root:Epoch[4] Batch [2440]	Speed: 663.11 samples/sec	acc=0.131836
+INFO:root:Epoch[4] Batch [2460]	Speed: 664.21 samples/sec	acc=0.135937
+INFO:root:Epoch[4] Batch [2480]	Speed: 664.47 samples/sec	acc=0.133594
+INFO:root:Epoch[4] Batch [2500]	Speed: 664.19 samples/sec	acc=0.133887
+INFO:root:Epoch[4] Batch [2520]	Speed: 664.20 samples/sec	acc=0.138379
+INFO:root:Epoch[4] Batch [2540]	Speed: 663.68 samples/sec	acc=0.131250
+INFO:root:Epoch[4] Batch [2560]	Speed: 664.52 samples/sec	acc=0.138672
+INFO:root:Epoch[4] Batch [2580]	Speed: 663.96 samples/sec	acc=0.137500
+INFO:root:Epoch[4] Batch [2600]	Speed: 663.75 samples/sec	acc=0.136719
+INFO:root:Epoch[4] Batch [2620]	Speed: 663.78 samples/sec	acc=0.133301
+INFO:root:Epoch[4] Batch [2640]	Speed: 663.47 samples/sec	acc=0.140137
+INFO:root:Epoch[4] Batch [2660]	Speed: 664.15 samples/sec	acc=0.126172
+INFO:root:Epoch[4] Batch [2680]	Speed: 664.19 samples/sec	acc=0.126270
+INFO:root:Epoch[4] Batch [2700]	Speed: 664.57 samples/sec	acc=0.128027
+INFO:root:Epoch[4] Batch [2720]	Speed: 663.94 samples/sec	acc=0.134473
+INFO:root:Epoch[4] Batch [2740]	Speed: 664.51 samples/sec	acc=0.131934
+INFO:root:Epoch[4] Batch [2760]	Speed: 663.85 samples/sec	acc=0.131250
+INFO:root:Epoch[4] Batch [2780]	Speed: 664.24 samples/sec	acc=0.129004
+INFO:root:Epoch[4] Batch [2800]	Speed: 665.26 samples/sec	acc=0.131445
+INFO:root:Epoch[4] Batch [2820]	Speed: 663.78 samples/sec	acc=0.134668
+INFO:root:Epoch[4] Batch [2840]	Speed: 664.51 samples/sec	acc=0.129102
+INFO:root:Epoch[4] Batch [2860]	Speed: 664.75 samples/sec	acc=0.134180
+INFO:root:Epoch[4] Batch [2880]	Speed: 664.51 samples/sec	acc=0.130078
+INFO:root:Epoch[4] Batch [2900]	Speed: 665.17 samples/sec	acc=0.132617
+INFO:root:Epoch[4] Batch [2920]	Speed: 663.81 samples/sec	acc=0.129004
+lr-batch-epoch: 0.1 2923 4
+INFO:root:Epoch[4] Batch [2940]	Speed: 665.69 samples/sec	acc=0.138184
+INFO:root:Epoch[4] Batch [2960]	Speed: 663.00 samples/sec	acc=0.135254
+INFO:root:Epoch[4] Batch [2980]	Speed: 664.51 samples/sec	acc=0.132520
+INFO:root:Epoch[4] Batch [3000]	Speed: 665.11 samples/sec	acc=0.139648
+INFO:root:Epoch[4] Batch [3020]	Speed: 664.24 samples/sec	acc=0.130957
+INFO:root:Epoch[4] Batch [3040]	Speed: 664.43 samples/sec	acc=0.133496
+INFO:root:Epoch[4] Batch [3060]	Speed: 665.05 samples/sec	acc=0.135352
+INFO:root:Epoch[4] Batch [3080]	Speed: 663.79 samples/sec	acc=0.135937
+INFO:root:Epoch[4] Batch [3100]	Speed: 664.52 samples/sec	acc=0.133691
+INFO:root:Epoch[4] Batch [3120]	Speed: 663.78 samples/sec	acc=0.128516
+INFO:root:Epoch[4] Batch [3140]	Speed: 663.99 samples/sec	acc=0.134375
+INFO:root:Epoch[4] Batch [3160]	Speed: 676.10 samples/sec	acc=0.136719
+INFO:root:Epoch[4] Batch [3180]	Speed: 664.20 samples/sec	acc=0.134961
+INFO:root:Epoch[4] Batch [3200]	Speed: 664.30 samples/sec	acc=0.142480
+INFO:root:Epoch[4] Batch [3220]	Speed: 664.11 samples/sec	acc=0.140039
+INFO:root:Epoch[4] Batch [3240]	Speed: 664.07 samples/sec	acc=0.139941
+INFO:root:Epoch[4] Batch [3260]	Speed: 664.08 samples/sec	acc=0.130664
+INFO:root:Epoch[4] Batch [3280]	Speed: 664.14 samples/sec	acc=0.135742
+INFO:root:Epoch[4] Batch [3300]	Speed: 664.48 samples/sec	acc=0.133496
+INFO:root:Epoch[4] Batch [3320]	Speed: 664.40 samples/sec	acc=0.137012
+INFO:root:Epoch[4] Batch [3340]	Speed: 664.75 samples/sec	acc=0.133301
+INFO:root:Epoch[4] Batch [3360]	Speed: 664.32 samples/sec	acc=0.133984
+INFO:root:Epoch[4] Batch [3380]	Speed: 664.18 samples/sec	acc=0.132910
+INFO:root:Epoch[4] Batch [3400]	Speed: 664.52 samples/sec	acc=0.134961
+INFO:root:Epoch[4] Batch [3420]	Speed: 664.84 samples/sec	acc=0.135254
+INFO:root:Epoch[4] Batch [3440]	Speed: 663.48 samples/sec	acc=0.135937
+INFO:root:Epoch[4] Batch [3460]	Speed: 664.85 samples/sec	acc=0.134570
+INFO:root:Epoch[4] Batch [3480]	Speed: 664.66 samples/sec	acc=0.135449
+INFO:root:Epoch[4] Batch [3500]	Speed: 663.44 samples/sec	acc=0.132324
+INFO:root:Epoch[4] Batch [3520]	Speed: 664.61 samples/sec	acc=0.134668
+INFO:root:Epoch[4] Batch [3540]	Speed: 664.72 samples/sec	acc=0.140137
+INFO:root:Epoch[4] Batch [3560]	Speed: 663.91 samples/sec	acc=0.130469
+INFO:root:Epoch[4] Batch [3580]	Speed: 664.69 samples/sec	acc=0.130566
+INFO:root:Epoch[4] Batch [3600]	Speed: 664.72 samples/sec	acc=0.132227
+INFO:root:Epoch[4] Batch [3620]	Speed: 664.70 samples/sec	acc=0.134961
+INFO:root:Epoch[4] Batch [3640]	Speed: 664.29 samples/sec	acc=0.133203
+INFO:root:Epoch[4] Batch [3660]	Speed: 664.37 samples/sec	acc=0.135840
+INFO:root:Epoch[4] Batch [3680]	Speed: 665.23 samples/sec	acc=0.133594
+INFO:root:Epoch[4] Batch [3700]	Speed: 664.60 samples/sec	acc=0.133105
+INFO:root:Epoch[4] Batch [3720]	Speed: 664.21 samples/sec	acc=0.130469
+INFO:root:Epoch[4] Batch [3740]	Speed: 664.38 samples/sec	acc=0.134473
+INFO:root:Epoch[4] Batch [3760]	Speed: 664.03 samples/sec	acc=0.136719
+INFO:root:Epoch[4] Batch [3780]	Speed: 665.02 samples/sec	acc=0.135059
+INFO:root:Epoch[4] Batch [3800]	Speed: 664.89 samples/sec	acc=0.140234
+INFO:root:Epoch[4] Batch [3820]	Speed: 664.25 samples/sec	acc=0.134277
+INFO:root:Epoch[4] Batch [3840]	Speed: 664.22 samples/sec	acc=0.139551
+INFO:root:Epoch[4] Batch [3860]	Speed: 664.00 samples/sec	acc=0.133105
+INFO:root:Epoch[4] Batch [3880]	Speed: 664.00 samples/sec	acc=0.134180
+INFO:root:Epoch[4] Batch [3900]	Speed: 664.55 samples/sec	acc=0.132715
+INFO:root:Epoch[4] Batch [3920]	Speed: 664.51 samples/sec	acc=0.129199
+[34000][MARGIN]0.665160,0.233647
+lr-batch-epoch: 0.1 3923 4
+testing verification..
+(14000, 512)
+infer time 20.705559
+[cfp_ff][34000]XNorm: 21.942060
+[cfp_ff][34000]Accuracy-Flip: 0.99429+-0.00319
+testing verification..
+(14000, 512)
+infer time 24.046523
+[cfp_fp][34000]XNorm: 17.491616
+[cfp_fp][34000]Accuracy-Flip: 0.83771+-0.01322
+testing verification..
+(12000, 512)
+infer time 18.027092
+[agedb_30][34000]XNorm: 21.578602
+[agedb_30][34000]Accuracy-Flip: 0.94017+-0.01031
+testing verification..
+(12000, 512)
+infer time 15.701564
+[lfw][34000]XNorm: 21.812679
+[lfw][34000]Accuracy-Flip: 0.99267+-0.00403
+[34000]Accuracy-Highest: 0.99283
+INFO:root:Epoch[4] Batch [3940]	Speed: 92.10 samples/sec	acc=0.136035
+INFO:root:Epoch[4] Batch [3960]	Speed: 674.59 samples/sec	acc=0.138477
+INFO:root:Epoch[4] Batch [3980]	Speed: 685.53 samples/sec	acc=0.128711
+INFO:root:Epoch[4] Batch [4000]	Speed: 663.08 samples/sec	acc=0.133594
+INFO:root:Epoch[4] Batch [4020]	Speed: 663.84 samples/sec	acc=0.128613
+INFO:root:Epoch[4] Batch [4040]	Speed: 663.94 samples/sec	acc=0.137793
+INFO:root:Epoch[4] Batch [4060]	Speed: 665.00 samples/sec	acc=0.137598
+INFO:root:Epoch[4] Batch [4080]	Speed: 664.74 samples/sec	acc=0.132715
+INFO:root:Epoch[4] Batch [4100]	Speed: 664.42 samples/sec	acc=0.132812
+INFO:root:Epoch[4] Batch [4120]	Speed: 664.66 samples/sec	acc=0.135352
+INFO:root:Epoch[4] Batch [4140]	Speed: 664.08 samples/sec	acc=0.133691
+INFO:root:Epoch[4] Batch [4160]	Speed: 665.17 samples/sec	acc=0.136719
+INFO:root:Epoch[4] Batch [4180]	Speed: 664.43 samples/sec	acc=0.133496
+INFO:root:Epoch[4] Batch [4200]	Speed: 664.08 samples/sec	acc=0.132910
+INFO:root:Epoch[4] Batch [4220]	Speed: 664.07 samples/sec	acc=0.131738
+INFO:root:Epoch[4] Batch [4240]	Speed: 665.29 samples/sec	acc=0.132129
+INFO:root:Epoch[4] Batch [4260]	Speed: 663.19 samples/sec	acc=0.138477
+INFO:root:Epoch[4] Batch [4280]	Speed: 664.24 samples/sec	acc=0.141699
+INFO:root:Epoch[4] Batch [4300]	Speed: 663.71 samples/sec	acc=0.136621
+INFO:root:Epoch[4] Batch [4320]	Speed: 664.06 samples/sec	acc=0.144336
+INFO:root:Epoch[4] Batch [4340]	Speed: 664.00 samples/sec	acc=0.134961
+INFO:root:Epoch[4] Batch [4360]	Speed: 664.02 samples/sec	acc=0.134668
+INFO:root:Epoch[4] Batch [4380]	Speed: 664.18 samples/sec	acc=0.137109
+INFO:root:Epoch[4] Batch [4400]	Speed: 664.32 samples/sec	acc=0.129590
+INFO:root:Epoch[4] Batch [4420]	Speed: 664.66 samples/sec	acc=0.135156
+INFO:root:Epoch[4] Batch [4440]	Speed: 664.54 samples/sec	acc=0.133008
+INFO:root:Epoch[4] Batch [4460]	Speed: 663.62 samples/sec	acc=0.133887
+INFO:root:Epoch[4] Batch [4480]	Speed: 664.57 samples/sec	acc=0.135449
+INFO:root:Epoch[4] Batch [4500]	Speed: 664.19 samples/sec	acc=0.135547
+INFO:root:Epoch[4] Batch [4520]	Speed: 664.83 samples/sec	acc=0.133105
+INFO:root:Epoch[4] Batch [4540]	Speed: 664.35 samples/sec	acc=0.136230
+INFO:root:Epoch[4] Batch [4560]	Speed: 664.76 samples/sec	acc=0.136230
+INFO:root:Epoch[4] Batch [4580]	Speed: 664.14 samples/sec	acc=0.130664
+INFO:root:Epoch[4] Batch [4600]	Speed: 664.79 samples/sec	acc=0.136621
+INFO:root:Epoch[4] Batch [4620]	Speed: 664.17 samples/sec	acc=0.134375
+INFO:root:Epoch[4] Batch [4640]	Speed: 664.96 samples/sec	acc=0.136914
+INFO:root:Epoch[4] Batch [4660]	Speed: 664.42 samples/sec	acc=0.135547
+INFO:root:Epoch[4] Batch [4680]	Speed: 665.15 samples/sec	acc=0.126953
+INFO:root:Epoch[4] Batch [4700]	Speed: 665.06 samples/sec	acc=0.137109
+INFO:root:Epoch[4] Batch [4720]	Speed: 664.71 samples/sec	acc=0.128027
+INFO:root:Epoch[4] Batch [4740]	Speed: 662.65 samples/sec	acc=0.135547
+INFO:root:Epoch[4] Batch [4760]	Speed: 679.88 samples/sec	acc=0.133105
+INFO:root:Epoch[4] Batch [4780]	Speed: 664.17 samples/sec	acc=0.138965
+INFO:root:Epoch[4] Batch [4800]	Speed: 665.00 samples/sec	acc=0.139648
+INFO:root:Epoch[4] Batch [4820]	Speed: 664.83 samples/sec	acc=0.127344
+INFO:root:Epoch[4] Batch [4840]	Speed: 664.51 samples/sec	acc=0.136523
+INFO:root:Epoch[4] Batch [4860]	Speed: 664.51 samples/sec	acc=0.135645
+INFO:root:Epoch[4] Batch [4880]	Speed: 664.12 samples/sec	acc=0.135254
+INFO:root:Epoch[4] Batch [4900]	Speed: 664.45 samples/sec	acc=0.136914
+INFO:root:Epoch[4] Batch [4920]	Speed: 664.58 samples/sec	acc=0.138770
+lr-batch-epoch: 0.1 4923 4
+INFO:root:Epoch[4] Batch [4940]	Speed: 664.65 samples/sec	acc=0.140625
+INFO:root:Epoch[4] Batch [4960]	Speed: 664.46 samples/sec	acc=0.130078
+INFO:root:Epoch[4] Batch [4980]	Speed: 664.99 samples/sec	acc=0.131641
+INFO:root:Epoch[4] Batch [5000]	Speed: 664.16 samples/sec	acc=0.132715
+INFO:root:Epoch[4] Batch [5020]	Speed: 664.60 samples/sec	acc=0.132227
+INFO:root:Epoch[4] Batch [5040]	Speed: 664.37 samples/sec	acc=0.133789
+INFO:root:Epoch[4] Batch [5060]	Speed: 664.67 samples/sec	acc=0.139258
+INFO:root:Epoch[4] Batch [5080]	Speed: 663.77 samples/sec	acc=0.137793
+INFO:root:Epoch[4] Batch [5100]	Speed: 664.68 samples/sec	acc=0.135937
+INFO:root:Epoch[4] Batch [5120]	Speed: 664.37 samples/sec	acc=0.137305
+INFO:root:Epoch[4] Batch [5140]	Speed: 664.57 samples/sec	acc=0.132520
+INFO:root:Epoch[4] Batch [5160]	Speed: 664.06 samples/sec	acc=0.132031
+INFO:root:Epoch[4] Batch [5180]	Speed: 664.06 samples/sec	acc=0.133301
+INFO:root:Epoch[4] Batch [5200]	Speed: 664.69 samples/sec	acc=0.134473
+INFO:root:Epoch[4] Batch [5220]	Speed: 664.23 samples/sec	acc=0.136523
+INFO:root:Epoch[4] Batch [5240]	Speed: 663.82 samples/sec	acc=0.129590
+INFO:root:Epoch[4] Batch [5260]	Speed: 664.67 samples/sec	acc=0.135059
+INFO:root:Epoch[4] Batch [5280]	Speed: 664.65 samples/sec	acc=0.131543
+INFO:root:Epoch[4] Batch [5300]	Speed: 664.45 samples/sec	acc=0.138672
+INFO:root:Epoch[4] Batch [5320]	Speed: 664.67 samples/sec	acc=0.132324
+INFO:root:Epoch[4] Batch [5340]	Speed: 664.22 samples/sec	acc=0.136621
+INFO:root:Epoch[4] Batch [5360]	Speed: 664.42 samples/sec	acc=0.133301
+INFO:root:Epoch[4] Batch [5380]	Speed: 664.16 samples/sec	acc=0.139258
+INFO:root:Epoch[4] Batch [5400]	Speed: 664.97 samples/sec	acc=0.141211
+INFO:root:Epoch[4] Batch [5420]	Speed: 664.40 samples/sec	acc=0.137988
+INFO:root:Epoch[4] Batch [5440]	Speed: 664.17 samples/sec	acc=0.135352
+INFO:root:Epoch[4] Batch [5460]	Speed: 664.70 samples/sec	acc=0.133398
+INFO:root:Epoch[4] Batch [5480]	Speed: 664.55 samples/sec	acc=0.133984
+INFO:root:Epoch[4] Batch [5500]	Speed: 664.81 samples/sec	acc=0.141699
+INFO:root:Epoch[4] Batch [5520]	Speed: 664.80 samples/sec	acc=0.135547
+INFO:root:Epoch[4] Batch [5540]	Speed: 670.28 samples/sec	acc=0.135352
+INFO:root:Epoch[4] Batch [5560]	Speed: 676.06 samples/sec	acc=0.141699
+INFO:root:Epoch[4] Batch [5580]	Speed: 664.74 samples/sec	acc=0.138281
+INFO:root:Epoch[4] Batch [5600]	Speed: 664.49 samples/sec	acc=0.140039
+INFO:root:Epoch[4] Batch [5620]	Speed: 664.72 samples/sec	acc=0.138477
+INFO:root:Epoch[4] Batch [5640]	Speed: 664.31 samples/sec	acc=0.138379
+INFO:root:Epoch[4] Batch [5660]	Speed: 664.81 samples/sec	acc=0.137891
+INFO:root:Epoch[4] Batch [5680]	Speed: 664.42 samples/sec	acc=0.140430
+INFO:root:Epoch[4] Batch [5700]	Speed: 664.92 samples/sec	acc=0.139453
+INFO:root:Epoch[4] Batch [5720]	Speed: 664.49 samples/sec	acc=0.138184
+INFO:root:Epoch[4] Batch [5740]	Speed: 664.19 samples/sec	acc=0.140820
+INFO:root:Epoch[4] Batch [5760]	Speed: 663.58 samples/sec	acc=0.139355
+INFO:root:Epoch[4] Batch [5780]	Speed: 664.84 samples/sec	acc=0.136523
+INFO:root:Epoch[4] Batch [5800]	Speed: 664.59 samples/sec	acc=0.137109
+INFO:root:Epoch[4] Batch [5820]	Speed: 664.72 samples/sec	acc=0.136719
+INFO:root:Epoch[4] Batch [5840]	Speed: 664.29 samples/sec	acc=0.137891
+INFO:root:Epoch[4] Batch [5860]	Speed: 664.58 samples/sec	acc=0.135156
+INFO:root:Epoch[4] Batch [5880]	Speed: 665.22 samples/sec	acc=0.132324
+INFO:root:Epoch[4] Batch [5900]	Speed: 664.37 samples/sec	acc=0.138281
+INFO:root:Epoch[4] Batch [5920]	Speed: 664.57 samples/sec	acc=0.136523
+[36000][MARGIN]0.645507,0.208008
+lr-batch-epoch: 0.1 5923 4
+testing verification..
+(14000, 512)
+infer time 20.000716
+[cfp_ff][36000]XNorm: 22.755588
+[cfp_ff][36000]Accuracy-Flip: 0.99329+-0.00332
+testing verification..
+(14000, 512)
+infer time 19.54085
+[cfp_fp][36000]XNorm: 19.190483
+[cfp_fp][36000]Accuracy-Flip: 0.84314+-0.01479
+testing verification..
+(12000, 512)
+infer time 15.789011
+[agedb_30][36000]XNorm: 22.199309
+[agedb_30][36000]Accuracy-Flip: 0.94867+-0.00799
+testing verification..
+(12000, 512)
+infer time 16.240729
+[lfw][36000]XNorm: 22.511726
+[lfw][36000]Accuracy-Flip: 0.99267+-0.00382
+[36000]Accuracy-Highest: 0.99283
+INFO:root:Epoch[4] Batch [5940]	Speed: 97.95 samples/sec	acc=0.139648
+INFO:root:Epoch[4] Batch [5960]	Speed: 662.60 samples/sec	acc=0.138379
+INFO:root:Epoch[4] Batch [5980]	Speed: 663.66 samples/sec	acc=0.137891
+INFO:root:Epoch[4] Batch [6000]	Speed: 663.55 samples/sec	acc=0.134570
+INFO:root:Epoch[4] Batch [6020]	Speed: 664.27 samples/sec	acc=0.142676
+INFO:root:Epoch[4] Batch [6040]	Speed: 664.79 samples/sec	acc=0.139355
+INFO:root:Epoch[4] Batch [6060]	Speed: 664.88 samples/sec	acc=0.134863
+INFO:root:Epoch[4] Batch [6080]	Speed: 664.20 samples/sec	acc=0.137598
+INFO:root:Epoch[4] Batch [6100]	Speed: 664.54 samples/sec	acc=0.136621
+INFO:root:Epoch[4] Batch [6120]	Speed: 664.35 samples/sec	acc=0.141211
+INFO:root:Epoch[4] Batch [6140]	Speed: 665.22 samples/sec	acc=0.129590
+INFO:root:Epoch[4] Batch [6160]	Speed: 664.82 samples/sec	acc=0.135156
+INFO:root:Epoch[4] Batch [6180]	Speed: 664.03 samples/sec	acc=0.124902
+INFO:root:Epoch[4] Batch [6200]	Speed: 664.97 samples/sec	acc=0.140527
+INFO:root:Epoch[4] Batch [6220]	Speed: 665.33 samples/sec	acc=0.133789
+INFO:root:Epoch[4] Batch [6240]	Speed: 663.84 samples/sec	acc=0.134180
+INFO:root:Epoch[4] Batch [6260]	Speed: 665.07 samples/sec	acc=0.137695
+INFO:root:Epoch[4] Batch [6280]	Speed: 664.55 samples/sec	acc=0.130664
+INFO:root:Epoch[4] Batch [6300]	Speed: 663.94 samples/sec	acc=0.133008
+INFO:root:Epoch[4] Batch [6320]	Speed: 664.43 samples/sec	acc=0.138379
+INFO:root:Epoch[4] Batch [6340]	Speed: 661.38 samples/sec	acc=0.132422
+INFO:root:Epoch[4] Batch [6360]	Speed: 683.76 samples/sec	acc=0.136035
+INFO:root:Epoch[4] Batch [6380]	Speed: 681.21 samples/sec	acc=0.136035
+INFO:root:Epoch[4] Batch [6400]	Speed: 664.48 samples/sec	acc=0.137402
+INFO:root:Epoch[4] Batch [6420]	Speed: 663.51 samples/sec	acc=0.135352
+INFO:root:Epoch[4] Batch [6440]	Speed: 664.72 samples/sec	acc=0.142187
+INFO:root:Epoch[4] Batch [6460]	Speed: 664.49 samples/sec	acc=0.135352
+INFO:root:Epoch[4] Batch [6480]	Speed: 664.50 samples/sec	acc=0.128027
+INFO:root:Epoch[4] Batch [6500]	Speed: 664.78 samples/sec	acc=0.133301
+INFO:root:Epoch[4] Batch [6520]	Speed: 664.70 samples/sec	acc=0.139355
+INFO:root:Epoch[4] Batch [6540]	Speed: 664.90 samples/sec	acc=0.139063
+INFO:root:Epoch[4] Batch [6560]	Speed: 663.38 samples/sec	acc=0.137500
+INFO:root:Epoch[4] Batch [6580]	Speed: 663.54 samples/sec	acc=0.140918
+INFO:root:Epoch[4] Batch [6600]	Speed: 664.60 samples/sec	acc=0.134766
+INFO:root:Epoch[4] Batch [6620]	Speed: 664.66 samples/sec	acc=0.135254
+INFO:root:Epoch[4] Batch [6640]	Speed: 663.58 samples/sec	acc=0.133008
+INFO:root:Epoch[4] Batch [6660]	Speed: 664.11 samples/sec	acc=0.138672
+INFO:root:Epoch[4] Batch [6680]	Speed: 664.11 samples/sec	acc=0.137109
+INFO:root:Epoch[4] Batch [6700]	Speed: 664.94 samples/sec	acc=0.131445
+INFO:root:Epoch[4] Batch [6720]	Speed: 664.73 samples/sec	acc=0.138086
+INFO:root:Epoch[4] Batch [6740]	Speed: 665.07 samples/sec	acc=0.137305
+INFO:root:Epoch[4] Batch [6760]	Speed: 663.99 samples/sec	acc=0.138965
+INFO:root:Epoch[4] Batch [6780]	Speed: 664.90 samples/sec	acc=0.134082
+INFO:root:Epoch[4] Batch [6800]	Speed: 664.52 samples/sec	acc=0.136035
+INFO:root:Epoch[4] Batch [6820]	Speed: 665.04 samples/sec	acc=0.140039
+INFO:root:Epoch[4] Batch [6840]	Speed: 664.37 samples/sec	acc=0.141699
+INFO:root:Epoch[4] Batch [6860]	Speed: 663.94 samples/sec	acc=0.134961
+INFO:root:Epoch[4] Batch [6880]	Speed: 664.10 samples/sec	acc=0.139063
+INFO:root:Epoch[4] Batch [6900]	Speed: 664.04 samples/sec	acc=0.132617
+INFO:root:Epoch[4] Batch [6920]	Speed: 664.35 samples/sec	acc=0.131445
+lr-batch-epoch: 0.1 6923 4
+INFO:root:Epoch[4] Batch [6940]	Speed: 663.91 samples/sec	acc=0.142090
+INFO:root:Epoch[4] Batch [6960]	Speed: 664.03 samples/sec	acc=0.140430
+INFO:root:Epoch[4] Batch [6980]	Speed: 664.41 samples/sec	acc=0.135840
+INFO:root:Epoch[4] Batch [7000]	Speed: 664.84 samples/sec	acc=0.134570
+INFO:root:Epoch[4] Batch [7020]	Speed: 664.88 samples/sec	acc=0.131152
+INFO:root:Epoch[4] Batch [7040]	Speed: 664.72 samples/sec	acc=0.136133
+INFO:root:Epoch[4] Batch [7060]	Speed: 664.12 samples/sec	acc=0.138770
+INFO:root:Epoch[4] Batch [7080]	Speed: 664.15 samples/sec	acc=0.138770
+INFO:root:Epoch[4] Batch [7100]	Speed: 664.44 samples/sec	acc=0.134180
+INFO:root:Epoch[4] Batch [7120]	Speed: 664.46 samples/sec	acc=0.140625
+INFO:root:Epoch[4] Batch [7140]	Speed: 672.42 samples/sec	acc=0.135547
+INFO:root:Epoch[4] Batch [7160]	Speed: 663.99 samples/sec	acc=0.140137
+INFO:root:Epoch[4] Batch [7180]	Speed: 664.67 samples/sec	acc=0.137109
+INFO:root:Epoch[4] Batch [7200]	Speed: 664.55 samples/sec	acc=0.143652
+INFO:root:Epoch[4] Batch [7220]	Speed: 664.70 samples/sec	acc=0.136133
+INFO:root:Epoch[4] Batch [7240]	Speed: 663.62 samples/sec	acc=0.137109
+INFO:root:Epoch[4] Batch [7260]	Speed: 664.10 samples/sec	acc=0.131934
+INFO:root:Epoch[4] Batch [7280]	Speed: 664.57 samples/sec	acc=0.135937
+INFO:root:Epoch[4] Batch [7300]	Speed: 664.42 samples/sec	acc=0.136719
+INFO:root:Epoch[4] Batch [7320]	Speed: 663.67 samples/sec	acc=0.136426
+INFO:root:Epoch[4] Batch [7340]	Speed: 665.36 samples/sec	acc=0.133398
+INFO:root:Epoch[4] Batch [7360]	Speed: 665.25 samples/sec	acc=0.135547
+INFO:root:Epoch[4] Batch [7380]	Speed: 663.89 samples/sec	acc=0.137109
+INFO:root:Epoch[4] Batch [7400]	Speed: 664.44 samples/sec	acc=0.133203
+INFO:root:Epoch[4] Batch [7420]	Speed: 664.29 samples/sec	acc=0.137793
+INFO:root:Epoch[4] Batch [7440]	Speed: 665.10 samples/sec	acc=0.140039
+INFO:root:Epoch[4] Batch [7460]	Speed: 664.15 samples/sec	acc=0.134961
+INFO:root:Epoch[4] Batch [7480]	Speed: 664.55 samples/sec	acc=0.142187
+INFO:root:Epoch[4] Batch [7500]	Speed: 663.73 samples/sec	acc=0.142773
+INFO:root:Epoch[4] Train-acc=0.134440
+INFO:root:Epoch[4] Time cost=6079.750
+call reset()
+INFO:root:Epoch[5] Batch [20]	Speed: 663.62 samples/sec	acc=0.166109
+INFO:root:Epoch[5] Batch [40]	Speed: 664.27 samples/sec	acc=0.161230
+INFO:root:Epoch[5] Batch [60]	Speed: 664.47 samples/sec	acc=0.165039
+INFO:root:Epoch[5] Batch [80]	Speed: 664.35 samples/sec	acc=0.163086
+INFO:root:Epoch[5] Batch [100]	Speed: 654.32 samples/sec	acc=0.162207
+INFO:root:Epoch[5] Batch [120]	Speed: 664.08 samples/sec	acc=0.152832
+INFO:root:Epoch[5] Batch [140]	Speed: 661.47 samples/sec	acc=0.155469
+INFO:root:Epoch[5] Batch [160]	Speed: 664.58 samples/sec	acc=0.154883
+INFO:root:Epoch[5] Batch [180]	Speed: 664.20 samples/sec	acc=0.154688
+INFO:root:Epoch[5] Batch [200]	Speed: 663.64 samples/sec	acc=0.158105
+INFO:root:Epoch[5] Batch [220]	Speed: 664.81 samples/sec	acc=0.145898
+INFO:root:Epoch[5] Batch [240]	Speed: 663.89 samples/sec	acc=0.151074
+INFO:root:Epoch[5] Batch [260]	Speed: 663.86 samples/sec	acc=0.147168
+INFO:root:Epoch[5] Batch [280]	Speed: 665.24 samples/sec	acc=0.149512
+INFO:root:Epoch[5] Batch [300]	Speed: 664.04 samples/sec	acc=0.141992
+INFO:root:Epoch[5] Batch [320]	Speed: 663.53 samples/sec	acc=0.149414
+INFO:root:Epoch[5] Batch [340]	Speed: 664.68 samples/sec	acc=0.140430
+INFO:root:Epoch[5] Batch [360]	Speed: 664.66 samples/sec	acc=0.146875
+INFO:root:Epoch[5] Batch [380]	Speed: 663.84 samples/sec	acc=0.141699
+INFO:root:Epoch[5] Batch [400]	Speed: 660.39 samples/sec	acc=0.145996
+[38000][MARGIN]0.674743,0.245232
+lr-batch-epoch: 0.1 404 5
+testing verification..
+(14000, 512)
+infer time 24.081838
+[cfp_ff][38000]XNorm: 20.832129
+[cfp_ff][38000]Accuracy-Flip: 0.99443+-0.00274
+testing verification..
+(14000, 512)
+infer time 20.217836
+[cfp_fp][38000]XNorm: 16.794967
+[cfp_fp][38000]Accuracy-Flip: 0.82929+-0.01178
+testing verification..
+(12000, 512)
+infer time 16.184396
+[agedb_30][38000]XNorm: 20.580146
+[agedb_30][38000]Accuracy-Flip: 0.94733+-0.01003
+testing verification..
+(12000, 512)
+infer time 19.499908
+[lfw][38000]XNorm: 20.892786
+[lfw][38000]Accuracy-Flip: 0.99283+-0.00358
+[38000]Accuracy-Highest: 0.99283
+INFO:root:Epoch[5] Batch [420]	Speed: 89.99 samples/sec	acc=0.141113
+INFO:root:Epoch[5] Batch [440]	Speed: 662.99 samples/sec	acc=0.140723
+INFO:root:Epoch[5] Batch [460]	Speed: 663.46 samples/sec	acc=0.144531
+INFO:root:Epoch[5] Batch [480]	Speed: 663.60 samples/sec	acc=0.141602
+INFO:root:Epoch[5] Batch [500]	Speed: 663.94 samples/sec	acc=0.151562
+INFO:root:Epoch[5] Batch [520]	Speed: 664.22 samples/sec	acc=0.143652
+INFO:root:Epoch[5] Batch [540]	Speed: 664.63 samples/sec	acc=0.149316
+INFO:root:Epoch[5] Batch [560]	Speed: 663.84 samples/sec	acc=0.141797
+INFO:root:Epoch[5] Batch [580]	Speed: 664.94 samples/sec	acc=0.143848
+INFO:root:Epoch[5] Batch [600]	Speed: 663.69 samples/sec	acc=0.138281
+INFO:root:Epoch[5] Batch [620]	Speed: 664.76 samples/sec	acc=0.139258
+INFO:root:Epoch[5] Batch [640]	Speed: 664.76 samples/sec	acc=0.145703
+INFO:root:Epoch[5] Batch [660]	Speed: 664.68 samples/sec	acc=0.140430
+INFO:root:Epoch[5] Batch [680]	Speed: 664.21 samples/sec	acc=0.142090
+INFO:root:Epoch[5] Batch [700]	Speed: 665.14 samples/sec	acc=0.143945
+INFO:root:Epoch[5] Batch [720]	Speed: 663.31 samples/sec	acc=0.138574
+INFO:root:Epoch[5] Batch [740]	Speed: 663.76 samples/sec	acc=0.144824
+INFO:root:Epoch[5] Batch [760]	Speed: 665.33 samples/sec	acc=0.149805
+INFO:root:Epoch[5] Batch [780]	Speed: 664.25 samples/sec	acc=0.141309
+INFO:root:Epoch[5] Batch [800]	Speed: 664.17 samples/sec	acc=0.134375
+INFO:root:Epoch[5] Batch [820]	Speed: 664.50 samples/sec	acc=0.138672
+INFO:root:Epoch[5] Batch [840]	Speed: 664.42 samples/sec	acc=0.148926
+INFO:root:Epoch[5] Batch [860]	Speed: 662.96 samples/sec	acc=0.145996
+INFO:root:Epoch[5] Batch [880]	Speed: 664.84 samples/sec	acc=0.138086
+INFO:root:Epoch[5] Batch [900]	Speed: 664.76 samples/sec	acc=0.137500
+INFO:root:Epoch[5] Batch [920]	Speed: 663.49 samples/sec	acc=0.136523
+INFO:root:Epoch[5] Batch [940]	Speed: 664.41 samples/sec	acc=0.138281
+INFO:root:Epoch[5] Batch [960]	Speed: 663.75 samples/sec	acc=0.134180
+INFO:root:Epoch[5] Batch [980]	Speed: 663.83 samples/sec	acc=0.138379
+INFO:root:Epoch[5] Batch [1000]	Speed: 664.62 samples/sec	acc=0.141309
+INFO:root:Epoch[5] Batch [1020]	Speed: 664.05 samples/sec	acc=0.142383
+INFO:root:Epoch[5] Batch [1040]	Speed: 664.07 samples/sec	acc=0.132422
+INFO:root:Epoch[5] Batch [1060]	Speed: 663.73 samples/sec	acc=0.149121
+INFO:root:Epoch[5] Batch [1080]	Speed: 664.80 samples/sec	acc=0.142969
+INFO:root:Epoch[5] Batch [1100]	Speed: 664.45 samples/sec	acc=0.142969
+INFO:root:Epoch[5] Batch [1120]	Speed: 664.61 samples/sec	acc=0.138965
+INFO:root:Epoch[5] Batch [1140]	Speed: 664.27 samples/sec	acc=0.136230
+INFO:root:Epoch[5] Batch [1160]	Speed: 663.53 samples/sec	acc=0.143652
+INFO:root:Epoch[5] Batch [1180]	Speed: 668.77 samples/sec	acc=0.140039
+INFO:root:Epoch[5] Batch [1200]	Speed: 668.72 samples/sec	acc=0.138281
+INFO:root:Epoch[5] Batch [1220]	Speed: 664.03 samples/sec	acc=0.147070
+INFO:root:Epoch[5] Batch [1240]	Speed: 665.04 samples/sec	acc=0.141016
+INFO:root:Epoch[5] Batch [1260]	Speed: 663.91 samples/sec	acc=0.135645
+INFO:root:Epoch[5] Batch [1280]	Speed: 663.42 samples/sec	acc=0.138574
+INFO:root:Epoch[5] Batch [1300]	Speed: 665.14 samples/sec	acc=0.142578
+INFO:root:Epoch[5] Batch [1320]	Speed: 664.59 samples/sec	acc=0.140430
+INFO:root:Epoch[5] Batch [1340]	Speed: 663.83 samples/sec	acc=0.138184
+INFO:root:Epoch[5] Batch [1360]	Speed: 663.44 samples/sec	acc=0.136914
+INFO:root:Epoch[5] Batch [1380]	Speed: 664.78 samples/sec	acc=0.137305
+INFO:root:Epoch[5] Batch [1400]	Speed: 663.30 samples/sec	acc=0.137695
+lr-batch-epoch: 0.1 1404 5
+INFO:root:Epoch[5] Batch [1420]	Speed: 664.21 samples/sec	acc=0.140332
+INFO:root:Epoch[5] Batch [1440]	Speed: 664.92 samples/sec	acc=0.135937
+INFO:root:Epoch[5] Batch [1460]	Speed: 663.71 samples/sec	acc=0.141504
+INFO:root:Epoch[5] Batch [1480]	Speed: 664.19 samples/sec	acc=0.138770
+INFO:root:Epoch[5] Batch [1500]	Speed: 664.02 samples/sec	acc=0.135645
+INFO:root:Epoch[5] Batch [1520]	Speed: 663.46 samples/sec	acc=0.145215
+INFO:root:Epoch[5] Batch [1540]	Speed: 664.59 samples/sec	acc=0.134180
+INFO:root:Epoch[5] Batch [1560]	Speed: 664.48 samples/sec	acc=0.144727
+INFO:root:Epoch[5] Batch [1580]	Speed: 662.96 samples/sec	acc=0.140039
+INFO:root:Epoch[5] Batch [1600]	Speed: 665.13 samples/sec	acc=0.132520
+INFO:root:Epoch[5] Batch [1620]	Speed: 664.44 samples/sec	acc=0.134277
+INFO:root:Epoch[5] Batch [1640]	Speed: 663.46 samples/sec	acc=0.131055
+INFO:root:Epoch[5] Batch [1660]	Speed: 664.37 samples/sec	acc=0.138184
+INFO:root:Epoch[5] Batch [1680]	Speed: 664.78 samples/sec	acc=0.140723
+INFO:root:Epoch[5] Batch [1700]	Speed: 663.94 samples/sec	acc=0.139063
+INFO:root:Epoch[5] Batch [1720]	Speed: 665.22 samples/sec	acc=0.143652
+INFO:root:Epoch[5] Batch [1740]	Speed: 663.99 samples/sec	acc=0.140137
+INFO:root:Epoch[5] Batch [1760]	Speed: 662.98 samples/sec	acc=0.134863
+INFO:root:Epoch[5] Batch [1780]	Speed: 664.87 samples/sec	acc=0.138770
+INFO:root:Epoch[5] Batch [1800]	Speed: 663.63 samples/sec	acc=0.140527
+INFO:root:Epoch[5] Batch [1820]	Speed: 663.02 samples/sec	acc=0.141113
+INFO:root:Epoch[5] Batch [1840]	Speed: 665.15 samples/sec	acc=0.136426
+INFO:root:Epoch[5] Batch [1860]	Speed: 663.57 samples/sec	acc=0.141016
+INFO:root:Epoch[5] Batch [1880]	Speed: 663.35 samples/sec	acc=0.136621
+INFO:root:Epoch[5] Batch [1900]	Speed: 664.57 samples/sec	acc=0.137988
+INFO:root:Epoch[5] Batch [1920]	Speed: 664.58 samples/sec	acc=0.139453
+INFO:root:Epoch[5] Batch [1940]	Speed: 663.83 samples/sec	acc=0.135937
+INFO:root:Epoch[5] Batch [1960]	Speed: 665.31 samples/sec	acc=0.133594
+INFO:root:Epoch[5] Batch [1980]	Speed: 658.59 samples/sec	acc=0.134961
+INFO:root:Epoch[5] Batch [2000]	Speed: 670.41 samples/sec	acc=0.137695
+INFO:root:Epoch[5] Batch [2020]	Speed: 664.72 samples/sec	acc=0.136621
+INFO:root:Epoch[5] Batch [2040]	Speed: 663.46 samples/sec	acc=0.137402
+INFO:root:Epoch[5] Batch [2060]	Speed: 663.82 samples/sec	acc=0.139453
+INFO:root:Epoch[5] Batch [2080]	Speed: 665.01 samples/sec	acc=0.142676
+INFO:root:Epoch[5] Batch [2100]	Speed: 663.95 samples/sec	acc=0.142285
+INFO:root:Epoch[5] Batch [2120]	Speed: 661.91 samples/sec	acc=0.140430
+INFO:root:Epoch[5] Batch [2140]	Speed: 664.59 samples/sec	acc=0.134961
+INFO:root:Epoch[5] Batch [2160]	Speed: 664.50 samples/sec	acc=0.141406
+INFO:root:Epoch[5] Batch [2180]	Speed: 664.63 samples/sec	acc=0.140234
+INFO:root:Epoch[5] Batch [2200]	Speed: 664.21 samples/sec	acc=0.138965
+INFO:root:Epoch[5] Batch [2220]	Speed: 664.70 samples/sec	acc=0.137207
+INFO:root:Epoch[5] Batch [2240]	Speed: 663.67 samples/sec	acc=0.140820
+INFO:root:Epoch[5] Batch [2260]	Speed: 665.07 samples/sec	acc=0.137793
+INFO:root:Epoch[5] Batch [2280]	Speed: 663.87 samples/sec	acc=0.136133
+INFO:root:Epoch[5] Batch [2300]	Speed: 662.99 samples/sec	acc=0.145117
+INFO:root:Epoch[5] Batch [2320]	Speed: 666.01 samples/sec	acc=0.136426
+INFO:root:Epoch[5] Batch [2340]	Speed: 664.30 samples/sec	acc=0.136621
+INFO:root:Epoch[5] Batch [2360]	Speed: 662.86 samples/sec	acc=0.138184
+INFO:root:Epoch[5] Batch [2380]	Speed: 664.91 samples/sec	acc=0.138965
+INFO:root:Epoch[5] Batch [2400]	Speed: 663.83 samples/sec	acc=0.142090
+[40000][MARGIN]0.655353,0.218351
+lr-batch-epoch: 0.1 2404 5
+testing verification..
+(14000, 512)
+infer time 17.79534
+[cfp_ff][40000]XNorm: 22.351214
+[cfp_ff][40000]Accuracy-Flip: 0.99414+-0.00370
+testing verification..
+(14000, 512)
+infer time 23.573196
+[cfp_fp][40000]XNorm: 18.387278
+[cfp_fp][40000]Accuracy-Flip: 0.83371+-0.01716
+testing verification..
+(12000, 512)
+infer time 18.960611
+[agedb_30][40000]XNorm: 21.789830
+[agedb_30][40000]Accuracy-Flip: 0.95050+-0.01088
+testing verification..
+(12000, 512)
+infer time 18.417199
+[lfw][40000]XNorm: 22.354413
+[lfw][40000]Accuracy-Flip: 0.99233+-0.00442
+[40000]Accuracy-Highest: 0.99283
+INFO:root:Epoch[5] Batch [2420]	Speed: 91.45 samples/sec	acc=0.135352
+INFO:root:Epoch[5] Batch [2440]	Speed: 663.59 samples/sec	acc=0.129004
+INFO:root:Epoch[5] Batch [2460]	Speed: 661.14 samples/sec	acc=0.136914
+INFO:root:Epoch[5] Batch [2480]	Speed: 663.48 samples/sec	acc=0.136328
+INFO:root:Epoch[5] Batch [2500]	Speed: 664.00 samples/sec	acc=0.143164
+INFO:root:Epoch[5] Batch [2520]	Speed: 663.39 samples/sec	acc=0.142773
+INFO:root:Epoch[5] Batch [2540]	Speed: 663.66 samples/sec	acc=0.137012
+INFO:root:Epoch[5] Batch [2560]	Speed: 664.64 samples/sec	acc=0.145703
+INFO:root:Epoch[5] Batch [2580]	Speed: 663.41 samples/sec	acc=0.141602
+INFO:root:Epoch[5] Batch [2600]	Speed: 663.93 samples/sec	acc=0.141895
+INFO:root:Epoch[5] Batch [2620]	Speed: 663.51 samples/sec	acc=0.138672
+INFO:root:Epoch[5] Batch [2640]	Speed: 664.87 samples/sec	acc=0.145410
+INFO:root:Epoch[5] Batch [2660]	Speed: 662.66 samples/sec	acc=0.139648
+INFO:root:Epoch[5] Batch [2680]	Speed: 664.42 samples/sec	acc=0.135742
+INFO:root:Epoch[5] Batch [2700]	Speed: 663.91 samples/sec	acc=0.139160
+INFO:root:Epoch[5] Batch [2720]	Speed: 663.33 samples/sec	acc=0.148047
+INFO:root:Epoch[5] Batch [2740]	Speed: 664.85 samples/sec	acc=0.139551
+INFO:root:Epoch[5] Batch [2760]	Speed: 663.97 samples/sec	acc=0.141309
+INFO:root:Epoch[5] Batch [2780]	Speed: 663.62 samples/sec	acc=0.139844
+INFO:root:Epoch[5] Batch [2800]	Speed: 667.05 samples/sec	acc=0.133008
+INFO:root:Epoch[5] Batch [2820]	Speed: 679.24 samples/sec	acc=0.140625
+INFO:root:Epoch[5] Batch [2840]	Speed: 677.24 samples/sec	acc=0.144824
+INFO:root:Epoch[5] Batch [2860]	Speed: 678.42 samples/sec	acc=0.143555
+INFO:root:Epoch[5] Batch [2880]	Speed: 672.38 samples/sec	acc=0.139844
+INFO:root:Epoch[5] Batch [2900]	Speed: 663.84 samples/sec	acc=0.140332
+INFO:root:Epoch[5] Batch [2920]	Speed: 665.11 samples/sec	acc=0.134082
+INFO:root:Epoch[5] Batch [2940]	Speed: 664.44 samples/sec	acc=0.140039
+INFO:root:Epoch[5] Batch [2960]	Speed: 664.03 samples/sec	acc=0.140625
+INFO:root:Epoch[5] Batch [2980]	Speed: 664.50 samples/sec	acc=0.143750
+INFO:root:Epoch[5] Batch [3000]	Speed: 663.73 samples/sec	acc=0.139355
+INFO:root:Epoch[5] Batch [3020]	Speed: 663.63 samples/sec	acc=0.134277
+INFO:root:Epoch[5] Batch [3040]	Speed: 663.70 samples/sec	acc=0.133691
+INFO:root:Epoch[5] Batch [3060]	Speed: 664.19 samples/sec	acc=0.135742
+INFO:root:Epoch[5] Batch [3080]	Speed: 664.15 samples/sec	acc=0.144629
+INFO:root:Epoch[5] Batch [3100]	Speed: 665.04 samples/sec	acc=0.133984
+INFO:root:Epoch[5] Batch [3120]	Speed: 663.98 samples/sec	acc=0.142969
+INFO:root:Epoch[5] Batch [3140]	Speed: 663.58 samples/sec	acc=0.137402
+INFO:root:Epoch[5] Batch [3160]	Speed: 664.99 samples/sec	acc=0.143848
+INFO:root:Epoch[5] Batch [3180]	Speed: 663.77 samples/sec	acc=0.134570
+INFO:root:Epoch[5] Batch [3200]	Speed: 663.87 samples/sec	acc=0.138477
+INFO:root:Epoch[5] Batch [3220]	Speed: 664.80 samples/sec	acc=0.146289
+INFO:root:Epoch[5] Batch [3240]	Speed: 664.14 samples/sec	acc=0.135449
+INFO:root:Epoch[5] Batch [3260]	Speed: 662.59 samples/sec	acc=0.138867
+INFO:root:Epoch[5] Batch [3280]	Speed: 665.15 samples/sec	acc=0.138574
+INFO:root:Epoch[5] Batch [3300]	Speed: 664.25 samples/sec	acc=0.136914
+INFO:root:Epoch[5] Batch [3320]	Speed: 663.58 samples/sec	acc=0.135059
+INFO:root:Epoch[5] Batch [3340]	Speed: 663.94 samples/sec	acc=0.138379
+INFO:root:Epoch[5] Batch [3360]	Speed: 664.14 samples/sec	acc=0.140820
+INFO:root:Epoch[5] Batch [3380]	Speed: 662.51 samples/sec	acc=0.139258
+INFO:root:Epoch[5] Batch [3400]	Speed: 664.20 samples/sec	acc=0.137109
+lr-batch-epoch: 0.1 3404 5
+INFO:root:Epoch[5] Batch [3420]	Speed: 663.72 samples/sec	acc=0.138965
+INFO:root:Epoch[5] Batch [3440]	Speed: 663.61 samples/sec	acc=0.143359
+INFO:root:Epoch[5] Batch [3460]	Speed: 665.30 samples/sec	acc=0.142383
+INFO:root:Epoch[5] Batch [3480]	Speed: 664.14 samples/sec	acc=0.144043
+INFO:root:Epoch[5] Batch [3500]	Speed: 663.99 samples/sec	acc=0.143262
+INFO:root:Epoch[5] Batch [3520]	Speed: 664.94 samples/sec	acc=0.139355
+INFO:root:Epoch[5] Batch [3540]	Speed: 664.22 samples/sec	acc=0.145508
+INFO:root:Epoch[5] Batch [3560]	Speed: 663.89 samples/sec	acc=0.138965
+INFO:root:Epoch[5] Batch [3580]	Speed: 665.22 samples/sec	acc=0.137598
+INFO:root:Epoch[5] Batch [3600]	Speed: 664.23 samples/sec	acc=0.141895
+INFO:root:Epoch[5] Batch [3620]	Speed: 664.12 samples/sec	acc=0.137305
+INFO:root:Epoch[5] Batch [3640]	Speed: 664.70 samples/sec	acc=0.135937
+INFO:root:Epoch[5] Batch [3660]	Speed: 664.32 samples/sec	acc=0.134277
+INFO:root:Epoch[5] Batch [3680]	Speed: 663.33 samples/sec	acc=0.139941
+INFO:root:Epoch[5] Batch [3700]	Speed: 664.60 samples/sec	acc=0.140332
+INFO:root:Epoch[5] Batch [3720]	Speed: 664.25 samples/sec	acc=0.137305
+INFO:root:Epoch[5] Batch [3740]	Speed: 663.31 samples/sec	acc=0.141113
+INFO:root:Epoch[5] Batch [3760]	Speed: 664.45 samples/sec	acc=0.139063
+INFO:root:Epoch[5] Batch [3780]	Speed: 663.92 samples/sec	acc=0.140723
+INFO:root:Epoch[5] Batch [3800]	Speed: 664.14 samples/sec	acc=0.141113
+INFO:root:Epoch[5] Batch [3820]	Speed: 665.16 samples/sec	acc=0.144336
+INFO:root:Epoch[5] Batch [3840]	Speed: 663.53 samples/sec	acc=0.136621
+INFO:root:Epoch[5] Batch [3860]	Speed: 663.64 samples/sec	acc=0.141895
+INFO:root:Epoch[5] Batch [3880]	Speed: 664.70 samples/sec	acc=0.142383
+INFO:root:Epoch[5] Batch [3900]	Speed: 663.63 samples/sec	acc=0.145313
+INFO:root:Epoch[5] Batch [3920]	Speed: 663.99 samples/sec	acc=0.136426
+INFO:root:Epoch[5] Batch [3940]	Speed: 665.28 samples/sec	acc=0.139160
+INFO:root:Epoch[5] Batch [3960]	Speed: 664.01 samples/sec	acc=0.143652
+INFO:root:Epoch[5] Batch [3980]	Speed: 663.47 samples/sec	acc=0.139648
+INFO:root:Epoch[5] Batch [4000]	Speed: 664.92 samples/sec	acc=0.136426
+INFO:root:Epoch[5] Batch [4020]	Speed: 664.63 samples/sec	acc=0.138281
+INFO:root:Epoch[5] Batch [4040]	Speed: 663.66 samples/sec	acc=0.137891
+INFO:root:Epoch[5] Batch [4060]	Speed: 665.09 samples/sec	acc=0.141113
+INFO:root:Epoch[5] Batch [4080]	Speed: 664.93 samples/sec	acc=0.141797
+INFO:root:Epoch[5] Batch [4100]	Speed: 664.06 samples/sec	acc=0.133789
+INFO:root:Epoch[5] Batch [4120]	Speed: 665.31 samples/sec	acc=0.142187
+INFO:root:Epoch[5] Batch [4140]	Speed: 664.33 samples/sec	acc=0.137500
+INFO:root:Epoch[5] Batch [4160]	Speed: 663.90 samples/sec	acc=0.144141
+INFO:root:Epoch[5] Batch [4180]	Speed: 663.52 samples/sec	acc=0.146484
+INFO:root:Epoch[5] Batch [4200]	Speed: 663.22 samples/sec	acc=0.142480
+INFO:root:Epoch[5] Batch [4220]	Speed: 664.41 samples/sec	acc=0.140332
+INFO:root:Epoch[5] Batch [4240]	Speed: 664.94 samples/sec	acc=0.143164
+INFO:root:Epoch[5] Batch [4260]	Speed: 664.87 samples/sec	acc=0.138086
+INFO:root:Epoch[5] Batch [4280]	Speed: 663.42 samples/sec	acc=0.143555
+INFO:root:Epoch[5] Batch [4300]	Speed: 663.76 samples/sec	acc=0.140039
+INFO:root:Epoch[5] Batch [4320]	Speed: 664.04 samples/sec	acc=0.143555
+INFO:root:Epoch[5] Batch [4340]	Speed: 663.19 samples/sec	acc=0.136133
+INFO:root:Epoch[5] Batch [4360]	Speed: 664.42 samples/sec	acc=0.144141
+INFO:root:Epoch[5] Batch [4380]	Speed: 674.18 samples/sec	acc=0.139648
+INFO:root:Epoch[5] Batch [4400]	Speed: 663.61 samples/sec	acc=0.137402
+[42000][MARGIN]0.649567,0.212691
+lr-batch-epoch: 0.1 4404 5
+testing verification..
+(14000, 512)
+infer time 20.639069
+[cfp_ff][42000]XNorm: 21.065195
+[cfp_ff][42000]Accuracy-Flip: 0.99371+-0.00294
+testing verification..
+(14000, 512)
+infer time 23.804205
+[cfp_fp][42000]XNorm: 17.375979
+[cfp_fp][42000]Accuracy-Flip: 0.83529+-0.01796
+testing verification..
+(12000, 512)
+infer time 17.203058
+[agedb_30][42000]XNorm: 21.139623
+[agedb_30][42000]Accuracy-Flip: 0.94650+-0.01277
+testing verification..
+(12000, 512)
+infer time 18.317502
+[lfw][42000]XNorm: 21.489225
+[lfw][42000]Accuracy-Flip: 0.99317+-0.00337
+saving 21
+INFO:root:Saved checkpoint to "../model-r50-am-lfw/model-0021.params"
+[42000]Accuracy-Highest: 0.99317
+INFO:root:Epoch[5] Batch [4420]	Speed: 87.85 samples/sec	acc=0.141309
+INFO:root:Epoch[5] Batch [4440]	Speed: 663.33 samples/sec	acc=0.143262
+INFO:root:Epoch[5] Batch [4460]	Speed: 661.54 samples/sec	acc=0.141211
+INFO:root:Epoch[5] Batch [4480]	Speed: 649.46 samples/sec	acc=0.143555
+INFO:root:Epoch[5] Batch [4500]	Speed: 662.12 samples/sec	acc=0.141406
+INFO:root:Epoch[5] Batch [4520]	Speed: 661.52 samples/sec	acc=0.145703
+INFO:root:Epoch[5] Batch [4540]	Speed: 664.12 samples/sec	acc=0.131250
+INFO:root:Epoch[5] Batch [4560]	Speed: 663.76 samples/sec	acc=0.140820
+INFO:root:Epoch[5] Batch [4580]	Speed: 663.64 samples/sec	acc=0.141504
+INFO:root:Epoch[5] Batch [4600]	Speed: 655.64 samples/sec	acc=0.143457
+INFO:root:Epoch[5] Batch [4620]	Speed: 664.14 samples/sec	acc=0.143945
+INFO:root:Epoch[5] Batch [4640]	Speed: 663.61 samples/sec	acc=0.133496
+INFO:root:Epoch[5] Batch [4660]	Speed: 664.78 samples/sec	acc=0.145410
+INFO:root:Epoch[5] Batch [4680]	Speed: 664.46 samples/sec	acc=0.136426
+INFO:root:Epoch[5] Batch [4700]	Speed: 664.31 samples/sec	acc=0.141016
+INFO:root:Epoch[5] Batch [4720]	Speed: 664.74 samples/sec	acc=0.137500
+INFO:root:Epoch[5] Batch [4740]	Speed: 663.75 samples/sec	acc=0.140918
+INFO:root:Epoch[5] Batch [4760]	Speed: 662.78 samples/sec	acc=0.142383
+INFO:root:Epoch[5] Batch [4780]	Speed: 664.63 samples/sec	acc=0.139648
+INFO:root:Epoch[5] Batch [4800]	Speed: 664.86 samples/sec	acc=0.137793
+INFO:root:Epoch[5] Batch [4820]	Speed: 663.84 samples/sec	acc=0.139355
+INFO:root:Epoch[5] Batch [4840]	Speed: 664.22 samples/sec	acc=0.132129
+INFO:root:Epoch[5] Batch [4860]	Speed: 662.86 samples/sec	acc=0.138867
+INFO:root:Epoch[5] Batch [4880]	Speed: 664.17 samples/sec	acc=0.143652
+INFO:root:Epoch[5] Batch [4900]	Speed: 664.26 samples/sec	acc=0.138086
+INFO:root:Epoch[5] Batch [4920]	Speed: 664.34 samples/sec	acc=0.141895
+INFO:root:Epoch[5] Batch [4940]	Speed: 662.41 samples/sec	acc=0.140137
+INFO:root:Epoch[5] Batch [4960]	Speed: 663.66 samples/sec	acc=0.143945
+INFO:root:Epoch[5] Batch [4980]	Speed: 664.03 samples/sec	acc=0.136719
+INFO:root:Epoch[5] Batch [5000]	Speed: 663.58 samples/sec	acc=0.140039
+INFO:root:Epoch[5] Batch [5020]	Speed: 664.20 samples/sec	acc=0.139648
+INFO:root:Epoch[5] Batch [5040]	Speed: 664.79 samples/sec	acc=0.137598
+INFO:root:Epoch[5] Batch [5060]	Speed: 663.25 samples/sec	acc=0.141309
+INFO:root:Epoch[5] Batch [5080]	Speed: 664.25 samples/sec	acc=0.142871
+INFO:root:Epoch[5] Batch [5100]	Speed: 665.24 samples/sec	acc=0.142773
+INFO:root:Epoch[5] Batch [5120]	Speed: 664.04 samples/sec	acc=0.146973
+INFO:root:Epoch[5] Batch [5140]	Speed: 664.69 samples/sec	acc=0.136914
+INFO:root:Epoch[5] Batch [5160]	Speed: 663.48 samples/sec	acc=0.141113
+INFO:root:Epoch[5] Batch [5180]	Speed: 662.54 samples/sec	acc=0.139844
+INFO:root:Epoch[5] Batch [5200]	Speed: 664.93 samples/sec	acc=0.143652
+INFO:root:Epoch[5] Batch [5220]	Speed: 664.31 samples/sec	acc=0.138965
+INFO:root:Epoch[5] Batch [5240]	Speed: 663.49 samples/sec	acc=0.137012
+INFO:root:Epoch[5] Batch [5260]	Speed: 664.44 samples/sec	acc=0.136914
+INFO:root:Epoch[5] Batch [5280]	Speed: 664.07 samples/sec	acc=0.138184
+INFO:root:Epoch[5] Batch [5300]	Speed: 663.11 samples/sec	acc=0.144043
+INFO:root:Epoch[5] Batch [5320]	Speed: 664.79 samples/sec	acc=0.143945
+INFO:root:Epoch[5] Batch [5340]	Speed: 663.83 samples/sec	acc=0.140430
+INFO:root:Epoch[5] Batch [5360]	Speed: 663.35 samples/sec	acc=0.137695
+INFO:root:Epoch[5] Batch [5380]	Speed: 664.62 samples/sec	acc=0.136035
+INFO:root:Epoch[5] Batch [5400]	Speed: 664.29 samples/sec	acc=0.141211
+lr-batch-epoch: 0.1 5404 5
+INFO:root:Epoch[5] Batch [5420]	Speed: 662.08 samples/sec	acc=0.139941
+INFO:root:Epoch[5] Batch [5440]	Speed: 665.13 samples/sec	acc=0.138770
+INFO:root:Epoch[5] Batch [5460]	Speed: 664.00 samples/sec	acc=0.137500
+INFO:root:Epoch[5] Batch [5480]	Speed: 663.09 samples/sec	acc=0.139258
+INFO:root:Epoch[5] Batch [5500]	Speed: 664.19 samples/sec	acc=0.138086
+INFO:root:Epoch[5] Batch [5520]	Speed: 663.79 samples/sec	acc=0.142187
+INFO:root:Epoch[5] Batch [5540]	Speed: 662.85 samples/sec	acc=0.143652
+INFO:root:Epoch[5] Batch [5560]	Speed: 652.96 samples/sec	acc=0.141406
+INFO:root:Epoch[5] Batch [5580]	Speed: 665.66 samples/sec	acc=0.139844
+INFO:root:Epoch[5] Batch [5600]	Speed: 663.40 samples/sec	acc=0.140527
+INFO:root:Epoch[5] Batch [5620]	Speed: 664.40 samples/sec	acc=0.140332
+INFO:root:Epoch[5] Batch [5640]	Speed: 664.85 samples/sec	acc=0.141016
+INFO:root:Epoch[5] Batch [5660]	Speed: 664.22 samples/sec	acc=0.142187
+INFO:root:Epoch[5] Batch [5680]	Speed: 664.72 samples/sec	acc=0.139551
+INFO:root:Epoch[5] Batch [5700]	Speed: 664.22 samples/sec	acc=0.142285
+INFO:root:Epoch[5] Batch [5720]	Speed: 663.39 samples/sec	acc=0.141797
+INFO:root:Epoch[5] Batch [5740]	Speed: 664.20 samples/sec	acc=0.143848
+INFO:root:Epoch[5] Batch [5760]	Speed: 662.64 samples/sec	acc=0.149609
+INFO:root:Epoch[5] Batch [5780]	Speed: 664.10 samples/sec	acc=0.140430
+INFO:root:Epoch[5] Batch [5800]	Speed: 664.61 samples/sec	acc=0.141016
+INFO:root:Epoch[5] Batch [5820]	Speed: 663.71 samples/sec	acc=0.142383
+INFO:root:Epoch[5] Batch [5840]	Speed: 662.78 samples/sec	acc=0.139355
+INFO:root:Epoch[5] Batch [5860]	Speed: 664.34 samples/sec	acc=0.141406
+INFO:root:Epoch[5] Batch [5880]	Speed: 663.50 samples/sec	acc=0.139941
+INFO:root:Epoch[5] Batch [5900]	Speed: 668.71 samples/sec	acc=0.139551
+INFO:root:Epoch[5] Batch [5920]	Speed: 663.03 samples/sec	acc=0.143750
+INFO:root:Epoch[5] Batch [5940]	Speed: 663.94 samples/sec	acc=0.133789
+INFO:root:Epoch[5] Batch [5960]	Speed: 663.10 samples/sec	acc=0.143164
+INFO:root:Epoch[5] Batch [5980]	Speed: 665.03 samples/sec	acc=0.139551
+INFO:root:Epoch[5] Batch [6000]	Speed: 663.63 samples/sec	acc=0.141406
+INFO:root:Epoch[5] Batch [6020]	Speed: 662.55 samples/sec	acc=0.140234
+INFO:root:Epoch[5] Batch [6040]	Speed: 664.72 samples/sec	acc=0.138770
+INFO:root:Epoch[5] Batch [6060]	Speed: 663.19 samples/sec	acc=0.142383
+INFO:root:Epoch[5] Batch [6080]	Speed: 663.71 samples/sec	acc=0.139453
+INFO:root:Epoch[5] Batch [6100]	Speed: 664.06 samples/sec	acc=0.137500
+INFO:root:Epoch[5] Batch [6120]	Speed: 663.20 samples/sec	acc=0.139453
+INFO:root:Epoch[5] Batch [6140]	Speed: 664.21 samples/sec	acc=0.135742
+INFO:root:Epoch[5] Batch [6160]	Speed: 664.02 samples/sec	acc=0.140723
+INFO:root:Epoch[5] Batch [6180]	Speed: 664.25 samples/sec	acc=0.146582
+INFO:root:Epoch[5] Batch [6200]	Speed: 662.77 samples/sec	acc=0.145703
+INFO:root:Epoch[5] Batch [6220]	Speed: 664.51 samples/sec	acc=0.139160
+INFO:root:Epoch[5] Batch [6240]	Speed: 664.71 samples/sec	acc=0.143555
+INFO:root:Epoch[5] Batch [6260]	Speed: 657.63 samples/sec	acc=0.144336
+INFO:root:Epoch[5] Batch [6280]	Speed: 671.55 samples/sec	acc=0.137207
+INFO:root:Epoch[5] Batch [6300]	Speed: 663.84 samples/sec	acc=0.142578
+INFO:root:Epoch[5] Batch [6320]	Speed: 662.12 samples/sec	acc=0.145801
+INFO:root:Epoch[5] Batch [6340]	Speed: 664.68 samples/sec	acc=0.136523
+INFO:root:Epoch[5] Batch [6360]	Speed: 663.53 samples/sec	acc=0.147266
+INFO:root:Epoch[5] Batch [6380]	Speed: 663.06 samples/sec	acc=0.140430
+INFO:root:Epoch[5] Batch [6400]	Speed: 664.70 samples/sec	acc=0.142773
+[44000][MARGIN]0.647738,0.209751
+lr-batch-epoch: 0.1 6404 5
+testing verification..
+(14000, 512)
+infer time 19.897472
+[cfp_ff][44000]XNorm: 21.254379
+[cfp_ff][44000]Accuracy-Flip: 0.99471+-0.00362
+testing verification..
+(14000, 512)
+infer time 21.493764
+[cfp_fp][44000]XNorm: 16.929554
+[cfp_fp][44000]Accuracy-Flip: 0.84329+-0.01483
+testing verification..
+(12000, 512)
+infer time 17.169521
+[agedb_30][44000]XNorm: 21.279628
+[agedb_30][44000]Accuracy-Flip: 0.94500+-0.00771
+testing verification..
+(12000, 512)
+infer time 20.694546
+[lfw][44000]XNorm: 21.324684
+[lfw][44000]Accuracy-Flip: 0.99350+-0.00462
+saving 22
+INFO:root:Saved checkpoint to "../model-r50-am-lfw/model-0022.params"
+[44000]Accuracy-Highest: 0.99350
+INFO:root:Epoch[5] Batch [6420]	Speed: 86.97 samples/sec	acc=0.145996
+INFO:root:Epoch[5] Batch [6440]	Speed: 662.84 samples/sec	acc=0.137109
+INFO:root:Epoch[5] Batch [6460]	Speed: 664.30 samples/sec	acc=0.147070
+INFO:root:Epoch[5] Batch [6480]	Speed: 664.11 samples/sec	acc=0.142676
+INFO:root:Epoch[5] Batch [6500]	Speed: 663.76 samples/sec	acc=0.141602
+INFO:root:Epoch[5] Batch [6520]	Speed: 663.39 samples/sec	acc=0.141504
+INFO:root:Epoch[5] Batch [6540]	Speed: 663.41 samples/sec	acc=0.140430
+INFO:root:Epoch[5] Batch [6560]	Speed: 662.88 samples/sec	acc=0.145996
+INFO:root:Epoch[5] Batch [6580]	Speed: 644.82 samples/sec	acc=0.140332
+INFO:root:Epoch[5] Batch [6600]	Speed: 664.77 samples/sec	acc=0.143359
+INFO:root:Epoch[5] Batch [6620]	Speed: 664.02 samples/sec	acc=0.142090
+INFO:root:Epoch[5] Batch [6640]	Speed: 665.13 samples/sec	acc=0.143652
+INFO:root:Epoch[5] Batch [6660]	Speed: 663.57 samples/sec	acc=0.141406
+INFO:root:Epoch[5] Batch [6680]	Speed: 663.94 samples/sec	acc=0.141113
+INFO:root:Epoch[5] Batch [6700]	Speed: 662.27 samples/sec	acc=0.138867
+INFO:root:Epoch[5] Batch [6720]	Speed: 664.19 samples/sec	acc=0.141016
+INFO:root:Epoch[5] Batch [6740]	Speed: 664.16 samples/sec	acc=0.148438
+INFO:root:Epoch[5] Batch [6760]	Speed: 664.51 samples/sec	acc=0.135937
+INFO:root:Epoch[5] Batch [6780]	Speed: 664.74 samples/sec	acc=0.139648
+INFO:root:Epoch[5] Batch [6800]	Speed: 664.05 samples/sec	acc=0.138965
+INFO:root:Epoch[5] Batch [6820]	Speed: 664.56 samples/sec	acc=0.142676
+INFO:root:Epoch[5] Batch [6840]	Speed: 663.77 samples/sec	acc=0.140234
+INFO:root:Epoch[5] Batch [6860]	Speed: 663.26 samples/sec	acc=0.139453
+INFO:root:Epoch[5] Batch [6880]	Speed: 664.40 samples/sec	acc=0.136719
+INFO:root:Epoch[5] Batch [6900]	Speed: 664.28 samples/sec	acc=0.139551
+INFO:root:Epoch[5] Batch [6920]	Speed: 664.14 samples/sec	acc=0.141211
+INFO:root:Epoch[5] Batch [6940]	Speed: 664.51 samples/sec	acc=0.138770
+INFO:root:Epoch[5] Batch [6960]	Speed: 663.96 samples/sec	acc=0.136914
+INFO:root:Epoch[5] Batch [6980]	Speed: 663.14 samples/sec	acc=0.134375
+INFO:root:Epoch[5] Batch [7000]	Speed: 664.34 samples/sec	acc=0.142480
+INFO:root:Epoch[5] Batch [7020]	Speed: 664.31 samples/sec	acc=0.143457
+INFO:root:Epoch[5] Batch [7040]	Speed: 663.75 samples/sec	acc=0.145703
+INFO:root:Epoch[5] Batch [7060]	Speed: 664.19 samples/sec	acc=0.140234
+INFO:root:Epoch[5] Batch [7080]	Speed: 664.91 samples/sec	acc=0.143652
+INFO:root:Epoch[5] Batch [7100]	Speed: 663.76 samples/sec	acc=0.146484
+INFO:root:Epoch[5] Batch [7120]	Speed: 664.68 samples/sec	acc=0.145605
+INFO:root:Epoch[5] Batch [7140]	Speed: 664.94 samples/sec	acc=0.142480
+INFO:root:Epoch[5] Batch [7160]	Speed: 662.88 samples/sec	acc=0.144141
+INFO:root:Epoch[5] Batch [7180]	Speed: 664.68 samples/sec	acc=0.145508
+INFO:root:Epoch[5] Batch [7200]	Speed: 663.77 samples/sec	acc=0.135156
+INFO:root:Epoch[5] Batch [7220]	Speed: 663.80 samples/sec	acc=0.144238
+INFO:root:Epoch[5] Batch [7240]	Speed: 663.88 samples/sec	acc=0.143652
+INFO:root:Epoch[5] Batch [7260]	Speed: 663.73 samples/sec	acc=0.146484
+INFO:root:Epoch[5] Batch [7280]	Speed: 664.01 samples/sec	acc=0.145605
+INFO:root:Epoch[5] Batch [7300]	Speed: 664.18 samples/sec	acc=0.147070
+INFO:root:Epoch[5] Batch [7320]	Speed: 663.87 samples/sec	acc=0.141699
+INFO:root:Epoch[5] Batch [7340]	Speed: 663.77 samples/sec	acc=0.139258
+INFO:root:Epoch[5] Batch [7360]	Speed: 665.46 samples/sec	acc=0.145313
+INFO:root:Epoch[5] Batch [7380]	Speed: 664.04 samples/sec	acc=0.141602
+INFO:root:Epoch[5] Batch [7400]	Speed: 663.55 samples/sec	acc=0.145703
+lr-batch-epoch: 0.1 7404 5
+INFO:root:Epoch[5] Batch [7420]	Speed: 664.02 samples/sec	acc=0.146680
+INFO:root:Epoch[5] Batch [7440]	Speed: 663.70 samples/sec	acc=0.147559
+INFO:root:Epoch[5] Batch [7460]	Speed: 662.97 samples/sec	acc=0.142285
+INFO:root:Epoch[5] Batch [7480]	Speed: 664.53 samples/sec	acc=0.142285
+INFO:root:Epoch[5] Batch [7500]	Speed: 663.38 samples/sec	acc=0.140332
+INFO:root:Epoch[5] Train-acc=0.142036
+INFO:root:Epoch[5] Time cost=6196.751
+call reset()
+INFO:root:Epoch[6] Batch [20]	Speed: 662.29 samples/sec	acc=0.174014
+INFO:root:Epoch[6] Batch [40]	Speed: 662.44 samples/sec	acc=0.167969
+INFO:root:Epoch[6] Batch [60]	Speed: 663.39 samples/sec	acc=0.163770
+INFO:root:Epoch[6] Batch [80]	Speed: 663.85 samples/sec	acc=0.168359
+INFO:root:Epoch[6] Batch [100]	Speed: 664.30 samples/sec	acc=0.159766
+INFO:root:Epoch[6] Batch [120]	Speed: 663.88 samples/sec	acc=0.160449
+INFO:root:Epoch[6] Batch [140]	Speed: 663.93 samples/sec	acc=0.161816
+INFO:root:Epoch[6] Batch [160]	Speed: 664.02 samples/sec	acc=0.168750
+INFO:root:Epoch[6] Batch [180]	Speed: 663.41 samples/sec	acc=0.160059
+INFO:root:Epoch[6] Batch [200]	Speed: 663.83 samples/sec	acc=0.158496
+INFO:root:Epoch[6] Batch [220]	Speed: 663.52 samples/sec	acc=0.154590
+INFO:root:Epoch[6] Batch [240]	Speed: 670.53 samples/sec	acc=0.160840
+INFO:root:Epoch[6] Batch [260]	Speed: 663.52 samples/sec	acc=0.155371
+INFO:root:Epoch[6] Batch [280]	Speed: 664.40 samples/sec	acc=0.156250
+INFO:root:Epoch[6] Batch [300]	Speed: 664.41 samples/sec	acc=0.145996
+INFO:root:Epoch[6] Batch [320]	Speed: 664.04 samples/sec	acc=0.154004
+INFO:root:Epoch[6] Batch [340]	Speed: 663.87 samples/sec	acc=0.155371
+INFO:root:Epoch[6] Batch [360]	Speed: 663.85 samples/sec	acc=0.153906
+INFO:root:Epoch[6] Batch [380]	Speed: 664.61 samples/sec	acc=0.153418
+INFO:root:Epoch[6] Batch [400]	Speed: 664.03 samples/sec	acc=0.148242
+INFO:root:Epoch[6] Batch [420]	Speed: 664.00 samples/sec	acc=0.150488
+INFO:root:Epoch[6] Batch [440]	Speed: 663.86 samples/sec	acc=0.148633
+INFO:root:Epoch[6] Batch [460]	Speed: 664.19 samples/sec	acc=0.146484
+INFO:root:Epoch[6] Batch [480]	Speed: 663.86 samples/sec	acc=0.151562
+INFO:root:Epoch[6] Batch [500]	Speed: 664.20 samples/sec	acc=0.144922
+INFO:root:Epoch[6] Batch [520]	Speed: 663.98 samples/sec	acc=0.146680
+INFO:root:Epoch[6] Batch [540]	Speed: 664.51 samples/sec	acc=0.145898
+INFO:root:Epoch[6] Batch [560]	Speed: 663.90 samples/sec	acc=0.152539
+INFO:root:Epoch[6] Batch [580]	Speed: 663.67 samples/sec	acc=0.146777
+INFO:root:Epoch[6] Batch [600]	Speed: 663.76 samples/sec	acc=0.151953
+INFO:root:Epoch[6] Batch [620]	Speed: 664.55 samples/sec	acc=0.149121
+INFO:root:Epoch[6] Batch [640]	Speed: 663.42 samples/sec	acc=0.134570
+INFO:root:Epoch[6] Batch [660]	Speed: 662.98 samples/sec	acc=0.141992
+INFO:root:Epoch[6] Batch [680]	Speed: 663.58 samples/sec	acc=0.143164
+INFO:root:Epoch[6] Batch [700]	Speed: 664.35 samples/sec	acc=0.146582
+INFO:root:Epoch[6] Batch [720]	Speed: 664.09 samples/sec	acc=0.145801
+INFO:root:Epoch[6] Batch [740]	Speed: 663.95 samples/sec	acc=0.148535
+INFO:root:Epoch[6] Batch [760]	Speed: 663.34 samples/sec	acc=0.145703
+INFO:root:Epoch[6] Batch [780]	Speed: 663.74 samples/sec	acc=0.148438
+INFO:root:Epoch[6] Batch [800]	Speed: 664.31 samples/sec	acc=0.148145
+INFO:root:Epoch[6] Batch [820]	Speed: 664.28 samples/sec	acc=0.146875
+INFO:root:Epoch[6] Batch [840]	Speed: 663.79 samples/sec	acc=0.141797
+INFO:root:Epoch[6] Batch [860]	Speed: 663.27 samples/sec	acc=0.141797
+INFO:root:Epoch[6] Batch [880]	Speed: 664.06 samples/sec	acc=0.141602
+[46000][MARGIN]0.642032,0.204640
+lr-batch-epoch: 0.1 885 6
+testing verification..
+(14000, 512)
+infer time 19.069164
+[cfp_ff][46000]XNorm: 21.353451
+[cfp_ff][46000]Accuracy-Flip: 0.99457+-0.00343
+testing verification..
+(14000, 512)
+infer time 22.772816
+[cfp_fp][46000]XNorm: 17.712679
+[cfp_fp][46000]Accuracy-Flip: 0.85229+-0.02026
+testing verification..
+(12000, 512)
+infer time 18.130649
+[agedb_30][46000]XNorm: 21.008399
+[agedb_30][46000]Accuracy-Flip: 0.94267+-0.01158
+testing verification..
+(12000, 512)
+infer time 18.063264
+[lfw][46000]XNorm: 21.172505
+[lfw][46000]Accuracy-Flip: 0.99183+-0.00383
+[46000]Accuracy-Highest: 0.99350
+INFO:root:Epoch[6] Batch [900]	Speed: 92.75 samples/sec	acc=0.143945
+INFO:root:Epoch[6] Batch [920]	Speed: 662.55 samples/sec	acc=0.143848
+INFO:root:Epoch[6] Batch [940]	Speed: 662.12 samples/sec	acc=0.147754
+INFO:root:Epoch[6] Batch [960]	Speed: 663.18 samples/sec	acc=0.145020
+INFO:root:Epoch[6] Batch [980]	Speed: 662.61 samples/sec	acc=0.144336
+INFO:root:Epoch[6] Batch [1000]	Speed: 663.46 samples/sec	acc=0.139355
+INFO:root:Epoch[6] Batch [1020]	Speed: 663.76 samples/sec	acc=0.145117
+INFO:root:Epoch[6] Batch [1040]	Speed: 663.61 samples/sec	acc=0.142773
+INFO:root:Epoch[6] Batch [1060]	Speed: 681.50 samples/sec	acc=0.146484
+INFO:root:Epoch[6] Batch [1080]	Speed: 672.49 samples/sec	acc=0.143457
+INFO:root:Epoch[6] Batch [1100]	Speed: 664.28 samples/sec	acc=0.146289
+INFO:root:Epoch[6] Batch [1120]	Speed: 664.24 samples/sec	acc=0.140332
+INFO:root:Epoch[6] Batch [1140]	Speed: 663.65 samples/sec	acc=0.147070
+INFO:root:Epoch[6] Batch [1160]	Speed: 664.59 samples/sec	acc=0.147559
+INFO:root:Epoch[6] Batch [1180]	Speed: 663.48 samples/sec	acc=0.141699
+INFO:root:Epoch[6] Batch [1200]	Speed: 664.38 samples/sec	acc=0.137793
+INFO:root:Epoch[6] Batch [1220]	Speed: 664.34 samples/sec	acc=0.136719
+INFO:root:Epoch[6] Batch [1240]	Speed: 664.15 samples/sec	acc=0.142285
+INFO:root:Epoch[6] Batch [1260]	Speed: 662.74 samples/sec	acc=0.145996
+INFO:root:Epoch[6] Batch [1280]	Speed: 664.45 samples/sec	acc=0.138379
+INFO:root:Epoch[6] Batch [1300]	Speed: 664.42 samples/sec	acc=0.140527
+INFO:root:Epoch[6] Batch [1320]	Speed: 663.64 samples/sec	acc=0.143750
+INFO:root:Epoch[6] Batch [1340]	Speed: 663.89 samples/sec	acc=0.131445
+INFO:root:Epoch[6] Batch [1360]	Speed: 663.15 samples/sec	acc=0.138184
+INFO:root:Epoch[6] Batch [1380]	Speed: 663.79 samples/sec	acc=0.148730
+INFO:root:Epoch[6] Batch [1400]	Speed: 664.57 samples/sec	acc=0.152051
+INFO:root:Epoch[6] Batch [1420]	Speed: 664.19 samples/sec	acc=0.139746
+INFO:root:Epoch[6] Batch [1440]	Speed: 662.39 samples/sec	acc=0.140918
+INFO:root:Epoch[6] Batch [1460]	Speed: 663.26 samples/sec	acc=0.139453
+INFO:root:Epoch[6] Batch [1480]	Speed: 663.11 samples/sec	acc=0.142285
+INFO:root:Epoch[6] Batch [1500]	Speed: 664.47 samples/sec	acc=0.145117
+INFO:root:Epoch[6] Batch [1520]	Speed: 664.42 samples/sec	acc=0.135059
+INFO:root:Epoch[6] Batch [1540]	Speed: 663.77 samples/sec	acc=0.143262
+INFO:root:Epoch[6] Batch [1560]	Speed: 663.89 samples/sec	acc=0.143164
+INFO:root:Epoch[6] Batch [1580]	Speed: 664.31 samples/sec	acc=0.138965
+INFO:root:Epoch[6] Batch [1600]	Speed: 664.42 samples/sec	acc=0.140332
+INFO:root:Epoch[6] Batch [1620]	Speed: 663.47 samples/sec	acc=0.142676
+INFO:root:Epoch[6] Batch [1640]	Speed: 663.39 samples/sec	acc=0.142969
+INFO:root:Epoch[6] Batch [1660]	Speed: 664.44 samples/sec	acc=0.144727
+INFO:root:Epoch[6] Batch [1680]	Speed: 663.32 samples/sec	acc=0.147461
+INFO:root:Epoch[6] Batch [1700]	Speed: 664.47 samples/sec	acc=0.141309
+INFO:root:Epoch[6] Batch [1720]	Speed: 664.69 samples/sec	acc=0.144727
+INFO:root:Epoch[6] Batch [1740]	Speed: 663.18 samples/sec	acc=0.146875
+INFO:root:Epoch[6] Batch [1760]	Speed: 663.96 samples/sec	acc=0.144141
+INFO:root:Epoch[6] Batch [1780]	Speed: 663.76 samples/sec	acc=0.142187
+INFO:root:Epoch[6] Batch [1800]	Speed: 664.06 samples/sec	acc=0.141113
+INFO:root:Epoch[6] Batch [1820]	Speed: 662.88 samples/sec	acc=0.141309
+INFO:root:Epoch[6] Batch [1840]	Speed: 664.29 samples/sec	acc=0.138672
+INFO:root:Epoch[6] Batch [1860]	Speed: 663.40 samples/sec	acc=0.145996
+INFO:root:Epoch[6] Batch [1880]	Speed: 664.21 samples/sec	acc=0.137598
+lr-batch-epoch: 0.1 1885 6
+INFO:root:Epoch[6] Batch [1900]	Speed: 664.62 samples/sec	acc=0.142676
+INFO:root:Epoch[6] Batch [1920]	Speed: 664.06 samples/sec	acc=0.140332
+INFO:root:Epoch[6] Batch [1940]	Speed: 664.00 samples/sec	acc=0.151367
+INFO:root:Epoch[6] Batch [1960]	Speed: 663.72 samples/sec	acc=0.142480
+INFO:root:Epoch[6] Batch [1980]	Speed: 664.23 samples/sec	acc=0.142578
+INFO:root:Epoch[6] Batch [2000]	Speed: 663.22 samples/sec	acc=0.142578
+INFO:root:Epoch[6] Batch [2020]	Speed: 663.53 samples/sec	acc=0.145605
+INFO:root:Epoch[6] Batch [2040]	Speed: 663.54 samples/sec	acc=0.143945
+INFO:root:Epoch[6] Batch [2060]	Speed: 663.74 samples/sec	acc=0.136035
+INFO:root:Epoch[6] Batch [2080]	Speed: 665.14 samples/sec	acc=0.144336
+INFO:root:Epoch[6] Batch [2100]	Speed: 663.37 samples/sec	acc=0.145313
+INFO:root:Epoch[6] Batch [2120]	Speed: 664.34 samples/sec	acc=0.138477
+INFO:root:Epoch[6] Batch [2140]	Speed: 663.86 samples/sec	acc=0.146484
+INFO:root:Epoch[6] Batch [2160]	Speed: 663.84 samples/sec	acc=0.144824
+INFO:root:Epoch[6] Batch [2180]	Speed: 663.19 samples/sec	acc=0.143359
+INFO:root:Epoch[6] Batch [2200]	Speed: 663.75 samples/sec	acc=0.149121
+INFO:root:Epoch[6] Batch [2220]	Speed: 664.50 samples/sec	acc=0.145117
+INFO:root:Epoch[6] Batch [2240]	Speed: 664.77 samples/sec	acc=0.142578
+INFO:root:Epoch[6] Batch [2260]	Speed: 664.15 samples/sec	acc=0.141016
+INFO:root:Epoch[6] Batch [2280]	Speed: 663.72 samples/sec	acc=0.138867
+INFO:root:Epoch[6] Batch [2300]	Speed: 664.22 samples/sec	acc=0.139648
+INFO:root:Epoch[6] Batch [2320]	Speed: 656.90 samples/sec	acc=0.145703
+INFO:root:Epoch[6] Batch [2340]	Speed: 662.47 samples/sec	acc=0.147461
+INFO:root:Epoch[6] Batch [2360]	Speed: 658.40 samples/sec	acc=0.148535
+INFO:root:Epoch[6] Batch [2380]	Speed: 661.27 samples/sec	acc=0.145410
+INFO:root:Epoch[6] Batch [2400]	Speed: 663.12 samples/sec	acc=0.151367
+INFO:root:Epoch[6] Batch [2420]	Speed: 663.54 samples/sec	acc=0.142285
+INFO:root:Epoch[6] Batch [2440]	Speed: 662.92 samples/sec	acc=0.138672
+INFO:root:Epoch[6] Batch [2460]	Speed: 663.02 samples/sec	acc=0.140527
+INFO:root:Epoch[6] Batch [2480]	Speed: 663.40 samples/sec	acc=0.140625
+INFO:root:Epoch[6] Batch [2500]	Speed: 662.60 samples/sec	acc=0.146094
+INFO:root:Epoch[6] Batch [2520]	Speed: 663.47 samples/sec	acc=0.146777
+INFO:root:Epoch[6] Batch [2540]	Speed: 662.83 samples/sec	acc=0.142383
+INFO:root:Epoch[6] Batch [2560]	Speed: 662.33 samples/sec	acc=0.147852
+INFO:root:Epoch[6] Batch [2580]	Speed: 663.48 samples/sec	acc=0.141309
+INFO:root:Epoch[6] Batch [2600]	Speed: 662.30 samples/sec	acc=0.135449
+INFO:root:Epoch[6] Batch [2620]	Speed: 660.78 samples/sec	acc=0.144727
+INFO:root:Epoch[6] Batch [2640]	Speed: 677.00 samples/sec	acc=0.144141
+INFO:root:Epoch[6] Batch [2660]	Speed: 662.99 samples/sec	acc=0.144434
+INFO:root:Epoch[6] Batch [2680]	Speed: 663.08 samples/sec	acc=0.146582
+INFO:root:Epoch[6] Batch [2700]	Speed: 662.79 samples/sec	acc=0.142969
+INFO:root:Epoch[6] Batch [2720]	Speed: 662.02 samples/sec	acc=0.148828
+INFO:root:Epoch[6] Batch [2740]	Speed: 663.42 samples/sec	acc=0.141016
+INFO:root:Epoch[6] Batch [2760]	Speed: 663.46 samples/sec	acc=0.141309
+INFO:root:Epoch[6] Batch [2780]	Speed: 662.95 samples/sec	acc=0.146777
+INFO:root:Epoch[6] Batch [2800]	Speed: 663.11 samples/sec	acc=0.144922
+INFO:root:Epoch[6] Batch [2820]	Speed: 663.00 samples/sec	acc=0.146582
+INFO:root:Epoch[6] Batch [2840]	Speed: 663.41 samples/sec	acc=0.146094
+INFO:root:Epoch[6] Batch [2860]	Speed: 662.67 samples/sec	acc=0.142676
+INFO:root:Epoch[6] Batch [2880]	Speed: 663.64 samples/sec	acc=0.149512
+[48000][MARGIN]0.650225,0.212011
+lr-batch-epoch: 0.1 2885 6
+testing verification..
+(14000, 512)
+infer time 21.576396
+[cfp_ff][48000]XNorm: 19.684269
+[cfp_ff][48000]Accuracy-Flip: 0.99271+-0.00296
+testing verification..
+(14000, 512)
+infer time 26.514342
+[cfp_fp][48000]XNorm: 15.737945
+[cfp_fp][48000]Accuracy-Flip: 0.84371+-0.01511
+testing verification..
+(12000, 512)
+infer time 19.399874
+[agedb_30][48000]XNorm: 19.425554
+[agedb_30][48000]Accuracy-Flip: 0.94800+-0.00945
+testing verification..
+(12000, 512)
+infer time 17.305142
+[lfw][48000]XNorm: 19.654909
+[lfw][48000]Accuracy-Flip: 0.99267+-0.00523
+[48000]Accuracy-Highest: 0.99350
+INFO:root:Epoch[6] Batch [2900]	Speed: 85.54 samples/sec	acc=0.138477
+INFO:root:Epoch[6] Batch [2920]	Speed: 662.61 samples/sec	acc=0.147168
+INFO:root:Epoch[6] Batch [2940]	Speed: 662.29 samples/sec	acc=0.146973
+INFO:root:Epoch[6] Batch [2960]	Speed: 661.81 samples/sec	acc=0.146094
+INFO:root:Epoch[6] Batch [2980]	Speed: 662.29 samples/sec	acc=0.136523
+INFO:root:Epoch[6] Batch [3000]	Speed: 662.23 samples/sec	acc=0.139160
+INFO:root:Epoch[6] Batch [3020]	Speed: 662.76 samples/sec	acc=0.142187
+INFO:root:Epoch[6] Batch [3040]	Speed: 662.81 samples/sec	acc=0.140234
+INFO:root:Epoch[6] Batch [3060]	Speed: 662.97 samples/sec	acc=0.141406
+INFO:root:Epoch[6] Batch [3080]	Speed: 662.93 samples/sec	acc=0.142871
+INFO:root:Epoch[6] Batch [3100]	Speed: 662.54 samples/sec	acc=0.143262
+INFO:root:Epoch[6] Batch [3120]	Speed: 662.54 samples/sec	acc=0.146875
+INFO:root:Epoch[6] Batch [3140]	Speed: 662.60 samples/sec	acc=0.143359
+INFO:root:Epoch[6] Batch [3160]	Speed: 663.18 samples/sec	acc=0.143164
+INFO:root:Epoch[6] Batch [3180]	Speed: 662.85 samples/sec	acc=0.142578
+INFO:root:Epoch[6] Batch [3200]	Speed: 662.26 samples/sec	acc=0.144629
+INFO:root:Epoch[6] Batch [3220]	Speed: 663.10 samples/sec	acc=0.144629
+INFO:root:Epoch[6] Batch [3240]	Speed: 663.06 samples/sec	acc=0.145703
+INFO:root:Epoch[6] Batch [3260]	Speed: 662.07 samples/sec	acc=0.144434
+INFO:root:Epoch[6] Batch [3280]	Speed: 663.05 samples/sec	acc=0.139551
+INFO:root:Epoch[6] Batch [3300]	Speed: 662.41 samples/sec	acc=0.146973
+INFO:root:Epoch[6] Batch [3320]	Speed: 663.01 samples/sec	acc=0.146094
+INFO:root:Epoch[6] Batch [3340]	Speed: 663.48 samples/sec	acc=0.148438
+INFO:root:Epoch[6] Batch [3360]	Speed: 662.62 samples/sec	acc=0.144629
+INFO:root:Epoch[6] Batch [3380]	Speed: 663.01 samples/sec	acc=0.143457
+INFO:root:Epoch[6] Batch [3400]	Speed: 663.14 samples/sec	acc=0.144434
+INFO:root:Epoch[6] Batch [3420]	Speed: 667.89 samples/sec	acc=0.137402
+INFO:root:Epoch[6] Batch [3440]	Speed: 662.94 samples/sec	acc=0.140039
+INFO:root:Epoch[6] Batch [3460]	Speed: 662.01 samples/sec	acc=0.143750
+INFO:root:Epoch[6] Batch [3480]	Speed: 662.62 samples/sec	acc=0.146289
+INFO:root:Epoch[6] Batch [3500]	Speed: 663.05 samples/sec	acc=0.147070
+INFO:root:Epoch[6] Batch [3520]	Speed: 661.63 samples/sec	acc=0.148535
+INFO:root:Epoch[6] Batch [3540]	Speed: 662.67 samples/sec	acc=0.140918
+INFO:root:Epoch[6] Batch [3560]	Speed: 662.30 samples/sec	acc=0.140723
+INFO:root:Epoch[6] Batch [3580]	Speed: 662.38 samples/sec	acc=0.144141
+INFO:root:Epoch[6] Batch [3600]	Speed: 662.21 samples/sec	acc=0.150391
+INFO:root:Epoch[6] Batch [3620]	Speed: 662.30 samples/sec	acc=0.143457
+INFO:root:Epoch[6] Batch [3640]	Speed: 663.17 samples/sec	acc=0.139648
+INFO:root:Epoch[6] Batch [3660]	Speed: 663.06 samples/sec	acc=0.144727
+INFO:root:Epoch[6] Batch [3680]	Speed: 663.17 samples/sec	acc=0.142480
+INFO:root:Epoch[6] Batch [3700]	Speed: 661.97 samples/sec	acc=0.148145
+INFO:root:Epoch[6] Batch [3720]	Speed: 662.74 samples/sec	acc=0.142480
+INFO:root:Epoch[6] Batch [3740]	Speed: 662.58 samples/sec	acc=0.143164
+INFO:root:Epoch[6] Batch [3760]	Speed: 662.09 samples/sec	acc=0.149023
+INFO:root:Epoch[6] Batch [3780]	Speed: 662.58 samples/sec	acc=0.147754
+INFO:root:Epoch[6] Batch [3800]	Speed: 662.42 samples/sec	acc=0.146289
+INFO:root:Epoch[6] Batch [3820]	Speed: 662.73 samples/sec	acc=0.150977
+INFO:root:Epoch[6] Batch [3840]	Speed: 662.65 samples/sec	acc=0.135840
+INFO:root:Epoch[6] Batch [3860]	Speed: 661.90 samples/sec	acc=0.141016
+INFO:root:Epoch[6] Batch [3880]	Speed: 662.53 samples/sec	acc=0.137012
+lr-batch-epoch: 0.1 3885 6
+INFO:root:Epoch[6] Batch [3900]	Speed: 662.18 samples/sec	acc=0.147070
+INFO:root:Epoch[6] Batch [3920]	Speed: 661.96 samples/sec	acc=0.145508
+INFO:root:Epoch[6] Batch [3940]	Speed: 662.30 samples/sec	acc=0.152051
+INFO:root:Epoch[6] Batch [3960]	Speed: 662.66 samples/sec	acc=0.144336
+INFO:root:Epoch[6] Batch [3980]	Speed: 662.56 samples/sec	acc=0.143359
+INFO:root:Epoch[6] Batch [4000]	Speed: 663.18 samples/sec	acc=0.146191
+INFO:root:Epoch[6] Batch [4020]	Speed: 663.72 samples/sec	acc=0.149805
+INFO:root:Epoch[6] Batch [4040]	Speed: 663.32 samples/sec	acc=0.147461
+INFO:root:Epoch[6] Batch [4060]	Speed: 663.35 samples/sec	acc=0.143945
+INFO:root:Epoch[6] Batch [4080]	Speed: 663.21 samples/sec	acc=0.144434
+INFO:root:Epoch[6] Batch [4100]	Speed: 663.13 samples/sec	acc=0.144922
+INFO:root:Epoch[6] Batch [4120]	Speed: 663.12 samples/sec	acc=0.150488
+INFO:root:Epoch[6] Batch [4140]	Speed: 662.99 samples/sec	acc=0.148926
+INFO:root:Epoch[6] Batch [4160]	Speed: 662.67 samples/sec	acc=0.145801
+INFO:root:Epoch[6] Batch [4180]	Speed: 663.13 samples/sec	acc=0.138574
+INFO:root:Epoch[6] Batch [4200]	Speed: 663.08 samples/sec	acc=0.144238
+INFO:root:Epoch[6] Batch [4220]	Speed: 665.64 samples/sec	acc=0.139941
+INFO:root:Epoch[6] Batch [4240]	Speed: 662.62 samples/sec	acc=0.144824
+INFO:root:Epoch[6] Batch [4260]	Speed: 662.86 samples/sec	acc=0.142676
+INFO:root:Epoch[6] Batch [4280]	Speed: 663.09 samples/sec	acc=0.144141
+INFO:root:Epoch[6] Batch [4300]	Speed: 662.95 samples/sec	acc=0.146289
+INFO:root:Epoch[6] Batch [4320]	Speed: 663.10 samples/sec	acc=0.142676
+INFO:root:Epoch[6] Batch [4340]	Speed: 662.35 samples/sec	acc=0.146191
+INFO:root:Epoch[6] Batch [4360]	Speed: 662.78 samples/sec	acc=0.137500
+INFO:root:Epoch[6] Batch [4380]	Speed: 662.40 samples/sec	acc=0.142578
+INFO:root:Epoch[6] Batch [4400]	Speed: 662.31 samples/sec	acc=0.150586
+INFO:root:Epoch[6] Batch [4420]	Speed: 662.28 samples/sec	acc=0.145508
+INFO:root:Epoch[6] Batch [4440]	Speed: 663.28 samples/sec	acc=0.151758
+INFO:root:Epoch[6] Batch [4460]	Speed: 662.63 samples/sec	acc=0.144531
+INFO:root:Epoch[6] Batch [4480]	Speed: 662.45 samples/sec	acc=0.143359
+INFO:root:Epoch[6] Batch [4500]	Speed: 661.36 samples/sec	acc=0.143457
+INFO:root:Epoch[6] Batch [4520]	Speed: 662.74 samples/sec	acc=0.148828
+INFO:root:Epoch[6] Batch [4540]	Speed: 662.27 samples/sec	acc=0.138867
+INFO:root:Epoch[6] Batch [4560]	Speed: 662.26 samples/sec	acc=0.146191
+INFO:root:Epoch[6] Batch [4580]	Speed: 662.78 samples/sec	acc=0.141895
+INFO:root:Epoch[6] Batch [4600]	Speed: 662.50 samples/sec	acc=0.145703
+INFO:root:Epoch[6] Batch [4620]	Speed: 661.88 samples/sec	acc=0.147168
+INFO:root:Epoch[6] Batch [4640]	Speed: 662.62 samples/sec	acc=0.144824
+INFO:root:Epoch[6] Batch [4660]	Speed: 661.89 samples/sec	acc=0.136230
+INFO:root:Epoch[6] Batch [4680]	Speed: 661.81 samples/sec	acc=0.145020
+INFO:root:Epoch[6] Batch [4700]	Speed: 662.03 samples/sec	acc=0.149414
+INFO:root:Epoch[6] Batch [4720]	Speed: 662.78 samples/sec	acc=0.146191
+INFO:root:Epoch[6] Batch [4740]	Speed: 662.65 samples/sec	acc=0.142676
+INFO:root:Epoch[6] Batch [4760]	Speed: 662.85 samples/sec	acc=0.147852
+INFO:root:Epoch[6] Batch [4780]	Speed: 662.85 samples/sec	acc=0.149121
+INFO:root:Epoch[6] Batch [4800]	Speed: 662.89 samples/sec	acc=0.145508
+INFO:root:Epoch[6] Batch [4820]	Speed: 663.18 samples/sec	acc=0.147461
+INFO:root:Epoch[6] Batch [4840]	Speed: 663.00 samples/sec	acc=0.140039
+INFO:root:Epoch[6] Batch [4860]	Speed: 661.96 samples/sec	acc=0.144141
+INFO:root:Epoch[6] Batch [4880]	Speed: 662.72 samples/sec	acc=0.146484
+[50000][MARGIN]0.644800,0.205978
+lr change to 0.01
+lr-batch-epoch: 0.01 4885 6
+testing verification..
+(14000, 512)
+infer time 20.111636
+[cfp_ff][50000]XNorm: 20.538264
+[cfp_ff][50000]Accuracy-Flip: 0.99443+-0.00370
+testing verification..
+(14000, 512)
+infer time 21.836273
+[cfp_fp][50000]XNorm: 16.234446
+[cfp_fp][50000]Accuracy-Flip: 0.83786+-0.01698
+testing verification..
+(12000, 512)
+infer time 20.022741
+[agedb_30][50000]XNorm: 20.046532
+[agedb_30][50000]Accuracy-Flip: 0.94617+-0.00882
+testing verification..
+(12000, 512)
+infer time 17.533046
+[lfw][50000]XNorm: 20.479102
+[lfw][50000]Accuracy-Flip: 0.99467+-0.00420
+saving 25
+INFO:root:Saved checkpoint to "../model-r50-am-lfw/model-0025.params"
+[50000]Accuracy-Highest: 0.99467
+INFO:root:Epoch[6] Batch [4900]	Speed: 86.64 samples/sec	acc=0.115137
+INFO:root:Epoch[6] Batch [4920]	Speed: 670.62 samples/sec	acc=0.133691
+INFO:root:Epoch[6] Batch [4940]	Speed: 662.34 samples/sec	acc=0.167285
+INFO:root:Epoch[6] Batch [4960]	Speed: 662.80 samples/sec	acc=0.188867
+INFO:root:Epoch[6] Batch [4980]	Speed: 662.64 samples/sec	acc=0.198242
+INFO:root:Epoch[6] Batch [5000]	Speed: 663.02 samples/sec	acc=0.206348
+INFO:root:Epoch[6] Batch [5020]	Speed: 662.72 samples/sec	acc=0.206934
+INFO:root:Epoch[6] Batch [5040]	Speed: 662.40 samples/sec	acc=0.216211
+INFO:root:Epoch[6] Batch [5060]	Speed: 662.95 samples/sec	acc=0.211230
+INFO:root:Epoch[6] Batch [5080]	Speed: 662.37 samples/sec	acc=0.233789
+INFO:root:Epoch[6] Batch [5100]	Speed: 663.51 samples/sec	acc=0.226660
+INFO:root:Epoch[6] Batch [5120]	Speed: 662.94 samples/sec	acc=0.241113
+INFO:root:Epoch[6] Batch [5140]	Speed: 662.56 samples/sec	acc=0.242090
+INFO:root:Epoch[6] Batch [5160]	Speed: 663.08 samples/sec	acc=0.246777
+INFO:root:Epoch[6] Batch [5180]	Speed: 662.62 samples/sec	acc=0.255371
+INFO:root:Epoch[6] Batch [5200]	Speed: 662.65 samples/sec	acc=0.242480
+INFO:root:Epoch[6] Batch [5220]	Speed: 662.60 samples/sec	acc=0.261328
+INFO:root:Epoch[6] Batch [5240]	Speed: 662.80 samples/sec	acc=0.266895
+INFO:root:Epoch[6] Batch [5260]	Speed: 662.61 samples/sec	acc=0.261035
+INFO:root:Epoch[6] Batch [5280]	Speed: 663.17 samples/sec	acc=0.272168
+INFO:root:Epoch[6] Batch [5300]	Speed: 662.79 samples/sec	acc=0.267285
+INFO:root:Epoch[6] Batch [5320]	Speed: 662.95 samples/sec	acc=0.270703
+INFO:root:Epoch[6] Batch [5340]	Speed: 663.26 samples/sec	acc=0.270898
+INFO:root:Epoch[6] Batch [5360]	Speed: 662.69 samples/sec	acc=0.276855
+INFO:root:Epoch[6] Batch [5380]	Speed: 662.18 samples/sec	acc=0.285742
+INFO:root:Epoch[6] Batch [5400]	Speed: 662.56 samples/sec	acc=0.280078
+INFO:root:Epoch[6] Batch [5420]	Speed: 662.15 samples/sec	acc=0.286719
+INFO:root:Epoch[6] Batch [5440]	Speed: 663.39 samples/sec	acc=0.293262
+INFO:root:Epoch[6] Batch [5460]	Speed: 662.71 samples/sec	acc=0.290527
+INFO:root:Epoch[6] Batch [5480]	Speed: 662.32 samples/sec	acc=0.295898
+INFO:root:Epoch[6] Batch [5500]	Speed: 662.77 samples/sec	acc=0.296875
+INFO:root:Epoch[6] Batch [5520]	Speed: 663.21 samples/sec	acc=0.302246
+INFO:root:Epoch[6] Batch [5540]	Speed: 663.61 samples/sec	acc=0.305566
+INFO:root:Epoch[6] Batch [5560]	Speed: 662.96 samples/sec	acc=0.305664
+INFO:root:Epoch[6] Batch [5580]	Speed: 662.10 samples/sec	acc=0.300195
+INFO:root:Epoch[6] Batch [5600]	Speed: 663.28 samples/sec	acc=0.311426
+INFO:root:Epoch[6] Batch [5620]	Speed: 662.87 samples/sec	acc=0.304688
+INFO:root:Epoch[6] Batch [5640]	Speed: 663.11 samples/sec	acc=0.305371
+INFO:root:Epoch[6] Batch [5660]	Speed: 662.46 samples/sec	acc=0.319824
+INFO:root:Epoch[6] Batch [5680]	Speed: 663.26 samples/sec	acc=0.325684
+INFO:root:Epoch[6] Batch [5700]	Speed: 663.42 samples/sec	acc=0.315137
+INFO:root:Epoch[6] Batch [5720]	Speed: 663.85 samples/sec	acc=0.331543
+INFO:root:Epoch[6] Batch [5740]	Speed: 663.76 samples/sec	acc=0.327051
+INFO:root:Epoch[6] Batch [5760]	Speed: 667.94 samples/sec	acc=0.331836
+INFO:root:Epoch[6] Batch [5780]	Speed: 663.71 samples/sec	acc=0.326074
+INFO:root:Epoch[6] Batch [5800]	Speed: 662.27 samples/sec	acc=0.329395
+INFO:root:Epoch[6] Batch [5820]	Speed: 663.85 samples/sec	acc=0.336230
+INFO:root:Epoch[6] Batch [5840]	Speed: 662.37 samples/sec	acc=0.335547
+INFO:root:Epoch[6] Batch [5860]	Speed: 662.94 samples/sec	acc=0.340820
+INFO:root:Epoch[6] Batch [5880]	Speed: 663.35 samples/sec	acc=0.332227
+lr-batch-epoch: 0.01 5885 6
+INFO:root:Epoch[6] Batch [5900]	Speed: 663.53 samples/sec	acc=0.342090
+INFO:root:Epoch[6] Batch [5920]	Speed: 663.25 samples/sec	acc=0.355469
+INFO:root:Epoch[6] Batch [5940]	Speed: 663.21 samples/sec	acc=0.337891
+INFO:root:Epoch[6] Batch [5960]	Speed: 662.47 samples/sec	acc=0.351465
+INFO:root:Epoch[6] Batch [5980]	Speed: 662.16 samples/sec	acc=0.347363
+INFO:root:Epoch[6] Batch [6000]	Speed: 663.61 samples/sec	acc=0.346777
+INFO:root:Epoch[6] Batch [6020]	Speed: 662.37 samples/sec	acc=0.356055
+INFO:root:Epoch[6] Batch [6040]	Speed: 662.63 samples/sec	acc=0.350488
+INFO:root:Epoch[6] Batch [6060]	Speed: 662.13 samples/sec	acc=0.349805
+INFO:root:Epoch[6] Batch [6080]	Speed: 663.49 samples/sec	acc=0.362500
+INFO:root:Epoch[6] Batch [6100]	Speed: 662.45 samples/sec	acc=0.358398
+INFO:root:Epoch[6] Batch [6120]	Speed: 663.36 samples/sec	acc=0.359766
+INFO:root:Epoch[6] Batch [6140]	Speed: 663.26 samples/sec	acc=0.375781
+INFO:root:Epoch[6] Batch [6160]	Speed: 662.71 samples/sec	acc=0.370117
+INFO:root:Epoch[6] Batch [6180]	Speed: 663.07 samples/sec	acc=0.363965
+INFO:root:Epoch[6] Batch [6200]	Speed: 663.55 samples/sec	acc=0.368848
+INFO:root:Epoch[6] Batch [6220]	Speed: 662.87 samples/sec	acc=0.367090
+INFO:root:Epoch[6] Batch [6240]	Speed: 662.68 samples/sec	acc=0.369629
+INFO:root:Epoch[6] Batch [6260]	Speed: 662.97 samples/sec	acc=0.379492
+INFO:root:Epoch[6] Batch [6280]	Speed: 662.98 samples/sec	acc=0.381348
+INFO:root:Epoch[6] Batch [6300]	Speed: 662.60 samples/sec	acc=0.373926
+INFO:root:Epoch[6] Batch [6320]	Speed: 663.18 samples/sec	acc=0.377734
+INFO:root:Epoch[6] Batch [6340]	Speed: 663.05 samples/sec	acc=0.384766
+INFO:root:Epoch[6] Batch [6360]	Speed: 663.17 samples/sec	acc=0.380566
+INFO:root:Epoch[6] Batch [6380]	Speed: 662.61 samples/sec	acc=0.387500
+INFO:root:Epoch[6] Batch [6400]	Speed: 662.98 samples/sec	acc=0.378418
+INFO:root:Epoch[6] Batch [6420]	Speed: 663.25 samples/sec	acc=0.391309
+INFO:root:Epoch[6] Batch [6440]	Speed: 663.24 samples/sec	acc=0.386328
+INFO:root:Epoch[6] Batch [6460]	Speed: 662.99 samples/sec	acc=0.401562
+INFO:root:Epoch[6] Batch [6480]	Speed: 662.61 samples/sec	acc=0.388965
+INFO:root:Epoch[6] Batch [6500]	Speed: 663.38 samples/sec	acc=0.389648
+INFO:root:Epoch[6] Batch [6520]	Speed: 662.88 samples/sec	acc=0.400391
+INFO:root:Epoch[6] Batch [6540]	Speed: 665.61 samples/sec	acc=0.400391
+INFO:root:Epoch[6] Batch [6560]	Speed: 662.97 samples/sec	acc=0.394727
+INFO:root:Epoch[6] Batch [6580]	Speed: 662.53 samples/sec	acc=0.401367
+INFO:root:Epoch[6] Batch [6600]	Speed: 663.05 samples/sec	acc=0.395605
+INFO:root:Epoch[6] Batch [6620]	Speed: 662.09 samples/sec	acc=0.397852
+INFO:root:Epoch[6] Batch [6640]	Speed: 663.59 samples/sec	acc=0.405371
+INFO:root:Epoch[6] Batch [6660]	Speed: 663.33 samples/sec	acc=0.408496
+INFO:root:Epoch[6] Batch [6680]	Speed: 663.26 samples/sec	acc=0.399414
+INFO:root:Epoch[6] Batch [6700]	Speed: 663.49 samples/sec	acc=0.406836
+INFO:root:Epoch[6] Batch [6720]	Speed: 662.72 samples/sec	acc=0.411719
+INFO:root:Epoch[6] Batch [6740]	Speed: 662.40 samples/sec	acc=0.407422
+INFO:root:Epoch[6] Batch [6760]	Speed: 662.93 samples/sec	acc=0.416309
+INFO:root:Epoch[6] Batch [6780]	Speed: 663.07 samples/sec	acc=0.414844
+INFO:root:Epoch[6] Batch [6800]	Speed: 662.97 samples/sec	acc=0.415137
+INFO:root:Epoch[6] Batch [6820]	Speed: 663.15 samples/sec	acc=0.418945
+INFO:root:Epoch[6] Batch [6840]	Speed: 662.41 samples/sec	acc=0.405176
+INFO:root:Epoch[6] Batch [6860]	Speed: 663.37 samples/sec	acc=0.412695
+INFO:root:Epoch[6] Batch [6880]	Speed: 662.82 samples/sec	acc=0.411719
+[52000][MARGIN]0.728075,0.315939
+lr-batch-epoch: 0.01 6885 6
+testing verification..
+(14000, 512)
+infer time 18.323727
+[cfp_ff][52000]XNorm: 22.220222
+[cfp_ff][52000]Accuracy-Flip: 0.99671+-0.00313
+testing verification..
+(14000, 512)
+infer time 16.814555
+[cfp_fp][52000]XNorm: 18.628988
+[cfp_fp][52000]Accuracy-Flip: 0.89243+-0.01752
+testing verification..
+(12000, 512)
+infer time 14.207714
+[agedb_30][52000]XNorm: 22.086024
+[agedb_30][52000]Accuracy-Flip: 0.96650+-0.00728
+testing verification..
+(12000, 512)
+infer time 13.266452
+[lfw][52000]XNorm: 22.244432
+[lfw][52000]Accuracy-Flip: 0.99617+-0.00259
+saving 26
+INFO:root:Saved checkpoint to "../model-r50-am-lfw/model-0026.params"
+[52000]Accuracy-Highest: 0.99617
+INFO:root:Epoch[6] Batch [6900]	Speed: 96.11 samples/sec	acc=0.415625
+INFO:root:Epoch[6] Batch [6920]	Speed: 662.27 samples/sec	acc=0.417773
+INFO:root:Epoch[6] Batch [6940]	Speed: 662.22 samples/sec	acc=0.412695
+INFO:root:Epoch[6] Batch [6960]	Speed: 662.51 samples/sec	acc=0.421973
+INFO:root:Epoch[6] Batch [6980]	Speed: 662.38 samples/sec	acc=0.420117
+INFO:root:Epoch[6] Batch [7000]	Speed: 662.57 samples/sec	acc=0.424121
+INFO:root:Epoch[6] Batch [7020]	Speed: 662.19 samples/sec	acc=0.424219
+INFO:root:Epoch[6] Batch [7040]	Speed: 662.63 samples/sec	acc=0.437207
+INFO:root:Epoch[6] Batch [7060]	Speed: 663.33 samples/sec	acc=0.432422
+INFO:root:Epoch[6] Batch [7080]	Speed: 662.37 samples/sec	acc=0.423926
+INFO:root:Epoch[6] Batch [7100]	Speed: 663.58 samples/sec	acc=0.425391
+INFO:root:Epoch[6] Batch [7120]	Speed: 662.68 samples/sec	acc=0.429395
+INFO:root:Epoch[6] Batch [7140]	Speed: 662.70 samples/sec	acc=0.440332
+INFO:root:Epoch[6] Batch [7160]	Speed: 663.03 samples/sec	acc=0.431250
+INFO:root:Epoch[6] Batch [7180]	Speed: 663.07 samples/sec	acc=0.436621
+INFO:root:Epoch[6] Batch [7200]	Speed: 662.73 samples/sec	acc=0.434180
+INFO:root:Epoch[6] Batch [7220]	Speed: 662.90 samples/sec	acc=0.440527
+INFO:root:Epoch[6] Batch [7240]	Speed: 672.57 samples/sec	acc=0.440234
+INFO:root:Epoch[6] Batch [7260]	Speed: 673.00 samples/sec	acc=0.433887
+INFO:root:Epoch[6] Batch [7280]	Speed: 662.85 samples/sec	acc=0.441309
+INFO:root:Epoch[6] Batch [7300]	Speed: 662.86 samples/sec	acc=0.443750
+INFO:root:Epoch[6] Batch [7320]	Speed: 663.38 samples/sec	acc=0.442676
+INFO:root:Epoch[6] Batch [7340]	Speed: 663.13 samples/sec	acc=0.435352
+INFO:root:Epoch[6] Batch [7360]	Speed: 662.60 samples/sec	acc=0.448047
+INFO:root:Epoch[6] Batch [7380]	Speed: 663.53 samples/sec	acc=0.446777
+INFO:root:Epoch[6] Batch [7400]	Speed: 663.37 samples/sec	acc=0.446387
+INFO:root:Epoch[6] Batch [7420]	Speed: 664.19 samples/sec	acc=0.454688
+INFO:root:Epoch[6] Batch [7440]	Speed: 662.77 samples/sec	acc=0.459766
+INFO:root:Epoch[6] Batch [7460]	Speed: 663.38 samples/sec	acc=0.445020
+INFO:root:Epoch[6] Batch [7480]	Speed: 663.61 samples/sec	acc=0.449023
+INFO:root:Epoch[6] Batch [7500]	Speed: 663.36 samples/sec	acc=0.450293
+INFO:root:Epoch[6] Train-acc=0.447266
+INFO:root:Epoch[6] Time cost=6197.176
+call reset()
+INFO:root:Epoch[7] Batch [20]	Speed: 678.15 samples/sec	acc=0.518601
+INFO:root:Epoch[7] Batch [40]	Speed: 662.57 samples/sec	acc=0.509570
+INFO:root:Epoch[7] Batch [60]	Speed: 662.93 samples/sec	acc=0.513574
+INFO:root:Epoch[7] Batch [80]	Speed: 662.33 samples/sec	acc=0.519824
+INFO:root:Epoch[7] Batch [100]	Speed: 662.92 samples/sec	acc=0.524512
+INFO:root:Epoch[7] Batch [120]	Speed: 662.60 samples/sec	acc=0.512207
+INFO:root:Epoch[7] Batch [140]	Speed: 662.62 samples/sec	acc=0.516113
+INFO:root:Epoch[7] Batch [160]	Speed: 662.45 samples/sec	acc=0.528320
+INFO:root:Epoch[7] Batch [180]	Speed: 663.33 samples/sec	acc=0.515723
+INFO:root:Epoch[7] Batch [200]	Speed: 662.14 samples/sec	acc=0.514844
+INFO:root:Epoch[7] Batch [220]	Speed: 663.23 samples/sec	acc=0.517480
+INFO:root:Epoch[7] Batch [240]	Speed: 662.47 samples/sec	acc=0.521484
+INFO:root:Epoch[7] Batch [260]	Speed: 662.32 samples/sec	acc=0.523828
+INFO:root:Epoch[7] Batch [280]	Speed: 662.67 samples/sec	acc=0.518750
+INFO:root:Epoch[7] Batch [300]	Speed: 662.91 samples/sec	acc=0.520898
+INFO:root:Epoch[7] Batch [320]	Speed: 662.64 samples/sec	acc=0.520215
+INFO:root:Epoch[7] Batch [340]	Speed: 663.27 samples/sec	acc=0.531641
+INFO:root:Epoch[7] Batch [360]	Speed: 662.94 samples/sec	acc=0.520020
+lr-batch-epoch: 0.01 366 7
+INFO:root:Epoch[7] Batch [380]	Speed: 662.31 samples/sec	acc=0.522168
+INFO:root:Epoch[7] Batch [400]	Speed: 663.03 samples/sec	acc=0.535742
+INFO:root:Epoch[7] Batch [420]	Speed: 662.55 samples/sec	acc=0.531836
+INFO:root:Epoch[7] Batch [440]	Speed: 663.08 samples/sec	acc=0.523730
+INFO:root:Epoch[7] Batch [460]	Speed: 678.53 samples/sec	acc=0.528418
+INFO:root:Epoch[7] Batch [480]	Speed: 662.93 samples/sec	acc=0.522559
+INFO:root:Epoch[7] Batch [500]	Speed: 662.50 samples/sec	acc=0.519336
+INFO:root:Epoch[7] Batch [520]	Speed: 663.41 samples/sec	acc=0.534668
+INFO:root:Epoch[7] Batch [540]	Speed: 662.82 samples/sec	acc=0.530859
+INFO:root:Epoch[7] Batch [560]	Speed: 663.39 samples/sec	acc=0.530664
+INFO:root:Epoch[7] Batch [580]	Speed: 662.86 samples/sec	acc=0.527930
+INFO:root:Epoch[7] Batch [600]	Speed: 663.42 samples/sec	acc=0.527734
+INFO:root:Epoch[7] Batch [620]	Speed: 662.39 samples/sec	acc=0.523047
+INFO:root:Epoch[7] Batch [640]	Speed: 662.75 samples/sec	acc=0.534668
+INFO:root:Epoch[7] Batch [660]	Speed: 663.35 samples/sec	acc=0.529102
+INFO:root:Epoch[7] Batch [680]	Speed: 661.91 samples/sec	acc=0.528906
+INFO:root:Epoch[7] Batch [700]	Speed: 663.22 samples/sec	acc=0.531641
+INFO:root:Epoch[7] Batch [720]	Speed: 662.41 samples/sec	acc=0.531445
+INFO:root:Epoch[7] Batch [740]	Speed: 663.06 samples/sec	acc=0.530566
+INFO:root:Epoch[7] Batch [760]	Speed: 663.04 samples/sec	acc=0.535547
+INFO:root:Epoch[7] Batch [780]	Speed: 661.68 samples/sec	acc=0.529492
+INFO:root:Epoch[7] Batch [800]	Speed: 663.62 samples/sec	acc=0.529590
+INFO:root:Epoch[7] Batch [820]	Speed: 662.37 samples/sec	acc=0.525195
+INFO:root:Epoch[7] Batch [840]	Speed: 661.92 samples/sec	acc=0.526660
+INFO:root:Epoch[7] Batch [860]	Speed: 662.73 samples/sec	acc=0.530664
+INFO:root:Epoch[7] Batch [880]	Speed: 663.21 samples/sec	acc=0.530371
+INFO:root:Epoch[7] Batch [900]	Speed: 661.90 samples/sec	acc=0.524023
+INFO:root:Epoch[7] Batch [920]	Speed: 663.07 samples/sec	acc=0.530762
+INFO:root:Epoch[7] Batch [940]	Speed: 663.84 samples/sec	acc=0.531250
+INFO:root:Epoch[7] Batch [960]	Speed: 662.83 samples/sec	acc=0.540723
+INFO:root:Epoch[7] Batch [980]	Speed: 662.47 samples/sec	acc=0.538477
+INFO:root:Epoch[7] Batch [1000]	Speed: 663.60 samples/sec	acc=0.534863
+INFO:root:Epoch[7] Batch [1020]	Speed: 662.45 samples/sec	acc=0.530957
+INFO:root:Epoch[7] Batch [1040]	Speed: 662.32 samples/sec	acc=0.539258
+INFO:root:Epoch[7] Batch [1060]	Speed: 663.14 samples/sec	acc=0.536133
+INFO:root:Epoch[7] Batch [1080]	Speed: 663.39 samples/sec	acc=0.523633
+INFO:root:Epoch[7] Batch [1100]	Speed: 663.21 samples/sec	acc=0.535547
+INFO:root:Epoch[7] Batch [1120]	Speed: 662.40 samples/sec	acc=0.536426
+INFO:root:Epoch[7] Batch [1140]	Speed: 663.16 samples/sec	acc=0.533496
+INFO:root:Epoch[7] Batch [1160]	Speed: 663.36 samples/sec	acc=0.536816
+INFO:root:Epoch[7] Batch [1180]	Speed: 662.86 samples/sec	acc=0.535352
+INFO:root:Epoch[7] Batch [1200]	Speed: 663.11 samples/sec	acc=0.537695
+INFO:root:Epoch[7] Batch [1220]	Speed: 662.92 samples/sec	acc=0.534863
+INFO:root:Epoch[7] Batch [1240]	Speed: 662.70 samples/sec	acc=0.532715
+INFO:root:Epoch[7] Batch [1260]	Speed: 664.53 samples/sec	acc=0.534375
+INFO:root:Epoch[7] Batch [1280]	Speed: 683.48 samples/sec	acc=0.534766
+INFO:root:Epoch[7] Batch [1300]	Speed: 663.26 samples/sec	acc=0.534277
+INFO:root:Epoch[7] Batch [1320]	Speed: 662.50 samples/sec	acc=0.550098
+INFO:root:Epoch[7] Batch [1340]	Speed: 662.82 samples/sec	acc=0.545410
+INFO:root:Epoch[7] Batch [1360]	Speed: 662.96 samples/sec	acc=0.543066
+[54000][MARGIN]0.775398,0.382445
+lr-batch-epoch: 0.01 1366 7
+testing verification..
+(14000, 512)
+infer time 16.71862
+[cfp_ff][54000]XNorm: 21.741931
+[cfp_ff][54000]Accuracy-Flip: 0.99600+-0.00262
+testing verification..
+(14000, 512)
+infer time 20.062862
+[cfp_fp][54000]XNorm: 18.540964
+[cfp_fp][54000]Accuracy-Flip: 0.89271+-0.01541
+testing verification..
+(12000, 512)
+infer time 16.860989
+[agedb_30][54000]XNorm: 21.926127
+[agedb_30][54000]Accuracy-Flip: 0.96850+-0.00693
+testing verification..
+(12000, 512)
+infer time 16.569762
+[lfw][54000]XNorm: 21.878332
+[lfw][54000]Accuracy-Flip: 0.99683+-0.00217
+saving 27
+INFO:root:Saved checkpoint to "../model-r50-am-lfw/model-0027.params"
+[54000]Accuracy-Highest: 0.99683
+INFO:root:Epoch[7] Batch [1380]	Speed: 91.55 samples/sec	acc=0.540039
+INFO:root:Epoch[7] Batch [1400]	Speed: 661.47 samples/sec	acc=0.544922
+INFO:root:Epoch[7] Batch [1420]	Speed: 663.12 samples/sec	acc=0.545996
+INFO:root:Epoch[7] Batch [1440]	Speed: 662.23 samples/sec	acc=0.541895
+INFO:root:Epoch[7] Batch [1460]	Speed: 661.89 samples/sec	acc=0.544434
+INFO:root:Epoch[7] Batch [1480]	Speed: 663.06 samples/sec	acc=0.531934
+INFO:root:Epoch[7] Batch [1500]	Speed: 657.32 samples/sec	acc=0.543555
+INFO:root:Epoch[7] Batch [1520]	Speed: 662.67 samples/sec	acc=0.543750
+INFO:root:Epoch[7] Batch [1540]	Speed: 663.48 samples/sec	acc=0.538867
+INFO:root:Epoch[7] Batch [1560]	Speed: 662.85 samples/sec	acc=0.544336
+INFO:root:Epoch[7] Batch [1580]	Speed: 661.97 samples/sec	acc=0.552148
+INFO:root:Epoch[7] Batch [1600]	Speed: 657.65 samples/sec	acc=0.537988
+INFO:root:Epoch[7] Batch [1620]	Speed: 662.40 samples/sec	acc=0.546973
+INFO:root:Epoch[7] Batch [1640]	Speed: 662.50 samples/sec	acc=0.548145
+INFO:root:Epoch[7] Batch [1660]	Speed: 657.64 samples/sec	acc=0.551758
+INFO:root:Epoch[7] Batch [1680]	Speed: 657.15 samples/sec	acc=0.538184
+INFO:root:Epoch[7] Batch [1700]	Speed: 663.33 samples/sec	acc=0.543848
+INFO:root:Epoch[7] Batch [1720]	Speed: 664.19 samples/sec	acc=0.548438
+INFO:root:Epoch[7] Batch [1740]	Speed: 663.08 samples/sec	acc=0.538867
+INFO:root:Epoch[7] Batch [1760]	Speed: 662.19 samples/sec	acc=0.543555
+INFO:root:Epoch[7] Batch [1780]	Speed: 663.11 samples/sec	acc=0.546094
+INFO:root:Epoch[7] Batch [1800]	Speed: 663.32 samples/sec	acc=0.540430
+INFO:root:Epoch[7] Batch [1820]	Speed: 662.51 samples/sec	acc=0.538086
+INFO:root:Epoch[7] Batch [1840]	Speed: 663.27 samples/sec	acc=0.548828
+INFO:root:Epoch[7] Batch [1860]	Speed: 662.65 samples/sec	acc=0.539844
+INFO:root:Epoch[7] Batch [1880]	Speed: 662.60 samples/sec	acc=0.543457
+INFO:root:Epoch[7] Batch [1900]	Speed: 663.18 samples/sec	acc=0.543652
+INFO:root:Epoch[7] Batch [1920]	Speed: 662.38 samples/sec	acc=0.554395
+INFO:root:Epoch[7] Batch [1940]	Speed: 663.46 samples/sec	acc=0.543945
+INFO:root:Epoch[7] Batch [1960]	Speed: 662.48 samples/sec	acc=0.552344
+INFO:root:Epoch[7] Batch [1980]	Speed: 665.00 samples/sec	acc=0.554785
+INFO:root:Epoch[7] Batch [2000]	Speed: 662.69 samples/sec	acc=0.551758
+INFO:root:Epoch[7] Batch [2020]	Speed: 662.52 samples/sec	acc=0.547559
+INFO:root:Epoch[7] Batch [2040]	Speed: 662.63 samples/sec	acc=0.547363
+INFO:root:Epoch[7] Batch [2060]	Speed: 663.16 samples/sec	acc=0.546875
+INFO:root:Epoch[7] Batch [2080]	Speed: 663.27 samples/sec	acc=0.545996
+INFO:root:Epoch[7] Batch [2100]	Speed: 663.18 samples/sec	acc=0.551270
+INFO:root:Epoch[7] Batch [2120]	Speed: 662.36 samples/sec	acc=0.552246
+INFO:root:Epoch[7] Batch [2140]	Speed: 662.86 samples/sec	acc=0.546973
+INFO:root:Epoch[7] Batch [2160]	Speed: 662.17 samples/sec	acc=0.553125
+INFO:root:Epoch[7] Batch [2180]	Speed: 662.46 samples/sec	acc=0.553516
+INFO:root:Epoch[7] Batch [2200]	Speed: 663.52 samples/sec	acc=0.553027
+INFO:root:Epoch[7] Batch [2220]	Speed: 662.72 samples/sec	acc=0.549316
+INFO:root:Epoch[7] Batch [2240]	Speed: 662.83 samples/sec	acc=0.543750
+INFO:root:Epoch[7] Batch [2260]	Speed: 663.20 samples/sec	acc=0.553027
+INFO:root:Epoch[7] Batch [2280]	Speed: 662.76 samples/sec	acc=0.559473
+INFO:root:Epoch[7] Batch [2300]	Speed: 662.83 samples/sec	acc=0.552148
+INFO:root:Epoch[7] Batch [2320]	Speed: 663.47 samples/sec	acc=0.553516
+INFO:root:Epoch[7] Batch [2340]	Speed: 662.57 samples/sec	acc=0.546875
+INFO:root:Epoch[7] Batch [2360]	Speed: 662.05 samples/sec	acc=0.556738
+lr-batch-epoch: 0.01 2366 7
+INFO:root:Epoch[7] Batch [2380]	Speed: 662.67 samples/sec	acc=0.548730
+INFO:root:Epoch[7] Batch [2400]	Speed: 662.44 samples/sec	acc=0.553223
+INFO:root:Epoch[7] Batch [2420]	Speed: 662.69 samples/sec	acc=0.552148
+INFO:root:Epoch[7] Batch [2440]	Speed: 662.60 samples/sec	acc=0.548828
+INFO:root:Epoch[7] Batch [2460]	Speed: 662.86 samples/sec	acc=0.552637
+INFO:root:Epoch[7] Batch [2480]	Speed: 662.85 samples/sec	acc=0.554395
+INFO:root:Epoch[7] Batch [2500]	Speed: 663.20 samples/sec	acc=0.557910
+INFO:root:Epoch[7] Batch [2520]	Speed: 662.90 samples/sec	acc=0.548145
+INFO:root:Epoch[7] Batch [2540]	Speed: 662.80 samples/sec	acc=0.555957
+INFO:root:Epoch[7] Batch [2560]	Speed: 661.57 samples/sec	acc=0.552930
+INFO:root:Epoch[7] Batch [2580]	Speed: 663.33 samples/sec	acc=0.547363
+INFO:root:Epoch[7] Batch [2600]	Speed: 661.64 samples/sec	acc=0.559961
+INFO:root:Epoch[7] Batch [2620]	Speed: 663.17 samples/sec	acc=0.558008
+INFO:root:Epoch[7] Batch [2640]	Speed: 662.61 samples/sec	acc=0.549121
+INFO:root:Epoch[7] Batch [2660]	Speed: 662.87 samples/sec	acc=0.548242
+INFO:root:Epoch[7] Batch [2680]	Speed: 663.34 samples/sec	acc=0.560059
+INFO:root:Epoch[7] Batch [2700]	Speed: 663.18 samples/sec	acc=0.560840
+INFO:root:Epoch[7] Batch [2720]	Speed: 662.61 samples/sec	acc=0.555957
+INFO:root:Epoch[7] Batch [2740]	Speed: 663.05 samples/sec	acc=0.548730
+INFO:root:Epoch[7] Batch [2760]	Speed: 662.94 samples/sec	acc=0.546875
+INFO:root:Epoch[7] Batch [2780]	Speed: 663.00 samples/sec	acc=0.547461
+INFO:root:Epoch[7] Batch [2800]	Speed: 663.56 samples/sec	acc=0.543750
+INFO:root:Epoch[7] Batch [2820]	Speed: 663.41 samples/sec	acc=0.565430
+INFO:root:Epoch[7] Batch [2840]	Speed: 662.85 samples/sec	acc=0.554688
+INFO:root:Epoch[7] Batch [2860]	Speed: 663.34 samples/sec	acc=0.554785
+INFO:root:Epoch[7] Batch [2880]	Speed: 662.57 samples/sec	acc=0.555762
+INFO:root:Epoch[7] Batch [2900]	Speed: 662.86 samples/sec	acc=0.561426
+INFO:root:Epoch[7] Batch [2920]	Speed: 662.44 samples/sec	acc=0.566211
+INFO:root:Epoch[7] Batch [2940]	Speed: 662.34 samples/sec	acc=0.551367
+INFO:root:Epoch[7] Batch [2960]	Speed: 662.46 samples/sec	acc=0.555469
+INFO:root:Epoch[7] Batch [2980]	Speed: 662.83 samples/sec	acc=0.559863
+INFO:root:Epoch[7] Batch [3000]	Speed: 662.79 samples/sec	acc=0.556348
+INFO:root:Epoch[7] Batch [3020]	Speed: 661.89 samples/sec	acc=0.565137
+INFO:root:Epoch[7] Batch [3040]	Speed: 662.90 samples/sec	acc=0.560254
+INFO:root:Epoch[7] Batch [3060]	Speed: 662.85 samples/sec	acc=0.555957
+INFO:root:Epoch[7] Batch [3080]	Speed: 662.76 samples/sec	acc=0.551953
+INFO:root:Epoch[7] Batch [3100]	Speed: 662.42 samples/sec	acc=0.559082
+INFO:root:Epoch[7] Batch [3120]	Speed: 662.71 samples/sec	acc=0.551660
+INFO:root:Epoch[7] Batch [3140]	Speed: 663.09 samples/sec	acc=0.564453
+INFO:root:Epoch[7] Batch [3160]	Speed: 662.15 samples/sec	acc=0.559961
+INFO:root:Epoch[7] Batch [3180]	Speed: 662.73 samples/sec	acc=0.558984
+INFO:root:Epoch[7] Batch [3200]	Speed: 662.85 samples/sec	acc=0.550977
+INFO:root:Epoch[7] Batch [3220]	Speed: 661.50 samples/sec	acc=0.560156
+INFO:root:Epoch[7] Batch [3240]	Speed: 663.03 samples/sec	acc=0.552441
+INFO:root:Epoch[7] Batch [3260]	Speed: 662.21 samples/sec	acc=0.560840
+INFO:root:Epoch[7] Batch [3280]	Speed: 662.11 samples/sec	acc=0.557031
+INFO:root:Epoch[7] Batch [3300]	Speed: 662.16 samples/sec	acc=0.553125
+INFO:root:Epoch[7] Batch [3320]	Speed: 662.30 samples/sec	acc=0.557520
+INFO:root:Epoch[7] Batch [3340]	Speed: 662.42 samples/sec	acc=0.560352
+INFO:root:Epoch[7] Batch [3360]	Speed: 662.02 samples/sec	acc=0.559766
+[56000][MARGIN]0.766567,0.369890
+lr-batch-epoch: 0.01 3366 7
+testing verification..
+(14000, 512)
+infer time 16.51119
+[cfp_ff][56000]XNorm: 22.216972
+[cfp_ff][56000]Accuracy-Flip: 0.99714+-0.00286
+testing verification..
+(14000, 512)
+infer time 16.271374
+[cfp_fp][56000]XNorm: 19.052099
+[cfp_fp][56000]Accuracy-Flip: 0.89857+-0.01664
+testing verification..
+(12000, 512)
+infer time 13.25264
+[agedb_30][56000]XNorm: 22.501946
+[agedb_30][56000]Accuracy-Flip: 0.96817+-0.00769
+testing verification..
+(12000, 512)
+infer time 12.692111
+[lfw][56000]XNorm: 22.469747
+[lfw][56000]Accuracy-Flip: 0.99683+-0.00273
+[56000]Accuracy-Highest: 0.99683
+INFO:root:Epoch[7] Batch [3380]	Speed: 109.23 samples/sec	acc=0.564941
+INFO:root:Epoch[7] Batch [3400]	Speed: 682.15 samples/sec	acc=0.564746
+INFO:root:Epoch[7] Batch [3420]	Speed: 679.39 samples/sec	acc=0.559473
+INFO:root:Epoch[7] Batch [3440]	Speed: 663.74 samples/sec	acc=0.569043
+INFO:root:Epoch[7] Batch [3460]	Speed: 663.19 samples/sec	acc=0.555371
+INFO:root:Epoch[7] Batch [3480]	Speed: 662.48 samples/sec	acc=0.554590
+INFO:root:Epoch[7] Batch [3500]	Speed: 662.13 samples/sec	acc=0.560547
+INFO:root:Epoch[7] Batch [3520]	Speed: 658.64 samples/sec	acc=0.562500
+INFO:root:Epoch[7] Batch [3540]	Speed: 669.50 samples/sec	acc=0.555273
+INFO:root:Epoch[7] Batch [3560]	Speed: 663.13 samples/sec	acc=0.560742
+INFO:root:Epoch[7] Batch [3580]	Speed: 663.01 samples/sec	acc=0.569922
+INFO:root:Epoch[7] Batch [3600]	Speed: 662.88 samples/sec	acc=0.561328
+INFO:root:Epoch[7] Batch [3620]	Speed: 662.45 samples/sec	acc=0.560449
+INFO:root:Epoch[7] Batch [3640]	Speed: 662.72 samples/sec	acc=0.565527
+INFO:root:Epoch[7] Batch [3660]	Speed: 662.76 samples/sec	acc=0.563379
+INFO:root:Epoch[7] Batch [3680]	Speed: 663.08 samples/sec	acc=0.556152
+INFO:root:Epoch[7] Batch [3700]	Speed: 663.33 samples/sec	acc=0.554883
+INFO:root:Epoch[7] Batch [3720]	Speed: 662.68 samples/sec	acc=0.553711
+INFO:root:Epoch[7] Batch [3740]	Speed: 662.79 samples/sec	acc=0.563867
+INFO:root:Epoch[7] Batch [3760]	Speed: 663.54 samples/sec	acc=0.563770
+INFO:root:Epoch[7] Batch [3780]	Speed: 662.54 samples/sec	acc=0.565723
+INFO:root:Epoch[7] Batch [3800]	Speed: 663.19 samples/sec	acc=0.558398
+INFO:root:Epoch[7] Batch [3820]	Speed: 662.68 samples/sec	acc=0.555176
+INFO:root:Epoch[7] Batch [3840]	Speed: 662.74 samples/sec	acc=0.558887
+INFO:root:Epoch[7] Batch [3860]	Speed: 663.54 samples/sec	acc=0.563086
+INFO:root:Epoch[7] Batch [3880]	Speed: 663.39 samples/sec	acc=0.558594
+INFO:root:Epoch[7] Batch [3900]	Speed: 663.33 samples/sec	acc=0.561621
+INFO:root:Epoch[7] Batch [3920]	Speed: 663.21 samples/sec	acc=0.566211
+INFO:root:Epoch[7] Batch [3940]	Speed: 663.09 samples/sec	acc=0.564453
+INFO:root:Epoch[7] Batch [3960]	Speed: 662.81 samples/sec	acc=0.560254
+INFO:root:Epoch[7] Batch [3980]	Speed: 663.26 samples/sec	acc=0.569434
+INFO:root:Epoch[7] Batch [4000]	Speed: 663.27 samples/sec	acc=0.569141
+INFO:root:Epoch[7] Batch [4020]	Speed: 662.82 samples/sec	acc=0.573242
+INFO:root:Epoch[7] Batch [4040]	Speed: 662.83 samples/sec	acc=0.567676
+INFO:root:Epoch[7] Batch [4060]	Speed: 663.42 samples/sec	acc=0.558691
+INFO:root:Epoch[7] Batch [4080]	Speed: 663.15 samples/sec	acc=0.564844
+INFO:root:Epoch[7] Batch [4100]	Speed: 662.78 samples/sec	acc=0.561230
+INFO:root:Epoch[7] Batch [4120]	Speed: 663.27 samples/sec	acc=0.560840
+INFO:root:Epoch[7] Batch [4140]	Speed: 663.65 samples/sec	acc=0.567871
+INFO:root:Epoch[7] Batch [4160]	Speed: 662.98 samples/sec	acc=0.566797
+INFO:root:Epoch[7] Batch [4180]	Speed: 662.90 samples/sec	acc=0.563965
+INFO:root:Epoch[7] Batch [4200]	Speed: 663.56 samples/sec	acc=0.558594
+INFO:root:Epoch[7] Batch [4220]	Speed: 662.66 samples/sec	acc=0.565918
+INFO:root:Epoch[7] Batch [4240]	Speed: 662.53 samples/sec	acc=0.566895
+INFO:root:Epoch[7] Batch [4260]	Speed: 662.27 samples/sec	acc=0.568945
+INFO:root:Epoch[7] Batch [4280]	Speed: 663.32 samples/sec	acc=0.568848
+INFO:root:Epoch[7] Batch [4300]	Speed: 663.25 samples/sec	acc=0.568164
+INFO:root:Epoch[7] Batch [4320]	Speed: 666.19 samples/sec	acc=0.573340
+INFO:root:Epoch[7] Batch [4340]	Speed: 662.84 samples/sec	acc=0.569336
+INFO:root:Epoch[7] Batch [4360]	Speed: 662.21 samples/sec	acc=0.563867
+lr-batch-epoch: 0.01 4366 7
+INFO:root:Epoch[7] Batch [4380]	Speed: 663.10 samples/sec	acc=0.562012
+INFO:root:Epoch[7] Batch [4400]	Speed: 662.20 samples/sec	acc=0.559570
+INFO:root:Epoch[7] Batch [4420]	Speed: 663.43 samples/sec	acc=0.568652
+INFO:root:Epoch[7] Batch [4440]	Speed: 663.31 samples/sec	acc=0.562012
+INFO:root:Epoch[7] Batch [4460]	Speed: 662.77 samples/sec	acc=0.564453
+INFO:root:Epoch[7] Batch [4480]	Speed: 663.17 samples/sec	acc=0.572559
+INFO:root:Epoch[7] Batch [4500]	Speed: 662.90 samples/sec	acc=0.559277
+INFO:root:Epoch[7] Batch [4520]	Speed: 662.83 samples/sec	acc=0.556738
+INFO:root:Epoch[7] Batch [4540]	Speed: 662.90 samples/sec	acc=0.559766
+INFO:root:Epoch[7] Batch [4560]	Speed: 662.50 samples/sec	acc=0.562109
+INFO:root:Epoch[7] Batch [4580]	Speed: 662.70 samples/sec	acc=0.567383
+INFO:root:Epoch[7] Batch [4600]	Speed: 663.10 samples/sec	acc=0.562695
+INFO:root:Epoch[7] Batch [4620]	Speed: 663.27 samples/sec	acc=0.559277
+INFO:root:Epoch[7] Batch [4640]	Speed: 662.70 samples/sec	acc=0.569043
+INFO:root:Epoch[7] Batch [4660]	Speed: 663.18 samples/sec	acc=0.571680
+INFO:root:Epoch[7] Batch [4680]	Speed: 663.11 samples/sec	acc=0.561133
+INFO:root:Epoch[7] Batch [4700]	Speed: 663.14 samples/sec	acc=0.565820
+INFO:root:Epoch[7] Batch [4720]	Speed: 663.20 samples/sec	acc=0.558887
+INFO:root:Epoch[7] Batch [4740]	Speed: 662.66 samples/sec	acc=0.566113
+INFO:root:Epoch[7] Batch [4760]	Speed: 663.27 samples/sec	acc=0.573828
+INFO:root:Epoch[7] Batch [4780]	Speed: 662.69 samples/sec	acc=0.561719
+INFO:root:Epoch[7] Batch [4800]	Speed: 663.29 samples/sec	acc=0.566406
+INFO:root:Epoch[7] Batch [4820]	Speed: 662.53 samples/sec	acc=0.556055
+INFO:root:Epoch[7] Batch [4840]	Speed: 662.92 samples/sec	acc=0.566406
+INFO:root:Epoch[7] Batch [4860]	Speed: 663.60 samples/sec	acc=0.551660
+INFO:root:Epoch[7] Batch [4880]	Speed: 662.14 samples/sec	acc=0.564648
+INFO:root:Epoch[7] Batch [4900]	Speed: 663.08 samples/sec	acc=0.563184
+INFO:root:Epoch[7] Batch [4920]	Speed: 663.13 samples/sec	acc=0.562402
+INFO:root:Epoch[7] Batch [4940]	Speed: 663.14 samples/sec	acc=0.561621
+INFO:root:Epoch[7] Batch [4960]	Speed: 663.71 samples/sec	acc=0.566992
+INFO:root:Epoch[7] Batch [4980]	Speed: 663.25 samples/sec	acc=0.564941
+INFO:root:Epoch[7] Batch [5000]	Speed: 663.42 samples/sec	acc=0.564746
+INFO:root:Epoch[7] Batch [5020]	Speed: 662.01 samples/sec	acc=0.564746
+INFO:root:Epoch[7] Batch [5040]	Speed: 662.94 samples/sec	acc=0.563477
+INFO:root:Epoch[7] Batch [5060]	Speed: 662.70 samples/sec	acc=0.568555
+INFO:root:Epoch[7] Batch [5080]	Speed: 663.13 samples/sec	acc=0.561035
+INFO:root:Epoch[7] Batch [5100]	Speed: 663.31 samples/sec	acc=0.561133
+INFO:root:Epoch[7] Batch [5120]	Speed: 662.60 samples/sec	acc=0.560254
+INFO:root:Epoch[7] Batch [5140]	Speed: 663.65 samples/sec	acc=0.567578
+INFO:root:Epoch[7] Batch [5160]	Speed: 662.53 samples/sec	acc=0.563477
+INFO:root:Epoch[7] Batch [5180]	Speed: 663.18 samples/sec	acc=0.567969
+INFO:root:Epoch[7] Batch [5200]	Speed: 663.63 samples/sec	acc=0.574219
+INFO:root:Epoch[7] Batch [5220]	Speed: 664.17 samples/sec	acc=0.564551
+INFO:root:Epoch[7] Batch [5240]	Speed: 663.71 samples/sec	acc=0.561621
+INFO:root:Epoch[7] Batch [5260]	Speed: 663.66 samples/sec	acc=0.566504
+INFO:root:Epoch[7] Batch [5280]	Speed: 663.67 samples/sec	acc=0.565527
+INFO:root:Epoch[7] Batch [5300]	Speed: 663.17 samples/sec	acc=0.557422
+INFO:root:Epoch[7] Batch [5320]	Speed: 662.68 samples/sec	acc=0.567090
+INFO:root:Epoch[7] Batch [5340]	Speed: 662.38 samples/sec	acc=0.574023
+INFO:root:Epoch[7] Batch [5360]	Speed: 663.03 samples/sec	acc=0.565820
+[58000][MARGIN]0.770447,0.375042
+lr-batch-epoch: 0.01 5366 7
+testing verification..
+(14000, 512)
+infer time 15.049288
+[cfp_ff][58000]XNorm: 22.002785
+[cfp_ff][58000]Accuracy-Flip: 0.99643+-0.00280
+testing verification..
+(14000, 512)
+infer time 16.034628
+[cfp_fp][58000]XNorm: 19.186742
+[cfp_fp][58000]Accuracy-Flip: 0.90157+-0.01607
+testing verification..
+(12000, 512)
+infer time 12.954738
+[agedb_30][58000]XNorm: 22.275353
+[agedb_30][58000]Accuracy-Flip: 0.96983+-0.00524
+testing verification..
+(12000, 512)
+infer time 12.860242
+[lfw][58000]XNorm: 22.386572
+[lfw][58000]Accuracy-Flip: 0.99533+-0.00356
+[58000]Accuracy-Highest: 0.99683
+INFO:root:Epoch[7] Batch [5380]	Speed: 112.08 samples/sec	acc=0.566992
+INFO:root:Epoch[7] Batch [5400]	Speed: 662.09 samples/sec	acc=0.567090
+INFO:root:Epoch[7] Batch [5420]	Speed: 662.21 samples/sec	acc=0.570020
+INFO:root:Epoch[7] Batch [5440]	Speed: 662.80 samples/sec	acc=0.559375
+INFO:root:Epoch[7] Batch [5460]	Speed: 662.08 samples/sec	acc=0.567871
+INFO:root:Epoch[7] Batch [5480]	Speed: 662.28 samples/sec	acc=0.560254
+INFO:root:Epoch[7] Batch [5500]	Speed: 662.19 samples/sec	acc=0.571680
+INFO:root:Epoch[7] Batch [5520]	Speed: 662.29 samples/sec	acc=0.555664
+INFO:root:Epoch[7] Batch [5540]	Speed: 662.10 samples/sec	acc=0.562988
+INFO:root:Epoch[7] Batch [5560]	Speed: 662.11 samples/sec	acc=0.569043
+INFO:root:Epoch[7] Batch [5580]	Speed: 662.40 samples/sec	acc=0.566992
+INFO:root:Epoch[7] Batch [5600]	Speed: 662.02 samples/sec	acc=0.564160
+INFO:root:Epoch[7] Batch [5620]	Speed: 662.04 samples/sec	acc=0.565820
+INFO:root:Epoch[7] Batch [5640]	Speed: 661.81 samples/sec	acc=0.562305
+INFO:root:Epoch[7] Batch [5660]	Speed: 662.65 samples/sec	acc=0.562305
+INFO:root:Epoch[7] Batch [5680]	Speed: 663.26 samples/sec	acc=0.567871
+INFO:root:Epoch[7] Batch [5700]	Speed: 663.16 samples/sec	acc=0.563281
+INFO:root:Epoch[7] Batch [5720]	Speed: 662.99 samples/sec	acc=0.561621
+INFO:root:Epoch[7] Batch [5740]	Speed: 662.33 samples/sec	acc=0.563281
+INFO:root:Epoch[7] Batch [5760]	Speed: 663.01 samples/sec	acc=0.558887
+INFO:root:Epoch[7] Batch [5780]	Speed: 662.58 samples/sec	acc=0.563477
+INFO:root:Epoch[7] Batch [5800]	Speed: 663.19 samples/sec	acc=0.562305
+INFO:root:Epoch[7] Batch [5820]	Speed: 662.43 samples/sec	acc=0.563965
+INFO:root:Epoch[7] Batch [5840]	Speed: 663.04 samples/sec	acc=0.565527
+INFO:root:Epoch[7] Batch [5860]	Speed: 663.36 samples/sec	acc=0.559180
+INFO:root:Epoch[7] Batch [5880]	Speed: 663.33 samples/sec	acc=0.560156
+INFO:root:Epoch[7] Batch [5900]	Speed: 662.62 samples/sec	acc=0.565918
+INFO:root:Epoch[7] Batch [5920]	Speed: 662.16 samples/sec	acc=0.555762
+INFO:root:Epoch[7] Batch [5940]	Speed: 677.51 samples/sec	acc=0.561426
+INFO:root:Epoch[7] Batch [5960]	Speed: 662.67 samples/sec	acc=0.558008
+INFO:root:Epoch[7] Batch [5980]	Speed: 662.93 samples/sec	acc=0.567676
+INFO:root:Epoch[7] Batch [6000]	Speed: 662.95 samples/sec	acc=0.566016
+INFO:root:Epoch[7] Batch [6020]	Speed: 662.33 samples/sec	acc=0.562695
+INFO:root:Epoch[7] Batch [6040]	Speed: 655.95 samples/sec	acc=0.559082
+INFO:root:Epoch[7] Batch [6060]	Speed: 661.90 samples/sec	acc=0.571973
+INFO:root:Epoch[7] Batch [6080]	Speed: 661.77 samples/sec	acc=0.561133
+INFO:root:Epoch[7] Batch [6100]	Speed: 662.14 samples/sec	acc=0.566406
+INFO:root:Epoch[7] Batch [6120]	Speed: 662.32 samples/sec	acc=0.562695
+INFO:root:Epoch[7] Batch [6140]	Speed: 662.70 samples/sec	acc=0.566797
+INFO:root:Epoch[7] Batch [6160]	Speed: 662.76 samples/sec	acc=0.559570
+INFO:root:Epoch[7] Batch [6180]	Speed: 663.09 samples/sec	acc=0.559375
+INFO:root:Epoch[7] Batch [6200]	Speed: 662.51 samples/sec	acc=0.568848
+INFO:root:Epoch[7] Batch [6220]	Speed: 663.49 samples/sec	acc=0.570801
+INFO:root:Epoch[7] Batch [6240]	Speed: 662.45 samples/sec	acc=0.562598
+INFO:root:Epoch[7] Batch [6260]	Speed: 662.56 samples/sec	acc=0.563867
+INFO:root:Epoch[7] Batch [6280]	Speed: 662.48 samples/sec	acc=0.554297
+INFO:root:Epoch[7] Batch [6300]	Speed: 663.36 samples/sec	acc=0.559863
+INFO:root:Epoch[7] Batch [6320]	Speed: 662.34 samples/sec	acc=0.553027
+INFO:root:Epoch[7] Batch [6340]	Speed: 662.78 samples/sec	acc=0.573242
+INFO:root:Epoch[7] Batch [6360]	Speed: 663.15 samples/sec	acc=0.553613
+lr-batch-epoch: 0.01 6366 7
+INFO:root:Epoch[7] Batch [6380]	Speed: 662.90 samples/sec	acc=0.564551
+INFO:root:Epoch[7] Batch [6400]	Speed: 662.54 samples/sec	acc=0.563770
+INFO:root:Epoch[7] Batch [6420]	Speed: 662.67 samples/sec	acc=0.563477
+INFO:root:Epoch[7] Batch [6440]	Speed: 663.15 samples/sec	acc=0.570508
+INFO:root:Epoch[7] Batch [6460]	Speed: 662.33 samples/sec	acc=0.566602
+INFO:root:Epoch[7] Batch [6480]	Speed: 663.59 samples/sec	acc=0.555469
+INFO:root:Epoch[7] Batch [6500]	Speed: 662.00 samples/sec	acc=0.566016
+INFO:root:Epoch[7] Batch [6520]	Speed: 663.49 samples/sec	acc=0.551270
+INFO:root:Epoch[7] Batch [6540]	Speed: 662.65 samples/sec	acc=0.559180
+INFO:root:Epoch[7] Batch [6560]	Speed: 662.63 samples/sec	acc=0.566406
+INFO:root:Epoch[7] Batch [6580]	Speed: 662.58 samples/sec	acc=0.554785
+INFO:root:Epoch[7] Batch [6600]	Speed: 662.78 samples/sec	acc=0.564844
+INFO:root:Epoch[7] Batch [6620]	Speed: 662.69 samples/sec	acc=0.567480
+INFO:root:Epoch[7] Batch [6640]	Speed: 662.88 samples/sec	acc=0.569238
+INFO:root:Epoch[7] Batch [6660]	Speed: 662.59 samples/sec	acc=0.559766
+INFO:root:Epoch[7] Batch [6680]	Speed: 663.37 samples/sec	acc=0.568164
+INFO:root:Epoch[7] Batch [6700]	Speed: 658.12 samples/sec	acc=0.567480
+INFO:root:Epoch[7] Batch [6720]	Speed: 667.62 samples/sec	acc=0.561523
+INFO:root:Epoch[7] Batch [6740]	Speed: 663.36 samples/sec	acc=0.567090
+INFO:root:Epoch[7] Batch [6760]	Speed: 663.23 samples/sec	acc=0.559961
+INFO:root:Epoch[7] Batch [6780]	Speed: 663.26 samples/sec	acc=0.556543
+INFO:root:Epoch[7] Batch [6800]	Speed: 662.93 samples/sec	acc=0.567090
+INFO:root:Epoch[7] Batch [6820]	Speed: 662.99 samples/sec	acc=0.563867
+INFO:root:Epoch[7] Batch [6840]	Speed: 661.98 samples/sec	acc=0.560547
+INFO:root:Epoch[7] Batch [6860]	Speed: 663.24 samples/sec	acc=0.566699
+INFO:root:Epoch[7] Batch [6880]	Speed: 663.39 samples/sec	acc=0.568457
+INFO:root:Epoch[7] Batch [6900]	Speed: 663.01 samples/sec	acc=0.560156
+INFO:root:Epoch[7] Batch [6920]	Speed: 662.36 samples/sec	acc=0.562305
+INFO:root:Epoch[7] Batch [6940]	Speed: 664.53 samples/sec	acc=0.556055
+INFO:root:Epoch[7] Batch [6960]	Speed: 663.67 samples/sec	acc=0.569629
+INFO:root:Epoch[7] Batch [6980]	Speed: 660.94 samples/sec	acc=0.564648
+INFO:root:Epoch[7] Batch [7000]	Speed: 664.06 samples/sec	acc=0.558301
+INFO:root:Epoch[7] Batch [7020]	Speed: 662.75 samples/sec	acc=0.564941
+INFO:root:Epoch[7] Batch [7040]	Speed: 662.92 samples/sec	acc=0.563281
+INFO:root:Epoch[7] Batch [7060]	Speed: 662.19 samples/sec	acc=0.558008
+INFO:root:Epoch[7] Batch [7080]	Speed: 663.18 samples/sec	acc=0.573047
+INFO:root:Epoch[7] Batch [7100]	Speed: 661.87 samples/sec	acc=0.566211
+INFO:root:Epoch[7] Batch [7120]	Speed: 663.65 samples/sec	acc=0.572070
+INFO:root:Epoch[7] Batch [7140]	Speed: 663.38 samples/sec	acc=0.567090
+INFO:root:Epoch[7] Batch [7160]	Speed: 662.94 samples/sec	acc=0.562402
+INFO:root:Epoch[7] Batch [7180]	Speed: 662.58 samples/sec	acc=0.551855
+INFO:root:Epoch[7] Batch [7200]	Speed: 662.51 samples/sec	acc=0.566602
+INFO:root:Epoch[7] Batch [7220]	Speed: 662.50 samples/sec	acc=0.563086
+INFO:root:Epoch[7] Batch [7240]	Speed: 661.71 samples/sec	acc=0.567285
+INFO:root:Epoch[7] Batch [7260]	Speed: 663.15 samples/sec	acc=0.569141
+INFO:root:Epoch[7] Batch [7280]	Speed: 661.54 samples/sec	acc=0.561230
+INFO:root:Epoch[7] Batch [7300]	Speed: 663.32 samples/sec	acc=0.555078
+INFO:root:Epoch[7] Batch [7320]	Speed: 661.87 samples/sec	acc=0.561328
+INFO:root:Epoch[7] Batch [7340]	Speed: 662.68 samples/sec	acc=0.552832
+INFO:root:Epoch[7] Batch [7360]	Speed: 662.83 samples/sec	acc=0.557227
+[60000][MARGIN]0.769198,0.372772
+lr-batch-epoch: 0.01 7366 7
+testing verification..
+(14000, 512)
+infer time 16.559355
+[cfp_ff][60000]XNorm: 21.482792
+[cfp_ff][60000]Accuracy-Flip: 0.99700+-0.00216
+testing verification..
+(14000, 512)
+infer time 16.091501
+[cfp_fp][60000]XNorm: 18.637491
+[cfp_fp][60000]Accuracy-Flip: 0.89843+-0.01690
+testing verification..
+(12000, 512)
+infer time 14.041797
+[agedb_30][60000]XNorm: 21.774649
+[agedb_30][60000]Accuracy-Flip: 0.97100+-0.00696
+testing verification..
+(12000, 512)
+infer time 12.773174
+[lfw][60000]XNorm: 21.725741
+[lfw][60000]Accuracy-Flip: 0.99650+-0.00337
+[60000]Accuracy-Highest: 0.99683
+INFO:root:Epoch[7] Batch [7380]	Speed: 108.26 samples/sec	acc=0.570703
+INFO:root:Epoch[7] Batch [7400]	Speed: 662.95 samples/sec	acc=0.553418
+INFO:root:Epoch[7] Batch [7420]	Speed: 662.13 samples/sec	acc=0.560059
+INFO:root:Epoch[7] Batch [7440]	Speed: 661.59 samples/sec	acc=0.558008
+INFO:root:Epoch[7] Batch [7460]	Speed: 662.32 samples/sec	acc=0.567969
+INFO:root:Epoch[7] Batch [7480]	Speed: 662.22 samples/sec	acc=0.562402
+INFO:root:Epoch[7] Batch [7500]	Speed: 662.55 samples/sec	acc=0.557324
+INFO:root:Epoch[7] Train-acc=0.556966
+INFO:root:Epoch[7] Time cost=6136.256
+call reset()
+INFO:root:Epoch[8] Batch [20]	Speed: 661.37 samples/sec	acc=0.621280
+INFO:root:Epoch[8] Batch [40]	Speed: 662.02 samples/sec	acc=0.629395
+INFO:root:Epoch[8] Batch [60]	Speed: 661.46 samples/sec	acc=0.623437
+INFO:root:Epoch[8] Batch [80]	Speed: 661.96 samples/sec	acc=0.627734
+INFO:root:Epoch[8] Batch [100]	Speed: 662.54 samples/sec	acc=0.627930
+INFO:root:Epoch[8] Batch [120]	Speed: 662.20 samples/sec	acc=0.626172
+INFO:root:Epoch[8] Batch [140]	Speed: 661.02 samples/sec	acc=0.625195
+INFO:root:Epoch[8] Batch [160]	Speed: 661.96 samples/sec	acc=0.625391
+INFO:root:Epoch[8] Batch [180]	Speed: 660.91 samples/sec	acc=0.626172
+INFO:root:Epoch[8] Batch [200]	Speed: 662.66 samples/sec	acc=0.615039
+INFO:root:Epoch[8] Batch [220]	Speed: 661.65 samples/sec	acc=0.616016
+INFO:root:Epoch[8] Batch [240]	Speed: 661.94 samples/sec	acc=0.626855
+INFO:root:Epoch[8] Batch [260]	Speed: 661.40 samples/sec	acc=0.621191
+INFO:root:Epoch[8] Batch [280]	Speed: 659.96 samples/sec	acc=0.622852
+INFO:root:Epoch[8] Batch [300]	Speed: 667.73 samples/sec	acc=0.621680
+INFO:root:Epoch[8] Batch [320]	Speed: 661.89 samples/sec	acc=0.616992
+INFO:root:Epoch[8] Batch [340]	Speed: 662.09 samples/sec	acc=0.609668
+INFO:root:Epoch[8] Batch [360]	Speed: 662.16 samples/sec	acc=0.613379
+INFO:root:Epoch[8] Batch [380]	Speed: 661.77 samples/sec	acc=0.614258
+INFO:root:Epoch[8] Batch [400]	Speed: 661.18 samples/sec	acc=0.621484
+INFO:root:Epoch[8] Batch [420]	Speed: 662.60 samples/sec	acc=0.611621
+INFO:root:Epoch[8] Batch [440]	Speed: 661.92 samples/sec	acc=0.613477
+INFO:root:Epoch[8] Batch [460]	Speed: 660.99 samples/sec	acc=0.616113
+INFO:root:Epoch[8] Batch [480]	Speed: 661.69 samples/sec	acc=0.605176
+INFO:root:Epoch[8] Batch [500]	Speed: 661.82 samples/sec	acc=0.609961
+INFO:root:Epoch[8] Batch [520]	Speed: 662.37 samples/sec	acc=0.617090
+INFO:root:Epoch[8] Batch [540]	Speed: 662.26 samples/sec	acc=0.619043
+INFO:root:Epoch[8] Batch [560]	Speed: 661.58 samples/sec	acc=0.609277
+INFO:root:Epoch[8] Batch [580]	Speed: 662.40 samples/sec	acc=0.605762
+INFO:root:Epoch[8] Batch [600]	Speed: 662.98 samples/sec	acc=0.610156
+INFO:root:Epoch[8] Batch [620]	Speed: 662.04 samples/sec	acc=0.610840
+INFO:root:Epoch[8] Batch [640]	Speed: 662.83 samples/sec	acc=0.599805
+INFO:root:Epoch[8] Batch [660]	Speed: 661.74 samples/sec	acc=0.609473
+INFO:root:Epoch[8] Batch [680]	Speed: 662.21 samples/sec	acc=0.608789
+INFO:root:Epoch[8] Batch [700]	Speed: 662.47 samples/sec	acc=0.597949
+INFO:root:Epoch[8] Batch [720]	Speed: 663.14 samples/sec	acc=0.610059
+INFO:root:Epoch[8] Batch [740]	Speed: 663.06 samples/sec	acc=0.600684
+INFO:root:Epoch[8] Batch [760]	Speed: 662.34 samples/sec	acc=0.604980
+INFO:root:Epoch[8] Batch [780]	Speed: 663.13 samples/sec	acc=0.602734
+INFO:root:Epoch[8] Batch [800]	Speed: 662.39 samples/sec	acc=0.605176
+INFO:root:Epoch[8] Batch [820]	Speed: 662.50 samples/sec	acc=0.600586
+INFO:root:Epoch[8] Batch [840]	Speed: 662.41 samples/sec	acc=0.600781
+lr-batch-epoch: 0.01 847 8
+INFO:root:Epoch[8] Batch [860]	Speed: 663.16 samples/sec	acc=0.594824
+INFO:root:Epoch[8] Batch [880]	Speed: 662.29 samples/sec	acc=0.597852
+INFO:root:Epoch[8] Batch [900]	Speed: 662.97 samples/sec	acc=0.585156
+INFO:root:Epoch[8] Batch [920]	Speed: 661.87 samples/sec	acc=0.601953
+INFO:root:Epoch[8] Batch [940]	Speed: 663.31 samples/sec	acc=0.595996
+INFO:root:Epoch[8] Batch [960]	Speed: 662.19 samples/sec	acc=0.597266
+INFO:root:Epoch[8] Batch [980]	Speed: 662.90 samples/sec	acc=0.588867
+INFO:root:Epoch[8] Batch [1000]	Speed: 662.16 samples/sec	acc=0.592480
+INFO:root:Epoch[8] Batch [1020]	Speed: 662.76 samples/sec	acc=0.603320
+INFO:root:Epoch[8] Batch [1040]	Speed: 662.69 samples/sec	acc=0.598926
+INFO:root:Epoch[8] Batch [1060]	Speed: 662.92 samples/sec	acc=0.602148
+INFO:root:Epoch[8] Batch [1080]	Speed: 662.64 samples/sec	acc=0.590527
+INFO:root:Epoch[8] Batch [1100]	Speed: 662.91 samples/sec	acc=0.593164
+INFO:root:Epoch[8] Batch [1120]	Speed: 662.30 samples/sec	acc=0.594922
+INFO:root:Epoch[8] Batch [1140]	Speed: 663.00 samples/sec	acc=0.589258
+INFO:root:Epoch[8] Batch [1160]	Speed: 664.21 samples/sec	acc=0.601465
+INFO:root:Epoch[8] Batch [1180]	Speed: 662.96 samples/sec	acc=0.590625
+INFO:root:Epoch[8] Batch [1200]	Speed: 663.02 samples/sec	acc=0.591797
+INFO:root:Epoch[8] Batch [1220]	Speed: 664.01 samples/sec	acc=0.597559
+INFO:root:Epoch[8] Batch [1240]	Speed: 663.73 samples/sec	acc=0.592090
+INFO:root:Epoch[8] Batch [1260]	Speed: 662.97 samples/sec	acc=0.578809
+INFO:root:Epoch[8] Batch [1280]	Speed: 662.84 samples/sec	acc=0.587598
+INFO:root:Epoch[8] Batch [1300]	Speed: 662.32 samples/sec	acc=0.595215
+INFO:root:Epoch[8] Batch [1320]	Speed: 662.54 samples/sec	acc=0.589355
+INFO:root:Epoch[8] Batch [1340]	Speed: 663.29 samples/sec	acc=0.587793
+INFO:root:Epoch[8] Batch [1360]	Speed: 662.83 samples/sec	acc=0.585059
+INFO:root:Epoch[8] Batch [1380]	Speed: 662.78 samples/sec	acc=0.585840
+INFO:root:Epoch[8] Batch [1400]	Speed: 662.38 samples/sec	acc=0.596289
+INFO:root:Epoch[8] Batch [1420]	Speed: 663.13 samples/sec	acc=0.583789
+INFO:root:Epoch[8] Batch [1440]	Speed: 661.78 samples/sec	acc=0.577637
+INFO:root:Epoch[8] Batch [1460]	Speed: 663.16 samples/sec	acc=0.585645
+INFO:root:Epoch[8] Batch [1480]	Speed: 662.54 samples/sec	acc=0.586328
+INFO:root:Epoch[8] Batch [1500]	Speed: 662.35 samples/sec	acc=0.585742
+INFO:root:Epoch[8] Batch [1520]	Speed: 664.16 samples/sec	acc=0.583496
+INFO:root:Epoch[8] Batch [1540]	Speed: 670.63 samples/sec	acc=0.575879
+INFO:root:Epoch[8] Batch [1560]	Speed: 663.11 samples/sec	acc=0.574316
+INFO:root:Epoch[8] Batch [1580]	Speed: 662.89 samples/sec	acc=0.584375
+INFO:root:Epoch[8] Batch [1600]	Speed: 662.28 samples/sec	acc=0.587109
+INFO:root:Epoch[8] Batch [1620]	Speed: 661.61 samples/sec	acc=0.578223
+INFO:root:Epoch[8] Batch [1640]	Speed: 662.98 samples/sec	acc=0.595020
+INFO:root:Epoch[8] Batch [1660]	Speed: 661.34 samples/sec	acc=0.584961
+INFO:root:Epoch[8] Batch [1680]	Speed: 663.11 samples/sec	acc=0.571973
+INFO:root:Epoch[8] Batch [1700]	Speed: 662.06 samples/sec	acc=0.579004
+INFO:root:Epoch[8] Batch [1720]	Speed: 662.32 samples/sec	acc=0.569336
+INFO:root:Epoch[8] Batch [1740]	Speed: 661.97 samples/sec	acc=0.571094
+INFO:root:Epoch[8] Batch [1760]	Speed: 662.63 samples/sec	acc=0.581738
+INFO:root:Epoch[8] Batch [1780]	Speed: 663.27 samples/sec	acc=0.579297
+INFO:root:Epoch[8] Batch [1800]	Speed: 662.86 samples/sec	acc=0.573242
+INFO:root:Epoch[8] Batch [1820]	Speed: 662.43 samples/sec	acc=0.574902
+INFO:root:Epoch[8] Batch [1840]	Speed: 662.56 samples/sec	acc=0.572949
+[62000][MARGIN]0.773695,0.380082
+lr-batch-epoch: 0.01 1847 8
+testing verification..
+(14000, 512)
+infer time 18.252385
+[cfp_ff][62000]XNorm: 22.598683
+[cfp_ff][62000]Accuracy-Flip: 0.99657+-0.00287
+testing verification..
+(14000, 512)
+infer time 19.152856
+[cfp_fp][62000]XNorm: 19.481398
+[cfp_fp][62000]Accuracy-Flip: 0.91057+-0.01518
+testing verification..
+(12000, 512)
+infer time 14.422528
+[agedb_30][62000]XNorm: 22.709066
+[agedb_30][62000]Accuracy-Flip: 0.96800+-0.00694
+testing verification..
+(12000, 512)
+infer time 13.125019
+[lfw][62000]XNorm: 22.687945
+[lfw][62000]Accuracy-Flip: 0.99650+-0.00273
+[62000]Accuracy-Highest: 0.99683
+INFO:root:Epoch[8] Batch [1860]	Speed: 100.11 samples/sec	acc=0.576562
+INFO:root:Epoch[8] Batch [1880]	Speed: 673.88 samples/sec	acc=0.581543
+INFO:root:Epoch[8] Batch [1900]	Speed: 661.95 samples/sec	acc=0.572363
+INFO:root:Epoch[8] Batch [1920]	Speed: 661.93 samples/sec	acc=0.561230
+INFO:root:Epoch[8] Batch [1940]	Speed: 661.48 samples/sec	acc=0.580273
+INFO:root:Epoch[8] Batch [1960]	Speed: 662.28 samples/sec	acc=0.576660
+INFO:root:Epoch[8] Batch [1980]	Speed: 660.65 samples/sec	acc=0.574902
+INFO:root:Epoch[8] Batch [2000]	Speed: 662.51 samples/sec	acc=0.570410
+INFO:root:Epoch[8] Batch [2020]	Speed: 662.49 samples/sec	acc=0.567285
+INFO:root:Epoch[8] Batch [2040]	Speed: 662.99 samples/sec	acc=0.565137
+INFO:root:Epoch[8] Batch [2060]	Speed: 662.68 samples/sec	acc=0.568945
+INFO:root:Epoch[8] Batch [2080]	Speed: 662.68 samples/sec	acc=0.571094
+INFO:root:Epoch[8] Batch [2100]	Speed: 662.51 samples/sec	acc=0.577734
+INFO:root:Epoch[8] Batch [2120]	Speed: 662.47 samples/sec	acc=0.580957
+INFO:root:Epoch[8] Batch [2140]	Speed: 662.47 samples/sec	acc=0.575098
+INFO:root:Epoch[8] Batch [2160]	Speed: 663.03 samples/sec	acc=0.573828
+INFO:root:Epoch[8] Batch [2180]	Speed: 662.67 samples/sec	acc=0.573242
+INFO:root:Epoch[8] Batch [2200]	Speed: 662.66 samples/sec	acc=0.570898
+INFO:root:Epoch[8] Batch [2220]	Speed: 662.59 samples/sec	acc=0.563965
+INFO:root:Epoch[8] Batch [2240]	Speed: 662.75 samples/sec	acc=0.574512
+INFO:root:Epoch[8] Batch [2260]	Speed: 661.11 samples/sec	acc=0.567676
+INFO:root:Epoch[8] Batch [2280]	Speed: 661.77 samples/sec	acc=0.556055
+INFO:root:Epoch[8] Batch [2300]	Speed: 672.93 samples/sec	acc=0.571582
+INFO:root:Epoch[8] Batch [2320]	Speed: 661.91 samples/sec	acc=0.576953
+INFO:root:Epoch[8] Batch [2340]	Speed: 662.78 samples/sec	acc=0.572363
+INFO:root:Epoch[8] Batch [2360]	Speed: 662.40 samples/sec	acc=0.573633
+INFO:root:Epoch[8] Batch [2380]	Speed: 662.78 samples/sec	acc=0.567480
+INFO:root:Epoch[8] Batch [2400]	Speed: 662.30 samples/sec	acc=0.563281
+INFO:root:Epoch[8] Batch [2420]	Speed: 662.82 samples/sec	acc=0.567578
+INFO:root:Epoch[8] Batch [2440]	Speed: 662.74 samples/sec	acc=0.568359
+INFO:root:Epoch[8] Batch [2460]	Speed: 662.31 samples/sec	acc=0.560937
+INFO:root:Epoch[8] Batch [2480]	Speed: 663.23 samples/sec	acc=0.562012
+INFO:root:Epoch[8] Batch [2500]	Speed: 662.24 samples/sec	acc=0.570215
+INFO:root:Epoch[8] Batch [2520]	Speed: 662.91 samples/sec	acc=0.574805
+INFO:root:Epoch[8] Batch [2540]	Speed: 663.08 samples/sec	acc=0.569238
+INFO:root:Epoch[8] Batch [2560]	Speed: 662.90 samples/sec	acc=0.562305
+INFO:root:Epoch[8] Batch [2580]	Speed: 662.96 samples/sec	acc=0.570508
+INFO:root:Epoch[8] Batch [2600]	Speed: 662.28 samples/sec	acc=0.564355
+INFO:root:Epoch[8] Batch [2620]	Speed: 663.15 samples/sec	acc=0.554492
+INFO:root:Epoch[8] Batch [2640]	Speed: 662.47 samples/sec	acc=0.563184
+INFO:root:Epoch[8] Batch [2660]	Speed: 662.35 samples/sec	acc=0.563574
+INFO:root:Epoch[8] Batch [2680]	Speed: 663.41 samples/sec	acc=0.559961
+INFO:root:Epoch[8] Batch [2700]	Speed: 662.38 samples/sec	acc=0.572852
+INFO:root:Epoch[8] Batch [2720]	Speed: 662.53 samples/sec	acc=0.557715
+INFO:root:Epoch[8] Batch [2740]	Speed: 663.01 samples/sec	acc=0.561914
+INFO:root:Epoch[8] Batch [2760]	Speed: 661.74 samples/sec	acc=0.556641
+INFO:root:Epoch[8] Batch [2780]	Speed: 662.34 samples/sec	acc=0.559082
+INFO:root:Epoch[8] Batch [2800]	Speed: 661.84 samples/sec	acc=0.560840
+INFO:root:Epoch[8] Batch [2820]	Speed: 662.49 samples/sec	acc=0.568066
+INFO:root:Epoch[8] Batch [2840]	Speed: 662.17 samples/sec	acc=0.569922
+lr-batch-epoch: 0.01 2847 8
+INFO:root:Epoch[8] Batch [2860]	Speed: 663.15 samples/sec	acc=0.563086
+INFO:root:Epoch[8] Batch [2880]	Speed: 663.10 samples/sec	acc=0.561621
+INFO:root:Epoch[8] Batch [2900]	Speed: 663.82 samples/sec	acc=0.566113
+INFO:root:Epoch[8] Batch [2920]	Speed: 663.19 samples/sec	acc=0.557715
+INFO:root:Epoch[8] Batch [2940]	Speed: 663.51 samples/sec	acc=0.557129
+INFO:root:Epoch[8] Batch [2960]	Speed: 663.43 samples/sec	acc=0.553809
+INFO:root:Epoch[8] Batch [2980]	Speed: 662.56 samples/sec	acc=0.570410
+INFO:root:Epoch[8] Batch [3000]	Speed: 662.38 samples/sec	acc=0.566406
+INFO:root:Epoch[8] Batch [3020]	Speed: 662.95 samples/sec	acc=0.562109
+INFO:root:Epoch[8] Batch [3040]	Speed: 662.20 samples/sec	acc=0.550391
+INFO:root:Epoch[8] Batch [3060]	Speed: 662.24 samples/sec	acc=0.557813
+INFO:root:Epoch[8] Batch [3080]	Speed: 665.51 samples/sec	acc=0.556152
+INFO:root:Epoch[8] Batch [3100]	Speed: 662.16 samples/sec	acc=0.546387
+INFO:root:Epoch[8] Batch [3120]	Speed: 663.27 samples/sec	acc=0.550098
+INFO:root:Epoch[8] Batch [3140]	Speed: 662.79 samples/sec	acc=0.554492
+INFO:root:Epoch[8] Batch [3160]	Speed: 662.66 samples/sec	acc=0.560449
+INFO:root:Epoch[8] Batch [3180]	Speed: 662.74 samples/sec	acc=0.551758
+INFO:root:Epoch[8] Batch [3200]	Speed: 662.66 samples/sec	acc=0.550195
+INFO:root:Epoch[8] Batch [3220]	Speed: 662.52 samples/sec	acc=0.549121
+INFO:root:Epoch[8] Batch [3240]	Speed: 662.36 samples/sec	acc=0.559766
+INFO:root:Epoch[8] Batch [3260]	Speed: 662.29 samples/sec	acc=0.565039
+INFO:root:Epoch[8] Batch [3280]	Speed: 663.23 samples/sec	acc=0.554004
+INFO:root:Epoch[8] Batch [3300]	Speed: 662.22 samples/sec	acc=0.564160
+INFO:root:Epoch[8] Batch [3320]	Speed: 663.11 samples/sec	acc=0.557129
+INFO:root:Epoch[8] Batch [3340]	Speed: 662.52 samples/sec	acc=0.545703
+INFO:root:Epoch[8] Batch [3360]	Speed: 662.15 samples/sec	acc=0.551855
+INFO:root:Epoch[8] Batch [3380]	Speed: 662.77 samples/sec	acc=0.557910
+INFO:root:Epoch[8] Batch [3400]	Speed: 662.50 samples/sec	acc=0.558789
+INFO:root:Epoch[8] Batch [3420]	Speed: 662.37 samples/sec	acc=0.553320
+INFO:root:Epoch[8] Batch [3440]	Speed: 662.41 samples/sec	acc=0.549707
+INFO:root:Epoch[8] Batch [3460]	Speed: 662.75 samples/sec	acc=0.552344
+INFO:root:Epoch[8] Batch [3480]	Speed: 661.67 samples/sec	acc=0.555859
+INFO:root:Epoch[8] Batch [3500]	Speed: 663.11 samples/sec	acc=0.550195
+INFO:root:Epoch[8] Batch [3520]	Speed: 662.13 samples/sec	acc=0.551270
+INFO:root:Epoch[8] Batch [3540]	Speed: 662.49 samples/sec	acc=0.548047
+INFO:root:Epoch[8] Batch [3560]	Speed: 662.86 samples/sec	acc=0.542578
+INFO:root:Epoch[8] Batch [3580]	Speed: 662.68 samples/sec	acc=0.559082
+INFO:root:Epoch[8] Batch [3600]	Speed: 662.54 samples/sec	acc=0.558594
+INFO:root:Epoch[8] Batch [3620]	Speed: 662.55 samples/sec	acc=0.550000
+INFO:root:Epoch[8] Batch [3640]	Speed: 662.30 samples/sec	acc=0.549414
+INFO:root:Epoch[8] Batch [3660]	Speed: 662.70 samples/sec	acc=0.542285
+INFO:root:Epoch[8] Batch [3680]	Speed: 662.84 samples/sec	acc=0.542969
+INFO:root:Epoch[8] Batch [3700]	Speed: 662.94 samples/sec	acc=0.546387
+INFO:root:Epoch[8] Batch [3720]	Speed: 663.06 samples/sec	acc=0.555078
+INFO:root:Epoch[8] Batch [3740]	Speed: 662.32 samples/sec	acc=0.547852
+INFO:root:Epoch[8] Batch [3760]	Speed: 662.77 samples/sec	acc=0.545020
+INFO:root:Epoch[8] Batch [3780]	Speed: 662.96 samples/sec	acc=0.547754
+INFO:root:Epoch[8] Batch [3800]	Speed: 662.71 samples/sec	acc=0.554980
+INFO:root:Epoch[8] Batch [3820]	Speed: 662.42 samples/sec	acc=0.553418
+INFO:root:Epoch[8] Batch [3840]	Speed: 662.83 samples/sec	acc=0.547363
+[64000][MARGIN]0.768907,0.373218
+lr-batch-epoch: 0.01 3847 8
+testing verification..
+(14000, 512)
+infer time 18.847999
+[cfp_ff][64000]XNorm: 22.131319
+[cfp_ff][64000]Accuracy-Flip: 0.99643+-0.00308
+testing verification..
+(14000, 512)
+infer time 15.98134
+[cfp_fp][64000]XNorm: 19.149032
+[cfp_fp][64000]Accuracy-Flip: 0.90443+-0.01668
+testing verification..
+(12000, 512)
+infer time 14.032172
+[agedb_30][64000]XNorm: 22.305228
+[agedb_30][64000]Accuracy-Flip: 0.97133+-0.00552
+testing verification..
+(12000, 512)
+infer time 13.825691
+[lfw][64000]XNorm: 22.346278
+[lfw][64000]Accuracy-Flip: 0.99700+-0.00256
+saving 32
+INFO:root:Saved checkpoint to "../model-r50-am-lfw/model-0032.params"
+[64000]Accuracy-Highest: 0.99700
+INFO:root:Epoch[8] Batch [3860]	Speed: 99.33 samples/sec	acc=0.552930
+INFO:root:Epoch[8] Batch [3880]	Speed: 662.10 samples/sec	acc=0.553516
+INFO:root:Epoch[8] Batch [3900]	Speed: 659.01 samples/sec	acc=0.550586
+INFO:root:Epoch[8] Batch [3920]	Speed: 682.55 samples/sec	acc=0.545410
+INFO:root:Epoch[8] Batch [3940]	Speed: 661.49 samples/sec	acc=0.548145
+INFO:root:Epoch[8] Batch [3960]	Speed: 662.42 samples/sec	acc=0.547852
+INFO:root:Epoch[8] Batch [3980]	Speed: 660.63 samples/sec	acc=0.547852
+INFO:root:Epoch[8] Batch [4000]	Speed: 661.70 samples/sec	acc=0.543945
+INFO:root:Epoch[8] Batch [4020]	Speed: 662.08 samples/sec	acc=0.543359
+INFO:root:Epoch[8] Batch [4040]	Speed: 662.71 samples/sec	acc=0.547754
+INFO:root:Epoch[8] Batch [4060]	Speed: 661.52 samples/sec	acc=0.537500
+INFO:root:Epoch[8] Batch [4080]	Speed: 662.18 samples/sec	acc=0.538086
+INFO:root:Epoch[8] Batch [4100]	Speed: 661.93 samples/sec	acc=0.541699
+INFO:root:Epoch[8] Batch [4120]	Speed: 661.88 samples/sec	acc=0.553027
+INFO:root:Epoch[8] Batch [4140]	Speed: 662.36 samples/sec	acc=0.540234
+INFO:root:Epoch[8] Batch [4160]	Speed: 662.05 samples/sec	acc=0.545508
+INFO:root:Epoch[8] Batch [4180]	Speed: 662.81 samples/sec	acc=0.545117
+INFO:root:Epoch[8] Batch [4200]	Speed: 662.65 samples/sec	acc=0.541406
+INFO:root:Epoch[8] Batch [4220]	Speed: 662.79 samples/sec	acc=0.545020
+INFO:root:Epoch[8] Batch [4240]	Speed: 662.36 samples/sec	acc=0.550293
+INFO:root:Epoch[8] Batch [4260]	Speed: 662.43 samples/sec	acc=0.550488
+INFO:root:Epoch[8] Batch [4280]	Speed: 663.45 samples/sec	acc=0.550098
+INFO:root:Epoch[8] Batch [4300]	Speed: 662.83 samples/sec	acc=0.543164
+INFO:root:Epoch[8] Batch [4320]	Speed: 662.35 samples/sec	acc=0.544141
+INFO:root:Epoch[8] Batch [4340]	Speed: 662.80 samples/sec	acc=0.539453
+INFO:root:Epoch[8] Batch [4360]	Speed: 663.19 samples/sec	acc=0.544238
+INFO:root:Epoch[8] Batch [4380]	Speed: 662.74 samples/sec	acc=0.539551
+INFO:root:Epoch[8] Batch [4400]	Speed: 662.32 samples/sec	acc=0.541113
+INFO:root:Epoch[8] Batch [4420]	Speed: 663.58 samples/sec	acc=0.541797
+INFO:root:Epoch[8] Batch [4440]	Speed: 662.52 samples/sec	acc=0.544727
+INFO:root:Epoch[8] Batch [4460]	Speed: 663.27 samples/sec	acc=0.541309
+INFO:root:Epoch[8] Batch [4480]	Speed: 662.25 samples/sec	acc=0.556055
+INFO:root:Epoch[8] Batch [4500]	Speed: 662.99 samples/sec	acc=0.544141
+INFO:root:Epoch[8] Batch [4520]	Speed: 661.80 samples/sec	acc=0.541406
+INFO:root:Epoch[8] Batch [4540]	Speed: 662.64 samples/sec	acc=0.546191
+INFO:root:Epoch[8] Batch [4560]	Speed: 663.28 samples/sec	acc=0.545605
+INFO:root:Epoch[8] Batch [4580]	Speed: 662.77 samples/sec	acc=0.538379
+INFO:root:Epoch[8] Batch [4600]	Speed: 663.23 samples/sec	acc=0.542090
+INFO:root:Epoch[8] Batch [4620]	Speed: 662.88 samples/sec	acc=0.534473
+INFO:root:Epoch[8] Batch [4640]	Speed: 662.99 samples/sec	acc=0.541016
+INFO:root:Epoch[8] Batch [4660]	Speed: 663.14 samples/sec	acc=0.545117
+INFO:root:Epoch[8] Batch [4680]	Speed: 663.12 samples/sec	acc=0.539062
+INFO:root:Epoch[8] Batch [4700]	Speed: 660.96 samples/sec	acc=0.542188
+INFO:root:Epoch[8] Batch [4720]	Speed: 675.86 samples/sec	acc=0.533203
+INFO:root:Epoch[8] Batch [4740]	Speed: 663.70 samples/sec	acc=0.531543
+INFO:root:Epoch[8] Batch [4760]	Speed: 662.14 samples/sec	acc=0.534863
+INFO:root:Epoch[8] Batch [4780]	Speed: 662.59 samples/sec	acc=0.532520
+INFO:root:Epoch[8] Batch [4800]	Speed: 662.37 samples/sec	acc=0.538672
+INFO:root:Epoch[8] Batch [4820]	Speed: 662.58 samples/sec	acc=0.544238
+INFO:root:Epoch[8] Batch [4840]	Speed: 662.62 samples/sec	acc=0.537891
+lr-batch-epoch: 0.01 4847 8
+INFO:root:Epoch[8] Batch [4860]	Speed: 662.82 samples/sec	acc=0.549512
+INFO:root:Epoch[8] Batch [4880]	Speed: 662.81 samples/sec	acc=0.532617
+INFO:root:Epoch[8] Batch [4900]	Speed: 662.83 samples/sec	acc=0.544141
+INFO:root:Epoch[8] Batch [4920]	Speed: 662.27 samples/sec	acc=0.538574
+INFO:root:Epoch[8] Batch [4940]	Speed: 662.30 samples/sec	acc=0.539844
+INFO:root:Epoch[8] Batch [4960]	Speed: 662.64 samples/sec	acc=0.532227
+INFO:root:Epoch[8] Batch [4980]	Speed: 662.82 samples/sec	acc=0.545898
+INFO:root:Epoch[8] Batch [5000]	Speed: 662.74 samples/sec	acc=0.535254
+INFO:root:Epoch[8] Batch [5020]	Speed: 663.17 samples/sec	acc=0.527930
+INFO:root:Epoch[8] Batch [5040]	Speed: 662.83 samples/sec	acc=0.531445
+INFO:root:Epoch[8] Batch [5060]	Speed: 662.55 samples/sec	acc=0.535840
+INFO:root:Epoch[8] Batch [5080]	Speed: 662.17 samples/sec	acc=0.542285
+INFO:root:Epoch[8] Batch [5100]	Speed: 662.99 samples/sec	acc=0.542773
+INFO:root:Epoch[8] Batch [5120]	Speed: 662.50 samples/sec	acc=0.535547
+INFO:root:Epoch[8] Batch [5140]	Speed: 661.90 samples/sec	acc=0.539551
+INFO:root:Epoch[8] Batch [5160]	Speed: 662.93 samples/sec	acc=0.530078
+INFO:root:Epoch[8] Batch [5180]	Speed: 662.14 samples/sec	acc=0.546875
+INFO:root:Epoch[8] Batch [5200]	Speed: 663.12 samples/sec	acc=0.540430
+INFO:root:Epoch[8] Batch [5220]	Speed: 662.71 samples/sec	acc=0.536328
+INFO:root:Epoch[8] Batch [5240]	Speed: 661.76 samples/sec	acc=0.531738
+INFO:root:Epoch[8] Batch [5260]	Speed: 662.63 samples/sec	acc=0.536426
+INFO:root:Epoch[8] Batch [5280]	Speed: 662.88 samples/sec	acc=0.532227
+INFO:root:Epoch[8] Batch [5300]	Speed: 662.28 samples/sec	acc=0.529687
+INFO:root:Epoch[8] Batch [5320]	Speed: 661.40 samples/sec	acc=0.534082
+INFO:root:Epoch[8] Batch [5340]	Speed: 662.14 samples/sec	acc=0.538574
+INFO:root:Epoch[8] Batch [5360]	Speed: 662.35 samples/sec	acc=0.536719
+INFO:root:Epoch[8] Batch [5380]	Speed: 662.62 samples/sec	acc=0.542578
+INFO:root:Epoch[8] Batch [5400]	Speed: 662.29 samples/sec	acc=0.531641
+INFO:root:Epoch[8] Batch [5420]	Speed: 662.81 samples/sec	acc=0.540137
+INFO:root:Epoch[8] Batch [5440]	Speed: 661.97 samples/sec	acc=0.546289
+INFO:root:Epoch[8] Batch [5460]	Speed: 663.01 samples/sec	acc=0.530566
+INFO:root:Epoch[8] Batch [5480]	Speed: 663.19 samples/sec	acc=0.536816
+INFO:root:Epoch[8] Batch [5500]	Speed: 673.49 samples/sec	acc=0.525293
+INFO:root:Epoch[8] Batch [5520]	Speed: 663.15 samples/sec	acc=0.538477
+INFO:root:Epoch[8] Batch [5540]	Speed: 662.00 samples/sec	acc=0.537598
+INFO:root:Epoch[8] Batch [5560]	Speed: 663.11 samples/sec	acc=0.536035
+INFO:root:Epoch[8] Batch [5580]	Speed: 662.38 samples/sec	acc=0.531641
+INFO:root:Epoch[8] Batch [5600]	Speed: 662.83 samples/sec	acc=0.532422
+INFO:root:Epoch[8] Batch [5620]	Speed: 662.47 samples/sec	acc=0.532422
+INFO:root:Epoch[8] Batch [5640]	Speed: 662.81 samples/sec	acc=0.530078
+INFO:root:Epoch[8] Batch [5660]	Speed: 661.91 samples/sec	acc=0.537793
+INFO:root:Epoch[8] Batch [5680]	Speed: 662.42 samples/sec	acc=0.529395
+INFO:root:Epoch[8] Batch [5700]	Speed: 662.99 samples/sec	acc=0.529785
+INFO:root:Epoch[8] Batch [5720]	Speed: 663.23 samples/sec	acc=0.529102
+INFO:root:Epoch[8] Batch [5740]	Speed: 662.18 samples/sec	acc=0.536914
+INFO:root:Epoch[8] Batch [5760]	Speed: 662.67 samples/sec	acc=0.540430
+INFO:root:Epoch[8] Batch [5780]	Speed: 662.91 samples/sec	acc=0.526855
+INFO:root:Epoch[8] Batch [5800]	Speed: 663.34 samples/sec	acc=0.536914
+INFO:root:Epoch[8] Batch [5820]	Speed: 661.92 samples/sec	acc=0.532031
+INFO:root:Epoch[8] Batch [5840]	Speed: 662.36 samples/sec	acc=0.528125
+[66000][MARGIN]0.766712,0.370095
+lr-batch-epoch: 0.01 5847 8
+testing verification..
+(14000, 512)
+infer time 20.762311
+[cfp_ff][66000]XNorm: 22.646185
+[cfp_ff][66000]Accuracy-Flip: 0.99714+-0.00313
+testing verification..
+(14000, 512)
+infer time 21.247468
+[cfp_fp][66000]XNorm: 19.670236
+[cfp_fp][66000]Accuracy-Flip: 0.90014+-0.01603
+testing verification..
+(12000, 512)
+infer time 13.983546
+[agedb_30][66000]XNorm: 22.844877
+[agedb_30][66000]Accuracy-Flip: 0.97333+-0.00624
+testing verification..
+(12000, 512)
+infer time 15.502448
+[lfw][66000]XNorm: 22.865621
+[lfw][66000]Accuracy-Flip: 0.99717+-0.00236
+saving 33
+INFO:root:Saved checkpoint to "../model-r50-am-lfw/model-0033.params"
+[66000]Accuracy-Highest: 0.99717
+INFO:root:Epoch[8] Batch [5860]	Speed: 90.48 samples/sec	acc=0.532617
+INFO:root:Epoch[8] Batch [5880]	Speed: 661.43 samples/sec	acc=0.529687
+INFO:root:Epoch[8] Batch [5900]	Speed: 662.62 samples/sec	acc=0.537305
+INFO:root:Epoch[8] Batch [5920]	Speed: 661.57 samples/sec	acc=0.534375
+INFO:root:Epoch[8] Batch [5940]	Speed: 662.01 samples/sec	acc=0.528809
+INFO:root:Epoch[8] Batch [5960]	Speed: 661.92 samples/sec	acc=0.534961
+INFO:root:Epoch[8] Batch [5980]	Speed: 662.58 samples/sec	acc=0.530078
+INFO:root:Epoch[8] Batch [6000]	Speed: 662.10 samples/sec	acc=0.531250
+INFO:root:Epoch[8] Batch [6020]	Speed: 662.08 samples/sec	acc=0.529004
+INFO:root:Epoch[8] Batch [6040]	Speed: 663.01 samples/sec	acc=0.535742
+INFO:root:Epoch[8] Batch [6060]	Speed: 663.00 samples/sec	acc=0.515430
+INFO:root:Epoch[8] Batch [6080]	Speed: 662.07 samples/sec	acc=0.534473
+INFO:root:Epoch[8] Batch [6100]	Speed: 662.07 samples/sec	acc=0.522656
+INFO:root:Epoch[8] Batch [6120]	Speed: 663.23 samples/sec	acc=0.533301
+INFO:root:Epoch[8] Batch [6140]	Speed: 662.00 samples/sec	acc=0.530078
+INFO:root:Epoch[8] Batch [6160]	Speed: 662.43 samples/sec	acc=0.530957
+INFO:root:Epoch[8] Batch [6180]	Speed: 662.49 samples/sec	acc=0.532031
+INFO:root:Epoch[8] Batch [6200]	Speed: 680.07 samples/sec	acc=0.522949
+INFO:root:Epoch[8] Batch [6220]	Speed: 662.70 samples/sec	acc=0.531738
+INFO:root:Epoch[8] Batch [6240]	Speed: 661.72 samples/sec	acc=0.532715
+INFO:root:Epoch[8] Batch [6260]	Speed: 662.62 samples/sec	acc=0.529590
+INFO:root:Epoch[8] Batch [6280]	Speed: 662.80 samples/sec	acc=0.532031
+INFO:root:Epoch[8] Batch [6300]	Speed: 662.60 samples/sec	acc=0.527051
+INFO:root:Epoch[8] Batch [6320]	Speed: 662.27 samples/sec	acc=0.528711
+INFO:root:Epoch[8] Batch [6340]	Speed: 661.90 samples/sec	acc=0.529297
+INFO:root:Epoch[8] Batch [6360]	Speed: 663.17 samples/sec	acc=0.526172
+INFO:root:Epoch[8] Batch [6380]	Speed: 663.73 samples/sec	acc=0.530762
+INFO:root:Epoch[8] Batch [6400]	Speed: 663.35 samples/sec	acc=0.521973
+INFO:root:Epoch[8] Batch [6420]	Speed: 662.75 samples/sec	acc=0.544727
+INFO:root:Epoch[8] Batch [6440]	Speed: 663.62 samples/sec	acc=0.525586
+INFO:root:Epoch[8] Batch [6460]	Speed: 663.89 samples/sec	acc=0.528320
+INFO:root:Epoch[8] Batch [6480]	Speed: 662.90 samples/sec	acc=0.521484
+INFO:root:Epoch[8] Batch [6500]	Speed: 662.64 samples/sec	acc=0.518457
+INFO:root:Epoch[8] Batch [6520]	Speed: 663.21 samples/sec	acc=0.528125
+INFO:root:Epoch[8] Batch [6540]	Speed: 663.04 samples/sec	acc=0.530371
+INFO:root:Epoch[8] Batch [6560]	Speed: 662.59 samples/sec	acc=0.532617
+INFO:root:Epoch[8] Batch [6580]	Speed: 663.02 samples/sec	acc=0.528223
+INFO:root:Epoch[8] Batch [6600]	Speed: 662.53 samples/sec	acc=0.529687
+INFO:root:Epoch[8] Batch [6620]	Speed: 662.09 samples/sec	acc=0.522266
+INFO:root:Epoch[8] Batch [6640]	Speed: 663.10 samples/sec	acc=0.520508
+INFO:root:Epoch[8] Batch [6660]	Speed: 662.85 samples/sec	acc=0.530371
+INFO:root:Epoch[8] Batch [6680]	Speed: 662.07 samples/sec	acc=0.528418
+INFO:root:Epoch[8] Batch [6700]	Speed: 663.03 samples/sec	acc=0.528613
+INFO:root:Epoch[8] Batch [6720]	Speed: 662.89 samples/sec	acc=0.527539
+INFO:root:Epoch[8] Batch [6740]	Speed: 666.24 samples/sec	acc=0.521777
+INFO:root:Epoch[8] Batch [6760]	Speed: 663.10 samples/sec	acc=0.531738
+INFO:root:Epoch[8] Batch [6780]	Speed: 662.72 samples/sec	acc=0.527246
+INFO:root:Epoch[8] Batch [6800]	Speed: 662.72 samples/sec	acc=0.523242
+INFO:root:Epoch[8] Batch [6820]	Speed: 662.92 samples/sec	acc=0.516113
+INFO:root:Epoch[8] Batch [6840]	Speed: 662.77 samples/sec	acc=0.525000
+lr-batch-epoch: 0.01 6847 8
+INFO:root:Epoch[8] Batch [6860]	Speed: 663.16 samples/sec	acc=0.521680
+INFO:root:Epoch[8] Batch [6880]	Speed: 662.46 samples/sec	acc=0.527637
+INFO:root:Epoch[8] Batch [6900]	Speed: 663.05 samples/sec	acc=0.520508
+INFO:root:Epoch[8] Batch [6920]	Speed: 662.62 samples/sec	acc=0.520410
+INFO:root:Epoch[8] Batch [6940]	Speed: 662.76 samples/sec	acc=0.521484
+INFO:root:Epoch[8] Batch [6960]	Speed: 667.88 samples/sec	acc=0.521289
+INFO:root:Epoch[8] Batch [6980]	Speed: 675.75 samples/sec	acc=0.525586
+INFO:root:Epoch[8] Batch [7000]	Speed: 662.07 samples/sec	acc=0.524512
+INFO:root:Epoch[8] Batch [7020]	Speed: 663.91 samples/sec	acc=0.523535
+INFO:root:Epoch[8] Batch [7040]	Speed: 662.99 samples/sec	acc=0.529297
+INFO:root:Epoch[8] Batch [7060]	Speed: 662.86 samples/sec	acc=0.515039
+INFO:root:Epoch[8] Batch [7080]	Speed: 662.33 samples/sec	acc=0.523828
+INFO:root:Epoch[8] Batch [7100]	Speed: 662.64 samples/sec	acc=0.524512
+INFO:root:Epoch[8] Batch [7120]	Speed: 663.30 samples/sec	acc=0.527051
+INFO:root:Epoch[8] Batch [7140]	Speed: 662.98 samples/sec	acc=0.524414
+INFO:root:Epoch[8] Batch [7160]	Speed: 663.00 samples/sec	acc=0.526172
+INFO:root:Epoch[8] Batch [7180]	Speed: 662.78 samples/sec	acc=0.525977
+INFO:root:Epoch[8] Batch [7200]	Speed: 662.91 samples/sec	acc=0.523047
+INFO:root:Epoch[8] Batch [7220]	Speed: 663.00 samples/sec	acc=0.527051
+INFO:root:Epoch[8] Batch [7240]	Speed: 662.71 samples/sec	acc=0.521094
+INFO:root:Epoch[8] Batch [7260]	Speed: 662.93 samples/sec	acc=0.521387
+INFO:root:Epoch[8] Batch [7280]	Speed: 662.24 samples/sec	acc=0.523438
+INFO:root:Epoch[8] Batch [7300]	Speed: 663.24 samples/sec	acc=0.503223
+INFO:root:Epoch[8] Batch [7320]	Speed: 662.66 samples/sec	acc=0.527344
+INFO:root:Epoch[8] Batch [7340]	Speed: 662.69 samples/sec	acc=0.526074
+INFO:root:Epoch[8] Batch [7360]	Speed: 662.47 samples/sec	acc=0.508691
+INFO:root:Epoch[8] Batch [7380]	Speed: 662.35 samples/sec	acc=0.522559
+INFO:root:Epoch[8] Batch [7400]	Speed: 662.73 samples/sec	acc=0.524512
+INFO:root:Epoch[8] Batch [7420]	Speed: 662.89 samples/sec	acc=0.523242
+INFO:root:Epoch[8] Batch [7440]	Speed: 662.68 samples/sec	acc=0.531934
+INFO:root:Epoch[8] Batch [7460]	Speed: 662.82 samples/sec	acc=0.522070
+INFO:root:Epoch[8] Batch [7480]	Speed: 663.24 samples/sec	acc=0.530762
+INFO:root:Epoch[8] Batch [7500]	Speed: 663.06 samples/sec	acc=0.514941
+INFO:root:Epoch[8] Train-acc=0.515516
+INFO:root:Epoch[8] Time cost=6080.978
+call reset()
+INFO:root:Epoch[9] Batch [20]	Speed: 656.42 samples/sec	acc=0.588170
+INFO:root:Epoch[9] Batch [40]	Speed: 662.05 samples/sec	acc=0.585254
+INFO:root:Epoch[9] Batch [60]	Speed: 662.01 samples/sec	acc=0.588867
+INFO:root:Epoch[9] Batch [80]	Speed: 662.56 samples/sec	acc=0.577246
+INFO:root:Epoch[9] Batch [100]	Speed: 662.69 samples/sec	acc=0.593164
+INFO:root:Epoch[9] Batch [120]	Speed: 662.88 samples/sec	acc=0.586523
+INFO:root:Epoch[9] Batch [140]	Speed: 662.44 samples/sec	acc=0.588965
+INFO:root:Epoch[9] Batch [160]	Speed: 663.17 samples/sec	acc=0.581152
+INFO:root:Epoch[9] Batch [180]	Speed: 662.39 samples/sec	acc=0.579102
+INFO:root:Epoch[9] Batch [200]	Speed: 666.46 samples/sec	acc=0.578809
+INFO:root:Epoch[9] Batch [220]	Speed: 663.15 samples/sec	acc=0.585059
+INFO:root:Epoch[9] Batch [240]	Speed: 662.37 samples/sec	acc=0.586133
+INFO:root:Epoch[9] Batch [260]	Speed: 662.37 samples/sec	acc=0.582129
+INFO:root:Epoch[9] Batch [280]	Speed: 662.80 samples/sec	acc=0.583789
+INFO:root:Epoch[9] Batch [300]	Speed: 662.80 samples/sec	acc=0.578223
+INFO:root:Epoch[9] Batch [320]	Speed: 662.52 samples/sec	acc=0.567773
+[68000][MARGIN]0.767312,0.369418
+lr-batch-epoch: 0.01 328 9
+testing verification..
+(14000, 512)
+infer time 17.877052
+[cfp_ff][68000]XNorm: 21.374485
+[cfp_ff][68000]Accuracy-Flip: 0.99643+-0.00287
+testing verification..
+(14000, 512)
+infer time 16.813819
+[cfp_fp][68000]XNorm: 18.497272
+[cfp_fp][68000]Accuracy-Flip: 0.90357+-0.01978
+testing verification..
+(12000, 512)
+infer time 15.67068
+[agedb_30][68000]XNorm: 22.127080
+[agedb_30][68000]Accuracy-Flip: 0.97217+-0.00823
+testing verification..
+(12000, 512)
+infer time 16.458884
+[lfw][68000]XNorm: 21.713268
+[lfw][68000]Accuracy-Flip: 0.99600+-0.00271
+[68000]Accuracy-Highest: 0.99717
+INFO:root:Epoch[9] Batch [340]	Speed: 100.36 samples/sec	acc=0.574707
+INFO:root:Epoch[9] Batch [360]	Speed: 662.11 samples/sec	acc=0.574414
+INFO:root:Epoch[9] Batch [380]	Speed: 661.70 samples/sec	acc=0.577441
+INFO:root:Epoch[9] Batch [400]	Speed: 661.94 samples/sec	acc=0.578906
+INFO:root:Epoch[9] Batch [420]	Speed: 662.16 samples/sec	acc=0.565430
+INFO:root:Epoch[9] Batch [440]	Speed: 661.48 samples/sec	acc=0.575098
+INFO:root:Epoch[9] Batch [460]	Speed: 662.69 samples/sec	acc=0.573535
+INFO:root:Epoch[9] Batch [480]	Speed: 662.18 samples/sec	acc=0.572168
+INFO:root:Epoch[9] Batch [500]	Speed: 662.38 samples/sec	acc=0.570410
+INFO:root:Epoch[9] Batch [520]	Speed: 662.30 samples/sec	acc=0.564941
+INFO:root:Epoch[9] Batch [540]	Speed: 662.54 samples/sec	acc=0.571680
+INFO:root:Epoch[9] Batch [560]	Speed: 662.65 samples/sec	acc=0.557031
+INFO:root:Epoch[9] Batch [580]	Speed: 663.02 samples/sec	acc=0.566504
+INFO:root:Epoch[9] Batch [600]	Speed: 662.92 samples/sec	acc=0.563379
+INFO:root:Epoch[9] Batch [620]	Speed: 664.00 samples/sec	acc=0.566211
+INFO:root:Epoch[9] Batch [640]	Speed: 663.43 samples/sec	acc=0.559082
+INFO:root:Epoch[9] Batch [660]	Speed: 663.89 samples/sec	acc=0.571484
+INFO:root:Epoch[9] Batch [680]	Speed: 660.50 samples/sec	acc=0.564551
+INFO:root:Epoch[9] Batch [700]	Speed: 662.50 samples/sec	acc=0.564648
+INFO:root:Epoch[9] Batch [720]	Speed: 653.16 samples/sec	acc=0.554004
+INFO:root:Epoch[9] Batch [740]	Speed: 662.42 samples/sec	acc=0.553125
+INFO:root:Epoch[9] Batch [760]	Speed: 663.35 samples/sec	acc=0.565332
+INFO:root:Epoch[9] Batch [780]	Speed: 662.57 samples/sec	acc=0.558789
+INFO:root:Epoch[9] Batch [800]	Speed: 662.36 samples/sec	acc=0.558887
+INFO:root:Epoch[9] Batch [820]	Speed: 662.99 samples/sec	acc=0.554395
+INFO:root:Epoch[9] Batch [840]	Speed: 662.52 samples/sec	acc=0.545508
+INFO:root:Epoch[9] Batch [860]	Speed: 662.48 samples/sec	acc=0.550488
+INFO:root:Epoch[9] Batch [880]	Speed: 663.44 samples/sec	acc=0.561426
+INFO:root:Epoch[9] Batch [900]	Speed: 661.99 samples/sec	acc=0.551270
+INFO:root:Epoch[9] Batch [920]	Speed: 662.65 samples/sec	acc=0.563086
+INFO:root:Epoch[9] Batch [940]	Speed: 662.92 samples/sec	acc=0.551465
+INFO:root:Epoch[9] Batch [960]	Speed: 666.34 samples/sec	acc=0.561719
+INFO:root:Epoch[9] Batch [980]	Speed: 665.32 samples/sec	acc=0.549512
+INFO:root:Epoch[9] Batch [1000]	Speed: 661.90 samples/sec	acc=0.554492
+INFO:root:Epoch[9] Batch [1020]	Speed: 662.17 samples/sec	acc=0.557324
+INFO:root:Epoch[9] Batch [1040]	Speed: 661.93 samples/sec	acc=0.545312
+INFO:root:Epoch[9] Batch [1060]	Speed: 663.08 samples/sec	acc=0.550977
+INFO:root:Epoch[9] Batch [1080]	Speed: 661.79 samples/sec	acc=0.556348
+INFO:root:Epoch[9] Batch [1100]	Speed: 662.20 samples/sec	acc=0.550977
+INFO:root:Epoch[9] Batch [1120]	Speed: 661.90 samples/sec	acc=0.547949
+INFO:root:Epoch[9] Batch [1140]	Speed: 662.83 samples/sec	acc=0.544141
+INFO:root:Epoch[9] Batch [1160]	Speed: 662.01 samples/sec	acc=0.551855
+INFO:root:Epoch[9] Batch [1180]	Speed: 663.57 samples/sec	acc=0.548242
+INFO:root:Epoch[9] Batch [1200]	Speed: 662.99 samples/sec	acc=0.536914
+INFO:root:Epoch[9] Batch [1220]	Speed: 661.53 samples/sec	acc=0.544336
+INFO:root:Epoch[9] Batch [1240]	Speed: 663.69 samples/sec	acc=0.543457
+INFO:root:Epoch[9] Batch [1260]	Speed: 662.85 samples/sec	acc=0.546387
+INFO:root:Epoch[9] Batch [1280]	Speed: 662.51 samples/sec	acc=0.547070
+INFO:root:Epoch[9] Batch [1300]	Speed: 663.24 samples/sec	acc=0.538965
+INFO:root:Epoch[9] Batch [1320]	Speed: 663.33 samples/sec	acc=0.546289
+lr-batch-epoch: 0.01 1328 9
+INFO:root:Epoch[9] Batch [1340]	Speed: 662.55 samples/sec	acc=0.549902
+INFO:root:Epoch[9] Batch [1360]	Speed: 663.39 samples/sec	acc=0.548438
+INFO:root:Epoch[9] Batch [1380]	Speed: 662.75 samples/sec	acc=0.548047
+INFO:root:Epoch[9] Batch [1400]	Speed: 661.94 samples/sec	acc=0.537598
+INFO:root:Epoch[9] Batch [1420]	Speed: 662.57 samples/sec	acc=0.536523
+INFO:root:Epoch[9] Batch [1440]	Speed: 662.88 samples/sec	acc=0.542480
+INFO:root:Epoch[9] Batch [1460]	Speed: 661.88 samples/sec	acc=0.537988
+INFO:root:Epoch[9] Batch [1480]	Speed: 663.84 samples/sec	acc=0.534473
+INFO:root:Epoch[9] Batch [1500]	Speed: 662.25 samples/sec	acc=0.535937
+INFO:root:Epoch[9] Batch [1520]	Speed: 662.12 samples/sec	acc=0.541895
+INFO:root:Epoch[9] Batch [1540]	Speed: 663.11 samples/sec	acc=0.536816
+INFO:root:Epoch[9] Batch [1560]	Speed: 662.51 samples/sec	acc=0.548828
+INFO:root:Epoch[9] Batch [1580]	Speed: 662.10 samples/sec	acc=0.533105
+INFO:root:Epoch[9] Batch [1600]	Speed: 663.62 samples/sec	acc=0.536230
+INFO:root:Epoch[9] Batch [1620]	Speed: 662.23 samples/sec	acc=0.542480
+INFO:root:Epoch[9] Batch [1640]	Speed: 662.58 samples/sec	acc=0.539551
+INFO:root:Epoch[9] Batch [1660]	Speed: 663.12 samples/sec	acc=0.540918
+INFO:root:Epoch[9] Batch [1680]	Speed: 662.66 samples/sec	acc=0.524707
+INFO:root:Epoch[9] Batch [1700]	Speed: 662.67 samples/sec	acc=0.537793
+INFO:root:Epoch[9] Batch [1720]	Speed: 662.23 samples/sec	acc=0.538867
+INFO:root:Epoch[9] Batch [1740]	Speed: 662.24 samples/sec	acc=0.531055
+INFO:root:Epoch[9] Batch [1760]	Speed: 662.15 samples/sec	acc=0.534277
+INFO:root:Epoch[9] Batch [1780]	Speed: 662.73 samples/sec	acc=0.526660
+INFO:root:Epoch[9] Batch [1800]	Speed: 662.97 samples/sec	acc=0.525000
+INFO:root:Epoch[9] Batch [1820]	Speed: 662.26 samples/sec	acc=0.531250
+INFO:root:Epoch[9] Batch [1840]	Speed: 663.02 samples/sec	acc=0.545117
+INFO:root:Epoch[9] Batch [1860]	Speed: 662.53 samples/sec	acc=0.532422
+INFO:root:Epoch[9] Batch [1880]	Speed: 662.70 samples/sec	acc=0.539258
+INFO:root:Epoch[9] Batch [1900]	Speed: 662.50 samples/sec	acc=0.532227
+INFO:root:Epoch[9] Batch [1920]	Speed: 662.53 samples/sec	acc=0.518262
+INFO:root:Epoch[9] Batch [1940]	Speed: 661.63 samples/sec	acc=0.539746
+INFO:root:Epoch[9] Batch [1960]	Speed: 663.46 samples/sec	acc=0.523535
+INFO:root:Epoch[9] Batch [1980]	Speed: 662.62 samples/sec	acc=0.531738
+INFO:root:Epoch[9] Batch [2000]	Speed: 662.28 samples/sec	acc=0.534863
+INFO:root:Epoch[9] Batch [2020]	Speed: 662.93 samples/sec	acc=0.525293
+INFO:root:Epoch[9] Batch [2040]	Speed: 661.63 samples/sec	acc=0.522461
+INFO:root:Epoch[9] Batch [2060]	Speed: 662.51 samples/sec	acc=0.528027
+INFO:root:Epoch[9] Batch [2080]	Speed: 663.02 samples/sec	acc=0.524707
+INFO:root:Epoch[9] Batch [2100]	Speed: 662.54 samples/sec	acc=0.533789
+INFO:root:Epoch[9] Batch [2120]	Speed: 661.47 samples/sec	acc=0.520117
+INFO:root:Epoch[9] Batch [2140]	Speed: 663.56 samples/sec	acc=0.530664
+INFO:root:Epoch[9] Batch [2160]	Speed: 662.27 samples/sec	acc=0.536230
+INFO:root:Epoch[9] Batch [2180]	Speed: 661.78 samples/sec	acc=0.529883
+INFO:root:Epoch[9] Batch [2200]	Speed: 663.35 samples/sec	acc=0.534668
+INFO:root:Epoch[9] Batch [2220]	Speed: 662.67 samples/sec	acc=0.525586
+INFO:root:Epoch[9] Batch [2240]	Speed: 661.81 samples/sec	acc=0.524121
+INFO:root:Epoch[9] Batch [2260]	Speed: 662.75 samples/sec	acc=0.527930
+INFO:root:Epoch[9] Batch [2280]	Speed: 662.19 samples/sec	acc=0.523438
+INFO:root:Epoch[9] Batch [2300]	Speed: 662.35 samples/sec	acc=0.522266
+INFO:root:Epoch[9] Batch [2320]	Speed: 663.38 samples/sec	acc=0.522559
+[70000][MARGIN]0.765564,0.367320
+lr-batch-epoch: 0.01 2328 9
+testing verification..
+(14000, 512)
+infer time 27.911592
+[cfp_ff][70000]XNorm: 21.644177
+[cfp_ff][70000]Accuracy-Flip: 0.99771+-0.00265
+testing verification..
+(14000, 512)
+infer time 23.607039
+[cfp_fp][70000]XNorm: 18.785028
+[cfp_fp][70000]Accuracy-Flip: 0.90643+-0.01594
+testing verification..
+(12000, 512)
+infer time 23.314731
+[agedb_30][70000]XNorm: 21.917033
+[agedb_30][70000]Accuracy-Flip: 0.97333+-0.00799
+testing verification..
+(12000, 512)
+infer time 15.590005
+[lfw][70000]XNorm: 22.027640
+[lfw][70000]Accuracy-Flip: 0.99633+-0.00332
+[70000]Accuracy-Highest: 0.99717
+INFO:root:Epoch[9] Batch [2340]	Speed: 82.29 samples/sec	acc=0.523047
+INFO:root:Epoch[9] Batch [2360]	Speed: 662.43 samples/sec	acc=0.521191
+INFO:root:Epoch[9] Batch [2380]	Speed: 648.00 samples/sec	acc=0.527637
+INFO:root:Epoch[9] Batch [2400]	Speed: 662.46 samples/sec	acc=0.524512
+INFO:root:Epoch[9] Batch [2420]	Speed: 660.77 samples/sec	acc=0.531055
+INFO:root:Epoch[9] Batch [2440]	Speed: 661.77 samples/sec	acc=0.523730
+INFO:root:Epoch[9] Batch [2460]	Speed: 661.78 samples/sec	acc=0.518457
+INFO:root:Epoch[9] Batch [2480]	Speed: 661.47 samples/sec	acc=0.523828
+INFO:root:Epoch[9] Batch [2500]	Speed: 662.50 samples/sec	acc=0.522852
+INFO:root:Epoch[9] Batch [2520]	Speed: 659.43 samples/sec	acc=0.528906
+INFO:root:Epoch[9] Batch [2540]	Speed: 665.35 samples/sec	acc=0.524219
+INFO:root:Epoch[9] Batch [2560]	Speed: 654.97 samples/sec	acc=0.516699
+INFO:root:Epoch[9] Batch [2580]	Speed: 651.54 samples/sec	acc=0.524414
+INFO:root:Epoch[9] Batch [2600]	Speed: 662.40 samples/sec	acc=0.517773
+INFO:root:Epoch[9] Batch [2620]	Speed: 661.62 samples/sec	acc=0.526367
+INFO:root:Epoch[9] Batch [2640]	Speed: 656.55 samples/sec	acc=0.523340
+INFO:root:Epoch[9] Batch [2660]	Speed: 662.23 samples/sec	acc=0.520703
+INFO:root:Epoch[9] Batch [2680]	Speed: 662.40 samples/sec	acc=0.527148
+INFO:root:Epoch[9] Batch [2700]	Speed: 650.95 samples/sec	acc=0.515527
+INFO:root:Epoch[9] Batch [2720]	Speed: 660.91 samples/sec	acc=0.524707
+INFO:root:Epoch[9] Batch [2740]	Speed: 655.28 samples/sec	acc=0.522559
+INFO:root:Epoch[9] Batch [2760]	Speed: 662.50 samples/sec	acc=0.520215
+INFO:root:Epoch[9] Batch [2780]	Speed: 653.57 samples/sec	acc=0.516309
+INFO:root:Epoch[9] Batch [2800]	Speed: 662.41 samples/sec	acc=0.518457
+INFO:root:Epoch[9] Batch [2820]	Speed: 661.59 samples/sec	acc=0.513770
+INFO:root:Epoch[9] Batch [2840]	Speed: 661.27 samples/sec	acc=0.519141
+INFO:root:Epoch[9] Batch [2860]	Speed: 662.31 samples/sec	acc=0.525098
+INFO:root:Epoch[9] Batch [2880]	Speed: 661.77 samples/sec	acc=0.518848
+INFO:root:Epoch[9] Batch [2900]	Speed: 661.07 samples/sec	acc=0.525098
+INFO:root:Epoch[9] Batch [2920]	Speed: 662.76 samples/sec	acc=0.523340
+INFO:root:Epoch[9] Batch [2940]	Speed: 662.26 samples/sec	acc=0.518848
+INFO:root:Epoch[9] Batch [2960]	Speed: 661.94 samples/sec	acc=0.513086
+INFO:root:Epoch[9] Batch [2980]	Speed: 663.77 samples/sec	acc=0.518164
+INFO:root:Epoch[9] Batch [3000]	Speed: 662.63 samples/sec	acc=0.523242
+INFO:root:Epoch[9] Batch [3020]	Speed: 662.70 samples/sec	acc=0.522656
+INFO:root:Epoch[9] Batch [3040]	Speed: 662.61 samples/sec	acc=0.519336
+INFO:root:Epoch[9] Batch [3060]	Speed: 662.72 samples/sec	acc=0.519824
+INFO:root:Epoch[9] Batch [3080]	Speed: 661.95 samples/sec	acc=0.519434
+INFO:root:Epoch[9] Batch [3100]	Speed: 662.84 samples/sec	acc=0.520215
+INFO:root:Epoch[9] Batch [3120]	Speed: 661.70 samples/sec	acc=0.516309
+INFO:root:Epoch[9] Batch [3140]	Speed: 662.81 samples/sec	acc=0.509473
+INFO:root:Epoch[9] Batch [3160]	Speed: 663.25 samples/sec	acc=0.518750
+INFO:root:Epoch[9] Batch [3180]	Speed: 662.64 samples/sec	acc=0.513965
+INFO:root:Epoch[9] Batch [3200]	Speed: 661.91 samples/sec	acc=0.514941
+INFO:root:Epoch[9] Batch [3220]	Speed: 663.86 samples/sec	acc=0.517090
+INFO:root:Epoch[9] Batch [3240]	Speed: 662.60 samples/sec	acc=0.513379
+INFO:root:Epoch[9] Batch [3260]	Speed: 661.86 samples/sec	acc=0.513184
+INFO:root:Epoch[9] Batch [3280]	Speed: 663.02 samples/sec	acc=0.513672
+INFO:root:Epoch[9] Batch [3300]	Speed: 663.08 samples/sec	acc=0.517969
+INFO:root:Epoch[9] Batch [3320]	Speed: 661.09 samples/sec	acc=0.520801
+lr-batch-epoch: 0.01 3328 9
+INFO:root:Epoch[9] Batch [3340]	Speed: 663.25 samples/sec	acc=0.517871
+INFO:root:Epoch[9] Batch [3360]	Speed: 663.17 samples/sec	acc=0.522559
+INFO:root:Epoch[9] Batch [3380]	Speed: 660.95 samples/sec	acc=0.520898
+INFO:root:Epoch[9] Batch [3400]	Speed: 663.58 samples/sec	acc=0.516113
+INFO:root:Epoch[9] Batch [3420]	Speed: 662.31 samples/sec	acc=0.510547
+INFO:root:Epoch[9] Batch [3440]	Speed: 643.11 samples/sec	acc=0.516406
+INFO:root:Epoch[9] Batch [3460]	Speed: 661.78 samples/sec	acc=0.520801
+INFO:root:Epoch[9] Batch [3480]	Speed: 658.51 samples/sec	acc=0.524023
+INFO:root:Epoch[9] Batch [3500]	Speed: 661.26 samples/sec	acc=0.516016
+INFO:root:Epoch[9] Batch [3520]	Speed: 661.55 samples/sec	acc=0.516309
+INFO:root:Epoch[9] Batch [3540]	Speed: 661.00 samples/sec	acc=0.509180
+INFO:root:Epoch[9] Batch [3560]	Speed: 655.79 samples/sec	acc=0.508887
+INFO:root:Epoch[9] Batch [3580]	Speed: 662.92 samples/sec	acc=0.522168
+INFO:root:Epoch[9] Batch [3600]	Speed: 658.86 samples/sec	acc=0.517383
+INFO:root:Epoch[9] Batch [3620]	Speed: 661.58 samples/sec	acc=0.514746
+INFO:root:Epoch[9] Batch [3640]	Speed: 662.47 samples/sec	acc=0.512305
+INFO:root:Epoch[9] Batch [3660]	Speed: 662.36 samples/sec	acc=0.513477
+INFO:root:Epoch[9] Batch [3680]	Speed: 662.87 samples/sec	acc=0.510352
+INFO:root:Epoch[9] Batch [3700]	Speed: 662.48 samples/sec	acc=0.516113
+INFO:root:Epoch[9] Batch [3720]	Speed: 662.53 samples/sec	acc=0.512695
+INFO:root:Epoch[9] Batch [3740]	Speed: 657.88 samples/sec	acc=0.511328
+INFO:root:Epoch[9] Batch [3760]	Speed: 662.30 samples/sec	acc=0.509961
+INFO:root:Epoch[9] Batch [3780]	Speed: 661.73 samples/sec	acc=0.514648
+INFO:root:Epoch[9] Batch [3800]	Speed: 660.65 samples/sec	acc=0.514551
+INFO:root:Epoch[9] Batch [3820]	Speed: 662.21 samples/sec	acc=0.514844
+INFO:root:Epoch[9] Batch [3840]	Speed: 661.00 samples/sec	acc=0.516797
+INFO:root:Epoch[9] Batch [3860]	Speed: 654.57 samples/sec	acc=0.512207
+INFO:root:Epoch[9] Batch [3880]	Speed: 662.28 samples/sec	acc=0.511621
+INFO:root:Epoch[9] Batch [3900]	Speed: 657.76 samples/sec	acc=0.509668
+INFO:root:Epoch[9] Batch [3920]	Speed: 660.59 samples/sec	acc=0.518945
+INFO:root:Epoch[9] Batch [3940]	Speed: 662.61 samples/sec	acc=0.516113
+INFO:root:Epoch[9] Batch [3960]	Speed: 661.99 samples/sec	acc=0.520020
+INFO:root:Epoch[9] Batch [3980]	Speed: 660.75 samples/sec	acc=0.516699
+INFO:root:Epoch[9] Batch [4000]	Speed: 662.86 samples/sec	acc=0.513379
+INFO:root:Epoch[9] Batch [4020]	Speed: 661.95 samples/sec	acc=0.520410
+INFO:root:Epoch[9] Batch [4040]	Speed: 661.62 samples/sec	acc=0.512012
+INFO:root:Epoch[9] Batch [4060]	Speed: 662.03 samples/sec	acc=0.504883
+INFO:root:Epoch[9] Batch [4080]	Speed: 662.06 samples/sec	acc=0.503711
+INFO:root:Epoch[9] Batch [4100]	Speed: 655.99 samples/sec	acc=0.512402
+INFO:root:Epoch[9] Batch [4120]	Speed: 669.69 samples/sec	acc=0.511328
+INFO:root:Epoch[9] Batch [4140]	Speed: 661.86 samples/sec	acc=0.516699
+INFO:root:Epoch[9] Batch [4160]	Speed: 661.47 samples/sec	acc=0.509082
+INFO:root:Epoch[9] Batch [4180]	Speed: 663.27 samples/sec	acc=0.513574
+INFO:root:Epoch[9] Batch [4200]	Speed: 658.29 samples/sec	acc=0.516309
+INFO:root:Epoch[9] Batch [4220]	Speed: 661.75 samples/sec	acc=0.523535
+INFO:root:Epoch[9] Batch [4240]	Speed: 651.90 samples/sec	acc=0.513770
+INFO:root:Epoch[9] Batch [4260]	Speed: 662.95 samples/sec	acc=0.521094
+INFO:root:Epoch[9] Batch [4280]	Speed: 661.30 samples/sec	acc=0.517285
+INFO:root:Epoch[9] Batch [4300]	Speed: 662.20 samples/sec	acc=0.508789
+INFO:root:Epoch[9] Batch [4320]	Speed: 661.79 samples/sec	acc=0.510840
+[72000][MARGIN]0.751948,0.348412
+lr-batch-epoch: 0.01 4328 9
+testing verification..
+(14000, 512)
+infer time 24.142456
+[cfp_ff][72000]XNorm: 21.711886
+[cfp_ff][72000]Accuracy-Flip: 0.99643+-0.00308
+testing verification..
+(14000, 512)
+infer time 22.896593
+[cfp_fp][72000]XNorm: 18.571745
+[cfp_fp][72000]Accuracy-Flip: 0.89929+-0.01595
+testing verification..
+(12000, 512)
+infer time 18.760471
+[agedb_30][72000]XNorm: 21.802036
+[agedb_30][72000]Accuracy-Flip: 0.97317+-0.00639
+testing verification..
+(12000, 512)
+infer time 20.440505
+[lfw][72000]XNorm: 21.822876
+[lfw][72000]Accuracy-Flip: 0.99667+-0.00279
+[72000]Accuracy-Highest: 0.99717
+INFO:root:Epoch[9] Batch [4340]	Speed: 83.42 samples/sec	acc=0.510352
+INFO:root:Epoch[9] Batch [4360]	Speed: 662.56 samples/sec	acc=0.507227
+INFO:root:Epoch[9] Batch [4380]	Speed: 661.52 samples/sec	acc=0.507812
+INFO:root:Epoch[9] Batch [4400]	Speed: 661.25 samples/sec	acc=0.518555
+INFO:root:Epoch[9] Batch [4420]	Speed: 661.68 samples/sec	acc=0.514453
+INFO:root:Epoch[9] Batch [4440]	Speed: 662.03 samples/sec	acc=0.514844
+INFO:root:Epoch[9] Batch [4460]	Speed: 661.64 samples/sec	acc=0.502344
+INFO:root:Epoch[9] Batch [4480]	Speed: 661.63 samples/sec	acc=0.514746
+INFO:root:Epoch[9] Batch [4500]	Speed: 661.95 samples/sec	acc=0.517480
+INFO:root:Epoch[9] Batch [4520]	Speed: 662.30 samples/sec	acc=0.508984
+INFO:root:Epoch[9] Batch [4540]	Speed: 662.95 samples/sec	acc=0.514355
+INFO:root:Epoch[9] Batch [4560]	Speed: 663.05 samples/sec	acc=0.507227
+INFO:root:Epoch[9] Batch [4580]	Speed: 655.40 samples/sec	acc=0.504590
+INFO:root:Epoch[9] Batch [4600]	Speed: 661.93 samples/sec	acc=0.513281
+INFO:root:Epoch[9] Batch [4620]	Speed: 662.17 samples/sec	acc=0.515918
+INFO:root:Epoch[9] Batch [4640]	Speed: 661.89 samples/sec	acc=0.507324
+INFO:root:Epoch[9] Batch [4660]	Speed: 662.38 samples/sec	acc=0.506738
+INFO:root:Epoch[9] Batch [4680]	Speed: 662.63 samples/sec	acc=0.499609
+INFO:root:Epoch[9] Batch [4700]	Speed: 662.33 samples/sec	acc=0.513281
+INFO:root:Epoch[9] Batch [4720]	Speed: 651.96 samples/sec	acc=0.504785
+INFO:root:Epoch[9] Batch [4740]	Speed: 661.74 samples/sec	acc=0.515918
+INFO:root:Epoch[9] Batch [4760]	Speed: 661.80 samples/sec	acc=0.516504
+INFO:root:Epoch[9] Batch [4780]	Speed: 659.50 samples/sec	acc=0.501270
+INFO:root:Epoch[9] Batch [4800]	Speed: 662.41 samples/sec	acc=0.505078
+INFO:root:Epoch[9] Batch [4820]	Speed: 661.83 samples/sec	acc=0.511816
+INFO:root:Epoch[9] Batch [4840]	Speed: 662.11 samples/sec	acc=0.515039
+INFO:root:Epoch[9] Batch [4860]	Speed: 662.90 samples/sec	acc=0.503711
+INFO:root:Epoch[9] Batch [4880]	Speed: 662.31 samples/sec	acc=0.523535
+INFO:root:Epoch[9] Batch [4900]	Speed: 662.34 samples/sec	acc=0.510645
+INFO:root:Epoch[9] Batch [4920]	Speed: 659.97 samples/sec	acc=0.512305
+INFO:root:Epoch[9] Batch [4940]	Speed: 672.18 samples/sec	acc=0.508789
+INFO:root:Epoch[9] Batch [4960]	Speed: 662.94 samples/sec	acc=0.502051
+INFO:root:Epoch[9] Batch [4980]	Speed: 662.65 samples/sec	acc=0.508301
+INFO:root:Epoch[9] Batch [5000]	Speed: 662.61 samples/sec	acc=0.507324
+INFO:root:Epoch[9] Batch [5020]	Speed: 663.05 samples/sec	acc=0.505664
+INFO:root:Epoch[9] Batch [5040]	Speed: 662.17 samples/sec	acc=0.515332
+INFO:root:Epoch[9] Batch [5060]	Speed: 662.30 samples/sec	acc=0.509961
+INFO:root:Epoch[9] Batch [5080]	Speed: 663.25 samples/sec	acc=0.512402
+INFO:root:Epoch[9] Batch [5100]	Speed: 662.61 samples/sec	acc=0.507227
+INFO:root:Epoch[9] Batch [5120]	Speed: 662.87 samples/sec	acc=0.508887
+INFO:root:Epoch[9] Batch [5140]	Speed: 663.21 samples/sec	acc=0.513965
+INFO:root:Epoch[9] Batch [5160]	Speed: 663.15 samples/sec	acc=0.509668
+INFO:root:Epoch[9] Batch [5180]	Speed: 662.53 samples/sec	acc=0.511230
+INFO:root:Epoch[9] Batch [5200]	Speed: 662.27 samples/sec	acc=0.507031
+INFO:root:Epoch[9] Batch [5220]	Speed: 662.23 samples/sec	acc=0.508496
+INFO:root:Epoch[9] Batch [5240]	Speed: 661.91 samples/sec	acc=0.506836
+INFO:root:Epoch[9] Batch [5260]	Speed: 663.00 samples/sec	acc=0.505566
+INFO:root:Epoch[9] Batch [5280]	Speed: 662.85 samples/sec	acc=0.510645
+INFO:root:Epoch[9] Batch [5300]	Speed: 663.09 samples/sec	acc=0.510352
+INFO:root:Epoch[9] Batch [5320]	Speed: 663.47 samples/sec	acc=0.507031
+lr-batch-epoch: 0.01 5328 9
+INFO:root:Epoch[9] Batch [5340]	Speed: 662.02 samples/sec	acc=0.508105
+INFO:root:Epoch[9] Batch [5360]	Speed: 661.85 samples/sec	acc=0.500781
+INFO:root:Epoch[9] Batch [5380]	Speed: 663.58 samples/sec	acc=0.520508
+INFO:root:Epoch[9] Batch [5400]	Speed: 663.17 samples/sec	acc=0.523047
+INFO:root:Epoch[9] Batch [5420]	Speed: 662.52 samples/sec	acc=0.516895
+INFO:root:Epoch[9] Batch [5440]	Speed: 662.87 samples/sec	acc=0.516211
+INFO:root:Epoch[9] Batch [5460]	Speed: 665.00 samples/sec	acc=0.499902
+INFO:root:Epoch[9] Batch [5480]	Speed: 662.03 samples/sec	acc=0.515430
+INFO:root:Epoch[9] Batch [5500]	Speed: 663.76 samples/sec	acc=0.507227
+INFO:root:Epoch[9] Batch [5520]	Speed: 662.11 samples/sec	acc=0.508594
+INFO:root:Epoch[9] Batch [5540]	Speed: 662.86 samples/sec	acc=0.508301
+INFO:root:Epoch[9] Batch [5560]	Speed: 652.79 samples/sec	acc=0.512402
+INFO:root:Epoch[9] Batch [5580]	Speed: 661.73 samples/sec	acc=0.511426
+INFO:root:Epoch[9] Batch [5600]	Speed: 661.65 samples/sec	acc=0.511035
+INFO:root:Epoch[9] Batch [5620]	Speed: 672.23 samples/sec	acc=0.500684
+INFO:root:Epoch[9] Batch [5640]	Speed: 664.06 samples/sec	acc=0.505469
+INFO:root:Epoch[9] Batch [5660]	Speed: 661.38 samples/sec	acc=0.499023
+INFO:root:Epoch[9] Batch [5680]	Speed: 662.43 samples/sec	acc=0.512207
+INFO:root:Epoch[9] Batch [5700]	Speed: 661.49 samples/sec	acc=0.501465
+INFO:root:Epoch[9] Batch [5720]	Speed: 667.14 samples/sec	acc=0.512500
+INFO:root:Epoch[9] Batch [5740]	Speed: 668.91 samples/sec	acc=0.511719
+INFO:root:Epoch[9] Batch [5760]	Speed: 662.83 samples/sec	acc=0.508398
+INFO:root:Epoch[9] Batch [5780]	Speed: 660.89 samples/sec	acc=0.513477
+INFO:root:Epoch[9] Batch [5800]	Speed: 661.61 samples/sec	acc=0.515430
+INFO:root:Epoch[9] Batch [5820]	Speed: 662.08 samples/sec	acc=0.503223
+INFO:root:Epoch[9] Batch [5840]	Speed: 661.93 samples/sec	acc=0.511426
+INFO:root:Epoch[9] Batch [5860]	Speed: 661.95 samples/sec	acc=0.511328
+INFO:root:Epoch[9] Batch [5880]	Speed: 662.67 samples/sec	acc=0.509180
+INFO:root:Epoch[9] Batch [5900]	Speed: 661.47 samples/sec	acc=0.510254
+INFO:root:Epoch[9] Batch [5920]	Speed: 662.88 samples/sec	acc=0.505957
+INFO:root:Epoch[9] Batch [5940]	Speed: 662.11 samples/sec	acc=0.514941
+INFO:root:Epoch[9] Batch [5960]	Speed: 662.01 samples/sec	acc=0.503125
+INFO:root:Epoch[9] Batch [5980]	Speed: 662.48 samples/sec	acc=0.508398
+INFO:root:Epoch[9] Batch [6000]	Speed: 662.72 samples/sec	acc=0.509473
+INFO:root:Epoch[9] Batch [6020]	Speed: 661.06 samples/sec	acc=0.506348
+INFO:root:Epoch[9] Batch [6040]	Speed: 652.16 samples/sec	acc=0.508789
+INFO:root:Epoch[9] Batch [6060]	Speed: 661.07 samples/sec	acc=0.499512
+INFO:root:Epoch[9] Batch [6080]	Speed: 660.13 samples/sec	acc=0.508105
+INFO:root:Epoch[9] Batch [6100]	Speed: 662.20 samples/sec	acc=0.505566
+INFO:root:Epoch[9] Batch [6120]	Speed: 662.20 samples/sec	acc=0.509180
+INFO:root:Epoch[9] Batch [6140]	Speed: 661.70 samples/sec	acc=0.502930
+INFO:root:Epoch[9] Batch [6160]	Speed: 662.93 samples/sec	acc=0.509277
+INFO:root:Epoch[9] Batch [6180]	Speed: 661.12 samples/sec	acc=0.507031
+INFO:root:Epoch[9] Batch [6200]	Speed: 661.10 samples/sec	acc=0.508105
+INFO:root:Epoch[9] Batch [6220]	Speed: 662.87 samples/sec	acc=0.503613
+INFO:root:Epoch[9] Batch [6240]	Speed: 662.82 samples/sec	acc=0.512891
+INFO:root:Epoch[9] Batch [6260]	Speed: 654.14 samples/sec	acc=0.499121
+INFO:root:Epoch[9] Batch [6280]	Speed: 651.94 samples/sec	acc=0.507324
+INFO:root:Epoch[9] Batch [6300]	Speed: 658.09 samples/sec	acc=0.503906
+INFO:root:Epoch[9] Batch [6320]	Speed: 660.61 samples/sec	acc=0.515039
+[74000][MARGIN]0.766221,0.369203
+lr-batch-epoch: 0.01 6328 9
+testing verification..
+(14000, 512)
+infer time 20.714861
+[cfp_ff][74000]XNorm: 22.009638
+[cfp_ff][74000]Accuracy-Flip: 0.99614+-0.00384
+testing verification..
+(14000, 512)
+infer time 23.832139
+[cfp_fp][74000]XNorm: 18.695112
+[cfp_fp][74000]Accuracy-Flip: 0.89957+-0.01733
+testing verification..
+(12000, 512)
+infer time 19.248342
+[agedb_30][74000]XNorm: 22.202675
+[agedb_30][74000]Accuracy-Flip: 0.96933+-0.00834
+testing verification..
+(12000, 512)
+infer time 17.353995
+[lfw][74000]XNorm: 22.211637
+[lfw][74000]Accuracy-Flip: 0.99533+-0.00332
+[74000]Accuracy-Highest: 0.99717
+INFO:root:Epoch[9] Batch [6340]	Speed: 88.14 samples/sec	acc=0.506152
+INFO:root:Epoch[9] Batch [6360]	Speed: 678.09 samples/sec	acc=0.509863
+INFO:root:Epoch[9] Batch [6380]	Speed: 661.56 samples/sec	acc=0.519043
+INFO:root:Epoch[9] Batch [6400]	Speed: 662.25 samples/sec	acc=0.499609
+INFO:root:Epoch[9] Batch [6420]	Speed: 661.29 samples/sec	acc=0.507617
+INFO:root:Epoch[9] Batch [6440]	Speed: 659.65 samples/sec	acc=0.502344
+INFO:root:Epoch[9] Batch [6460]	Speed: 664.17 samples/sec	acc=0.513867
+INFO:root:Epoch[9] Batch [6480]	Speed: 661.81 samples/sec	acc=0.505273
+INFO:root:Epoch[9] Batch [6500]	Speed: 661.72 samples/sec	acc=0.516211
+INFO:root:Epoch[9] Batch [6520]	Speed: 658.98 samples/sec	acc=0.495898
+INFO:root:Epoch[9] Batch [6540]	Speed: 670.75 samples/sec	acc=0.499023
+INFO:root:Epoch[9] Batch [6560]	Speed: 662.01 samples/sec	acc=0.506348
+INFO:root:Epoch[9] Batch [6580]	Speed: 662.92 samples/sec	acc=0.501563
+INFO:root:Epoch[9] Batch [6600]	Speed: 662.90 samples/sec	acc=0.503516
+INFO:root:Epoch[9] Batch [6620]	Speed: 661.23 samples/sec	acc=0.502637
+INFO:root:Epoch[9] Batch [6640]	Speed: 663.16 samples/sec	acc=0.500781
+INFO:root:Epoch[9] Batch [6660]	Speed: 662.84 samples/sec	acc=0.518945
+INFO:root:Epoch[9] Batch [6680]	Speed: 662.72 samples/sec	acc=0.512402
+INFO:root:Epoch[9] Batch [6700]	Speed: 662.72 samples/sec	acc=0.502637
+INFO:root:Epoch[9] Batch [6720]	Speed: 662.78 samples/sec	acc=0.510059
+INFO:root:Epoch[9] Batch [6740]	Speed: 661.68 samples/sec	acc=0.506543
+INFO:root:Epoch[9] Batch [6760]	Speed: 663.19 samples/sec	acc=0.496484
+INFO:root:Epoch[9] Batch [6780]	Speed: 662.72 samples/sec	acc=0.506055
+INFO:root:Epoch[9] Batch [6800]	Speed: 662.32 samples/sec	acc=0.504492
+INFO:root:Epoch[9] Batch [6820]	Speed: 662.93 samples/sec	acc=0.508691
+INFO:root:Epoch[9] Batch [6840]	Speed: 662.26 samples/sec	acc=0.502637
+INFO:root:Epoch[9] Batch [6860]	Speed: 662.54 samples/sec	acc=0.501270
+INFO:root:Epoch[9] Batch [6880]	Speed: 663.21 samples/sec	acc=0.512109
+INFO:root:Epoch[9] Batch [6900]	Speed: 662.14 samples/sec	acc=0.500098
+INFO:root:Epoch[9] Batch [6920]	Speed: 649.15 samples/sec	acc=0.505859
+INFO:root:Epoch[9] Batch [6940]	Speed: 662.76 samples/sec	acc=0.508594
+INFO:root:Epoch[9] Batch [6960]	Speed: 661.61 samples/sec	acc=0.499805
+INFO:root:Epoch[9] Batch [6980]	Speed: 661.25 samples/sec	acc=0.512598
+INFO:root:Epoch[9] Batch [7000]	Speed: 661.99 samples/sec	acc=0.505664
+INFO:root:Epoch[9] Batch [7020]	Speed: 661.92 samples/sec	acc=0.508496
+INFO:root:Epoch[9] Batch [7040]	Speed: 661.09 samples/sec	acc=0.502051
+INFO:root:Epoch[9] Batch [7060]	Speed: 662.27 samples/sec	acc=0.504883
+INFO:root:Epoch[9] Batch [7080]	Speed: 661.38 samples/sec	acc=0.511719
+INFO:root:Epoch[9] Batch [7100]	Speed: 661.90 samples/sec	acc=0.508594
+INFO:root:Epoch[9] Batch [7120]	Speed: 661.90 samples/sec	acc=0.508984
+INFO:root:Epoch[9] Batch [7140]	Speed: 661.86 samples/sec	acc=0.508691
+INFO:root:Epoch[9] Batch [7160]	Speed: 662.13 samples/sec	acc=0.507031
+INFO:root:Epoch[9] Batch [7180]	Speed: 662.90 samples/sec	acc=0.508691
+INFO:root:Epoch[9] Batch [7200]	Speed: 662.94 samples/sec	acc=0.509961
+INFO:root:Epoch[9] Batch [7220]	Speed: 662.17 samples/sec	acc=0.503613
+INFO:root:Epoch[9] Batch [7240]	Speed: 661.95 samples/sec	acc=0.505957
+INFO:root:Epoch[9] Batch [7260]	Speed: 662.50 samples/sec	acc=0.504395
+INFO:root:Epoch[9] Batch [7280]	Speed: 662.35 samples/sec	acc=0.510254
+INFO:root:Epoch[9] Batch [7300]	Speed: 662.92 samples/sec	acc=0.507129
+INFO:root:Epoch[9] Batch [7320]	Speed: 663.28 samples/sec	acc=0.512207
+lr-batch-epoch: 0.01 7328 9
+INFO:root:Epoch[9] Batch [7340]	Speed: 673.43 samples/sec	acc=0.504395
+INFO:root:Epoch[9] Batch [7360]	Speed: 662.14 samples/sec	acc=0.507031
+INFO:root:Epoch[9] Batch [7380]	Speed: 662.52 samples/sec	acc=0.499512
+INFO:root:Epoch[9] Batch [7400]	Speed: 661.93 samples/sec	acc=0.504687
+INFO:root:Epoch[9] Batch [7420]	Speed: 663.24 samples/sec	acc=0.496191
+INFO:root:Epoch[9] Batch [7440]	Speed: 661.83 samples/sec	acc=0.507227
+INFO:root:Epoch[9] Batch [7460]	Speed: 662.80 samples/sec	acc=0.515625
+INFO:root:Epoch[9] Batch [7480]	Speed: 663.18 samples/sec	acc=0.500195
+INFO:root:Epoch[9] Batch [7500]	Speed: 662.56 samples/sec	acc=0.511523
+INFO:root:Epoch[9] Train-acc=0.494575
+INFO:root:Epoch[9] Time cost=6220.531
+call reset()
+INFO:root:Epoch[10] Batch [20]	Speed: 656.37 samples/sec	acc=0.562686
+INFO:root:Epoch[10] Batch [40]	Speed: 662.70 samples/sec	acc=0.573242
+INFO:root:Epoch[10] Batch [60]	Speed: 662.41 samples/sec	acc=0.568652
+INFO:root:Epoch[10] Batch [80]	Speed: 662.11 samples/sec	acc=0.573438
+INFO:root:Epoch[10] Batch [100]	Speed: 649.89 samples/sec	acc=0.571582
+INFO:root:Epoch[10] Batch [120]	Speed: 663.32 samples/sec	acc=0.569238
+INFO:root:Epoch[10] Batch [140]	Speed: 661.99 samples/sec	acc=0.567969
+INFO:root:Epoch[10] Batch [160]	Speed: 662.46 samples/sec	acc=0.570215
+INFO:root:Epoch[10] Batch [180]	Speed: 662.94 samples/sec	acc=0.569434
+INFO:root:Epoch[10] Batch [200]	Speed: 648.88 samples/sec	acc=0.568359
+INFO:root:Epoch[10] Batch [220]	Speed: 662.10 samples/sec	acc=0.570703
+INFO:root:Epoch[10] Batch [240]	Speed: 662.05 samples/sec	acc=0.556445
+INFO:root:Epoch[10] Batch [260]	Speed: 661.54 samples/sec	acc=0.555859
+INFO:root:Epoch[10] Batch [280]	Speed: 661.71 samples/sec	acc=0.565527
+INFO:root:Epoch[10] Batch [300]	Speed: 661.64 samples/sec	acc=0.559961
+INFO:root:Epoch[10] Batch [320]	Speed: 661.54 samples/sec	acc=0.554492
+INFO:root:Epoch[10] Batch [340]	Speed: 662.55 samples/sec	acc=0.562207
+INFO:root:Epoch[10] Batch [360]	Speed: 662.66 samples/sec	acc=0.570020
+INFO:root:Epoch[10] Batch [380]	Speed: 661.60 samples/sec	acc=0.557227
+INFO:root:Epoch[10] Batch [400]	Speed: 661.85 samples/sec	acc=0.558496
+INFO:root:Epoch[10] Batch [420]	Speed: 663.39 samples/sec	acc=0.559766
+INFO:root:Epoch[10] Batch [440]	Speed: 654.99 samples/sec	acc=0.564551
+INFO:root:Epoch[10] Batch [460]	Speed: 662.77 samples/sec	acc=0.556738
+INFO:root:Epoch[10] Batch [480]	Speed: 662.70 samples/sec	acc=0.553613
+INFO:root:Epoch[10] Batch [500]	Speed: 661.84 samples/sec	acc=0.558398
+INFO:root:Epoch[10] Batch [520]	Speed: 662.20 samples/sec	acc=0.557520
+INFO:root:Epoch[10] Batch [540]	Speed: 663.36 samples/sec	acc=0.549219
+INFO:root:Epoch[10] Batch [560]	Speed: 662.10 samples/sec	acc=0.550391
+INFO:root:Epoch[10] Batch [580]	Speed: 663.33 samples/sec	acc=0.550781
+INFO:root:Epoch[10] Batch [600]	Speed: 659.12 samples/sec	acc=0.552539
+INFO:root:Epoch[10] Batch [620]	Speed: 677.53 samples/sec	acc=0.550684
+INFO:root:Epoch[10] Batch [640]	Speed: 662.77 samples/sec	acc=0.543066
+INFO:root:Epoch[10] Batch [660]	Speed: 662.42 samples/sec	acc=0.545410
+INFO:root:Epoch[10] Batch [680]	Speed: 662.62 samples/sec	acc=0.543750
+INFO:root:Epoch[10] Batch [700]	Speed: 661.89 samples/sec	acc=0.543555
+INFO:root:Epoch[10] Batch [720]	Speed: 662.67 samples/sec	acc=0.551660
+INFO:root:Epoch[10] Batch [740]	Speed: 662.04 samples/sec	acc=0.538477
+INFO:root:Epoch[10] Batch [760]	Speed: 663.28 samples/sec	acc=0.545117
+INFO:root:Epoch[10] Batch [780]	Speed: 662.89 samples/sec	acc=0.541113
+INFO:root:Epoch[10] Batch [800]	Speed: 662.29 samples/sec	acc=0.542676
+[76000][MARGIN]0.766518,0.368607
+lr-batch-epoch: 0.01 809 10
+testing verification..
+(14000, 512)
+infer time 21.345826
+[cfp_ff][76000]XNorm: 22.003889
+[cfp_ff][76000]Accuracy-Flip: 0.99714+-0.00230
+testing verification..
+(14000, 512)
+infer time 24.937849
+[cfp_fp][76000]XNorm: 18.870976
+[cfp_fp][76000]Accuracy-Flip: 0.90100+-0.01268
+testing verification..
+(12000, 512)
+infer time 19.077448
+[agedb_30][76000]XNorm: 21.995266
+[agedb_30][76000]Accuracy-Flip: 0.97067+-0.00779
+testing verification..
+(12000, 512)
+infer time 19.046583
+[lfw][76000]XNorm: 22.148182
+[lfw][76000]Accuracy-Flip: 0.99683+-0.00293
+[76000]Accuracy-Highest: 0.99717
+INFO:root:Epoch[10] Batch [820]	Speed: 85.54 samples/sec	acc=0.548340
+INFO:root:Epoch[10] Batch [840]	Speed: 650.51 samples/sec	acc=0.546094
+INFO:root:Epoch[10] Batch [860]	Speed: 678.89 samples/sec	acc=0.546289
+INFO:root:Epoch[10] Batch [880]	Speed: 658.34 samples/sec	acc=0.539941
+INFO:root:Epoch[10] Batch [900]	Speed: 662.61 samples/sec	acc=0.538574
+INFO:root:Epoch[10] Batch [920]	Speed: 637.39 samples/sec	acc=0.548535
+INFO:root:Epoch[10] Batch [940]	Speed: 661.71 samples/sec	acc=0.550195
+INFO:root:Epoch[10] Batch [960]	Speed: 662.33 samples/sec	acc=0.539551
+INFO:root:Epoch[10] Batch [980]	Speed: 661.54 samples/sec	acc=0.539844
+INFO:root:Epoch[10] Batch [1000]	Speed: 656.81 samples/sec	acc=0.526172
+INFO:root:Epoch[10] Batch [1020]	Speed: 660.63 samples/sec	acc=0.534375
+INFO:root:Epoch[10] Batch [1040]	Speed: 662.14 samples/sec	acc=0.529785
+INFO:root:Epoch[10] Batch [1060]	Speed: 662.82 samples/sec	acc=0.537891
+INFO:root:Epoch[10] Batch [1080]	Speed: 661.59 samples/sec	acc=0.529687
+INFO:root:Epoch[10] Batch [1100]	Speed: 661.61 samples/sec	acc=0.540137
+INFO:root:Epoch[10] Batch [1120]	Speed: 654.03 samples/sec	acc=0.525781
+INFO:root:Epoch[10] Batch [1140]	Speed: 645.43 samples/sec	acc=0.523828
+INFO:root:Epoch[10] Batch [1160]	Speed: 655.88 samples/sec	acc=0.538184
+INFO:root:Epoch[10] Batch [1180]	Speed: 659.75 samples/sec	acc=0.537402
+INFO:root:Epoch[10] Batch [1200]	Speed: 662.69 samples/sec	acc=0.535645
+INFO:root:Epoch[10] Batch [1220]	Speed: 661.87 samples/sec	acc=0.532617
+INFO:root:Epoch[10] Batch [1240]	Speed: 642.59 samples/sec	acc=0.528223
+INFO:root:Epoch[10] Batch [1260]	Speed: 642.67 samples/sec	acc=0.531055
+INFO:root:Epoch[10] Batch [1280]	Speed: 661.76 samples/sec	acc=0.532129
+INFO:root:Epoch[10] Batch [1300]	Speed: 662.51 samples/sec	acc=0.523340
+INFO:root:Epoch[10] Batch [1320]	Speed: 652.05 samples/sec	acc=0.530176
+INFO:root:Epoch[10] Batch [1340]	Speed: 662.16 samples/sec	acc=0.532813
+INFO:root:Epoch[10] Batch [1360]	Speed: 657.98 samples/sec	acc=0.521582
+INFO:root:Epoch[10] Batch [1380]	Speed: 662.55 samples/sec	acc=0.528809
+INFO:root:Epoch[10] Batch [1400]	Speed: 662.14 samples/sec	acc=0.529687
+INFO:root:Epoch[10] Batch [1420]	Speed: 662.09 samples/sec	acc=0.523340
+INFO:root:Epoch[10] Batch [1440]	Speed: 663.28 samples/sec	acc=0.536719
+INFO:root:Epoch[10] Batch [1460]	Speed: 661.95 samples/sec	acc=0.524316
+INFO:root:Epoch[10] Batch [1480]	Speed: 628.56 samples/sec	acc=0.536133
+INFO:root:Epoch[10] Batch [1500]	Speed: 662.72 samples/sec	acc=0.524121
+INFO:root:Epoch[10] Batch [1520]	Speed: 657.61 samples/sec	acc=0.531738
+INFO:root:Epoch[10] Batch [1540]	Speed: 646.28 samples/sec	acc=0.531055
+INFO:root:Epoch[10] Batch [1560]	Speed: 663.10 samples/sec	acc=0.522070
+INFO:root:Epoch[10] Batch [1580]	Speed: 644.19 samples/sec	acc=0.523145
+INFO:root:Epoch[10] Batch [1600]	Speed: 668.17 samples/sec	acc=0.515820
+INFO:root:Epoch[10] Batch [1620]	Speed: 662.17 samples/sec	acc=0.527637
+INFO:root:Epoch[10] Batch [1640]	Speed: 662.61 samples/sec	acc=0.516992
+INFO:root:Epoch[10] Batch [1660]	Speed: 657.86 samples/sec	acc=0.525293
+INFO:root:Epoch[10] Batch [1680]	Speed: 663.19 samples/sec	acc=0.512109
+INFO:root:Epoch[10] Batch [1700]	Speed: 655.94 samples/sec	acc=0.536426
+INFO:root:Epoch[10] Batch [1720]	Speed: 662.02 samples/sec	acc=0.520117
+INFO:root:Epoch[10] Batch [1740]	Speed: 663.35 samples/sec	acc=0.524707
+INFO:root:Epoch[10] Batch [1760]	Speed: 661.70 samples/sec	acc=0.521973
+INFO:root:Epoch[10] Batch [1780]	Speed: 664.95 samples/sec	acc=0.514746
+INFO:root:Epoch[10] Batch [1800]	Speed: 662.84 samples/sec	acc=0.507812
+lr-batch-epoch: 0.01 1809 10
+INFO:root:Epoch[10] Batch [1820]	Speed: 663.07 samples/sec	acc=0.522070
+INFO:root:Epoch[10] Batch [1840]	Speed: 661.54 samples/sec	acc=0.523828
+INFO:root:Epoch[10] Batch [1860]	Speed: 656.46 samples/sec	acc=0.522852
+INFO:root:Epoch[10] Batch [1880]	Speed: 673.98 samples/sec	acc=0.522559
+INFO:root:Epoch[10] Batch [1900]	Speed: 674.91 samples/sec	acc=0.523438
+INFO:root:Epoch[10] Batch [1920]	Speed: 636.93 samples/sec	acc=0.525293
+INFO:root:Epoch[10] Batch [1940]	Speed: 639.67 samples/sec	acc=0.514844
+INFO:root:Epoch[10] Batch [1960]	Speed: 662.24 samples/sec	acc=0.522754
+INFO:root:Epoch[10] Batch [1980]	Speed: 663.18 samples/sec	acc=0.520801
+INFO:root:Epoch[10] Batch [2000]	Speed: 651.25 samples/sec	acc=0.513672
+INFO:root:Epoch[10] Batch [2020]	Speed: 661.46 samples/sec	acc=0.515527
+INFO:root:Epoch[10] Batch [2040]	Speed: 662.35 samples/sec	acc=0.527930
+INFO:root:Epoch[10] Batch [2060]	Speed: 648.15 samples/sec	acc=0.522754
+INFO:root:Epoch[10] Batch [2080]	Speed: 662.42 samples/sec	acc=0.515137
+INFO:root:Epoch[10] Batch [2100]	Speed: 662.42 samples/sec	acc=0.531250
+INFO:root:Epoch[10] Batch [2120]	Speed: 661.77 samples/sec	acc=0.524707
+INFO:root:Epoch[10] Batch [2140]	Speed: 661.41 samples/sec	acc=0.512305
+INFO:root:Epoch[10] Batch [2160]	Speed: 661.65 samples/sec	acc=0.512891
+INFO:root:Epoch[10] Batch [2180]	Speed: 662.07 samples/sec	acc=0.521875
+INFO:root:Epoch[10] Batch [2200]	Speed: 660.89 samples/sec	acc=0.526855
+INFO:root:Epoch[10] Batch [2220]	Speed: 675.81 samples/sec	acc=0.520312
+INFO:root:Epoch[10] Batch [2240]	Speed: 667.51 samples/sec	acc=0.514551
+INFO:root:Epoch[10] Batch [2260]	Speed: 661.34 samples/sec	acc=0.517676
+INFO:root:Epoch[10] Batch [2280]	Speed: 662.16 samples/sec	acc=0.525293
+INFO:root:Epoch[10] Batch [2300]	Speed: 660.90 samples/sec	acc=0.525195
+INFO:root:Epoch[10] Batch [2320]	Speed: 668.63 samples/sec	acc=0.521484
+INFO:root:Epoch[10] Batch [2340]	Speed: 661.97 samples/sec	acc=0.528711
+INFO:root:Epoch[10] Batch [2360]	Speed: 661.82 samples/sec	acc=0.515625
+INFO:root:Epoch[10] Batch [2380]	Speed: 662.06 samples/sec	acc=0.516309
+INFO:root:Epoch[10] Batch [2400]	Speed: 662.65 samples/sec	acc=0.511816
+INFO:root:Epoch[10] Batch [2420]	Speed: 661.43 samples/sec	acc=0.515820
+INFO:root:Epoch[10] Batch [2440]	Speed: 662.54 samples/sec	acc=0.515039
+INFO:root:Epoch[10] Batch [2460]	Speed: 662.99 samples/sec	acc=0.521289
+INFO:root:Epoch[10] Batch [2480]	Speed: 661.58 samples/sec	acc=0.517676
+INFO:root:Epoch[10] Batch [2500]	Speed: 663.06 samples/sec	acc=0.520898
+INFO:root:Epoch[10] Batch [2520]	Speed: 662.42 samples/sec	acc=0.513770
+INFO:root:Epoch[10] Batch [2540]	Speed: 661.06 samples/sec	acc=0.512402
+INFO:root:Epoch[10] Batch [2560]	Speed: 661.99 samples/sec	acc=0.507910
+INFO:root:Epoch[10] Batch [2580]	Speed: 663.06 samples/sec	acc=0.523633
+INFO:root:Epoch[10] Batch [2600]	Speed: 662.34 samples/sec	acc=0.521680
+INFO:root:Epoch[10] Batch [2620]	Speed: 662.66 samples/sec	acc=0.512207
+INFO:root:Epoch[10] Batch [2640]	Speed: 662.69 samples/sec	acc=0.517383
+INFO:root:Epoch[10] Batch [2660]	Speed: 661.49 samples/sec	acc=0.517188
+INFO:root:Epoch[10] Batch [2680]	Speed: 662.29 samples/sec	acc=0.515527
+INFO:root:Epoch[10] Batch [2700]	Speed: 661.23 samples/sec	acc=0.514551
+INFO:root:Epoch[10] Batch [2720]	Speed: 671.05 samples/sec	acc=0.512500
+INFO:root:Epoch[10] Batch [2740]	Speed: 649.85 samples/sec	acc=0.519043
+INFO:root:Epoch[10] Batch [2760]	Speed: 663.05 samples/sec	acc=0.512500
+INFO:root:Epoch[10] Batch [2780]	Speed: 655.80 samples/sec	acc=0.514648
+INFO:root:Epoch[10] Batch [2800]	Speed: 643.86 samples/sec	acc=0.518848
+[78000][MARGIN]0.763951,0.365564
+lr-batch-epoch: 0.01 2809 10
+testing verification..
+(14000, 512)
+infer time 22.305833
+[cfp_ff][78000]XNorm: 21.742797
+[cfp_ff][78000]Accuracy-Flip: 0.99743+-0.00246
+testing verification..
+(14000, 512)
+infer time 20.541787
+[cfp_fp][78000]XNorm: 18.882740
+[cfp_fp][78000]Accuracy-Flip: 0.90014+-0.01820
+testing verification..
+(12000, 512)
+infer time 16.556826
+[agedb_30][78000]XNorm: 21.761145
+[agedb_30][78000]Accuracy-Flip: 0.97050+-0.00683
+testing verification..
+(12000, 512)
+infer time 17.500644
+[lfw][78000]XNorm: 21.933259
+[lfw][78000]Accuracy-Flip: 0.99683+-0.00217
+[78000]Accuracy-Highest: 0.99717
+INFO:root:Epoch[10] Batch [2820]	Speed: 92.25 samples/sec	acc=0.520312
+INFO:root:Epoch[10] Batch [2840]	Speed: 661.65 samples/sec	acc=0.522559
+INFO:root:Epoch[10] Batch [2860]	Speed: 662.77 samples/sec	acc=0.515625
+INFO:root:Epoch[10] Batch [2880]	Speed: 662.95 samples/sec	acc=0.511426
+INFO:root:Epoch[10] Batch [2900]	Speed: 661.86 samples/sec	acc=0.516406
+INFO:root:Epoch[10] Batch [2920]	Speed: 662.70 samples/sec	acc=0.514160
+INFO:root:Epoch[10] Batch [2940]	Speed: 662.82 samples/sec	acc=0.510938
+INFO:root:Epoch[10] Batch [2960]	Speed: 662.02 samples/sec	acc=0.501660
+INFO:root:Epoch[10] Batch [2980]	Speed: 662.04 samples/sec	acc=0.512793
+INFO:root:Epoch[10] Batch [3000]	Speed: 662.37 samples/sec	acc=0.515527
+INFO:root:Epoch[10] Batch [3020]	Speed: 660.03 samples/sec	acc=0.514160
+INFO:root:Epoch[10] Batch [3040]	Speed: 666.67 samples/sec	acc=0.523047
+INFO:root:Epoch[10] Batch [3060]	Speed: 649.15 samples/sec	acc=0.520605
+INFO:root:Epoch[10] Batch [3080]	Speed: 661.78 samples/sec	acc=0.514258
+INFO:root:Epoch[10] Batch [3100]	Speed: 661.87 samples/sec	acc=0.519336
+INFO:root:Epoch[10] Batch [3120]	Speed: 662.36 samples/sec	acc=0.516992
+INFO:root:Epoch[10] Batch [3140]	Speed: 658.33 samples/sec	acc=0.513770
+INFO:root:Epoch[10] Batch [3160]	Speed: 662.19 samples/sec	acc=0.510840
+INFO:root:Epoch[10] Batch [3180]	Speed: 655.44 samples/sec	acc=0.511426
+INFO:root:Epoch[10] Batch [3200]	Speed: 678.06 samples/sec	acc=0.516797
+INFO:root:Epoch[10] Batch [3220]	Speed: 668.03 samples/sec	acc=0.507324
+INFO:root:Epoch[10] Batch [3240]	Speed: 662.59 samples/sec	acc=0.519531
+INFO:root:Epoch[10] Batch [3260]	Speed: 661.80 samples/sec	acc=0.513672
+INFO:root:Epoch[10] Batch [3280]	Speed: 662.00 samples/sec	acc=0.515625
+INFO:root:Epoch[10] Batch [3300]	Speed: 662.83 samples/sec	acc=0.520215
+INFO:root:Epoch[10] Batch [3320]	Speed: 661.45 samples/sec	acc=0.520312
+INFO:root:Epoch[10] Batch [3340]	Speed: 662.13 samples/sec	acc=0.507910
+INFO:root:Epoch[10] Batch [3360]	Speed: 662.98 samples/sec	acc=0.508594
+INFO:root:Epoch[10] Batch [3380]	Speed: 660.89 samples/sec	acc=0.515625
+INFO:root:Epoch[10] Batch [3400]	Speed: 662.11 samples/sec	acc=0.519629
+INFO:root:Epoch[10] Batch [3420]	Speed: 663.03 samples/sec	acc=0.517676
+INFO:root:Epoch[10] Batch [3440]	Speed: 661.58 samples/sec	acc=0.514941
+INFO:root:Epoch[10] Batch [3460]	Speed: 661.81 samples/sec	acc=0.514648
+INFO:root:Epoch[10] Batch [3480]	Speed: 662.38 samples/sec	acc=0.510254
+INFO:root:Epoch[10] Batch [3500]	Speed: 659.37 samples/sec	acc=0.515039
+INFO:root:Epoch[10] Batch [3520]	Speed: 664.48 samples/sec	acc=0.512109
+INFO:root:Epoch[10] Batch [3540]	Speed: 662.64 samples/sec	acc=0.502930
+INFO:root:Epoch[10] Batch [3560]	Speed: 662.15 samples/sec	acc=0.500977
+INFO:root:Epoch[10] Batch [3580]	Speed: 662.73 samples/sec	acc=0.521094
+INFO:root:Epoch[10] Batch [3600]	Speed: 663.40 samples/sec	acc=0.513086
+INFO:root:Epoch[10] Batch [3620]	Speed: 662.68 samples/sec	acc=0.514258
+INFO:root:Epoch[10] Batch [3640]	Speed: 662.94 samples/sec	acc=0.510742
+INFO:root:Epoch[10] Batch [3660]	Speed: 662.25 samples/sec	acc=0.506836
+INFO:root:Epoch[10] Batch [3680]	Speed: 661.13 samples/sec	acc=0.515039
+INFO:root:Epoch[10] Batch [3700]	Speed: 661.54 samples/sec	acc=0.507715
+INFO:root:Epoch[10] Batch [3720]	Speed: 662.39 samples/sec	acc=0.507129
+INFO:root:Epoch[10] Batch [3740]	Speed: 661.16 samples/sec	acc=0.509277
+INFO:root:Epoch[10] Batch [3760]	Speed: 662.50 samples/sec	acc=0.513770
+INFO:root:Epoch[10] Batch [3780]	Speed: 661.38 samples/sec	acc=0.510547
+INFO:root:Epoch[10] Batch [3800]	Speed: 668.29 samples/sec	acc=0.512207
+lr-batch-epoch: 0.01 3809 10
+INFO:root:Epoch[10] Batch [3820]	Speed: 661.98 samples/sec	acc=0.515332
+INFO:root:Epoch[10] Batch [3840]	Speed: 661.72 samples/sec	acc=0.509961
+INFO:root:Epoch[10] Batch [3860]	Speed: 662.31 samples/sec	acc=0.516504
+INFO:root:Epoch[10] Batch [3880]	Speed: 662.12 samples/sec	acc=0.516309
+INFO:root:Epoch[10] Batch [3900]	Speed: 661.82 samples/sec	acc=0.509668
+INFO:root:Epoch[10] Batch [3920]	Speed: 661.37 samples/sec	acc=0.505176
+INFO:root:Epoch[10] Batch [3940]	Speed: 661.32 samples/sec	acc=0.508594
+INFO:root:Epoch[10] Batch [3960]	Speed: 662.50 samples/sec	acc=0.513770
+INFO:root:Epoch[10] Batch [3980]	Speed: 661.19 samples/sec	acc=0.515234
+INFO:root:Epoch[10] Batch [4000]	Speed: 661.51 samples/sec	acc=0.513184
+INFO:root:Epoch[10] Batch [4020]	Speed: 662.09 samples/sec	acc=0.506836
+INFO:root:Epoch[10] Batch [4040]	Speed: 661.66 samples/sec	acc=0.502832
+INFO:root:Epoch[10] Batch [4060]	Speed: 660.81 samples/sec	acc=0.515625
+INFO:root:Epoch[10] Batch [4080]	Speed: 641.06 samples/sec	acc=0.513965
+INFO:root:Epoch[10] Batch [4100]	Speed: 661.29 samples/sec	acc=0.508887
+INFO:root:Epoch[10] Batch [4120]	Speed: 661.94 samples/sec	acc=0.518457
+INFO:root:Epoch[10] Batch [4140]	Speed: 662.06 samples/sec	acc=0.504785
+INFO:root:Epoch[10] Batch [4160]	Speed: 661.90 samples/sec	acc=0.508789
+INFO:root:Epoch[10] Batch [4180]	Speed: 662.09 samples/sec	acc=0.516406
+INFO:root:Epoch[10] Batch [4200]	Speed: 662.36 samples/sec	acc=0.506641
+INFO:root:Epoch[10] Batch [4220]	Speed: 649.50 samples/sec	acc=0.504492
+INFO:root:Epoch[10] Batch [4240]	Speed: 661.27 samples/sec	acc=0.503906
+INFO:root:Epoch[10] Batch [4260]	Speed: 662.82 samples/sec	acc=0.512988
+INFO:root:Epoch[10] Batch [4280]	Speed: 661.64 samples/sec	acc=0.508984
+INFO:root:Epoch[10] Batch [4300]	Speed: 662.20 samples/sec	acc=0.522852
+INFO:root:Epoch[10] Batch [4320]	Speed: 662.35 samples/sec	acc=0.514844
+INFO:root:Epoch[10] Batch [4340]	Speed: 661.69 samples/sec	acc=0.516309
+INFO:root:Epoch[10] Batch [4360]	Speed: 661.98 samples/sec	acc=0.519922
+INFO:root:Epoch[10] Batch [4380]	Speed: 662.40 samples/sec	acc=0.514160
+INFO:root:Epoch[10] Batch [4400]	Speed: 661.94 samples/sec	acc=0.514746
+INFO:root:Epoch[10] Batch [4420]	Speed: 661.95 samples/sec	acc=0.510254
+INFO:root:Epoch[10] Batch [4440]	Speed: 661.79 samples/sec	acc=0.505273
+INFO:root:Epoch[10] Batch [4460]	Speed: 661.85 samples/sec	acc=0.509473
+INFO:root:Epoch[10] Batch [4480]	Speed: 661.30 samples/sec	acc=0.515137
+INFO:root:Epoch[10] Batch [4500]	Speed: 662.09 samples/sec	acc=0.508887
+INFO:root:Epoch[10] Batch [4520]	Speed: 661.66 samples/sec	acc=0.512891
+INFO:root:Epoch[10] Batch [4540]	Speed: 661.62 samples/sec	acc=0.511523
+INFO:root:Epoch[10] Batch [4560]	Speed: 662.46 samples/sec	acc=0.512109
+INFO:root:Epoch[10] Batch [4580]	Speed: 657.30 samples/sec	acc=0.510254
+INFO:root:Epoch[10] Batch [4600]	Speed: 664.14 samples/sec	acc=0.513867
+INFO:root:Epoch[10] Batch [4620]	Speed: 662.11 samples/sec	acc=0.515918
+INFO:root:Epoch[10] Batch [4640]	Speed: 662.15 samples/sec	acc=0.504297
+INFO:root:Epoch[10] Batch [4660]	Speed: 662.22 samples/sec	acc=0.515918
+INFO:root:Epoch[10] Batch [4680]	Speed: 662.00 samples/sec	acc=0.509180
+INFO:root:Epoch[10] Batch [4700]	Speed: 656.05 samples/sec	acc=0.508887
+INFO:root:Epoch[10] Batch [4720]	Speed: 661.71 samples/sec	acc=0.511914
+INFO:root:Epoch[10] Batch [4740]	Speed: 662.36 samples/sec	acc=0.498730
+INFO:root:Epoch[10] Batch [4760]	Speed: 661.53 samples/sec	acc=0.511328
+INFO:root:Epoch[10] Batch [4780]	Speed: 662.24 samples/sec	acc=0.505762
+INFO:root:Epoch[10] Batch [4800]	Speed: 662.22 samples/sec	acc=0.513867
+[80000][MARGIN]0.757792,0.356637
+lr change to 0.001
+lr-batch-epoch: 0.001 4809 10
+testing verification..
+(14000, 512)
+infer time 20.193704
+[cfp_ff][80000]XNorm: 21.835066
+[cfp_ff][80000]Accuracy-Flip: 0.99686+-0.00246
+testing verification..
+(14000, 512)
+infer time 18.215044
+[cfp_fp][80000]XNorm: 19.073938
+[cfp_fp][80000]Accuracy-Flip: 0.90757+-0.01565
+testing verification..
+(12000, 512)
+infer time 14.828748
+[agedb_30][80000]XNorm: 22.301100
+[agedb_30][80000]Accuracy-Flip: 0.96883+-0.00771
+testing verification..
+(12000, 512)
+infer time 14.81863
+[lfw][80000]XNorm: 22.352697
+[lfw][80000]Accuracy-Flip: 0.99617+-0.00350
+[80000]Accuracy-Highest: 0.99717
+INFO:root:Epoch[10] Batch [4820]	Speed: 100.70 samples/sec	acc=0.485156
+INFO:root:Epoch[10] Batch [4840]	Speed: 661.51 samples/sec	acc=0.479785
+INFO:root:Epoch[10] Batch [4860]	Speed: 662.07 samples/sec	acc=0.509570
+INFO:root:Epoch[10] Batch [4880]	Speed: 661.71 samples/sec	acc=0.533594
+INFO:root:Epoch[10] Batch [4900]	Speed: 659.34 samples/sec	acc=0.533789
+INFO:root:Epoch[10] Batch [4920]	Speed: 661.61 samples/sec	acc=0.554004
+INFO:root:Epoch[10] Batch [4940]	Speed: 660.77 samples/sec	acc=0.545508
+INFO:root:Epoch[10] Batch [4960]	Speed: 662.51 samples/sec	acc=0.560742
+INFO:root:Epoch[10] Batch [4980]	Speed: 662.20 samples/sec	acc=0.569629
+INFO:root:Epoch[10] Batch [5000]	Speed: 660.65 samples/sec	acc=0.560742
+INFO:root:Epoch[10] Batch [5020]	Speed: 649.35 samples/sec	acc=0.566309
+INFO:root:Epoch[10] Batch [5040]	Speed: 651.36 samples/sec	acc=0.561914
+INFO:root:Epoch[10] Batch [5060]	Speed: 661.99 samples/sec	acc=0.562988
+INFO:root:Epoch[10] Batch [5080]	Speed: 661.57 samples/sec	acc=0.575684
+INFO:root:Epoch[10] Batch [5100]	Speed: 662.19 samples/sec	acc=0.577734
+INFO:root:Epoch[10] Batch [5120]	Speed: 661.94 samples/sec	acc=0.573438
+INFO:root:Epoch[10] Batch [5140]	Speed: 662.21 samples/sec	acc=0.579883
+INFO:root:Epoch[10] Batch [5160]	Speed: 662.53 samples/sec	acc=0.580859
+INFO:root:Epoch[10] Batch [5180]	Speed: 656.48 samples/sec	acc=0.572070
+INFO:root:Epoch[10] Batch [5200]	Speed: 660.39 samples/sec	acc=0.587305
+INFO:root:Epoch[10] Batch [5220]	Speed: 655.62 samples/sec	acc=0.578223
+INFO:root:Epoch[10] Batch [5240]	Speed: 661.59 samples/sec	acc=0.580176
+INFO:root:Epoch[10] Batch [5260]	Speed: 661.71 samples/sec	acc=0.586230
+INFO:root:Epoch[10] Batch [5280]	Speed: 657.49 samples/sec	acc=0.579785
+INFO:root:Epoch[10] Batch [5300]	Speed: 650.05 samples/sec	acc=0.582715
+INFO:root:Epoch[10] Batch [5320]	Speed: 662.82 samples/sec	acc=0.587891
+INFO:root:Epoch[10] Batch [5340]	Speed: 662.62 samples/sec	acc=0.578613
+INFO:root:Epoch[10] Batch [5360]	Speed: 642.86 samples/sec	acc=0.583203
+INFO:root:Epoch[10] Batch [5380]	Speed: 662.66 samples/sec	acc=0.586035
+INFO:root:Epoch[10] Batch [5400]	Speed: 665.24 samples/sec	acc=0.589648
+INFO:root:Epoch[10] Batch [5420]	Speed: 661.20 samples/sec	acc=0.584082
+INFO:root:Epoch[10] Batch [5440]	Speed: 661.75 samples/sec	acc=0.582715
+INFO:root:Epoch[10] Batch [5460]	Speed: 662.44 samples/sec	acc=0.591406
+INFO:root:Epoch[10] Batch [5480]	Speed: 661.14 samples/sec	acc=0.586621
+INFO:root:Epoch[10] Batch [5500]	Speed: 661.79 samples/sec	acc=0.591797
+INFO:root:Epoch[10] Batch [5520]	Speed: 648.42 samples/sec	acc=0.597754
+INFO:root:Epoch[10] Batch [5540]	Speed: 672.93 samples/sec	acc=0.599414
+INFO:root:Epoch[10] Batch [5560]	Speed: 662.42 samples/sec	acc=0.597461
+INFO:root:Epoch[10] Batch [5580]	Speed: 662.10 samples/sec	acc=0.597656
+INFO:root:Epoch[10] Batch [5600]	Speed: 661.51 samples/sec	acc=0.604102
+INFO:root:Epoch[10] Batch [5620]	Speed: 661.90 samples/sec	acc=0.586719
+INFO:root:Epoch[10] Batch [5640]	Speed: 662.20 samples/sec	acc=0.598730
+INFO:root:Epoch[10] Batch [5660]	Speed: 661.40 samples/sec	acc=0.598437
+INFO:root:Epoch[10] Batch [5680]	Speed: 662.36 samples/sec	acc=0.601953
+INFO:root:Epoch[10] Batch [5700]	Speed: 661.73 samples/sec	acc=0.604004
+INFO:root:Epoch[10] Batch [5720]	Speed: 661.01 samples/sec	acc=0.603027
+INFO:root:Epoch[10] Batch [5740]	Speed: 661.64 samples/sec	acc=0.605078
+INFO:root:Epoch[10] Batch [5760]	Speed: 653.45 samples/sec	acc=0.604590
+INFO:root:Epoch[10] Batch [5780]	Speed: 666.89 samples/sec	acc=0.601074
+INFO:root:Epoch[10] Batch [5800]	Speed: 652.31 samples/sec	acc=0.612305
+lr-batch-epoch: 0.001 5809 10
+INFO:root:Epoch[10] Batch [5820]	Speed: 670.74 samples/sec	acc=0.613184
+INFO:root:Epoch[10] Batch [5840]	Speed: 661.20 samples/sec	acc=0.600391
+INFO:root:Epoch[10] Batch [5860]	Speed: 662.13 samples/sec	acc=0.596582
+INFO:root:Epoch[10] Batch [5880]	Speed: 651.04 samples/sec	acc=0.603320
+INFO:root:Epoch[10] Batch [5900]	Speed: 661.14 samples/sec	acc=0.611426
+INFO:root:Epoch[10] Batch [5920]	Speed: 660.84 samples/sec	acc=0.610840
+INFO:root:Epoch[10] Batch [5940]	Speed: 659.71 samples/sec	acc=0.608301
+INFO:root:Epoch[10] Batch [5960]	Speed: 662.08 samples/sec	acc=0.609570
+INFO:root:Epoch[10] Batch [5980]	Speed: 661.93 samples/sec	acc=0.615332
+INFO:root:Epoch[10] Batch [6000]	Speed: 659.89 samples/sec	acc=0.614355
+INFO:root:Epoch[10] Batch [6020]	Speed: 661.95 samples/sec	acc=0.607227
+INFO:root:Epoch[10] Batch [6040]	Speed: 661.60 samples/sec	acc=0.610156
+INFO:root:Epoch[10] Batch [6060]	Speed: 662.72 samples/sec	acc=0.613965
+INFO:root:Epoch[10] Batch [6080]	Speed: 661.70 samples/sec	acc=0.612695
+INFO:root:Epoch[10] Batch [6100]	Speed: 661.85 samples/sec	acc=0.612695
+INFO:root:Epoch[10] Batch [6120]	Speed: 662.35 samples/sec	acc=0.620410
+INFO:root:Epoch[10] Batch [6140]	Speed: 661.86 samples/sec	acc=0.614648
+INFO:root:Epoch[10] Batch [6160]	Speed: 646.63 samples/sec	acc=0.614551
+INFO:root:Epoch[10] Batch [6180]	Speed: 652.10 samples/sec	acc=0.616113
+INFO:root:Epoch[10] Batch [6200]	Speed: 659.59 samples/sec	acc=0.612012
+INFO:root:Epoch[10] Batch [6220]	Speed: 663.49 samples/sec	acc=0.615332
+INFO:root:Epoch[10] Batch [6240]	Speed: 649.58 samples/sec	acc=0.615918
+INFO:root:Epoch[10] Batch [6260]	Speed: 661.33 samples/sec	acc=0.613379
+INFO:root:Epoch[10] Batch [6280]	Speed: 661.48 samples/sec	acc=0.612305
+INFO:root:Epoch[10] Batch [6300]	Speed: 662.48 samples/sec	acc=0.608789
+INFO:root:Epoch[10] Batch [6320]	Speed: 666.84 samples/sec	acc=0.625977
+INFO:root:Epoch[10] Batch [6340]	Speed: 661.62 samples/sec	acc=0.615039
+INFO:root:Epoch[10] Batch [6360]	Speed: 662.50 samples/sec	acc=0.618457
+INFO:root:Epoch[10] Batch [6380]	Speed: 658.99 samples/sec	acc=0.620996
+INFO:root:Epoch[10] Batch [6400]	Speed: 662.38 samples/sec	acc=0.624902
+INFO:root:Epoch[10] Batch [6420]	Speed: 662.59 samples/sec	acc=0.619043
+INFO:root:Epoch[10] Batch [6440]	Speed: 661.52 samples/sec	acc=0.624219
+INFO:root:Epoch[10] Batch [6460]	Speed: 661.98 samples/sec	acc=0.626465
+INFO:root:Epoch[10] Batch [6480]	Speed: 662.33 samples/sec	acc=0.624609
+INFO:root:Epoch[10] Batch [6500]	Speed: 661.43 samples/sec	acc=0.615625
+INFO:root:Epoch[10] Batch [6520]	Speed: 662.08 samples/sec	acc=0.617188
+INFO:root:Epoch[10] Batch [6540]	Speed: 662.44 samples/sec	acc=0.622852
+INFO:root:Epoch[10] Batch [6560]	Speed: 668.89 samples/sec	acc=0.623828
+INFO:root:Epoch[10] Batch [6580]	Speed: 657.25 samples/sec	acc=0.629297
+INFO:root:Epoch[10] Batch [6600]	Speed: 662.63 samples/sec	acc=0.619629
+INFO:root:Epoch[10] Batch [6620]	Speed: 661.48 samples/sec	acc=0.628320
+INFO:root:Epoch[10] Batch [6640]	Speed: 661.40 samples/sec	acc=0.622168
+INFO:root:Epoch[10] Batch [6660]	Speed: 661.12 samples/sec	acc=0.623145
+INFO:root:Epoch[10] Batch [6680]	Speed: 661.53 samples/sec	acc=0.629687
+INFO:root:Epoch[10] Batch [6700]	Speed: 661.27 samples/sec	acc=0.623340
+INFO:root:Epoch[10] Batch [6720]	Speed: 662.32 samples/sec	acc=0.629883
+INFO:root:Epoch[10] Batch [6740]	Speed: 661.56 samples/sec	acc=0.626074
+INFO:root:Epoch[10] Batch [6760]	Speed: 662.13 samples/sec	acc=0.630566
+INFO:root:Epoch[10] Batch [6780]	Speed: 662.29 samples/sec	acc=0.628906
+INFO:root:Epoch[10] Batch [6800]	Speed: 662.13 samples/sec	acc=0.625586
+[82000][MARGIN]0.795547,0.411048
+lr-batch-epoch: 0.001 6809 10
+testing verification..
+(14000, 512)
+infer time 23.51775
+[cfp_ff][82000]XNorm: 21.816110
+[cfp_ff][82000]Accuracy-Flip: 0.99814+-0.00222
+testing verification..
+(14000, 512)
+infer time 25.430244
+[cfp_fp][82000]XNorm: 19.249503
+[cfp_fp][82000]Accuracy-Flip: 0.92000+-0.01553
+testing verification..
+(12000, 512)
+infer time 20.465929
+[agedb_30][82000]XNorm: 22.299578
+[agedb_30][82000]Accuracy-Flip: 0.97650+-0.00664
+testing verification..
+(12000, 512)
+infer time 20.801524
+[lfw][82000]XNorm: 22.182999
+[lfw][82000]Accuracy-Flip: 0.99767+-0.00281
+saving 41
+INFO:root:Saved checkpoint to "../model-r50-am-lfw/model-0041.params"
+[82000]Accuracy-Highest: 0.99767
+INFO:root:Epoch[10] Batch [6820]	Speed: 78.65 samples/sec	acc=0.624414
+INFO:root:Epoch[10] Batch [6840]	Speed: 662.41 samples/sec	acc=0.634863
+INFO:root:Epoch[10] Batch [6860]	Speed: 661.77 samples/sec	acc=0.632715
+INFO:root:Epoch[10] Batch [6880]	Speed: 661.81 samples/sec	acc=0.620703
+INFO:root:Epoch[10] Batch [6900]	Speed: 665.16 samples/sec	acc=0.635059
+INFO:root:Epoch[10] Batch [6920]	Speed: 645.79 samples/sec	acc=0.630078
+INFO:root:Epoch[10] Batch [6940]	Speed: 651.71 samples/sec	acc=0.639062
+INFO:root:Epoch[10] Batch [6960]	Speed: 645.37 samples/sec	acc=0.639551
+INFO:root:Epoch[10] Batch [6980]	Speed: 669.54 samples/sec	acc=0.632812
+INFO:root:Epoch[10] Batch [7000]	Speed: 661.93 samples/sec	acc=0.631055
+INFO:root:Epoch[10] Batch [7020]	Speed: 663.21 samples/sec	acc=0.626465
+INFO:root:Epoch[10] Batch [7040]	Speed: 651.73 samples/sec	acc=0.632715
+INFO:root:Epoch[10] Batch [7060]	Speed: 627.33 samples/sec	acc=0.638672
+INFO:root:Epoch[10] Batch [7080]	Speed: 664.37 samples/sec	acc=0.638867
+INFO:root:Epoch[10] Batch [7100]	Speed: 657.87 samples/sec	acc=0.635938
+INFO:root:Epoch[10] Batch [7120]	Speed: 665.13 samples/sec	acc=0.629883
+INFO:root:Epoch[10] Batch [7140]	Speed: 661.72 samples/sec	acc=0.627344
+INFO:root:Epoch[10] Batch [7160]	Speed: 659.49 samples/sec	acc=0.645215
+INFO:root:Epoch[10] Batch [7180]	Speed: 653.52 samples/sec	acc=0.635254
+INFO:root:Epoch[10] Batch [7200]	Speed: 624.98 samples/sec	acc=0.644922
+INFO:root:Epoch[10] Batch [7220]	Speed: 654.33 samples/sec	acc=0.636035
+INFO:root:Epoch[10] Batch [7240]	Speed: 662.73 samples/sec	acc=0.639941
+INFO:root:Epoch[10] Batch [7260]	Speed: 662.53 samples/sec	acc=0.639355
+INFO:root:Epoch[10] Batch [7280]	Speed: 661.53 samples/sec	acc=0.636133
+INFO:root:Epoch[10] Batch [7300]	Speed: 662.77 samples/sec	acc=0.633789
+INFO:root:Epoch[10] Batch [7320]	Speed: 662.60 samples/sec	acc=0.642871
+INFO:root:Epoch[10] Batch [7340]	Speed: 661.51 samples/sec	acc=0.635742
+INFO:root:Epoch[10] Batch [7360]	Speed: 618.70 samples/sec	acc=0.633984
+INFO:root:Epoch[10] Batch [7380]	Speed: 646.58 samples/sec	acc=0.638867
+INFO:root:Epoch[10] Batch [7400]	Speed: 661.66 samples/sec	acc=0.642773
+INFO:root:Epoch[10] Batch [7420]	Speed: 660.14 samples/sec	acc=0.637012
+INFO:root:Epoch[10] Batch [7440]	Speed: 662.69 samples/sec	acc=0.635742
+INFO:root:Epoch[10] Batch [7460]	Speed: 661.37 samples/sec	acc=0.631152
+INFO:root:Epoch[10] Batch [7480]	Speed: 662.18 samples/sec	acc=0.640332
+INFO:root:Epoch[10] Batch [7500]	Speed: 662.74 samples/sec	acc=0.629004
+INFO:root:Epoch[10] Train-acc=0.643446
+INFO:root:Epoch[10] Time cost=6231.634
+call reset()
+INFO:root:Epoch[11] Batch [20]	Speed: 658.18 samples/sec	acc=0.713077
+INFO:root:Epoch[11] Batch [40]	Speed: 662.09 samples/sec	acc=0.717285
+INFO:root:Epoch[11] Batch [60]	Speed: 662.33 samples/sec	acc=0.723145
+INFO:root:Epoch[11] Batch [80]	Speed: 648.65 samples/sec	acc=0.722559
+INFO:root:Epoch[11] Batch [100]	Speed: 666.57 samples/sec	acc=0.711914
+INFO:root:Epoch[11] Batch [120]	Speed: 655.94 samples/sec	acc=0.723535
+INFO:root:Epoch[11] Batch [140]	Speed: 659.46 samples/sec	acc=0.714063
+INFO:root:Epoch[11] Batch [160]	Speed: 662.45 samples/sec	acc=0.720117
+INFO:root:Epoch[11] Batch [180]	Speed: 661.51 samples/sec	acc=0.714160
+INFO:root:Epoch[11] Batch [200]	Speed: 661.18 samples/sec	acc=0.718555
+INFO:root:Epoch[11] Batch [220]	Speed: 662.64 samples/sec	acc=0.719824
+INFO:root:Epoch[11] Batch [240]	Speed: 661.38 samples/sec	acc=0.718652
+INFO:root:Epoch[11] Batch [260]	Speed: 662.89 samples/sec	acc=0.727637
+INFO:root:Epoch[11] Batch [280]	Speed: 653.24 samples/sec	acc=0.721582
+lr-batch-epoch: 0.001 290 11
+INFO:root:Epoch[11] Batch [300]	Speed: 657.62 samples/sec	acc=0.721973
+INFO:root:Epoch[11] Batch [320]	Speed: 656.25 samples/sec	acc=0.725098
+INFO:root:Epoch[11] Batch [340]	Speed: 653.91 samples/sec	acc=0.720605
+INFO:root:Epoch[11] Batch [360]	Speed: 661.49 samples/sec	acc=0.716309
+INFO:root:Epoch[11] Batch [380]	Speed: 660.46 samples/sec	acc=0.720508
+INFO:root:Epoch[11] Batch [400]	Speed: 650.84 samples/sec	acc=0.712305
+INFO:root:Epoch[11] Batch [420]	Speed: 667.02 samples/sec	acc=0.726270
+INFO:root:Epoch[11] Batch [440]	Speed: 642.18 samples/sec	acc=0.711816
+INFO:root:Epoch[11] Batch [460]	Speed: 661.97 samples/sec	acc=0.717578
+INFO:root:Epoch[11] Batch [480]	Speed: 661.21 samples/sec	acc=0.708496
+INFO:root:Epoch[11] Batch [500]	Speed: 661.64 samples/sec	acc=0.720801
+INFO:root:Epoch[11] Batch [520]	Speed: 662.55 samples/sec	acc=0.729297
+INFO:root:Epoch[11] Batch [540]	Speed: 661.36 samples/sec	acc=0.713672
+INFO:root:Epoch[11] Batch [560]	Speed: 662.15 samples/sec	acc=0.733887
+INFO:root:Epoch[11] Batch [580]	Speed: 637.75 samples/sec	acc=0.717773
+INFO:root:Epoch[11] Batch [600]	Speed: 661.76 samples/sec	acc=0.724023
+INFO:root:Epoch[11] Batch [620]	Speed: 662.89 samples/sec	acc=0.718359
+INFO:root:Epoch[11] Batch [640]	Speed: 644.97 samples/sec	acc=0.720117
+INFO:root:Epoch[11] Batch [660]	Speed: 643.57 samples/sec	acc=0.721484
+INFO:root:Epoch[11] Batch [680]	Speed: 660.28 samples/sec	acc=0.716504
+INFO:root:Epoch[11] Batch [700]	Speed: 656.84 samples/sec	acc=0.718262
+INFO:root:Epoch[11] Batch [720]	Speed: 662.32 samples/sec	acc=0.726270
+INFO:root:Epoch[11] Batch [740]	Speed: 662.19 samples/sec	acc=0.720313
+INFO:root:Epoch[11] Batch [760]	Speed: 662.40 samples/sec	acc=0.721777
+INFO:root:Epoch[11] Batch [780]	Speed: 661.71 samples/sec	acc=0.721875
+INFO:root:Epoch[11] Batch [800]	Speed: 662.72 samples/sec	acc=0.724023
+INFO:root:Epoch[11] Batch [820]	Speed: 662.85 samples/sec	acc=0.718555
+INFO:root:Epoch[11] Batch [840]	Speed: 662.03 samples/sec	acc=0.715039
+INFO:root:Epoch[11] Batch [860]	Speed: 661.93 samples/sec	acc=0.721191
+INFO:root:Epoch[11] Batch [880]	Speed: 664.30 samples/sec	acc=0.724609
+INFO:root:Epoch[11] Batch [900]	Speed: 661.11 samples/sec	acc=0.716504
+INFO:root:Epoch[11] Batch [920]	Speed: 662.22 samples/sec	acc=0.718848
+INFO:root:Epoch[11] Batch [940]	Speed: 661.79 samples/sec	acc=0.721973
+INFO:root:Epoch[11] Batch [960]	Speed: 641.59 samples/sec	acc=0.732227
+INFO:root:Epoch[11] Batch [980]	Speed: 662.47 samples/sec	acc=0.716797
+INFO:root:Epoch[11] Batch [1000]	Speed: 643.29 samples/sec	acc=0.717383
+INFO:root:Epoch[11] Batch [1020]	Speed: 661.61 samples/sec	acc=0.724512
+INFO:root:Epoch[11] Batch [1040]	Speed: 644.83 samples/sec	acc=0.730273
+INFO:root:Epoch[11] Batch [1060]	Speed: 660.10 samples/sec	acc=0.720313
+INFO:root:Epoch[11] Batch [1080]	Speed: 661.61 samples/sec	acc=0.720215
+INFO:root:Epoch[11] Batch [1100]	Speed: 659.62 samples/sec	acc=0.715332
+INFO:root:Epoch[11] Batch [1120]	Speed: 659.91 samples/sec	acc=0.724805
+INFO:root:Epoch[11] Batch [1140]	Speed: 662.08 samples/sec	acc=0.718945
+INFO:root:Epoch[11] Batch [1160]	Speed: 661.37 samples/sec	acc=0.718945
+INFO:root:Epoch[11] Batch [1180]	Speed: 663.30 samples/sec	acc=0.716992
+INFO:root:Epoch[11] Batch [1200]	Speed: 661.13 samples/sec	acc=0.729980
+INFO:root:Epoch[11] Batch [1220]	Speed: 659.26 samples/sec	acc=0.733008
+INFO:root:Epoch[11] Batch [1240]	Speed: 663.69 samples/sec	acc=0.730859
+INFO:root:Epoch[11] Batch [1260]	Speed: 647.66 samples/sec	acc=0.718750
+INFO:root:Epoch[11] Batch [1280]	Speed: 662.49 samples/sec	acc=0.720410
+[84000][MARGIN]0.783003,0.393450
+lr-batch-epoch: 0.001 1290 11
+testing verification..
+(14000, 512)
+infer time 20.904889
+[cfp_ff][84000]XNorm: 21.706894
+[cfp_ff][84000]Accuracy-Flip: 0.99729+-0.00243
+testing verification..
+(14000, 512)
+infer time 24.969568
+[cfp_fp][84000]XNorm: 19.449380
+[cfp_fp][84000]Accuracy-Flip: 0.92157+-0.01396
+testing verification..
+(12000, 512)
+infer time 21.333209
+[agedb_30][84000]XNorm: 22.451793
+[agedb_30][84000]Accuracy-Flip: 0.97633+-0.00657
+testing verification..
+(12000, 512)
+infer time 21.245474
+[lfw][84000]XNorm: 22.198713
+[lfw][84000]Accuracy-Flip: 0.99767+-0.00281
+saving 42
+INFO:root:Saved checkpoint to "../model-r50-am-lfw/model-0042.params"
+[84000]Accuracy-Highest: 0.99767
+INFO:root:Epoch[11] Batch [1300]	Speed: 80.00 samples/sec	acc=0.716992
+INFO:root:Epoch[11] Batch [1320]	Speed: 653.05 samples/sec	acc=0.725391
+INFO:root:Epoch[11] Batch [1340]	Speed: 661.33 samples/sec	acc=0.718848
+INFO:root:Epoch[11] Batch [1360]	Speed: 661.98 samples/sec	acc=0.727051
+INFO:root:Epoch[11] Batch [1380]	Speed: 661.54 samples/sec	acc=0.723926
+INFO:root:Epoch[11] Batch [1400]	Speed: 661.65 samples/sec	acc=0.723145
+INFO:root:Epoch[11] Batch [1420]	Speed: 662.32 samples/sec	acc=0.722363
+INFO:root:Epoch[11] Batch [1440]	Speed: 650.98 samples/sec	acc=0.724316
+INFO:root:Epoch[11] Batch [1460]	Speed: 661.70 samples/sec	acc=0.728711
+INFO:root:Epoch[11] Batch [1480]	Speed: 661.87 samples/sec	acc=0.728516
+INFO:root:Epoch[11] Batch [1500]	Speed: 663.45 samples/sec	acc=0.728223
+INFO:root:Epoch[11] Batch [1520]	Speed: 661.14 samples/sec	acc=0.729492
+INFO:root:Epoch[11] Batch [1540]	Speed: 660.94 samples/sec	acc=0.717383
+INFO:root:Epoch[11] Batch [1560]	Speed: 658.77 samples/sec	acc=0.719824
+INFO:root:Epoch[11] Batch [1580]	Speed: 668.94 samples/sec	acc=0.720996
+INFO:root:Epoch[11] Batch [1600]	Speed: 662.47 samples/sec	acc=0.725488
+INFO:root:Epoch[11] Batch [1620]	Speed: 661.41 samples/sec	acc=0.724512
+INFO:root:Epoch[11] Batch [1640]	Speed: 662.32 samples/sec	acc=0.724707
+INFO:root:Epoch[11] Batch [1660]	Speed: 656.46 samples/sec	acc=0.728125
+INFO:root:Epoch[11] Batch [1680]	Speed: 679.16 samples/sec	acc=0.730469
+INFO:root:Epoch[11] Batch [1700]	Speed: 662.37 samples/sec	acc=0.728223
+INFO:root:Epoch[11] Batch [1720]	Speed: 662.30 samples/sec	acc=0.718555
+INFO:root:Epoch[11] Batch [1740]	Speed: 660.93 samples/sec	acc=0.721387
+INFO:root:Epoch[11] Batch [1760]	Speed: 661.97 samples/sec	acc=0.729199
+INFO:root:Epoch[11] Batch [1780]	Speed: 661.57 samples/sec	acc=0.728711
+INFO:root:Epoch[11] Batch [1800]	Speed: 661.51 samples/sec	acc=0.724219
+INFO:root:Epoch[11] Batch [1820]	Speed: 661.86 samples/sec	acc=0.729492
+INFO:root:Epoch[11] Batch [1840]	Speed: 646.59 samples/sec	acc=0.720117
+INFO:root:Epoch[11] Batch [1860]	Speed: 662.02 samples/sec	acc=0.722070
+INFO:root:Epoch[11] Batch [1880]	Speed: 632.21 samples/sec	acc=0.723047
+INFO:root:Epoch[11] Batch [1900]	Speed: 636.99 samples/sec	acc=0.730371
+INFO:root:Epoch[11] Batch [1920]	Speed: 661.67 samples/sec	acc=0.732129
+INFO:root:Epoch[11] Batch [1940]	Speed: 661.33 samples/sec	acc=0.735840
+INFO:root:Epoch[11] Batch [1960]	Speed: 661.84 samples/sec	acc=0.719434
+INFO:root:Epoch[11] Batch [1980]	Speed: 661.64 samples/sec	acc=0.730957
+INFO:root:Epoch[11] Batch [2000]	Speed: 661.97 samples/sec	acc=0.720117
+INFO:root:Epoch[11] Batch [2020]	Speed: 662.11 samples/sec	acc=0.730957
+INFO:root:Epoch[11] Batch [2040]	Speed: 660.18 samples/sec	acc=0.732031
+INFO:root:Epoch[11] Batch [2060]	Speed: 661.74 samples/sec	acc=0.727246
+INFO:root:Epoch[11] Batch [2080]	Speed: 656.76 samples/sec	acc=0.732520
+INFO:root:Epoch[11] Batch [2100]	Speed: 661.27 samples/sec	acc=0.730469
+INFO:root:Epoch[11] Batch [2120]	Speed: 658.52 samples/sec	acc=0.722363
+INFO:root:Epoch[11] Batch [2140]	Speed: 662.38 samples/sec	acc=0.723242
+INFO:root:Epoch[11] Batch [2160]	Speed: 661.00 samples/sec	acc=0.728418
+INFO:root:Epoch[11] Batch [2180]	Speed: 661.62 samples/sec	acc=0.729297
+INFO:root:Epoch[11] Batch [2200]	Speed: 662.05 samples/sec	acc=0.724512
+INFO:root:Epoch[11] Batch [2220]	Speed: 661.49 samples/sec	acc=0.734961
+INFO:root:Epoch[11] Batch [2240]	Speed: 661.92 samples/sec	acc=0.730371
+INFO:root:Epoch[11] Batch [2260]	Speed: 662.09 samples/sec	acc=0.731934
+INFO:root:Epoch[11] Batch [2280]	Speed: 661.26 samples/sec	acc=0.733594
+lr-batch-epoch: 0.001 2290 11
+INFO:root:Epoch[11] Batch [2300]	Speed: 661.32 samples/sec	acc=0.729492
+INFO:root:Epoch[11] Batch [2320]	Speed: 661.70 samples/sec	acc=0.723730
+INFO:root:Epoch[11] Batch [2340]	Speed: 661.63 samples/sec	acc=0.731348
+INFO:root:Epoch[11] Batch [2360]	Speed: 658.33 samples/sec	acc=0.731055
+INFO:root:Epoch[11] Batch [2380]	Speed: 658.91 samples/sec	acc=0.732129
+INFO:root:Epoch[11] Batch [2400]	Speed: 661.71 samples/sec	acc=0.732031
+INFO:root:Epoch[11] Batch [2420]	Speed: 661.92 samples/sec	acc=0.732227
+INFO:root:Epoch[11] Batch [2440]	Speed: 657.86 samples/sec	acc=0.724902
+INFO:root:Epoch[11] Batch [2460]	Speed: 682.65 samples/sec	acc=0.728418
+INFO:root:Epoch[11] Batch [2480]	Speed: 661.93 samples/sec	acc=0.733105
+INFO:root:Epoch[11] Batch [2500]	Speed: 662.24 samples/sec	acc=0.734961
+INFO:root:Epoch[11] Batch [2520]	Speed: 661.05 samples/sec	acc=0.733008
+INFO:root:Epoch[11] Batch [2540]	Speed: 661.44 samples/sec	acc=0.735254
+INFO:root:Epoch[11] Batch [2560]	Speed: 662.59 samples/sec	acc=0.725977
+INFO:root:Epoch[11] Batch [2580]	Speed: 661.05 samples/sec	acc=0.731543
+INFO:root:Epoch[11] Batch [2600]	Speed: 652.62 samples/sec	acc=0.724219
+INFO:root:Epoch[11] Batch [2620]	Speed: 663.81 samples/sec	acc=0.733398
+INFO:root:Epoch[11] Batch [2640]	Speed: 656.05 samples/sec	acc=0.728809
+INFO:root:Epoch[11] Batch [2660]	Speed: 662.27 samples/sec	acc=0.734668
+INFO:root:Epoch[11] Batch [2680]	Speed: 661.71 samples/sec	acc=0.734082
+INFO:root:Epoch[11] Batch [2700]	Speed: 661.47 samples/sec	acc=0.727441
+INFO:root:Epoch[11] Batch [2720]	Speed: 661.72 samples/sec	acc=0.730176
+INFO:root:Epoch[11] Batch [2740]	Speed: 662.38 samples/sec	acc=0.732129
+INFO:root:Epoch[11] Batch [2760]	Speed: 661.35 samples/sec	acc=0.729492
+INFO:root:Epoch[11] Batch [2780]	Speed: 662.21 samples/sec	acc=0.733008
+INFO:root:Epoch[11] Batch [2800]	Speed: 662.06 samples/sec	acc=0.735547
+INFO:root:Epoch[11] Batch [2820]	Speed: 661.69 samples/sec	acc=0.740527
+INFO:root:Epoch[11] Batch [2840]	Speed: 651.64 samples/sec	acc=0.733789
+INFO:root:Epoch[11] Batch [2860]	Speed: 662.39 samples/sec	acc=0.741113
+INFO:root:Epoch[11] Batch [2880]	Speed: 661.72 samples/sec	acc=0.737207
+INFO:root:Epoch[11] Batch [2900]	Speed: 661.59 samples/sec	acc=0.733887
+INFO:root:Epoch[11] Batch [2920]	Speed: 661.91 samples/sec	acc=0.731445
+INFO:root:Epoch[11] Batch [2940]	Speed: 661.61 samples/sec	acc=0.733105
+INFO:root:Epoch[11] Batch [2960]	Speed: 636.69 samples/sec	acc=0.734961
+INFO:root:Epoch[11] Batch [2980]	Speed: 662.45 samples/sec	acc=0.726758
+INFO:root:Epoch[11] Batch [3000]	Speed: 661.14 samples/sec	acc=0.734766
+INFO:root:Epoch[11] Batch [3020]	Speed: 660.52 samples/sec	acc=0.733691
+INFO:root:Epoch[11] Batch [3040]	Speed: 640.09 samples/sec	acc=0.734766
+INFO:root:Epoch[11] Batch [3060]	Speed: 661.77 samples/sec	acc=0.734668
+INFO:root:Epoch[11] Batch [3080]	Speed: 637.01 samples/sec	acc=0.723047
+INFO:root:Epoch[11] Batch [3100]	Speed: 653.97 samples/sec	acc=0.735352
+INFO:root:Epoch[11] Batch [3120]	Speed: 661.59 samples/sec	acc=0.733789
+INFO:root:Epoch[11] Batch [3140]	Speed: 654.05 samples/sec	acc=0.736523
+INFO:root:Epoch[11] Batch [3160]	Speed: 669.71 samples/sec	acc=0.734082
+INFO:root:Epoch[11] Batch [3180]	Speed: 660.87 samples/sec	acc=0.731836
+INFO:root:Epoch[11] Batch [3200]	Speed: 654.29 samples/sec	acc=0.734570
+INFO:root:Epoch[11] Batch [3220]	Speed: 661.74 samples/sec	acc=0.721875
+INFO:root:Epoch[11] Batch [3240]	Speed: 639.25 samples/sec	acc=0.730469
+INFO:root:Epoch[11] Batch [3260]	Speed: 637.49 samples/sec	acc=0.728320
+INFO:root:Epoch[11] Batch [3280]	Speed: 661.63 samples/sec	acc=0.730176
+[86000][MARGIN]0.789603,0.401927
+lr-batch-epoch: 0.001 3290 11
+testing verification..
+(14000, 512)
+infer time 24.048905
+[cfp_ff][86000]XNorm: 21.485136
+[cfp_ff][86000]Accuracy-Flip: 0.99771+-0.00232
+testing verification..
+(14000, 512)
+infer time 26.887102
+[cfp_fp][86000]XNorm: 19.288736
+[cfp_fp][86000]Accuracy-Flip: 0.91943+-0.01477
+testing verification..
+(12000, 512)
+infer time 23.028544
+[agedb_30][86000]XNorm: 22.311972
+[agedb_30][86000]Accuracy-Flip: 0.97717+-0.00742
+testing verification..
+(12000, 512)
+infer time 19.656507
+[lfw][86000]XNorm: 22.103961
+[lfw][86000]Accuracy-Flip: 0.99750+-0.00300
+[86000]Accuracy-Highest: 0.99767
+INFO:root:Epoch[11] Batch [3300]	Speed: 79.17 samples/sec	acc=0.726953
+INFO:root:Epoch[11] Batch [3320]	Speed: 669.48 samples/sec	acc=0.734570
+INFO:root:Epoch[11] Batch [3340]	Speed: 659.53 samples/sec	acc=0.729883
+INFO:root:Epoch[11] Batch [3360]	Speed: 672.55 samples/sec	acc=0.726270
+INFO:root:Epoch[11] Batch [3380]	Speed: 661.74 samples/sec	acc=0.733887
+INFO:root:Epoch[11] Batch [3400]	Speed: 659.75 samples/sec	acc=0.732129
+INFO:root:Epoch[11] Batch [3420]	Speed: 658.60 samples/sec	acc=0.730762
+INFO:root:Epoch[11] Batch [3440]	Speed: 659.50 samples/sec	acc=0.732910
+INFO:root:Epoch[11] Batch [3460]	Speed: 663.08 samples/sec	acc=0.736133
+INFO:root:Epoch[11] Batch [3480]	Speed: 661.39 samples/sec	acc=0.739551
+INFO:root:Epoch[11] Batch [3500]	Speed: 657.22 samples/sec	acc=0.735254
+INFO:root:Epoch[11] Batch [3520]	Speed: 662.30 samples/sec	acc=0.731836
+INFO:root:Epoch[11] Batch [3540]	Speed: 645.55 samples/sec	acc=0.734180
+INFO:root:Epoch[11] Batch [3560]	Speed: 661.63 samples/sec	acc=0.733984
+INFO:root:Epoch[11] Batch [3580]	Speed: 662.72 samples/sec	acc=0.734473
+INFO:root:Epoch[11] Batch [3600]	Speed: 661.90 samples/sec	acc=0.736426
+INFO:root:Epoch[11] Batch [3620]	Speed: 661.98 samples/sec	acc=0.732129
+INFO:root:Epoch[11] Batch [3640]	Speed: 662.60 samples/sec	acc=0.741797
+INFO:root:Epoch[11] Batch [3660]	Speed: 647.91 samples/sec	acc=0.735742
+INFO:root:Epoch[11] Batch [3680]	Speed: 651.55 samples/sec	acc=0.734375
+INFO:root:Epoch[11] Batch [3700]	Speed: 662.54 samples/sec	acc=0.739551
+INFO:root:Epoch[11] Batch [3720]	Speed: 661.66 samples/sec	acc=0.737109
+INFO:root:Epoch[11] Batch [3740]	Speed: 662.01 samples/sec	acc=0.731836
+INFO:root:Epoch[11] Batch [3760]	Speed: 661.91 samples/sec	acc=0.737598
+INFO:root:Epoch[11] Batch [3780]	Speed: 661.57 samples/sec	acc=0.739941
+INFO:root:Epoch[11] Batch [3800]	Speed: 662.41 samples/sec	acc=0.735547
+INFO:root:Epoch[11] Batch [3820]	Speed: 662.43 samples/sec	acc=0.731738
+INFO:root:Epoch[11] Batch [3840]	Speed: 642.35 samples/sec	acc=0.731543
+INFO:root:Epoch[11] Batch [3860]	Speed: 661.69 samples/sec	acc=0.731152
+INFO:root:Epoch[11] Batch [3880]	Speed: 654.56 samples/sec	acc=0.737012
+INFO:root:Epoch[11] Batch [3900]	Speed: 662.29 samples/sec	acc=0.731836
+INFO:root:Epoch[11] Batch [3920]	Speed: 662.00 samples/sec	acc=0.737988
+INFO:root:Epoch[11] Batch [3940]	Speed: 662.76 samples/sec	acc=0.735156
+INFO:root:Epoch[11] Batch [3960]	Speed: 661.84 samples/sec	acc=0.737109
+INFO:root:Epoch[11] Batch [3980]	Speed: 661.54 samples/sec	acc=0.737500
+INFO:root:Epoch[11] Batch [4000]	Speed: 656.51 samples/sec	acc=0.741895
+INFO:root:Epoch[11] Batch [4020]	Speed: 667.01 samples/sec	acc=0.727344
+INFO:root:Epoch[11] Batch [4040]	Speed: 645.09 samples/sec	acc=0.734863
+INFO:root:Epoch[11] Batch [4060]	Speed: 663.69 samples/sec	acc=0.735645
+INFO:root:Epoch[11] Batch [4080]	Speed: 659.10 samples/sec	acc=0.741211
+INFO:root:Epoch[11] Batch [4100]	Speed: 662.61 samples/sec	acc=0.735254
+INFO:root:Epoch[11] Batch [4120]	Speed: 659.04 samples/sec	acc=0.743555
+INFO:root:Epoch[11] Batch [4140]	Speed: 661.71 samples/sec	acc=0.733594
+INFO:root:Epoch[11] Batch [4160]	Speed: 645.73 samples/sec	acc=0.730762
+INFO:root:Epoch[11] Batch [4180]	Speed: 656.21 samples/sec	acc=0.733008
+INFO:root:Epoch[11] Batch [4200]	Speed: 662.37 samples/sec	acc=0.741602
+INFO:root:Epoch[11] Batch [4220]	Speed: 662.17 samples/sec	acc=0.733984
+INFO:root:Epoch[11] Batch [4240]	Speed: 662.48 samples/sec	acc=0.734766
+INFO:root:Epoch[11] Batch [4260]	Speed: 655.27 samples/sec	acc=0.735156
+INFO:root:Epoch[11] Batch [4280]	Speed: 666.01 samples/sec	acc=0.740820
+lr-batch-epoch: 0.001 4290 11
+INFO:root:Epoch[11] Batch [4300]	Speed: 663.31 samples/sec	acc=0.735938
+INFO:root:Epoch[11] Batch [4320]	Speed: 661.94 samples/sec	acc=0.739746
+INFO:root:Epoch[11] Batch [4340]	Speed: 660.34 samples/sec	acc=0.729590
+INFO:root:Epoch[11] Batch [4360]	Speed: 663.02 samples/sec	acc=0.733984
+INFO:root:Epoch[11] Batch [4380]	Speed: 662.01 samples/sec	acc=0.743066
+INFO:root:Epoch[11] Batch [4400]	Speed: 639.29 samples/sec	acc=0.732129
+INFO:root:Epoch[11] Batch [4420]	Speed: 663.29 samples/sec	acc=0.739355
+INFO:root:Epoch[11] Batch [4440]	Speed: 661.22 samples/sec	acc=0.740625
+INFO:root:Epoch[11] Batch [4460]	Speed: 658.87 samples/sec	acc=0.744531
+INFO:root:Epoch[11] Batch [4480]	Speed: 656.89 samples/sec	acc=0.737305
+INFO:root:Epoch[11] Batch [4500]	Speed: 638.92 samples/sec	acc=0.736621
+INFO:root:Epoch[11] Batch [4520]	Speed: 664.03 samples/sec	acc=0.736914
+INFO:root:Epoch[11] Batch [4540]	Speed: 662.87 samples/sec	acc=0.736719
+INFO:root:Epoch[11] Batch [4560]	Speed: 661.89 samples/sec	acc=0.746680
+INFO:root:Epoch[11] Batch [4580]	Speed: 658.52 samples/sec	acc=0.738184
+INFO:root:Epoch[11] Batch [4600]	Speed: 638.56 samples/sec	acc=0.735156
+INFO:root:Epoch[11] Batch [4620]	Speed: 661.60 samples/sec	acc=0.734180
+INFO:root:Epoch[11] Batch [4640]	Speed: 661.08 samples/sec	acc=0.741504
+INFO:root:Epoch[11] Batch [4660]	Speed: 661.79 samples/sec	acc=0.741406
+INFO:root:Epoch[11] Batch [4680]	Speed: 662.22 samples/sec	acc=0.737891
+INFO:root:Epoch[11] Batch [4700]	Speed: 634.13 samples/sec	acc=0.737402
+INFO:root:Epoch[11] Batch [4720]	Speed: 662.49 samples/sec	acc=0.740723
+INFO:root:Epoch[11] Batch [4740]	Speed: 661.47 samples/sec	acc=0.739355
+INFO:root:Epoch[11] Batch [4760]	Speed: 636.50 samples/sec	acc=0.739844
+INFO:root:Epoch[11] Batch [4780]	Speed: 664.18 samples/sec	acc=0.734668
+INFO:root:Epoch[11] Batch [4800]	Speed: 659.44 samples/sec	acc=0.743164
+INFO:root:Epoch[11] Batch [4820]	Speed: 677.97 samples/sec	acc=0.743359
+INFO:root:Epoch[11] Batch [4840]	Speed: 662.85 samples/sec	acc=0.737207
+INFO:root:Epoch[11] Batch [4860]	Speed: 642.66 samples/sec	acc=0.740039
+INFO:root:Epoch[11] Batch [4880]	Speed: 661.34 samples/sec	acc=0.744629
+INFO:root:Epoch[11] Batch [4900]	Speed: 654.49 samples/sec	acc=0.738379
+INFO:root:Epoch[11] Batch [4920]	Speed: 626.99 samples/sec	acc=0.729395
+INFO:root:Epoch[11] Batch [4940]	Speed: 670.26 samples/sec	acc=0.741113
+INFO:root:Epoch[11] Batch [4960]	Speed: 664.32 samples/sec	acc=0.742773
+INFO:root:Epoch[11] Batch [4980]	Speed: 650.18 samples/sec	acc=0.749805
+INFO:root:Epoch[11] Batch [5000]	Speed: 662.01 samples/sec	acc=0.739355
+INFO:root:Epoch[11] Batch [5020]	Speed: 649.19 samples/sec	acc=0.739941
+INFO:root:Epoch[11] Batch [5040]	Speed: 660.84 samples/sec	acc=0.745410
+INFO:root:Epoch[11] Batch [5060]	Speed: 662.38 samples/sec	acc=0.745801
+INFO:root:Epoch[11] Batch [5080]	Speed: 662.39 samples/sec	acc=0.743359
+INFO:root:Epoch[11] Batch [5100]	Speed: 661.41 samples/sec	acc=0.736719
+INFO:root:Epoch[11] Batch [5120]	Speed: 657.94 samples/sec	acc=0.739258
+INFO:root:Epoch[11] Batch [5140]	Speed: 662.43 samples/sec	acc=0.737793
+INFO:root:Epoch[11] Batch [5160]	Speed: 661.45 samples/sec	acc=0.738672
+INFO:root:Epoch[11] Batch [5180]	Speed: 661.66 samples/sec	acc=0.746680
+INFO:root:Epoch[11] Batch [5200]	Speed: 659.85 samples/sec	acc=0.741113
+INFO:root:Epoch[11] Batch [5220]	Speed: 659.06 samples/sec	acc=0.738086
+INFO:root:Epoch[11] Batch [5240]	Speed: 648.75 samples/sec	acc=0.737012
+INFO:root:Epoch[11] Batch [5260]	Speed: 662.55 samples/sec	acc=0.739355
+INFO:root:Epoch[11] Batch [5280]	Speed: 661.31 samples/sec	acc=0.738184
+[88000][MARGIN]0.800128,0.417858
+lr-batch-epoch: 0.001 5290 11
+testing verification..
+(14000, 512)
+infer time 21.09969
+[cfp_ff][88000]XNorm: 21.877948
+[cfp_ff][88000]Accuracy-Flip: 0.99814+-0.00231
+testing verification..
+(14000, 512)
+infer time 24.106134
+[cfp_fp][88000]XNorm: 19.718805
+[cfp_fp][88000]Accuracy-Flip: 0.92357+-0.01305
+testing verification..
+(12000, 512)
+infer time 20.448449
+[agedb_30][88000]XNorm: 22.722104
+[agedb_30][88000]Accuracy-Flip: 0.97783+-0.00650
+testing verification..
+(12000, 512)
+infer time 18.662641
+[lfw][88000]XNorm: 22.350139
+[lfw][88000]Accuracy-Flip: 0.99783+-0.00279
+saving 44
+INFO:root:Saved checkpoint to "../model-r50-am-lfw/model-0044.params"
+[88000]Accuracy-Highest: 0.99783
+INFO:root:Epoch[11] Batch [5300]	Speed: 82.98 samples/sec	acc=0.739551
+INFO:root:Epoch[11] Batch [5320]	Speed: 641.87 samples/sec	acc=0.743652
+INFO:root:Epoch[11] Batch [5340]	Speed: 667.66 samples/sec	acc=0.741211
+INFO:root:Epoch[11] Batch [5360]	Speed: 665.03 samples/sec	acc=0.742383
+INFO:root:Epoch[11] Batch [5380]	Speed: 660.35 samples/sec	acc=0.735156
+INFO:root:Epoch[11] Batch [5400]	Speed: 661.58 samples/sec	acc=0.739160
+INFO:root:Epoch[11] Batch [5420]	Speed: 651.04 samples/sec	acc=0.743945
+INFO:root:Epoch[11] Batch [5440]	Speed: 663.00 samples/sec	acc=0.742188
+INFO:root:Epoch[11] Batch [5460]	Speed: 654.73 samples/sec	acc=0.739355
+INFO:root:Epoch[11] Batch [5480]	Speed: 652.51 samples/sec	acc=0.742773
+INFO:root:Epoch[11] Batch [5500]	Speed: 656.31 samples/sec	acc=0.739453
+INFO:root:Epoch[11] Batch [5520]	Speed: 661.66 samples/sec	acc=0.735156
+INFO:root:Epoch[11] Batch [5540]	Speed: 661.88 samples/sec	acc=0.740625
+INFO:root:Epoch[11] Batch [5560]	Speed: 662.69 samples/sec	acc=0.745117
+INFO:root:Epoch[11] Batch [5580]	Speed: 629.39 samples/sec	acc=0.743652
+INFO:root:Epoch[11] Batch [5600]	Speed: 666.49 samples/sec	acc=0.743262
+INFO:root:Epoch[11] Batch [5620]	Speed: 663.09 samples/sec	acc=0.738379
+INFO:root:Epoch[11] Batch [5640]	Speed: 661.37 samples/sec	acc=0.741309
+INFO:root:Epoch[11] Batch [5660]	Speed: 661.69 samples/sec	acc=0.734961
+INFO:root:Epoch[11] Batch [5680]	Speed: 662.36 samples/sec	acc=0.736719
+INFO:root:Epoch[11] Batch [5700]	Speed: 661.70 samples/sec	acc=0.743262
+INFO:root:Epoch[11] Batch [5720]	Speed: 652.33 samples/sec	acc=0.731250
+INFO:root:Epoch[11] Batch [5740]	Speed: 661.65 samples/sec	acc=0.738477
+INFO:root:Epoch[11] Batch [5760]	Speed: 661.72 samples/sec	acc=0.737695
+INFO:root:Epoch[11] Batch [5780]	Speed: 662.20 samples/sec	acc=0.743848
+INFO:root:Epoch[11] Batch [5800]	Speed: 662.74 samples/sec	acc=0.740137
+INFO:root:Epoch[11] Batch [5820]	Speed: 656.33 samples/sec	acc=0.735645
+INFO:root:Epoch[11] Batch [5840]	Speed: 644.27 samples/sec	acc=0.742773
+INFO:root:Epoch[11] Batch [5860]	Speed: 660.99 samples/sec	acc=0.740625
+INFO:root:Epoch[11] Batch [5880]	Speed: 654.13 samples/sec	acc=0.741992
+INFO:root:Epoch[11] Batch [5900]	Speed: 654.64 samples/sec	acc=0.743652
+INFO:root:Epoch[11] Batch [5920]	Speed: 657.07 samples/sec	acc=0.748340
+INFO:root:Epoch[11] Batch [5940]	Speed: 661.80 samples/sec	acc=0.740039
+INFO:root:Epoch[11] Batch [5960]	Speed: 661.93 samples/sec	acc=0.745898
+INFO:root:Epoch[11] Batch [5980]	Speed: 662.66 samples/sec	acc=0.744824
+INFO:root:Epoch[11] Batch [6000]	Speed: 669.99 samples/sec	acc=0.747656
+INFO:root:Epoch[11] Batch [6020]	Speed: 661.92 samples/sec	acc=0.742285
+INFO:root:Epoch[11] Batch [6040]	Speed: 662.28 samples/sec	acc=0.746777
+INFO:root:Epoch[11] Batch [6060]	Speed: 662.11 samples/sec	acc=0.744238
+INFO:root:Epoch[11] Batch [6080]	Speed: 662.52 samples/sec	acc=0.738574
+INFO:root:Epoch[11] Batch [6100]	Speed: 640.49 samples/sec	acc=0.742773
+INFO:root:Epoch[11] Batch [6120]	Speed: 666.58 samples/sec	acc=0.744922
+INFO:root:Epoch[11] Batch [6140]	Speed: 665.33 samples/sec	acc=0.743945
+INFO:root:Epoch[11] Batch [6160]	Speed: 662.45 samples/sec	acc=0.740332
+INFO:root:Epoch[11] Batch [6180]	Speed: 661.67 samples/sec	acc=0.745605
+INFO:root:Epoch[11] Batch [6200]	Speed: 654.57 samples/sec	acc=0.739160
+INFO:root:Epoch[11] Batch [6220]	Speed: 662.67 samples/sec	acc=0.742090
+INFO:root:Epoch[11] Batch [6240]	Speed: 661.51 samples/sec	acc=0.747168
+INFO:root:Epoch[11] Batch [6260]	Speed: 662.50 samples/sec	acc=0.745215
+INFO:root:Epoch[11] Batch [6280]	Speed: 653.99 samples/sec	acc=0.749219
+lr-batch-epoch: 0.001 6290 11
+INFO:root:Epoch[11] Batch [6300]	Speed: 660.84 samples/sec	acc=0.737695
+INFO:root:Epoch[11] Batch [6320]	Speed: 662.02 samples/sec	acc=0.743555
+INFO:root:Epoch[11] Batch [6340]	Speed: 661.27 samples/sec	acc=0.744434
+INFO:root:Epoch[11] Batch [6360]	Speed: 661.85 samples/sec	acc=0.747852
+INFO:root:Epoch[11] Batch [6380]	Speed: 661.96 samples/sec	acc=0.754297
+INFO:root:Epoch[11] Batch [6400]	Speed: 668.44 samples/sec	acc=0.747656
+INFO:root:Epoch[11] Batch [6420]	Speed: 661.69 samples/sec	acc=0.743652
+INFO:root:Epoch[11] Batch [6440]	Speed: 662.23 samples/sec	acc=0.748730
+INFO:root:Epoch[11] Batch [6460]	Speed: 647.46 samples/sec	acc=0.744727
+INFO:root:Epoch[11] Batch [6480]	Speed: 656.54 samples/sec	acc=0.742871
+INFO:root:Epoch[11] Batch [6500]	Speed: 668.98 samples/sec	acc=0.744434
+INFO:root:Epoch[11] Batch [6520]	Speed: 643.74 samples/sec	acc=0.745703
+INFO:root:Epoch[11] Batch [6540]	Speed: 651.54 samples/sec	acc=0.740234
+INFO:root:Epoch[11] Batch [6560]	Speed: 662.72 samples/sec	acc=0.745703
+INFO:root:Epoch[11] Batch [6580]	Speed: 639.35 samples/sec	acc=0.748145
+INFO:root:Epoch[11] Batch [6600]	Speed: 652.21 samples/sec	acc=0.751270
+INFO:root:Epoch[11] Batch [6620]	Speed: 663.17 samples/sec	acc=0.747949
+INFO:root:Epoch[11] Batch [6640]	Speed: 650.31 samples/sec	acc=0.746973
+INFO:root:Epoch[11] Batch [6660]	Speed: 661.88 samples/sec	acc=0.743164
+INFO:root:Epoch[11] Batch [6680]	Speed: 662.26 samples/sec	acc=0.740625
+INFO:root:Epoch[11] Batch [6700]	Speed: 662.83 samples/sec	acc=0.739941
+INFO:root:Epoch[11] Batch [6720]	Speed: 661.80 samples/sec	acc=0.748828
+INFO:root:Epoch[11] Batch [6740]	Speed: 661.98 samples/sec	acc=0.746191
+INFO:root:Epoch[11] Batch [6760]	Speed: 662.37 samples/sec	acc=0.747168
+INFO:root:Epoch[11] Batch [6780]	Speed: 661.21 samples/sec	acc=0.747754
+INFO:root:Epoch[11] Batch [6800]	Speed: 659.13 samples/sec	acc=0.740918
+INFO:root:Epoch[11] Batch [6820]	Speed: 662.92 samples/sec	acc=0.744434
+INFO:root:Epoch[11] Batch [6840]	Speed: 661.45 samples/sec	acc=0.746289
+INFO:root:Epoch[11] Batch [6860]	Speed: 661.22 samples/sec	acc=0.745313
+INFO:root:Epoch[11] Batch [6880]	Speed: 659.62 samples/sec	acc=0.742188
+INFO:root:Epoch[11] Batch [6900]	Speed: 638.48 samples/sec	acc=0.749512
+INFO:root:Epoch[11] Batch [6920]	Speed: 662.37 samples/sec	acc=0.751367
+INFO:root:Epoch[11] Batch [6940]	Speed: 662.12 samples/sec	acc=0.745801
+INFO:root:Epoch[11] Batch [6960]	Speed: 661.83 samples/sec	acc=0.749316
+INFO:root:Epoch[11] Batch [6980]	Speed: 661.66 samples/sec	acc=0.746777
+INFO:root:Epoch[11] Batch [7000]	Speed: 662.44 samples/sec	acc=0.752051
+INFO:root:Epoch[11] Batch [7020]	Speed: 628.74 samples/sec	acc=0.749805
+INFO:root:Epoch[11] Batch [7040]	Speed: 645.77 samples/sec	acc=0.741797
+INFO:root:Epoch[11] Batch [7060]	Speed: 662.21 samples/sec	acc=0.742480
+INFO:root:Epoch[11] Batch [7080]	Speed: 661.97 samples/sec	acc=0.746387
+INFO:root:Epoch[11] Batch [7100]	Speed: 661.92 samples/sec	acc=0.747461
+INFO:root:Epoch[11] Batch [7120]	Speed: 662.78 samples/sec	acc=0.746973
+INFO:root:Epoch[11] Batch [7140]	Speed: 661.17 samples/sec	acc=0.747363
+INFO:root:Epoch[11] Batch [7160]	Speed: 661.81 samples/sec	acc=0.746289
+INFO:root:Epoch[11] Batch [7180]	Speed: 663.23 samples/sec	acc=0.742578
+INFO:root:Epoch[11] Batch [7200]	Speed: 661.44 samples/sec	acc=0.749512
+INFO:root:Epoch[11] Batch [7220]	Speed: 662.07 samples/sec	acc=0.749609
+INFO:root:Epoch[11] Batch [7240]	Speed: 661.97 samples/sec	acc=0.740723
+INFO:root:Epoch[11] Batch [7260]	Speed: 662.48 samples/sec	acc=0.742480
+INFO:root:Epoch[11] Batch [7280]	Speed: 658.09 samples/sec	acc=0.742578
+[90000][MARGIN]0.785767,0.397132
+lr-batch-epoch: 0.001 7290 11
+testing verification..
+(14000, 512)
+infer time 23.314756
+[cfp_ff][90000]XNorm: 21.591677
+[cfp_ff][90000]Accuracy-Flip: 0.99757+-0.00222
+testing verification..
+(14000, 512)
+infer time 22.06703
+[cfp_fp][90000]XNorm: 19.494454
+[cfp_fp][90000]Accuracy-Flip: 0.92300+-0.01378
+testing verification..
+(12000, 512)
+infer time 19.160242
+[agedb_30][90000]XNorm: 22.471320
+[agedb_30][90000]Accuracy-Flip: 0.97750+-0.00712
+testing verification..
+(12000, 512)
+infer time 18.699288
+[lfw][90000]XNorm: 22.201853
+[lfw][90000]Accuracy-Flip: 0.99733+-0.00300
+[90000]Accuracy-Highest: 0.99783
+INFO:root:Epoch[11] Batch [7300]	Speed: 86.67 samples/sec	acc=0.744141
+INFO:root:Epoch[11] Batch [7320]	Speed: 661.82 samples/sec	acc=0.750977
+INFO:root:Epoch[11] Batch [7340]	Speed: 662.09 samples/sec	acc=0.750586
+INFO:root:Epoch[11] Batch [7360]	Speed: 662.30 samples/sec	acc=0.754004
+INFO:root:Epoch[11] Batch [7380]	Speed: 661.01 samples/sec	acc=0.746875
+INFO:root:Epoch[11] Batch [7400]	Speed: 661.81 samples/sec	acc=0.744727
+INFO:root:Epoch[11] Batch [7420]	Speed: 661.42 samples/sec	acc=0.746289
+INFO:root:Epoch[11] Batch [7440]	Speed: 656.75 samples/sec	acc=0.745020
+INFO:root:Epoch[11] Batch [7460]	Speed: 662.21 samples/sec	acc=0.747461
+INFO:root:Epoch[11] Batch [7480]	Speed: 662.00 samples/sec	acc=0.741992
+INFO:root:Epoch[11] Batch [7500]	Speed: 661.58 samples/sec	acc=0.747656
+INFO:root:Epoch[11] Train-acc=0.744900
+INFO:root:Epoch[11] Time cost=6281.392
+call reset()
+INFO:root:Epoch[12] Batch [20]	Speed: 656.28 samples/sec	acc=0.787481
+INFO:root:Epoch[12] Batch [40]	Speed: 652.75 samples/sec	acc=0.782422
+INFO:root:Epoch[12] Batch [60]	Speed: 662.63 samples/sec	acc=0.787012
+INFO:root:Epoch[12] Batch [80]	Speed: 661.64 samples/sec	acc=0.779492
+INFO:root:Epoch[12] Batch [100]	Speed: 662.41 samples/sec	acc=0.781055
+INFO:root:Epoch[12] Batch [120]	Speed: 662.89 samples/sec	acc=0.781250
+INFO:root:Epoch[12] Batch [140]	Speed: 661.10 samples/sec	acc=0.781250
+INFO:root:Epoch[12] Batch [160]	Speed: 661.91 samples/sec	acc=0.783008
+INFO:root:Epoch[12] Batch [180]	Speed: 662.65 samples/sec	acc=0.789844
+INFO:root:Epoch[12] Batch [200]	Speed: 661.56 samples/sec	acc=0.790820
+INFO:root:Epoch[12] Batch [220]	Speed: 662.81 samples/sec	acc=0.777734
+INFO:root:Epoch[12] Batch [240]	Speed: 662.03 samples/sec	acc=0.786328
+INFO:root:Epoch[12] Batch [260]	Speed: 656.98 samples/sec	acc=0.781641
+INFO:root:Epoch[12] Batch [280]	Speed: 659.39 samples/sec	acc=0.780859
+INFO:root:Epoch[12] Batch [300]	Speed: 658.15 samples/sec	acc=0.778516
+INFO:root:Epoch[12] Batch [320]	Speed: 661.79 samples/sec	acc=0.787109
+INFO:root:Epoch[12] Batch [340]	Speed: 645.82 samples/sec	acc=0.779395
+INFO:root:Epoch[12] Batch [360]	Speed: 662.76 samples/sec	acc=0.778320
+INFO:root:Epoch[12] Batch [380]	Speed: 663.40 samples/sec	acc=0.780176
+INFO:root:Epoch[12] Batch [400]	Speed: 662.21 samples/sec	acc=0.780762
+INFO:root:Epoch[12] Batch [420]	Speed: 663.27 samples/sec	acc=0.779883
+INFO:root:Epoch[12] Batch [440]	Speed: 661.48 samples/sec	acc=0.780859
+INFO:root:Epoch[12] Batch [460]	Speed: 661.48 samples/sec	acc=0.784668
+INFO:root:Epoch[12] Batch [480]	Speed: 661.49 samples/sec	acc=0.781738
+INFO:root:Epoch[12] Batch [500]	Speed: 662.38 samples/sec	acc=0.783301
+INFO:root:Epoch[12] Batch [520]	Speed: 661.89 samples/sec	acc=0.778418
+INFO:root:Epoch[12] Batch [540]	Speed: 662.09 samples/sec	acc=0.783301
+INFO:root:Epoch[12] Batch [560]	Speed: 661.86 samples/sec	acc=0.776660
+INFO:root:Epoch[12] Batch [580]	Speed: 661.52 samples/sec	acc=0.782617
+INFO:root:Epoch[12] Batch [600]	Speed: 662.45 samples/sec	acc=0.781641
+INFO:root:Epoch[12] Batch [620]	Speed: 660.82 samples/sec	acc=0.781445
+INFO:root:Epoch[12] Batch [640]	Speed: 661.85 samples/sec	acc=0.776758
+INFO:root:Epoch[12] Batch [660]	Speed: 662.73 samples/sec	acc=0.771777
+INFO:root:Epoch[12] Batch [680]	Speed: 661.07 samples/sec	acc=0.782031
+INFO:root:Epoch[12] Batch [700]	Speed: 662.62 samples/sec	acc=0.776855
+INFO:root:Epoch[12] Batch [720]	Speed: 645.14 samples/sec	acc=0.773438
+INFO:root:Epoch[12] Batch [740]	Speed: 662.01 samples/sec	acc=0.775391
+INFO:root:Epoch[12] Batch [760]	Speed: 648.75 samples/sec	acc=0.782129
+lr-batch-epoch: 0.001 771 12
+INFO:root:Epoch[12] Batch [780]	Speed: 659.18 samples/sec	acc=0.782813
+INFO:root:Epoch[12] Batch [800]	Speed: 645.84 samples/sec	acc=0.784473
+INFO:root:Epoch[12] Batch [820]	Speed: 655.17 samples/sec	acc=0.777148
+INFO:root:Epoch[12] Batch [840]	Speed: 663.15 samples/sec	acc=0.779590
+INFO:root:Epoch[12] Batch [860]	Speed: 650.39 samples/sec	acc=0.780176
+INFO:root:Epoch[12] Batch [880]	Speed: 630.79 samples/sec	acc=0.775781
+INFO:root:Epoch[12] Batch [900]	Speed: 644.85 samples/sec	acc=0.781250
+INFO:root:Epoch[12] Batch [920]	Speed: 660.13 samples/sec	acc=0.783008
+INFO:root:Epoch[12] Batch [940]	Speed: 662.39 samples/sec	acc=0.781836
+INFO:root:Epoch[12] Batch [960]	Speed: 647.10 samples/sec	acc=0.781641
+INFO:root:Epoch[12] Batch [980]	Speed: 661.82 samples/sec	acc=0.780566
+INFO:root:Epoch[12] Batch [1000]	Speed: 661.99 samples/sec	acc=0.773828
+INFO:root:Epoch[12] Batch [1020]	Speed: 652.59 samples/sec	acc=0.781250
+INFO:root:Epoch[12] Batch [1040]	Speed: 661.01 samples/sec	acc=0.776367
+INFO:root:Epoch[12] Batch [1060]	Speed: 644.73 samples/sec	acc=0.781543
+INFO:root:Epoch[12] Batch [1080]	Speed: 639.26 samples/sec	acc=0.779687
+INFO:root:Epoch[12] Batch [1100]	Speed: 661.39 samples/sec	acc=0.777148
+INFO:root:Epoch[12] Batch [1120]	Speed: 661.63 samples/sec	acc=0.780762
+INFO:root:Epoch[12] Batch [1140]	Speed: 662.30 samples/sec	acc=0.777051
+INFO:root:Epoch[12] Batch [1160]	Speed: 661.74 samples/sec	acc=0.777148
+INFO:root:Epoch[12] Batch [1180]	Speed: 647.47 samples/sec	acc=0.782617
+INFO:root:Epoch[12] Batch [1200]	Speed: 659.21 samples/sec	acc=0.775391
+INFO:root:Epoch[12] Batch [1220]	Speed: 661.16 samples/sec	acc=0.784863
+INFO:root:Epoch[12] Batch [1240]	Speed: 651.28 samples/sec	acc=0.770312
+INFO:root:Epoch[12] Batch [1260]	Speed: 632.52 samples/sec	acc=0.784473
+INFO:root:Epoch[12] Batch [1280]	Speed: 629.13 samples/sec	acc=0.779492
+INFO:root:Epoch[12] Batch [1300]	Speed: 661.82 samples/sec	acc=0.780078
+INFO:root:Epoch[12] Batch [1320]	Speed: 662.76 samples/sec	acc=0.777148
+INFO:root:Epoch[12] Batch [1340]	Speed: 652.58 samples/sec	acc=0.781152
+INFO:root:Epoch[12] Batch [1360]	Speed: 661.81 samples/sec	acc=0.781738
+INFO:root:Epoch[12] Batch [1380]	Speed: 662.53 samples/sec	acc=0.778125
+INFO:root:Epoch[12] Batch [1400]	Speed: 659.76 samples/sec	acc=0.783398
+INFO:root:Epoch[12] Batch [1420]	Speed: 660.74 samples/sec	acc=0.778516
+INFO:root:Epoch[12] Batch [1440]	Speed: 662.26 samples/sec	acc=0.775879
+INFO:root:Epoch[12] Batch [1460]	Speed: 662.09 samples/sec	acc=0.775293
+INFO:root:Epoch[12] Batch [1480]	Speed: 642.34 samples/sec	acc=0.779297
+INFO:root:Epoch[12] Batch [1500]	Speed: 659.46 samples/sec	acc=0.780859
+INFO:root:Epoch[12] Batch [1520]	Speed: 661.78 samples/sec	acc=0.778809
+INFO:root:Epoch[12] Batch [1540]	Speed: 646.70 samples/sec	acc=0.779102
+INFO:root:Epoch[12] Batch [1560]	Speed: 662.90 samples/sec	acc=0.780371
+INFO:root:Epoch[12] Batch [1580]	Speed: 655.27 samples/sec	acc=0.776074
+INFO:root:Epoch[12] Batch [1600]	Speed: 661.74 samples/sec	acc=0.780762
+INFO:root:Epoch[12] Batch [1620]	Speed: 650.86 samples/sec	acc=0.779883
+INFO:root:Epoch[12] Batch [1640]	Speed: 660.06 samples/sec	acc=0.786719
+INFO:root:Epoch[12] Batch [1660]	Speed: 659.83 samples/sec	acc=0.771875
+INFO:root:Epoch[12] Batch [1680]	Speed: 661.99 samples/sec	acc=0.784668
+INFO:root:Epoch[12] Batch [1700]	Speed: 660.39 samples/sec	acc=0.779492
+INFO:root:Epoch[12] Batch [1720]	Speed: 662.18 samples/sec	acc=0.780859
+INFO:root:Epoch[12] Batch [1740]	Speed: 663.02 samples/sec	acc=0.778125
+INFO:root:Epoch[12] Batch [1760]	Speed: 661.68 samples/sec	acc=0.775684
+[92000][MARGIN]0.788903,0.402588
+lr-batch-epoch: 0.001 1771 12
+testing verification..
+(14000, 512)
+infer time 24.775264
+[cfp_ff][92000]XNorm: 21.401436
+[cfp_ff][92000]Accuracy-Flip: 0.99786+-0.00214
+testing verification..
+(14000, 512)
+infer time 20.701599
+[cfp_fp][92000]XNorm: 19.430333
+[cfp_fp][92000]Accuracy-Flip: 0.92543+-0.01364
+testing verification..
+(12000, 512)
+infer time 18.498203
+[agedb_30][92000]XNorm: 22.487997
+[agedb_30][92000]Accuracy-Flip: 0.97717+-0.00715
+testing verification..
+(12000, 512)
+infer time 20.977791
+[lfw][92000]XNorm: 22.113033
+[lfw][92000]Accuracy-Flip: 0.99717+-0.00289
+[92000]Accuracy-Highest: 0.99783
+INFO:root:Epoch[12] Batch [1780]	Speed: 85.09 samples/sec	acc=0.780762
+INFO:root:Epoch[12] Batch [1800]	Speed: 662.94 samples/sec	acc=0.773242
+INFO:root:Epoch[12] Batch [1820]	Speed: 661.95 samples/sec	acc=0.779102
+INFO:root:Epoch[12] Batch [1840]	Speed: 662.45 samples/sec	acc=0.780176
+INFO:root:Epoch[12] Batch [1860]	Speed: 663.29 samples/sec	acc=0.772949
+INFO:root:Epoch[12] Batch [1880]	Speed: 661.86 samples/sec	acc=0.778223
+INFO:root:Epoch[12] Batch [1900]	Speed: 663.03 samples/sec	acc=0.783789
+INFO:root:Epoch[12] Batch [1920]	Speed: 662.86 samples/sec	acc=0.783887
+INFO:root:Epoch[12] Batch [1940]	Speed: 659.11 samples/sec	acc=0.779785
+INFO:root:Epoch[12] Batch [1960]	Speed: 674.93 samples/sec	acc=0.769238
+INFO:root:Epoch[12] Batch [1980]	Speed: 662.86 samples/sec	acc=0.772168
+INFO:root:Epoch[12] Batch [2000]	Speed: 662.10 samples/sec	acc=0.775879
+INFO:root:Epoch[12] Batch [2020]	Speed: 659.88 samples/sec	acc=0.778516
+INFO:root:Epoch[12] Batch [2040]	Speed: 621.67 samples/sec	acc=0.778125
+INFO:root:Epoch[12] Batch [2060]	Speed: 659.80 samples/sec	acc=0.784180
+INFO:root:Epoch[12] Batch [2080]	Speed: 654.13 samples/sec	acc=0.769434
+INFO:root:Epoch[12] Batch [2100]	Speed: 662.70 samples/sec	acc=0.777051
+INFO:root:Epoch[12] Batch [2120]	Speed: 662.13 samples/sec	acc=0.779883
+INFO:root:Epoch[12] Batch [2140]	Speed: 662.66 samples/sec	acc=0.787012
+INFO:root:Epoch[12] Batch [2160]	Speed: 662.65 samples/sec	acc=0.777930
+INFO:root:Epoch[12] Batch [2180]	Speed: 652.25 samples/sec	acc=0.784473
+INFO:root:Epoch[12] Batch [2200]	Speed: 662.19 samples/sec	acc=0.769531
+INFO:root:Epoch[12] Batch [2220]	Speed: 662.74 samples/sec	acc=0.779492
+INFO:root:Epoch[12] Batch [2240]	Speed: 660.35 samples/sec	acc=0.768164
+INFO:root:Epoch[12] Batch [2260]	Speed: 662.89 samples/sec	acc=0.772852
+INFO:root:Epoch[12] Batch [2280]	Speed: 672.79 samples/sec	acc=0.776758
+INFO:root:Epoch[12] Batch [2300]	Speed: 644.20 samples/sec	acc=0.783105
+INFO:root:Epoch[12] Batch [2320]	Speed: 666.39 samples/sec	acc=0.774707
+INFO:root:Epoch[12] Batch [2340]	Speed: 656.59 samples/sec	acc=0.781543
+INFO:root:Epoch[12] Batch [2360]	Speed: 661.80 samples/sec	acc=0.777539
+INFO:root:Epoch[12] Batch [2380]	Speed: 661.74 samples/sec	acc=0.783203
+INFO:root:Epoch[12] Batch [2400]	Speed: 661.78 samples/sec	acc=0.782520
+INFO:root:Epoch[12] Batch [2420]	Speed: 661.93 samples/sec	acc=0.779395
+INFO:root:Epoch[12] Batch [2440]	Speed: 662.61 samples/sec	acc=0.780176
+INFO:root:Epoch[12] Batch [2460]	Speed: 662.04 samples/sec	acc=0.785645
+INFO:root:Epoch[12] Batch [2480]	Speed: 646.54 samples/sec	acc=0.779004
+INFO:root:Epoch[12] Batch [2500]	Speed: 673.72 samples/sec	acc=0.781445
+INFO:root:Epoch[12] Batch [2520]	Speed: 657.87 samples/sec	acc=0.782324
+INFO:root:Epoch[12] Batch [2540]	Speed: 662.78 samples/sec	acc=0.779980
+INFO:root:Epoch[12] Batch [2560]	Speed: 650.40 samples/sec	acc=0.779883
+INFO:root:Epoch[12] Batch [2580]	Speed: 659.89 samples/sec	acc=0.775293
+INFO:root:Epoch[12] Batch [2600]	Speed: 653.85 samples/sec	acc=0.776270
+INFO:root:Epoch[12] Batch [2620]	Speed: 662.54 samples/sec	acc=0.774316
+INFO:root:Epoch[12] Batch [2640]	Speed: 612.76 samples/sec	acc=0.778711
+INFO:root:Epoch[12] Batch [2660]	Speed: 661.81 samples/sec	acc=0.777539
+INFO:root:Epoch[12] Batch [2680]	Speed: 661.96 samples/sec	acc=0.770117
+INFO:root:Epoch[12] Batch [2700]	Speed: 662.30 samples/sec	acc=0.778125
+INFO:root:Epoch[12] Batch [2720]	Speed: 661.04 samples/sec	acc=0.771387
+INFO:root:Epoch[12] Batch [2740]	Speed: 662.37 samples/sec	acc=0.778320
+INFO:root:Epoch[12] Batch [2760]	Speed: 666.35 samples/sec	acc=0.773633
+lr-batch-epoch: 0.001 2771 12
+INFO:root:Epoch[12] Batch [2780]	Speed: 643.76 samples/sec	acc=0.784180
+INFO:root:Epoch[12] Batch [2800]	Speed: 661.62 samples/sec	acc=0.770605
+INFO:root:Epoch[12] Batch [2820]	Speed: 661.83 samples/sec	acc=0.778418
+INFO:root:Epoch[12] Batch [2840]	Speed: 661.17 samples/sec	acc=0.779980
+INFO:root:Epoch[12] Batch [2860]	Speed: 652.31 samples/sec	acc=0.777930
+INFO:root:Epoch[12] Batch [2880]	Speed: 662.03 samples/sec	acc=0.777832
+INFO:root:Epoch[12] Batch [2900]	Speed: 661.25 samples/sec	acc=0.774902
+INFO:root:Epoch[12] Batch [2920]	Speed: 640.87 samples/sec	acc=0.789160
+INFO:root:Epoch[12] Batch [2940]	Speed: 659.20 samples/sec	acc=0.771387
+INFO:root:Epoch[12] Batch [2960]	Speed: 660.91 samples/sec	acc=0.781836
+INFO:root:Epoch[12] Batch [2980]	Speed: 667.32 samples/sec	acc=0.784570
+INFO:root:Epoch[12] Batch [3000]	Speed: 636.51 samples/sec	acc=0.774121
+INFO:root:Epoch[12] Batch [3020]	Speed: 645.93 samples/sec	acc=0.772754
+INFO:root:Epoch[12] Batch [3040]	Speed: 647.84 samples/sec	acc=0.778418
+INFO:root:Epoch[12] Batch [3060]	Speed: 656.51 samples/sec	acc=0.783887
+INFO:root:Epoch[12] Batch [3080]	Speed: 664.42 samples/sec	acc=0.775293
+INFO:root:Epoch[12] Batch [3100]	Speed: 644.90 samples/sec	acc=0.780859
+INFO:root:Epoch[12] Batch [3120]	Speed: 662.45 samples/sec	acc=0.777148
+INFO:root:Epoch[12] Batch [3140]	Speed: 661.36 samples/sec	acc=0.772852
+INFO:root:Epoch[12] Batch [3160]	Speed: 651.11 samples/sec	acc=0.784863
+INFO:root:Epoch[12] Batch [3180]	Speed: 664.75 samples/sec	acc=0.777539
+INFO:root:Epoch[12] Batch [3200]	Speed: 661.79 samples/sec	acc=0.772949
+INFO:root:Epoch[12] Batch [3220]	Speed: 661.97 samples/sec	acc=0.781152
+INFO:root:Epoch[12] Batch [3240]	Speed: 662.29 samples/sec	acc=0.773438
+INFO:root:Epoch[12] Batch [3260]	Speed: 661.67 samples/sec	acc=0.773535
+INFO:root:Epoch[12] Batch [3280]	Speed: 661.95 samples/sec	acc=0.774316
+INFO:root:Epoch[12] Batch [3300]	Speed: 662.59 samples/sec	acc=0.777930
+INFO:root:Epoch[12] Batch [3320]	Speed: 652.38 samples/sec	acc=0.767773
+INFO:root:Epoch[12] Batch [3340]	Speed: 661.92 samples/sec	acc=0.771680
+INFO:root:Epoch[12] Batch [3360]	Speed: 662.55 samples/sec	acc=0.776172
+INFO:root:Epoch[12] Batch [3380]	Speed: 662.40 samples/sec	acc=0.772949
+INFO:root:Epoch[12] Batch [3400]	Speed: 646.09 samples/sec	acc=0.774512
+INFO:root:Epoch[12] Batch [3420]	Speed: 662.72 samples/sec	acc=0.779590
+INFO:root:Epoch[12] Batch [3440]	Speed: 661.37 samples/sec	acc=0.781152
+INFO:root:Epoch[12] Batch [3460]	Speed: 661.75 samples/sec	acc=0.774609
+INFO:root:Epoch[12] Batch [3480]	Speed: 640.27 samples/sec	acc=0.771191
+INFO:root:Epoch[12] Batch [3500]	Speed: 661.63 samples/sec	acc=0.770020
+INFO:root:Epoch[12] Batch [3520]	Speed: 655.62 samples/sec	acc=0.780566
+INFO:root:Epoch[12] Batch [3540]	Speed: 677.17 samples/sec	acc=0.768750
+INFO:root:Epoch[12] Batch [3560]	Speed: 669.49 samples/sec	acc=0.776465
+INFO:root:Epoch[12] Batch [3580]	Speed: 666.62 samples/sec	acc=0.776367
+INFO:root:Epoch[12] Batch [3600]	Speed: 662.96 samples/sec	acc=0.777734
+INFO:root:Epoch[12] Batch [3620]	Speed: 661.11 samples/sec	acc=0.778125
+INFO:root:Epoch[12] Batch [3640]	Speed: 662.16 samples/sec	acc=0.779590
+INFO:root:Epoch[12] Batch [3660]	Speed: 662.54 samples/sec	acc=0.775781
+INFO:root:Epoch[12] Batch [3680]	Speed: 661.90 samples/sec	acc=0.773633
+INFO:root:Epoch[12] Batch [3700]	Speed: 661.63 samples/sec	acc=0.773535
+INFO:root:Epoch[12] Batch [3720]	Speed: 662.21 samples/sec	acc=0.780469
+INFO:root:Epoch[12] Batch [3740]	Speed: 652.69 samples/sec	acc=0.774414
+INFO:root:Epoch[12] Batch [3760]	Speed: 655.94 samples/sec	acc=0.779883
+[94000][MARGIN]0.792813,0.407446
+lr-batch-epoch: 0.001 3771 12
+testing verification..
+(14000, 512)
+infer time 28.883321
+[cfp_ff][94000]XNorm: 21.517234
+[cfp_ff][94000]Accuracy-Flip: 0.99771+-0.00249
+testing verification..
+(14000, 512)
+infer time 27.832593
+[cfp_fp][94000]XNorm: 19.505244
+[cfp_fp][94000]Accuracy-Flip: 0.92429+-0.01311
+testing verification..
+(12000, 512)
+infer time 21.234582
+[agedb_30][94000]XNorm: 22.565056
+[agedb_30][94000]Accuracy-Flip: 0.97817+-0.00664
+testing verification..
+(12000, 512)
+infer time 18.56832
+[lfw][94000]XNorm: 22.137997
+[lfw][94000]Accuracy-Flip: 0.99800+-0.00277
+saving 47
+INFO:root:Saved checkpoint to "../model-r50-am-lfw/model-0047.params"
+[94000]Accuracy-Highest: 0.99800
+INFO:root:Epoch[12] Batch [3780]	Speed: 75.44 samples/sec	acc=0.778125
+INFO:root:Epoch[12] Batch [3800]	Speed: 644.26 samples/sec	acc=0.779199
+INFO:root:Epoch[12] Batch [3820]	Speed: 661.34 samples/sec	acc=0.783301
+INFO:root:Epoch[12] Batch [3840]	Speed: 659.50 samples/sec	acc=0.778809
+INFO:root:Epoch[12] Batch [3860]	Speed: 662.36 samples/sec	acc=0.773145
+INFO:root:Epoch[12] Batch [3880]	Speed: 661.73 samples/sec	acc=0.781445
+INFO:root:Epoch[12] Batch [3900]	Speed: 662.28 samples/sec	acc=0.781250
+INFO:root:Epoch[12] Batch [3920]	Speed: 664.27 samples/sec	acc=0.768945
+INFO:root:Epoch[12] Batch [3940]	Speed: 673.07 samples/sec	acc=0.774121
+INFO:root:Epoch[12] Batch [3960]	Speed: 663.49 samples/sec	acc=0.778320
+INFO:root:Epoch[12] Batch [3980]	Speed: 647.27 samples/sec	acc=0.777832
+INFO:root:Epoch[12] Batch [4000]	Speed: 662.56 samples/sec	acc=0.775879
+INFO:root:Epoch[12] Batch [4020]	Speed: 660.87 samples/sec	acc=0.776367
+INFO:root:Epoch[12] Batch [4040]	Speed: 661.68 samples/sec	acc=0.777051
+INFO:root:Epoch[12] Batch [4060]	Speed: 661.56 samples/sec	acc=0.775195
+INFO:root:Epoch[12] Batch [4080]	Speed: 662.28 samples/sec	acc=0.775684
+INFO:root:Epoch[12] Batch [4100]	Speed: 662.05 samples/sec	acc=0.778613
+INFO:root:Epoch[12] Batch [4120]	Speed: 658.03 samples/sec	acc=0.775098
+INFO:root:Epoch[12] Batch [4140]	Speed: 662.43 samples/sec	acc=0.777734
+INFO:root:Epoch[12] Batch [4160]	Speed: 661.57 samples/sec	acc=0.781738
+INFO:root:Epoch[12] Batch [4180]	Speed: 662.03 samples/sec	acc=0.777832
+INFO:root:Epoch[12] Batch [4200]	Speed: 652.55 samples/sec	acc=0.775488
+INFO:root:Epoch[12] Batch [4220]	Speed: 662.04 samples/sec	acc=0.766113
+INFO:root:Epoch[12] Batch [4240]	Speed: 656.44 samples/sec	acc=0.782031
+INFO:root:Epoch[12] Batch [4260]	Speed: 656.34 samples/sec	acc=0.781152
+INFO:root:Epoch[12] Batch [4280]	Speed: 662.20 samples/sec	acc=0.771875
+INFO:root:Epoch[12] Batch [4300]	Speed: 641.37 samples/sec	acc=0.774609
+INFO:root:Epoch[12] Batch [4320]	Speed: 621.41 samples/sec	acc=0.778125
+INFO:root:Epoch[12] Batch [4340]	Speed: 653.16 samples/sec	acc=0.774609
+INFO:root:Epoch[12] Batch [4360]	Speed: 648.57 samples/sec	acc=0.774414
+INFO:root:Epoch[12] Batch [4380]	Speed: 662.01 samples/sec	acc=0.783594
+INFO:root:Epoch[12] Batch [4400]	Speed: 661.49 samples/sec	acc=0.779395
+INFO:root:Epoch[12] Batch [4420]	Speed: 662.02 samples/sec	acc=0.773340
+INFO:root:Epoch[12] Batch [4440]	Speed: 624.40 samples/sec	acc=0.780957
+INFO:root:Epoch[12] Batch [4460]	Speed: 629.29 samples/sec	acc=0.779492
+INFO:root:Epoch[12] Batch [4480]	Speed: 642.27 samples/sec	acc=0.774902
+INFO:root:Epoch[12] Batch [4500]	Speed: 650.69 samples/sec	acc=0.770605
+INFO:root:Epoch[12] Batch [4520]	Speed: 660.87 samples/sec	acc=0.779785
+INFO:root:Epoch[12] Batch [4540]	Speed: 661.89 samples/sec	acc=0.779395
+INFO:root:Epoch[12] Batch [4560]	Speed: 661.48 samples/sec	acc=0.775195
+INFO:root:Epoch[12] Batch [4580]	Speed: 649.11 samples/sec	acc=0.774121
+INFO:root:Epoch[12] Batch [4600]	Speed: 660.74 samples/sec	acc=0.774512
+INFO:root:Epoch[12] Batch [4620]	Speed: 662.44 samples/sec	acc=0.777441
+INFO:root:Epoch[12] Batch [4640]	Speed: 657.31 samples/sec	acc=0.776172
+INFO:root:Epoch[12] Batch [4660]	Speed: 662.24 samples/sec	acc=0.782324
+INFO:root:Epoch[12] Batch [4680]	Speed: 660.89 samples/sec	acc=0.775000
+INFO:root:Epoch[12] Batch [4700]	Speed: 661.71 samples/sec	acc=0.776953
+INFO:root:Epoch[12] Batch [4720]	Speed: 658.59 samples/sec	acc=0.770898
+INFO:root:Epoch[12] Batch [4740]	Speed: 662.85 samples/sec	acc=0.771191
+INFO:root:Epoch[12] Batch [4760]	Speed: 661.33 samples/sec	acc=0.773535
+lr-batch-epoch: 0.001 4771 12
+INFO:root:Epoch[12] Batch [4780]	Speed: 653.32 samples/sec	acc=0.775488
+INFO:root:Epoch[12] Batch [4800]	Speed: 661.56 samples/sec	acc=0.777441
+INFO:root:Epoch[12] Batch [4820]	Speed: 624.38 samples/sec	acc=0.782324
+INFO:root:Epoch[12] Batch [4840]	Speed: 666.73 samples/sec	acc=0.776563
+INFO:root:Epoch[12] Batch [4860]	Speed: 661.66 samples/sec	acc=0.771973
+INFO:root:Epoch[12] Batch [4880]	Speed: 660.56 samples/sec	acc=0.774316
+INFO:root:Epoch[12] Batch [4900]	Speed: 661.43 samples/sec	acc=0.779004
+INFO:root:Epoch[12] Batch [4920]	Speed: 658.25 samples/sec	acc=0.775293
+INFO:root:Epoch[12] Batch [4940]	Speed: 660.82 samples/sec	acc=0.776367
+INFO:root:Epoch[12] Batch [4960]	Speed: 645.62 samples/sec	acc=0.776074
+INFO:root:Epoch[12] Batch [4980]	Speed: 657.53 samples/sec	acc=0.771582
+INFO:root:Epoch[12] Batch [5000]	Speed: 676.22 samples/sec	acc=0.775098
+INFO:root:Epoch[12] Batch [5020]	Speed: 675.92 samples/sec	acc=0.778027
+INFO:root:Epoch[12] Batch [5040]	Speed: 648.27 samples/sec	acc=0.773145
+INFO:root:Epoch[12] Batch [5060]	Speed: 653.57 samples/sec	acc=0.775488
+INFO:root:Epoch[12] Batch [5080]	Speed: 618.77 samples/sec	acc=0.774023
+INFO:root:Epoch[12] Batch [5100]	Speed: 659.73 samples/sec	acc=0.772168
+INFO:root:Epoch[12] Batch [5120]	Speed: 660.35 samples/sec	acc=0.775391
+INFO:root:Epoch[12] Batch [5140]	Speed: 633.74 samples/sec	acc=0.783887
+INFO:root:Epoch[12] Batch [5160]	Speed: 657.46 samples/sec	acc=0.775293
+INFO:root:Epoch[12] Batch [5180]	Speed: 661.13 samples/sec	acc=0.775000
+INFO:root:Epoch[12] Batch [5200]	Speed: 661.92 samples/sec	acc=0.775879
+INFO:root:Epoch[12] Batch [5220]	Speed: 652.96 samples/sec	acc=0.771289
+INFO:root:Epoch[12] Batch [5240]	Speed: 661.11 samples/sec	acc=0.771387
+INFO:root:Epoch[12] Batch [5260]	Speed: 661.25 samples/sec	acc=0.776660
+INFO:root:Epoch[12] Batch [5280]	Speed: 662.51 samples/sec	acc=0.779980
+INFO:root:Epoch[12] Batch [5300]	Speed: 646.93 samples/sec	acc=0.780469
+INFO:root:Epoch[12] Batch [5320]	Speed: 661.73 samples/sec	acc=0.772754
+INFO:root:Epoch[12] Batch [5340]	Speed: 641.38 samples/sec	acc=0.774121
+INFO:root:Epoch[12] Batch [5360]	Speed: 646.74 samples/sec	acc=0.769629
+INFO:root:Epoch[12] Batch [5380]	Speed: 655.19 samples/sec	acc=0.768750
+INFO:root:Epoch[12] Batch [5400]	Speed: 663.17 samples/sec	acc=0.778906
+INFO:root:Epoch[12] Batch [5420]	Speed: 661.31 samples/sec	acc=0.779004
+INFO:root:Epoch[12] Batch [5440]	Speed: 661.64 samples/sec	acc=0.782129
+INFO:root:Epoch[12] Batch [5460]	Speed: 656.27 samples/sec	acc=0.775195
+INFO:root:Epoch[12] Batch [5480]	Speed: 659.40 samples/sec	acc=0.778711
+INFO:root:Epoch[12] Batch [5500]	Speed: 661.60 samples/sec	acc=0.775977
+INFO:root:Epoch[12] Batch [5520]	Speed: 661.79 samples/sec	acc=0.776953
+INFO:root:Epoch[12] Batch [5540]	Speed: 642.72 samples/sec	acc=0.780273
+INFO:root:Epoch[12] Batch [5560]	Speed: 661.11 samples/sec	acc=0.776367
+INFO:root:Epoch[12] Batch [5580]	Speed: 662.06 samples/sec	acc=0.782129
+INFO:root:Epoch[12] Batch [5600]	Speed: 661.63 samples/sec	acc=0.781543
+INFO:root:Epoch[12] Batch [5620]	Speed: 661.15 samples/sec	acc=0.773438
+INFO:root:Epoch[12] Batch [5640]	Speed: 650.24 samples/sec	acc=0.781250
+INFO:root:Epoch[12] Batch [5660]	Speed: 641.97 samples/sec	acc=0.773633
+INFO:root:Epoch[12] Batch [5680]	Speed: 661.34 samples/sec	acc=0.773047
+INFO:root:Epoch[12] Batch [5700]	Speed: 661.42 samples/sec	acc=0.775586
+INFO:root:Epoch[12] Batch [5720]	Speed: 661.68 samples/sec	acc=0.774805
+INFO:root:Epoch[12] Batch [5740]	Speed: 629.66 samples/sec	acc=0.775977
+INFO:root:Epoch[12] Batch [5760]	Speed: 661.47 samples/sec	acc=0.783105
+[96000][MARGIN]0.794114,0.408994
+lr-batch-epoch: 0.001 5771 12
+testing verification..
+(14000, 512)
+infer time 24.707595
+[cfp_ff][96000]XNorm: 21.380046
+[cfp_ff][96000]Accuracy-Flip: 0.99771+-0.00232
+testing verification..
+(14000, 512)
+infer time 26.146211
+[cfp_fp][96000]XNorm: 19.539928
+[cfp_fp][96000]Accuracy-Flip: 0.92600+-0.01454
+testing verification..
+(12000, 512)
+infer time 19.633509
+[agedb_30][96000]XNorm: 22.444902
+[agedb_30][96000]Accuracy-Flip: 0.97750+-0.00647
+testing verification..
+(12000, 512)
+infer time 17.610915
+[lfw][96000]XNorm: 22.062007
+[lfw][96000]Accuracy-Flip: 0.99767+-0.00281
+[96000]Accuracy-Highest: 0.99800
+INFO:root:Epoch[12] Batch [5780]	Speed: 83.37 samples/sec	acc=0.778906
+INFO:root:Epoch[12] Batch [5800]	Speed: 641.45 samples/sec	acc=0.776660
+INFO:root:Epoch[12] Batch [5820]	Speed: 662.19 samples/sec	acc=0.778809
+INFO:root:Epoch[12] Batch [5840]	Speed: 661.06 samples/sec	acc=0.777637
+INFO:root:Epoch[12] Batch [5860]	Speed: 662.07 samples/sec	acc=0.784473
+INFO:root:Epoch[12] Batch [5880]	Speed: 662.67 samples/sec	acc=0.779980
+INFO:root:Epoch[12] Batch [5900]	Speed: 661.73 samples/sec	acc=0.778320
+INFO:root:Epoch[12] Batch [5920]	Speed: 655.19 samples/sec	acc=0.777344
+INFO:root:Epoch[12] Batch [5940]	Speed: 665.23 samples/sec	acc=0.779004
+INFO:root:Epoch[12] Batch [5960]	Speed: 657.28 samples/sec	acc=0.776563
+INFO:root:Epoch[12] Batch [5980]	Speed: 663.39 samples/sec	acc=0.771973
+INFO:root:Epoch[12] Batch [6000]	Speed: 663.21 samples/sec	acc=0.778027
+INFO:root:Epoch[12] Batch [6020]	Speed: 661.27 samples/sec	acc=0.776074
+INFO:root:Epoch[12] Batch [6040]	Speed: 650.88 samples/sec	acc=0.778320
+INFO:root:Epoch[12] Batch [6060]	Speed: 656.87 samples/sec	acc=0.775000
+INFO:root:Epoch[12] Batch [6080]	Speed: 655.13 samples/sec	acc=0.774512
+INFO:root:Epoch[12] Batch [6100]	Speed: 634.44 samples/sec	acc=0.769922
+INFO:root:Epoch[12] Batch [6120]	Speed: 661.66 samples/sec	acc=0.776074
+INFO:root:Epoch[12] Batch [6140]	Speed: 662.12 samples/sec	acc=0.781055
+INFO:root:Epoch[12] Batch [6160]	Speed: 662.01 samples/sec	acc=0.778223
+INFO:root:Epoch[12] Batch [6180]	Speed: 661.86 samples/sec	acc=0.779004
+INFO:root:Epoch[12] Batch [6200]	Speed: 652.68 samples/sec	acc=0.772559
+INFO:root:Epoch[12] Batch [6220]	Speed: 662.21 samples/sec	acc=0.784766
+INFO:root:Epoch[12] Batch [6240]	Speed: 662.43 samples/sec	acc=0.778809
+INFO:root:Epoch[12] Batch [6260]	Speed: 660.87 samples/sec	acc=0.774609
+INFO:root:Epoch[12] Batch [6280]	Speed: 662.16 samples/sec	acc=0.777539
+INFO:root:Epoch[12] Batch [6300]	Speed: 661.79 samples/sec	acc=0.782129
+INFO:root:Epoch[12] Batch [6320]	Speed: 660.72 samples/sec	acc=0.774316
+INFO:root:Epoch[12] Batch [6340]	Speed: 661.87 samples/sec	acc=0.778223
+INFO:root:Epoch[12] Batch [6360]	Speed: 662.24 samples/sec	acc=0.782422
+INFO:root:Epoch[12] Batch [6380]	Speed: 661.32 samples/sec	acc=0.774219
+INFO:root:Epoch[12] Batch [6400]	Speed: 658.23 samples/sec	acc=0.778906
+INFO:root:Epoch[12] Batch [6420]	Speed: 662.95 samples/sec	acc=0.774609
+INFO:root:Epoch[12] Batch [6440]	Speed: 659.14 samples/sec	acc=0.772559
+INFO:root:Epoch[12] Batch [6460]	Speed: 663.70 samples/sec	acc=0.777734
+INFO:root:Epoch[12] Batch [6480]	Speed: 661.93 samples/sec	acc=0.778613
+INFO:root:Epoch[12] Batch [6500]	Speed: 661.21 samples/sec	acc=0.773438
+INFO:root:Epoch[12] Batch [6520]	Speed: 652.94 samples/sec	acc=0.767480
+INFO:root:Epoch[12] Batch [6540]	Speed: 661.89 samples/sec	acc=0.776465
+INFO:root:Epoch[12] Batch [6560]	Speed: 661.01 samples/sec	acc=0.777734
+INFO:root:Epoch[12] Batch [6580]	Speed: 656.54 samples/sec	acc=0.774121
+INFO:root:Epoch[12] Batch [6600]	Speed: 662.62 samples/sec	acc=0.779785
+INFO:root:Epoch[12] Batch [6620]	Speed: 642.29 samples/sec	acc=0.779199
+INFO:root:Epoch[12] Batch [6640]	Speed: 661.45 samples/sec	acc=0.771582
+INFO:root:Epoch[12] Batch [6660]	Speed: 660.96 samples/sec	acc=0.779785
+INFO:root:Epoch[12] Batch [6680]	Speed: 661.46 samples/sec	acc=0.783398
+INFO:root:Epoch[12] Batch [6700]	Speed: 661.71 samples/sec	acc=0.770020
+INFO:root:Epoch[12] Batch [6720]	Speed: 660.89 samples/sec	acc=0.771680
+INFO:root:Epoch[12] Batch [6740]	Speed: 661.08 samples/sec	acc=0.770703
+INFO:root:Epoch[12] Batch [6760]	Speed: 653.36 samples/sec	acc=0.781641
+lr-batch-epoch: 0.001 6771 12
+INFO:root:Epoch[12] Batch [6780]	Speed: 662.72 samples/sec	acc=0.777051
+INFO:root:Epoch[12] Batch [6800]	Speed: 661.14 samples/sec	acc=0.776758
+INFO:root:Epoch[12] Batch [6820]	Speed: 662.25 samples/sec	acc=0.772949
+INFO:root:Epoch[12] Batch [6840]	Speed: 662.17 samples/sec	acc=0.780078
+INFO:root:Epoch[12] Batch [6860]	Speed: 661.31 samples/sec	acc=0.776563
+INFO:root:Epoch[12] Batch [6880]	Speed: 661.97 samples/sec	acc=0.768750
+INFO:root:Epoch[12] Batch [6900]	Speed: 661.87 samples/sec	acc=0.775000
+INFO:root:Epoch[12] Batch [6920]	Speed: 658.81 samples/sec	acc=0.776953
+INFO:root:Epoch[12] Batch [6940]	Speed: 655.35 samples/sec	acc=0.767285
+INFO:root:Epoch[12] Batch [6960]	Speed: 671.44 samples/sec	acc=0.775879
+INFO:root:Epoch[12] Batch [6980]	Speed: 661.03 samples/sec	acc=0.782031
+INFO:root:Epoch[12] Batch [7000]	Speed: 661.84 samples/sec	acc=0.773535
+INFO:root:Epoch[12] Batch [7020]	Speed: 662.24 samples/sec	acc=0.776172
+INFO:root:Epoch[12] Batch [7040]	Speed: 661.42 samples/sec	acc=0.775195
+INFO:root:Epoch[12] Batch [7060]	Speed: 641.51 samples/sec	acc=0.774609
+INFO:root:Epoch[12] Batch [7080]	Speed: 662.07 samples/sec	acc=0.773438
+INFO:root:Epoch[12] Batch [7100]	Speed: 661.32 samples/sec	acc=0.772070
+INFO:root:Epoch[12] Batch [7120]	Speed: 661.41 samples/sec	acc=0.780176
+INFO:root:Epoch[12] Batch [7140]	Speed: 662.43 samples/sec	acc=0.772559
+INFO:root:Epoch[12] Batch [7160]	Speed: 661.00 samples/sec	acc=0.781250
+INFO:root:Epoch[12] Batch [7180]	Speed: 661.25 samples/sec	acc=0.771484
+INFO:root:Epoch[12] Batch [7200]	Speed: 662.13 samples/sec	acc=0.770117
+INFO:root:Epoch[12] Batch [7220]	Speed: 660.99 samples/sec	acc=0.773926
+INFO:root:Epoch[12] Batch [7240]	Speed: 661.72 samples/sec	acc=0.774512
+INFO:root:Epoch[12] Batch [7260]	Speed: 649.74 samples/sec	acc=0.779297
+INFO:root:Epoch[12] Batch [7280]	Speed: 660.24 samples/sec	acc=0.777051
+INFO:root:Epoch[12] Batch [7300]	Speed: 656.93 samples/sec	acc=0.773926
+INFO:root:Epoch[12] Batch [7320]	Speed: 664.92 samples/sec	acc=0.777539
+INFO:root:Epoch[12] Batch [7340]	Speed: 656.65 samples/sec	acc=0.773242
+INFO:root:Epoch[12] Batch [7360]	Speed: 661.06 samples/sec	acc=0.780566
+INFO:root:Epoch[12] Batch [7380]	Speed: 662.24 samples/sec	acc=0.771582
+INFO:root:Epoch[12] Batch [7400]	Speed: 646.32 samples/sec	acc=0.777539
+INFO:root:Epoch[12] Batch [7420]	Speed: 657.46 samples/sec	acc=0.778418
+INFO:root:Epoch[12] Batch [7440]	Speed: 662.12 samples/sec	acc=0.769922
+INFO:root:Epoch[12] Batch [7460]	Speed: 661.02 samples/sec	acc=0.775391
+INFO:root:Epoch[12] Batch [7480]	Speed: 649.36 samples/sec	acc=0.775488
+INFO:root:Epoch[12] Batch [7500]	Speed: 668.19 samples/sec	acc=0.771191
+INFO:root:Epoch[12] Train-acc=0.773438
+INFO:root:Epoch[12] Time cost=6184.963
+call reset()
+INFO:root:Epoch[13] Batch [20]	Speed: 662.14 samples/sec	acc=0.810082
+INFO:root:Epoch[13] Batch [40]	Speed: 662.24 samples/sec	acc=0.816602
+INFO:root:Epoch[13] Batch [60]	Speed: 635.54 samples/sec	acc=0.804883
+INFO:root:Epoch[13] Batch [80]	Speed: 661.82 samples/sec	acc=0.818750
+INFO:root:Epoch[13] Batch [100]	Speed: 661.94 samples/sec	acc=0.813477
+INFO:root:Epoch[13] Batch [120]	Speed: 647.14 samples/sec	acc=0.812891
+INFO:root:Epoch[13] Batch [140]	Speed: 660.42 samples/sec	acc=0.814941
+INFO:root:Epoch[13] Batch [160]	Speed: 662.15 samples/sec	acc=0.810937
+INFO:root:Epoch[13] Batch [180]	Speed: 656.79 samples/sec	acc=0.816797
+INFO:root:Epoch[13] Batch [200]	Speed: 662.13 samples/sec	acc=0.803418
+INFO:root:Epoch[13] Batch [220]	Speed: 662.49 samples/sec	acc=0.806152
+INFO:root:Epoch[13] Batch [240]	Speed: 662.39 samples/sec	acc=0.816602
+[98000][MARGIN]0.809840,0.431885
+lr-batch-epoch: 0.001 252 13
+testing verification..
+(14000, 512)
+infer time 23.044673
+[cfp_ff][98000]XNorm: 21.540596
+[cfp_ff][98000]Accuracy-Flip: 0.99814+-0.00231
+testing verification..
+(14000, 512)
+infer time 24.11354
+[cfp_fp][98000]XNorm: 19.770752
+[cfp_fp][98000]Accuracy-Flip: 0.92786+-0.01449
+testing verification..
+(12000, 512)
+infer time 19.230186
+[agedb_30][98000]XNorm: 22.672518
+[agedb_30][98000]Accuracy-Flip: 0.97750+-0.00634
+testing verification..
+(12000, 512)
+infer time 20.331927
+[lfw][98000]XNorm: 22.276239
+[lfw][98000]Accuracy-Flip: 0.99767+-0.00271
+[98000]Accuracy-Highest: 0.99800
+INFO:root:Epoch[13] Batch [260]	Speed: 84.10 samples/sec	acc=0.817090
+INFO:root:Epoch[13] Batch [280]	Speed: 657.84 samples/sec	acc=0.806055
+INFO:root:Epoch[13] Batch [300]	Speed: 641.27 samples/sec	acc=0.814160
+INFO:root:Epoch[13] Batch [320]	Speed: 654.98 samples/sec	acc=0.810937
+INFO:root:Epoch[13] Batch [340]	Speed: 661.53 samples/sec	acc=0.808887
+INFO:root:Epoch[13] Batch [360]	Speed: 659.82 samples/sec	acc=0.814746
+INFO:root:Epoch[13] Batch [380]	Speed: 653.69 samples/sec	acc=0.803809
+INFO:root:Epoch[13] Batch [400]	Speed: 661.59 samples/sec	acc=0.815723
+INFO:root:Epoch[13] Batch [420]	Speed: 656.61 samples/sec	acc=0.804297
+INFO:root:Epoch[13] Batch [440]	Speed: 661.62 samples/sec	acc=0.811230
+INFO:root:Epoch[13] Batch [460]	Speed: 662.25 samples/sec	acc=0.811133
+INFO:root:Epoch[13] Batch [480]	Speed: 661.17 samples/sec	acc=0.809473
+INFO:root:Epoch[13] Batch [500]	Speed: 661.27 samples/sec	acc=0.806836
+INFO:root:Epoch[13] Batch [520]	Speed: 661.87 samples/sec	acc=0.805566
+INFO:root:Epoch[13] Batch [540]	Speed: 660.33 samples/sec	acc=0.802539
+INFO:root:Epoch[13] Batch [560]	Speed: 662.10 samples/sec	acc=0.814648
+INFO:root:Epoch[13] Batch [580]	Speed: 662.06 samples/sec	acc=0.813086
+INFO:root:Epoch[13] Batch [600]	Speed: 661.21 samples/sec	acc=0.809668
+INFO:root:Epoch[13] Batch [620]	Speed: 660.65 samples/sec	acc=0.808789
+INFO:root:Epoch[13] Batch [640]	Speed: 661.62 samples/sec	acc=0.802539
+INFO:root:Epoch[13] Batch [660]	Speed: 656.41 samples/sec	acc=0.811133
+INFO:root:Epoch[13] Batch [680]	Speed: 661.28 samples/sec	acc=0.806445
+INFO:root:Epoch[13] Batch [700]	Speed: 656.79 samples/sec	acc=0.803418
+INFO:root:Epoch[13] Batch [720]	Speed: 677.83 samples/sec	acc=0.808789
+INFO:root:Epoch[13] Batch [740]	Speed: 661.04 samples/sec	acc=0.803125
+INFO:root:Epoch[13] Batch [760]	Speed: 652.94 samples/sec	acc=0.804492
+INFO:root:Epoch[13] Batch [780]	Speed: 661.46 samples/sec	acc=0.801367
+INFO:root:Epoch[13] Batch [800]	Speed: 660.82 samples/sec	acc=0.800879
+INFO:root:Epoch[13] Batch [820]	Speed: 662.05 samples/sec	acc=0.809570
+INFO:root:Epoch[13] Batch [840]	Speed: 662.04 samples/sec	acc=0.802148
+INFO:root:Epoch[13] Batch [860]	Speed: 657.54 samples/sec	acc=0.804980
+INFO:root:Epoch[13] Batch [880]	Speed: 661.35 samples/sec	acc=0.802734
+INFO:root:Epoch[13] Batch [900]	Speed: 661.49 samples/sec	acc=0.798535
+INFO:root:Epoch[13] Batch [920]	Speed: 661.99 samples/sec	acc=0.809570
+INFO:root:Epoch[13] Batch [940]	Speed: 661.10 samples/sec	acc=0.804688
+INFO:root:Epoch[13] Batch [960]	Speed: 661.08 samples/sec	acc=0.805371
+INFO:root:Epoch[13] Batch [980]	Speed: 661.60 samples/sec	acc=0.804980
+INFO:root:Epoch[13] Batch [1000]	Speed: 647.51 samples/sec	acc=0.802637
+INFO:root:Epoch[13] Batch [1020]	Speed: 661.53 samples/sec	acc=0.803711
+INFO:root:Epoch[13] Batch [1040]	Speed: 661.68 samples/sec	acc=0.808105
+INFO:root:Epoch[13] Batch [1060]	Speed: 662.06 samples/sec	acc=0.806445
+INFO:root:Epoch[13] Batch [1080]	Speed: 655.45 samples/sec	acc=0.804785
+INFO:root:Epoch[13] Batch [1100]	Speed: 663.04 samples/sec	acc=0.811426
+INFO:root:Epoch[13] Batch [1120]	Speed: 661.30 samples/sec	acc=0.807129
+INFO:root:Epoch[13] Batch [1140]	Speed: 660.88 samples/sec	acc=0.808594
+INFO:root:Epoch[13] Batch [1160]	Speed: 662.09 samples/sec	acc=0.805371
+INFO:root:Epoch[13] Batch [1180]	Speed: 661.30 samples/sec	acc=0.808008
+INFO:root:Epoch[13] Batch [1200]	Speed: 661.11 samples/sec	acc=0.806543
+INFO:root:Epoch[13] Batch [1220]	Speed: 649.81 samples/sec	acc=0.794434
+INFO:root:Epoch[13] Batch [1240]	Speed: 660.49 samples/sec	acc=0.804199
+lr-batch-epoch: 0.001 1252 13
+INFO:root:Epoch[13] Batch [1260]	Speed: 672.85 samples/sec	acc=0.802539
+INFO:root:Epoch[13] Batch [1280]	Speed: 661.69 samples/sec	acc=0.806348
+INFO:root:Epoch[13] Batch [1300]	Speed: 668.28 samples/sec	acc=0.812500
+INFO:root:Epoch[13] Batch [1320]	Speed: 678.44 samples/sec	acc=0.800879
+INFO:root:Epoch[13] Batch [1340]	Speed: 661.02 samples/sec	acc=0.804102
+INFO:root:Epoch[13] Batch [1360]	Speed: 661.22 samples/sec	acc=0.802930
+INFO:root:Epoch[13] Batch [1380]	Speed: 659.16 samples/sec	acc=0.802148
+INFO:root:Epoch[13] Batch [1400]	Speed: 656.04 samples/sec	acc=0.805176
+INFO:root:Epoch[13] Batch [1420]	Speed: 662.75 samples/sec	acc=0.803613
+INFO:root:Epoch[13] Batch [1440]	Speed: 660.68 samples/sec	acc=0.800195
+INFO:root:Epoch[13] Batch [1460]	Speed: 662.18 samples/sec	acc=0.796875
+INFO:root:Epoch[13] Batch [1480]	Speed: 653.52 samples/sec	acc=0.800000
+INFO:root:Epoch[13] Batch [1500]	Speed: 661.59 samples/sec	acc=0.806055
+INFO:root:Epoch[13] Batch [1520]	Speed: 646.26 samples/sec	acc=0.806934
+INFO:root:Epoch[13] Batch [1540]	Speed: 649.35 samples/sec	acc=0.811230
+INFO:root:Epoch[13] Batch [1560]	Speed: 630.93 samples/sec	acc=0.802930
+INFO:root:Epoch[13] Batch [1580]	Speed: 662.41 samples/sec	acc=0.810840
+INFO:root:Epoch[13] Batch [1600]	Speed: 656.82 samples/sec	acc=0.806250
+INFO:root:Epoch[13] Batch [1620]	Speed: 661.21 samples/sec	acc=0.799805
+INFO:root:Epoch[13] Batch [1640]	Speed: 650.01 samples/sec	acc=0.804590
+INFO:root:Epoch[13] Batch [1660]	Speed: 662.20 samples/sec	acc=0.800586
+INFO:root:Epoch[13] Batch [1680]	Speed: 661.80 samples/sec	acc=0.810449
+INFO:root:Epoch[13] Batch [1700]	Speed: 662.23 samples/sec	acc=0.798926
+INFO:root:Epoch[13] Batch [1720]	Speed: 661.84 samples/sec	acc=0.807520
+INFO:root:Epoch[13] Batch [1740]	Speed: 661.87 samples/sec	acc=0.801758
+INFO:root:Epoch[13] Batch [1760]	Speed: 638.93 samples/sec	acc=0.810352
+INFO:root:Epoch[13] Batch [1780]	Speed: 655.01 samples/sec	acc=0.806055
+INFO:root:Epoch[13] Batch [1800]	Speed: 661.74 samples/sec	acc=0.805859
+INFO:root:Epoch[13] Batch [1820]	Speed: 662.06 samples/sec	acc=0.804395
+INFO:root:Epoch[13] Batch [1840]	Speed: 662.54 samples/sec	acc=0.805469
+INFO:root:Epoch[13] Batch [1860]	Speed: 661.99 samples/sec	acc=0.810352
+INFO:root:Epoch[13] Batch [1880]	Speed: 662.21 samples/sec	acc=0.803320
+INFO:root:Epoch[13] Batch [1900]	Speed: 661.57 samples/sec	acc=0.800977
+INFO:root:Epoch[13] Batch [1920]	Speed: 641.69 samples/sec	acc=0.803223
+INFO:root:Epoch[13] Batch [1940]	Speed: 659.53 samples/sec	acc=0.805176
+INFO:root:Epoch[13] Batch [1960]	Speed: 648.42 samples/sec	acc=0.808984
+INFO:root:Epoch[13] Batch [1980]	Speed: 655.89 samples/sec	acc=0.802246
+INFO:root:Epoch[13] Batch [2000]	Speed: 664.19 samples/sec	acc=0.806641
+INFO:root:Epoch[13] Batch [2020]	Speed: 638.49 samples/sec	acc=0.810254
+INFO:root:Epoch[13] Batch [2040]	Speed: 601.14 samples/sec	acc=0.805762
+INFO:root:Epoch[13] Batch [2060]	Speed: 652.61 samples/sec	acc=0.808887
+INFO:root:Epoch[13] Batch [2080]	Speed: 644.37 samples/sec	acc=0.801660
+INFO:root:Epoch[13] Batch [2100]	Speed: 661.35 samples/sec	acc=0.797949
+INFO:root:Epoch[13] Batch [2120]	Speed: 667.73 samples/sec	acc=0.799414
+INFO:root:Epoch[13] Batch [2140]	Speed: 675.87 samples/sec	acc=0.797656
+INFO:root:Epoch[13] Batch [2160]	Speed: 645.34 samples/sec	acc=0.801855
+INFO:root:Epoch[13] Batch [2180]	Speed: 661.10 samples/sec	acc=0.800098
+INFO:root:Epoch[13] Batch [2200]	Speed: 662.14 samples/sec	acc=0.804004
+INFO:root:Epoch[13] Batch [2220]	Speed: 661.34 samples/sec	acc=0.801172
+INFO:root:Epoch[13] Batch [2240]	Speed: 640.26 samples/sec	acc=0.804199
+[100000][MARGIN]0.801697,0.419335
+lr change to 0.0001
+lr-batch-epoch: 0.0001 2252 13
+testing verification..
+(14000, 512)
+infer time 26.93032
+[cfp_ff][100000]XNorm: 21.206661
+[cfp_ff][100000]Accuracy-Flip: 0.99829+-0.00229
+testing verification..
+(14000, 512)
+infer time 24.364064
+[cfp_fp][100000]XNorm: 19.494624
+[cfp_fp][100000]Accuracy-Flip: 0.92743+-0.01295
+testing verification..
+(12000, 512)
+infer time 22.174277
+[agedb_30][100000]XNorm: 22.312334
+[agedb_30][100000]Accuracy-Flip: 0.97767+-0.00700
+testing verification..
+(12000, 512)
+infer time 25.665378
+[lfw][100000]XNorm: 21.961490
+[lfw][100000]Accuracy-Flip: 0.99817+-0.00229
+saving 50
+INFO:root:Saved checkpoint to "../model-r50-am-lfw/model-0050.params"
+[100000]Accuracy-Highest: 0.99817
+INFO:root:Epoch[13] Batch [2260]	Speed: 72.44 samples/sec	acc=0.806543
+INFO:root:Epoch[13] Batch [2280]	Speed: 668.04 samples/sec	acc=0.801367
+INFO:root:Epoch[13] Batch [2300]	Speed: 652.38 samples/sec	acc=0.808496
+INFO:root:Epoch[13] Batch [2320]	Speed: 674.38 samples/sec	acc=0.800098
+INFO:root:Epoch[13] Batch [2340]	Speed: 663.41 samples/sec	acc=0.804590
+INFO:root:Epoch[13] Batch [2360]	Speed: 662.05 samples/sec	acc=0.812207
+INFO:root:Epoch[13] Batch [2380]	Speed: 661.66 samples/sec	acc=0.808301
+INFO:root:Epoch[13] Batch [2400]	Speed: 661.21 samples/sec	acc=0.803809
+INFO:root:Epoch[13] Batch [2420]	Speed: 665.95 samples/sec	acc=0.797949
+INFO:root:Epoch[13] Batch [2440]	Speed: 662.26 samples/sec	acc=0.805859
+INFO:root:Epoch[13] Batch [2460]	Speed: 661.76 samples/sec	acc=0.806152
+INFO:root:Epoch[13] Batch [2480]	Speed: 661.95 samples/sec	acc=0.804688
+INFO:root:Epoch[13] Batch [2500]	Speed: 662.82 samples/sec	acc=0.802246
+INFO:root:Epoch[13] Batch [2520]	Speed: 622.50 samples/sec	acc=0.800977
+INFO:root:Epoch[13] Batch [2540]	Speed: 627.07 samples/sec	acc=0.811719
+INFO:root:Epoch[13] Batch [2560]	Speed: 654.91 samples/sec	acc=0.810840
+INFO:root:Epoch[13] Batch [2580]	Speed: 652.78 samples/sec	acc=0.803418
+INFO:root:Epoch[13] Batch [2600]	Speed: 604.71 samples/sec	acc=0.804785
+INFO:root:Epoch[13] Batch [2620]	Speed: 644.27 samples/sec	acc=0.810352
+INFO:root:Epoch[13] Batch [2640]	Speed: 661.71 samples/sec	acc=0.805566
+INFO:root:Epoch[13] Batch [2660]	Speed: 637.08 samples/sec	acc=0.805566
+INFO:root:Epoch[13] Batch [2680]	Speed: 662.65 samples/sec	acc=0.806250
+INFO:root:Epoch[13] Batch [2700]	Speed: 622.05 samples/sec	acc=0.808594
+INFO:root:Epoch[13] Batch [2720]	Speed: 662.91 samples/sec	acc=0.805957
+INFO:root:Epoch[13] Batch [2740]	Speed: 661.81 samples/sec	acc=0.811328
+INFO:root:Epoch[13] Batch [2760]	Speed: 661.41 samples/sec	acc=0.807227
+INFO:root:Epoch[13] Batch [2780]	Speed: 658.69 samples/sec	acc=0.807129
+INFO:root:Epoch[13] Batch [2800]	Speed: 632.60 samples/sec	acc=0.804297
+INFO:root:Epoch[13] Batch [2820]	Speed: 651.17 samples/sec	acc=0.799414
+INFO:root:Epoch[13] Batch [2840]	Speed: 661.97 samples/sec	acc=0.804199
+INFO:root:Epoch[13] Batch [2860]	Speed: 662.45 samples/sec	acc=0.818555
+INFO:root:Epoch[13] Batch [2880]	Speed: 661.55 samples/sec	acc=0.818066
+INFO:root:Epoch[13] Batch [2900]	Speed: 661.82 samples/sec	acc=0.806738
+INFO:root:Epoch[13] Batch [2920]	Speed: 662.44 samples/sec	acc=0.808203
+INFO:root:Epoch[13] Batch [2940]	Speed: 648.67 samples/sec	acc=0.806152
+INFO:root:Epoch[13] Batch [2960]	Speed: 665.08 samples/sec	acc=0.807031
+INFO:root:Epoch[13] Batch [2980]	Speed: 660.09 samples/sec	acc=0.812109
+INFO:root:Epoch[13] Batch [3000]	Speed: 666.33 samples/sec	acc=0.809961
+INFO:root:Epoch[13] Batch [3020]	Speed: 640.95 samples/sec	acc=0.805176
+INFO:root:Epoch[13] Batch [3040]	Speed: 641.19 samples/sec	acc=0.803516
+INFO:root:Epoch[13] Batch [3060]	Speed: 659.65 samples/sec	acc=0.806250
+INFO:root:Epoch[13] Batch [3080]	Speed: 661.10 samples/sec	acc=0.804199
+INFO:root:Epoch[13] Batch [3100]	Speed: 657.21 samples/sec	acc=0.808008
+INFO:root:Epoch[13] Batch [3120]	Speed: 634.70 samples/sec	acc=0.812305
+INFO:root:Epoch[13] Batch [3140]	Speed: 674.40 samples/sec	acc=0.800781
+INFO:root:Epoch[13] Batch [3160]	Speed: 662.37 samples/sec	acc=0.812012
+INFO:root:Epoch[13] Batch [3180]	Speed: 661.53 samples/sec	acc=0.813281
+INFO:root:Epoch[13] Batch [3200]	Speed: 651.72 samples/sec	acc=0.806934
+INFO:root:Epoch[13] Batch [3220]	Speed: 641.70 samples/sec	acc=0.797461
+INFO:root:Epoch[13] Batch [3240]	Speed: 662.12 samples/sec	acc=0.805957
+lr-batch-epoch: 0.0001 3252 13
+INFO:root:Epoch[13] Batch [3260]	Speed: 615.46 samples/sec	acc=0.807910
+INFO:root:Epoch[13] Batch [3280]	Speed: 662.65 samples/sec	acc=0.811328
+INFO:root:Epoch[13] Batch [3300]	Speed: 661.85 samples/sec	acc=0.807031
+INFO:root:Epoch[13] Batch [3320]	Speed: 652.86 samples/sec	acc=0.802930
+INFO:root:Epoch[13] Batch [3340]	Speed: 629.09 samples/sec	acc=0.809473
+INFO:root:Epoch[13] Batch [3360]	Speed: 661.49 samples/sec	acc=0.804492
+INFO:root:Epoch[13] Batch [3380]	Speed: 654.63 samples/sec	acc=0.818945
+INFO:root:Epoch[13] Batch [3400]	Speed: 661.71 samples/sec	acc=0.805566
+INFO:root:Epoch[13] Batch [3420]	Speed: 661.22 samples/sec	acc=0.810840
+INFO:root:Epoch[13] Batch [3440]	Speed: 661.92 samples/sec	acc=0.806738
+INFO:root:Epoch[13] Batch [3460]	Speed: 653.05 samples/sec	acc=0.809082
+INFO:root:Epoch[13] Batch [3480]	Speed: 655.29 samples/sec	acc=0.804297
+INFO:root:Epoch[13] Batch [3500]	Speed: 657.45 samples/sec	acc=0.811328
+INFO:root:Epoch[13] Batch [3520]	Speed: 646.95 samples/sec	acc=0.810352
+INFO:root:Epoch[13] Batch [3540]	Speed: 661.73 samples/sec	acc=0.812891
+INFO:root:Epoch[13] Batch [3560]	Speed: 661.58 samples/sec	acc=0.805566
+INFO:root:Epoch[13] Batch [3580]	Speed: 662.72 samples/sec	acc=0.809863
+INFO:root:Epoch[13] Batch [3600]	Speed: 661.43 samples/sec	acc=0.800684
+INFO:root:Epoch[13] Batch [3620]	Speed: 662.03 samples/sec	acc=0.803613
+INFO:root:Epoch[13] Batch [3640]	Speed: 662.38 samples/sec	acc=0.812793
+INFO:root:Epoch[13] Batch [3660]	Speed: 643.43 samples/sec	acc=0.805957
+INFO:root:Epoch[13] Batch [3680]	Speed: 649.53 samples/sec	acc=0.808301
+INFO:root:Epoch[13] Batch [3700]	Speed: 656.30 samples/sec	acc=0.807422
+INFO:root:Epoch[13] Batch [3720]	Speed: 623.31 samples/sec	acc=0.807715
+INFO:root:Epoch[13] Batch [3740]	Speed: 638.47 samples/sec	acc=0.806934
+INFO:root:Epoch[13] Batch [3760]	Speed: 647.90 samples/sec	acc=0.806152
+INFO:root:Epoch[13] Batch [3780]	Speed: 661.29 samples/sec	acc=0.819336
+INFO:root:Epoch[13] Batch [3800]	Speed: 662.24 samples/sec	acc=0.806250
+INFO:root:Epoch[13] Batch [3820]	Speed: 670.34 samples/sec	acc=0.814160
+INFO:root:Epoch[13] Batch [3840]	Speed: 632.94 samples/sec	acc=0.809766
+INFO:root:Epoch[13] Batch [3860]	Speed: 662.68 samples/sec	acc=0.808691
+INFO:root:Epoch[13] Batch [3880]	Speed: 662.14 samples/sec	acc=0.809082
+INFO:root:Epoch[13] Batch [3900]	Speed: 641.89 samples/sec	acc=0.807715
+INFO:root:Epoch[13] Batch [3920]	Speed: 653.52 samples/sec	acc=0.807617
+INFO:root:Epoch[13] Batch [3940]	Speed: 636.86 samples/sec	acc=0.801855
+INFO:root:Epoch[13] Batch [3960]	Speed: 661.01 samples/sec	acc=0.810254
+INFO:root:Epoch[13] Batch [3980]	Speed: 637.00 samples/sec	acc=0.811133
+INFO:root:Epoch[13] Batch [4000]	Speed: 658.71 samples/sec	acc=0.807324
+INFO:root:Epoch[13] Batch [4020]	Speed: 661.31 samples/sec	acc=0.807813
+INFO:root:Epoch[13] Batch [4040]	Speed: 642.92 samples/sec	acc=0.805762
+INFO:root:Epoch[13] Batch [4060]	Speed: 662.15 samples/sec	acc=0.811914
+INFO:root:Epoch[13] Batch [4080]	Speed: 661.90 samples/sec	acc=0.810352
+INFO:root:Epoch[13] Batch [4100]	Speed: 662.47 samples/sec	acc=0.808789
+INFO:root:Epoch[13] Batch [4120]	Speed: 660.62 samples/sec	acc=0.813379
+INFO:root:Epoch[13] Batch [4140]	Speed: 638.23 samples/sec	acc=0.806641
+INFO:root:Epoch[13] Batch [4160]	Speed: 661.64 samples/sec	acc=0.806250
+INFO:root:Epoch[13] Batch [4180]	Speed: 662.08 samples/sec	acc=0.812598
+INFO:root:Epoch[13] Batch [4200]	Speed: 661.08 samples/sec	acc=0.802832
+INFO:root:Epoch[13] Batch [4220]	Speed: 635.06 samples/sec	acc=0.809961
+INFO:root:Epoch[13] Batch [4240]	Speed: 662.19 samples/sec	acc=0.812207
+[102000][MARGIN]0.802079,0.420633
+lr-batch-epoch: 0.0001 4252 13
+testing verification..
+(14000, 512)
+infer time 17.704362
+[cfp_ff][102000]XNorm: 21.493158
+[cfp_ff][102000]Accuracy-Flip: 0.99814+-0.00231
+testing verification..
+(14000, 512)
+infer time 18.302814
+[cfp_fp][102000]XNorm: 19.743070
+[cfp_fp][102000]Accuracy-Flip: 0.92757+-0.01447
+testing verification..
+(12000, 512)
+infer time 14.929935
+[agedb_30][102000]XNorm: 22.679150
+[agedb_30][102000]Accuracy-Flip: 0.97767+-0.00696
+testing verification..
+(12000, 512)
+infer time 14.35169
+[lfw][102000]XNorm: 22.254405
+[lfw][102000]Accuracy-Flip: 0.99800+-0.00267
+[102000]Accuracy-Highest: 0.99817
+INFO:root:Epoch[13] Batch [4260]	Speed: 102.97 samples/sec	acc=0.809570
+INFO:root:Epoch[13] Batch [4280]	Speed: 661.60 samples/sec	acc=0.812891
+INFO:root:Epoch[13] Batch [4300]	Speed: 662.28 samples/sec	acc=0.808691
+INFO:root:Epoch[13] Batch [4320]	Speed: 660.82 samples/sec	acc=0.810742
+INFO:root:Epoch[13] Batch [4340]	Speed: 662.71 samples/sec	acc=0.808691
+INFO:root:Epoch[13] Batch [4360]	Speed: 659.93 samples/sec	acc=0.817187
+INFO:root:Epoch[13] Batch [4380]	Speed: 636.39 samples/sec	acc=0.809766
+INFO:root:Epoch[13] Batch [4400]	Speed: 661.86 samples/sec	acc=0.810742
+INFO:root:Epoch[13] Batch [4420]	Speed: 642.09 samples/sec	acc=0.809766
+INFO:root:Epoch[13] Batch [4440]	Speed: 650.43 samples/sec	acc=0.806543
+INFO:root:Epoch[13] Batch [4460]	Speed: 644.42 samples/sec	acc=0.810937
+INFO:root:Epoch[13] Batch [4480]	Speed: 660.88 samples/sec	acc=0.800391
+INFO:root:Epoch[13] Batch [4500]	Speed: 638.75 samples/sec	acc=0.807031
+INFO:root:Epoch[13] Batch [4520]	Speed: 662.58 samples/sec	acc=0.804297
+INFO:root:Epoch[13] Batch [4540]	Speed: 662.54 samples/sec	acc=0.809961
+INFO:root:Epoch[13] Batch [4560]	Speed: 662.31 samples/sec	acc=0.808984
+INFO:root:Epoch[13] Batch [4580]	Speed: 662.20 samples/sec	acc=0.809180
+INFO:root:Epoch[13] Batch [4600]	Speed: 662.52 samples/sec	acc=0.796387
+INFO:root:Epoch[13] Batch [4620]	Speed: 657.23 samples/sec	acc=0.811621
+INFO:root:Epoch[13] Batch [4640]	Speed: 675.32 samples/sec	acc=0.804688
+INFO:root:Epoch[13] Batch [4660]	Speed: 642.32 samples/sec	acc=0.813379
+INFO:root:Epoch[13] Batch [4680]	Speed: 651.51 samples/sec	acc=0.809961
+INFO:root:Epoch[13] Batch [4700]	Speed: 612.78 samples/sec	acc=0.812207
+INFO:root:Epoch[13] Batch [4720]	Speed: 667.86 samples/sec	acc=0.814941
+INFO:root:Epoch[13] Batch [4740]	Speed: 659.61 samples/sec	acc=0.808301
+INFO:root:Epoch[13] Batch [4760]	Speed: 661.19 samples/sec	acc=0.812891
+INFO:root:Epoch[13] Batch [4780]	Speed: 662.02 samples/sec	acc=0.803125
+INFO:root:Epoch[13] Batch [4800]	Speed: 660.58 samples/sec	acc=0.806934
+INFO:root:Epoch[13] Batch [4820]	Speed: 661.61 samples/sec	acc=0.800488
+INFO:root:Epoch[13] Batch [4840]	Speed: 659.21 samples/sec	acc=0.808008
+INFO:root:Epoch[13] Batch [4860]	Speed: 657.96 samples/sec	acc=0.809668
+INFO:root:Epoch[13] Batch [4880]	Speed: 661.24 samples/sec	acc=0.816016
+INFO:root:Epoch[13] Batch [4900]	Speed: 662.58 samples/sec	acc=0.811035
+INFO:root:Epoch[13] Batch [4920]	Speed: 660.93 samples/sec	acc=0.805469
+INFO:root:Epoch[13] Batch [4940]	Speed: 662.40 samples/sec	acc=0.805176
+INFO:root:Epoch[13] Batch [4960]	Speed: 663.33 samples/sec	acc=0.817187
+INFO:root:Epoch[13] Batch [4980]	Speed: 640.34 samples/sec	acc=0.813867
+INFO:root:Epoch[13] Batch [5000]	Speed: 634.15 samples/sec	acc=0.812109
+INFO:root:Epoch[13] Batch [5020]	Speed: 662.98 samples/sec	acc=0.812598
+INFO:root:Epoch[13] Batch [5040]	Speed: 644.74 samples/sec	acc=0.812988
+INFO:root:Epoch[13] Batch [5060]	Speed: 653.28 samples/sec	acc=0.813477
+INFO:root:Epoch[13] Batch [5080]	Speed: 662.01 samples/sec	acc=0.803809
+INFO:root:Epoch[13] Batch [5100]	Speed: 661.96 samples/sec	acc=0.810156
+INFO:root:Epoch[13] Batch [5120]	Speed: 662.17 samples/sec	acc=0.808789
+INFO:root:Epoch[13] Batch [5140]	Speed: 642.05 samples/sec	acc=0.809082
+INFO:root:Epoch[13] Batch [5160]	Speed: 623.37 samples/sec	acc=0.808008
+INFO:root:Epoch[13] Batch [5180]	Speed: 665.08 samples/sec	acc=0.804199
+INFO:root:Epoch[13] Batch [5200]	Speed: 662.74 samples/sec	acc=0.814941
+INFO:root:Epoch[13] Batch [5220]	Speed: 632.47 samples/sec	acc=0.813770
+INFO:root:Epoch[13] Batch [5240]	Speed: 662.52 samples/sec	acc=0.812012
+lr-batch-epoch: 0.0001 5252 13
+INFO:root:Epoch[13] Batch [5260]	Speed: 659.63 samples/sec	acc=0.809570
+INFO:root:Epoch[13] Batch [5280]	Speed: 668.54 samples/sec	acc=0.811035
+INFO:root:Epoch[13] Batch [5300]	Speed: 633.15 samples/sec	acc=0.813477
+INFO:root:Epoch[13] Batch [5320]	Speed: 684.29 samples/sec	acc=0.801953
+INFO:root:Epoch[13] Batch [5340]	Speed: 661.54 samples/sec	acc=0.804199
+INFO:root:Epoch[13] Batch [5360]	Speed: 644.50 samples/sec	acc=0.804883
+INFO:root:Epoch[13] Batch [5380]	Speed: 662.20 samples/sec	acc=0.812305
+INFO:root:Epoch[13] Batch [5400]	Speed: 650.40 samples/sec	acc=0.813379
+INFO:root:Epoch[13] Batch [5420]	Speed: 662.35 samples/sec	acc=0.804199
+INFO:root:Epoch[13] Batch [5440]	Speed: 663.33 samples/sec	acc=0.807129
+INFO:root:Epoch[13] Batch [5460]	Speed: 662.57 samples/sec	acc=0.808496
+INFO:root:Epoch[13] Batch [5480]	Speed: 652.91 samples/sec	acc=0.806738
+INFO:root:Epoch[13] Batch [5500]	Speed: 667.66 samples/sec	acc=0.808789
+INFO:root:Epoch[13] Batch [5520]	Speed: 646.28 samples/sec	acc=0.807715
+INFO:root:Epoch[13] Batch [5540]	Speed: 663.58 samples/sec	acc=0.813867
+INFO:root:Epoch[13] Batch [5560]	Speed: 660.25 samples/sec	acc=0.810449
+INFO:root:Epoch[13] Batch [5580]	Speed: 647.36 samples/sec	acc=0.809766
+INFO:root:Epoch[13] Batch [5600]	Speed: 662.71 samples/sec	acc=0.809570
+INFO:root:Epoch[13] Batch [5620]	Speed: 655.96 samples/sec	acc=0.810059
+INFO:root:Epoch[13] Batch [5640]	Speed: 616.69 samples/sec	acc=0.811230
+INFO:root:Epoch[13] Batch [5660]	Speed: 661.15 samples/sec	acc=0.816797
+INFO:root:Epoch[13] Batch [5680]	Speed: 663.27 samples/sec	acc=0.811426
+INFO:root:Epoch[13] Batch [5700]	Speed: 661.79 samples/sec	acc=0.809668
+INFO:root:Epoch[13] Batch [5720]	Speed: 662.59 samples/sec	acc=0.808691
+INFO:root:Epoch[13] Batch [5740]	Speed: 662.07 samples/sec	acc=0.812207
+INFO:root:Epoch[13] Batch [5760]	Speed: 662.00 samples/sec	acc=0.811523
+INFO:root:Epoch[13] Batch [5780]	Speed: 662.14 samples/sec	acc=0.810449
+INFO:root:Epoch[13] Batch [5800]	Speed: 662.64 samples/sec	acc=0.811621
+INFO:root:Epoch[13] Batch [5820]	Speed: 661.78 samples/sec	acc=0.804395
+INFO:root:Epoch[13] Batch [5840]	Speed: 661.25 samples/sec	acc=0.814355
+INFO:root:Epoch[13] Batch [5860]	Speed: 676.80 samples/sec	acc=0.811035
+INFO:root:Epoch[13] Batch [5880]	Speed: 660.45 samples/sec	acc=0.810449
+INFO:root:Epoch[13] Batch [5900]	Speed: 655.29 samples/sec	acc=0.814160
+INFO:root:Epoch[13] Batch [5920]	Speed: 662.73 samples/sec	acc=0.817676
+INFO:root:Epoch[13] Batch [5940]	Speed: 661.03 samples/sec	acc=0.810254
+INFO:root:Epoch[13] Batch [5960]	Speed: 661.98 samples/sec	acc=0.813379
+INFO:root:Epoch[13] Batch [5980]	Speed: 659.42 samples/sec	acc=0.809473
+INFO:root:Epoch[13] Batch [6000]	Speed: 661.89 samples/sec	acc=0.808887
+INFO:root:Epoch[13] Batch [6020]	Speed: 661.24 samples/sec	acc=0.813574
+INFO:root:Epoch[13] Batch [6040]	Speed: 659.76 samples/sec	acc=0.805859
+INFO:root:Epoch[13] Batch [6060]	Speed: 671.72 samples/sec	acc=0.808105
+INFO:root:Epoch[13] Batch [6080]	Speed: 650.30 samples/sec	acc=0.808789
+INFO:root:Epoch[13] Batch [6100]	Speed: 653.91 samples/sec	acc=0.809277
+INFO:root:Epoch[13] Batch [6120]	Speed: 661.21 samples/sec	acc=0.814160
+INFO:root:Epoch[13] Batch [6140]	Speed: 661.81 samples/sec	acc=0.805273
+INFO:root:Epoch[13] Batch [6160]	Speed: 662.63 samples/sec	acc=0.807813
+INFO:root:Epoch[13] Batch [6180]	Speed: 662.43 samples/sec	acc=0.805078
+INFO:root:Epoch[13] Batch [6200]	Speed: 661.82 samples/sec	acc=0.819238
+INFO:root:Epoch[13] Batch [6220]	Speed: 655.09 samples/sec	acc=0.807715
+INFO:root:Epoch[13] Batch [6240]	Speed: 653.35 samples/sec	acc=0.813086
+[104000][MARGIN]0.799129,0.416642
+lr-batch-epoch: 0.0001 6252 13
+testing verification..
+(14000, 512)
+infer time 23.345104
+[cfp_ff][104000]XNorm: 21.440349
+[cfp_ff][104000]Accuracy-Flip: 0.99829+-0.00229
+testing verification..
+(14000, 512)
+infer time 26.054203
+[cfp_fp][104000]XNorm: 19.720415
+[cfp_fp][104000]Accuracy-Flip: 0.92700+-0.01354
+testing verification..
+(12000, 512)
+infer time 22.68381
+[agedb_30][104000]XNorm: 22.635694
+[agedb_30][104000]Accuracy-Flip: 0.97633+-0.00609
+testing verification..
+(12000, 512)
+infer time 21.415218
+[lfw][104000]XNorm: 22.193441
+[lfw][104000]Accuracy-Flip: 0.99733+-0.00291
+[104000]Accuracy-Highest: 0.99817
+INFO:root:Epoch[13] Batch [6260]	Speed: 78.84 samples/sec	acc=0.811621
+INFO:root:Epoch[13] Batch [6280]	Speed: 652.60 samples/sec	acc=0.810352
+INFO:root:Epoch[13] Batch [6300]	Speed: 661.39 samples/sec	acc=0.815332
+INFO:root:Epoch[13] Batch [6320]	Speed: 633.41 samples/sec	acc=0.810840
+INFO:root:Epoch[13] Batch [6340]	Speed: 662.06 samples/sec	acc=0.811035
+INFO:root:Epoch[13] Batch [6360]	Speed: 661.78 samples/sec	acc=0.811133
+INFO:root:Epoch[13] Batch [6380]	Speed: 661.74 samples/sec	acc=0.808496
+INFO:root:Epoch[13] Batch [6400]	Speed: 662.72 samples/sec	acc=0.804688
+INFO:root:Epoch[13] Batch [6420]	Speed: 661.83 samples/sec	acc=0.810352
+INFO:root:Epoch[13] Batch [6440]	Speed: 653.24 samples/sec	acc=0.808984
+INFO:root:Epoch[13] Batch [6460]	Speed: 662.01 samples/sec	acc=0.807031
+INFO:root:Epoch[13] Batch [6480]	Speed: 661.86 samples/sec	acc=0.807129
+INFO:root:Epoch[13] Batch [6500]	Speed: 661.57 samples/sec	acc=0.806348
+INFO:root:Epoch[13] Batch [6520]	Speed: 662.16 samples/sec	acc=0.811133
+INFO:root:Epoch[13] Batch [6540]	Speed: 661.37 samples/sec	acc=0.809961
+INFO:root:Epoch[13] Batch [6560]	Speed: 660.95 samples/sec	acc=0.805469
+INFO:root:Epoch[13] Batch [6580]	Speed: 662.45 samples/sec	acc=0.813672
+INFO:root:Epoch[13] Batch [6600]	Speed: 661.49 samples/sec	acc=0.810840
+INFO:root:Epoch[13] Batch [6620]	Speed: 661.70 samples/sec	acc=0.811426
+INFO:root:Epoch[13] Batch [6640]	Speed: 662.29 samples/sec	acc=0.805469
+INFO:root:Epoch[13] Batch [6660]	Speed: 661.36 samples/sec	acc=0.810449
+INFO:root:Epoch[13] Batch [6680]	Speed: 661.87 samples/sec	acc=0.812012
+INFO:root:Epoch[13] Batch [6700]	Speed: 662.44 samples/sec	acc=0.809277
+INFO:root:Epoch[13] Batch [6720]	Speed: 661.95 samples/sec	acc=0.805664
+INFO:root:Epoch[13] Batch [6740]	Speed: 662.32 samples/sec	acc=0.807715
+INFO:root:Epoch[13] Batch [6760]	Speed: 662.83 samples/sec	acc=0.807617
+INFO:root:Epoch[13] Batch [6780]	Speed: 667.13 samples/sec	acc=0.806934
+INFO:root:Epoch[13] Batch [6800]	Speed: 661.54 samples/sec	acc=0.806348
+INFO:root:Epoch[13] Batch [6820]	Speed: 644.86 samples/sec	acc=0.807910
+INFO:root:Epoch[13] Batch [6840]	Speed: 653.33 samples/sec	acc=0.807910
+INFO:root:Epoch[13] Batch [6860]	Speed: 654.18 samples/sec	acc=0.811035
+INFO:root:Epoch[13] Batch [6880]	Speed: 646.49 samples/sec	acc=0.811914
+INFO:root:Epoch[13] Batch [6900]	Speed: 650.03 samples/sec	acc=0.806641
+INFO:root:Epoch[13] Batch [6920]	Speed: 661.85 samples/sec	acc=0.811035
+INFO:root:Epoch[13] Batch [6940]	Speed: 646.81 samples/sec	acc=0.807520
+INFO:root:Epoch[13] Batch [6960]	Speed: 661.08 samples/sec	acc=0.812109
+INFO:root:Epoch[13] Batch [6980]	Speed: 661.77 samples/sec	acc=0.804297
+INFO:root:Epoch[13] Batch [7000]	Speed: 662.08 samples/sec	acc=0.809766
+INFO:root:Epoch[13] Batch [7020]	Speed: 662.30 samples/sec	acc=0.816113
+INFO:root:Epoch[13] Batch [7040]	Speed: 662.26 samples/sec	acc=0.809473
+INFO:root:Epoch[13] Batch [7060]	Speed: 662.05 samples/sec	acc=0.814258
+INFO:root:Epoch[13] Batch [7080]	Speed: 661.05 samples/sec	acc=0.812598
+INFO:root:Epoch[13] Batch [7100]	Speed: 662.43 samples/sec	acc=0.814258
+INFO:root:Epoch[13] Batch [7120]	Speed: 648.44 samples/sec	acc=0.812598
+INFO:root:Epoch[13] Batch [7140]	Speed: 637.28 samples/sec	acc=0.812695
+INFO:root:Epoch[13] Batch [7160]	Speed: 663.33 samples/sec	acc=0.805957
+INFO:root:Epoch[13] Batch [7180]	Speed: 660.09 samples/sec	acc=0.810840
+INFO:root:Epoch[13] Batch [7200]	Speed: 655.69 samples/sec	acc=0.811230
+INFO:root:Epoch[13] Batch [7220]	Speed: 662.98 samples/sec	acc=0.803809
+INFO:root:Epoch[13] Batch [7240]	Speed: 662.36 samples/sec	acc=0.814355
+lr-batch-epoch: 0.0001 7252 13
+INFO:root:Epoch[13] Batch [7260]	Speed: 661.05 samples/sec	acc=0.813086
+INFO:root:Epoch[13] Batch [7280]	Speed: 662.32 samples/sec	acc=0.811816
+INFO:root:Epoch[13] Batch [7300]	Speed: 661.94 samples/sec	acc=0.818652
+INFO:root:Epoch[13] Batch [7320]	Speed: 650.28 samples/sec	acc=0.804492
+INFO:root:Epoch[13] Batch [7340]	Speed: 663.79 samples/sec	acc=0.811621
+INFO:root:Epoch[13] Batch [7360]	Speed: 662.90 samples/sec	acc=0.812305
+INFO:root:Epoch[13] Batch [7380]	Speed: 660.87 samples/sec	acc=0.810059
+INFO:root:Epoch[13] Batch [7400]	Speed: 657.30 samples/sec	acc=0.812500
+INFO:root:Epoch[13] Batch [7420]	Speed: 642.06 samples/sec	acc=0.809961
+INFO:root:Epoch[13] Batch [7440]	Speed: 659.70 samples/sec	acc=0.811914
+INFO:root:Epoch[13] Batch [7460]	Speed: 642.02 samples/sec	acc=0.815918
+INFO:root:Epoch[13] Batch [7480]	Speed: 662.81 samples/sec	acc=0.810645
+INFO:root:Epoch[13] Batch [7500]	Speed: 661.44 samples/sec	acc=0.815039
+INFO:root:Epoch[13] Train-acc=0.818576
+INFO:root:Epoch[13] Time cost=6294.678
+call reset()
+INFO:root:Epoch[14] Batch [20]	Speed: 660.12 samples/sec	acc=0.821987
+INFO:root:Epoch[14] Batch [40]	Speed: 662.86 samples/sec	acc=0.828125
+INFO:root:Epoch[14] Batch [60]	Speed: 660.25 samples/sec	acc=0.825098
+INFO:root:Epoch[14] Batch [80]	Speed: 656.02 samples/sec	acc=0.825000
+INFO:root:Epoch[14] Batch [100]	Speed: 663.50 samples/sec	acc=0.825586
+INFO:root:Epoch[14] Batch [120]	Speed: 661.59 samples/sec	acc=0.832227
+INFO:root:Epoch[14] Batch [140]	Speed: 659.53 samples/sec	acc=0.823242
+INFO:root:Epoch[14] Batch [160]	Speed: 639.93 samples/sec	acc=0.825293
+INFO:root:Epoch[14] Batch [180]	Speed: 662.50 samples/sec	acc=0.822559
+INFO:root:Epoch[14] Batch [200]	Speed: 661.43 samples/sec	acc=0.830078
+INFO:root:Epoch[14] Batch [220]	Speed: 662.20 samples/sec	acc=0.830566
+INFO:root:Epoch[14] Batch [240]	Speed: 662.20 samples/sec	acc=0.824414
+INFO:root:Epoch[14] Batch [260]	Speed: 647.92 samples/sec	acc=0.823047
+INFO:root:Epoch[14] Batch [280]	Speed: 651.16 samples/sec	acc=0.826953
+INFO:root:Epoch[14] Batch [300]	Speed: 661.93 samples/sec	acc=0.819824
+INFO:root:Epoch[14] Batch [320]	Speed: 660.79 samples/sec	acc=0.832129
+INFO:root:Epoch[14] Batch [340]	Speed: 661.94 samples/sec	acc=0.833594
+INFO:root:Epoch[14] Batch [360]	Speed: 662.32 samples/sec	acc=0.829688
+INFO:root:Epoch[14] Batch [380]	Speed: 661.29 samples/sec	acc=0.826465
+INFO:root:Epoch[14] Batch [400]	Speed: 645.40 samples/sec	acc=0.826758
+INFO:root:Epoch[14] Batch [420]	Speed: 662.83 samples/sec	acc=0.832617
+INFO:root:Epoch[14] Batch [440]	Speed: 661.14 samples/sec	acc=0.829395
+INFO:root:Epoch[14] Batch [460]	Speed: 661.97 samples/sec	acc=0.830273
+INFO:root:Epoch[14] Batch [480]	Speed: 661.30 samples/sec	acc=0.830469
+INFO:root:Epoch[14] Batch [500]	Speed: 661.24 samples/sec	acc=0.824414
+INFO:root:Epoch[14] Batch [520]	Speed: 661.30 samples/sec	acc=0.828516
+INFO:root:Epoch[14] Batch [540]	Speed: 670.07 samples/sec	acc=0.826562
+INFO:root:Epoch[14] Batch [560]	Speed: 673.61 samples/sec	acc=0.824316
+INFO:root:Epoch[14] Batch [580]	Speed: 677.93 samples/sec	acc=0.824121
+INFO:root:Epoch[14] Batch [600]	Speed: 662.14 samples/sec	acc=0.821094
+INFO:root:Epoch[14] Batch [620]	Speed: 661.74 samples/sec	acc=0.829590
+INFO:root:Epoch[14] Batch [640]	Speed: 643.90 samples/sec	acc=0.829785
+INFO:root:Epoch[14] Batch [660]	Speed: 671.88 samples/sec	acc=0.837012
+INFO:root:Epoch[14] Batch [680]	Speed: 644.04 samples/sec	acc=0.825000
+INFO:root:Epoch[14] Batch [700]	Speed: 661.87 samples/sec	acc=0.820703
+INFO:root:Epoch[14] Batch [720]	Speed: 662.50 samples/sec	acc=0.826660
+[106000][MARGIN]0.802184,0.420316
+lr-batch-epoch: 0.0001 733 14
+testing verification..
+(14000, 512)
+infer time 21.774034
+[cfp_ff][106000]XNorm: 21.429996
+[cfp_ff][106000]Accuracy-Flip: 0.99814+-0.00231
+testing verification..
+(14000, 512)
+infer time 22.95958
+[cfp_fp][106000]XNorm: 19.773630
+[cfp_fp][106000]Accuracy-Flip: 0.92686+-0.01443
+testing verification..
+(12000, 512)
+infer time 19.419289
+[agedb_30][106000]XNorm: 22.650390
+[agedb_30][106000]Accuracy-Flip: 0.97817+-0.00677
+testing verification..
+(12000, 512)
+infer time 19.988042
+[lfw][106000]XNorm: 22.188714
+[lfw][106000]Accuracy-Flip: 0.99800+-0.00267
+[106000]Accuracy-Highest: 0.99817
+INFO:root:Epoch[14] Batch [740]	Speed: 85.99 samples/sec	acc=0.830664
+INFO:root:Epoch[14] Batch [760]	Speed: 662.19 samples/sec	acc=0.822949
+INFO:root:Epoch[14] Batch [780]	Speed: 662.82 samples/sec	acc=0.821973
+INFO:root:Epoch[14] Batch [800]	Speed: 642.73 samples/sec	acc=0.821484
+INFO:root:Epoch[14] Batch [820]	Speed: 663.08 samples/sec	acc=0.824609
+INFO:root:Epoch[14] Batch [840]	Speed: 677.05 samples/sec	acc=0.825000
+INFO:root:Epoch[14] Batch [860]	Speed: 661.38 samples/sec	acc=0.827246
+INFO:root:Epoch[14] Batch [880]	Speed: 662.02 samples/sec	acc=0.827246
+INFO:root:Epoch[14] Batch [900]	Speed: 646.12 samples/sec	acc=0.826855
+INFO:root:Epoch[14] Batch [920]	Speed: 636.14 samples/sec	acc=0.821289
+INFO:root:Epoch[14] Batch [940]	Speed: 653.08 samples/sec	acc=0.826953
+INFO:root:Epoch[14] Batch [960]	Speed: 661.49 samples/sec	acc=0.821973
+INFO:root:Epoch[14] Batch [980]	Speed: 661.10 samples/sec	acc=0.823926
+INFO:root:Epoch[14] Batch [1000]	Speed: 660.94 samples/sec	acc=0.828516
+INFO:root:Epoch[14] Batch [1020]	Speed: 661.42 samples/sec	acc=0.825391
+INFO:root:Epoch[14] Batch [1040]	Speed: 649.06 samples/sec	acc=0.825781
+INFO:root:Epoch[14] Batch [1060]	Speed: 661.82 samples/sec	acc=0.831055
+INFO:root:Epoch[14] Batch [1080]	Speed: 661.00 samples/sec	acc=0.828320
+INFO:root:Epoch[14] Batch [1100]	Speed: 661.29 samples/sec	acc=0.830176
+INFO:root:Epoch[14] Batch [1120]	Speed: 662.42 samples/sec	acc=0.825098
+INFO:root:Epoch[14] Batch [1140]	Speed: 662.24 samples/sec	acc=0.821973
+INFO:root:Epoch[14] Batch [1160]	Speed: 653.30 samples/sec	acc=0.822266
+INFO:root:Epoch[14] Batch [1180]	Speed: 661.64 samples/sec	acc=0.833105
+INFO:root:Epoch[14] Batch [1200]	Speed: 661.79 samples/sec	acc=0.825000
+INFO:root:Epoch[14] Batch [1220]	Speed: 639.05 samples/sec	acc=0.827734
+INFO:root:Epoch[14] Batch [1240]	Speed: 660.96 samples/sec	acc=0.824609
+INFO:root:Epoch[14] Batch [1260]	Speed: 662.59 samples/sec	acc=0.816113
+INFO:root:Epoch[14] Batch [1280]	Speed: 660.22 samples/sec	acc=0.824609
+INFO:root:Epoch[14] Batch [1300]	Speed: 662.29 samples/sec	acc=0.822852
+INFO:root:Epoch[14] Batch [1320]	Speed: 662.85 samples/sec	acc=0.830859
+INFO:root:Epoch[14] Batch [1340]	Speed: 661.75 samples/sec	acc=0.824414
+INFO:root:Epoch[14] Batch [1360]	Speed: 662.31 samples/sec	acc=0.825488
+INFO:root:Epoch[14] Batch [1380]	Speed: 662.77 samples/sec	acc=0.824707
+INFO:root:Epoch[14] Batch [1400]	Speed: 660.20 samples/sec	acc=0.826270
+INFO:root:Epoch[14] Batch [1420]	Speed: 661.79 samples/sec	acc=0.824707
+INFO:root:Epoch[14] Batch [1440]	Speed: 661.72 samples/sec	acc=0.825977
+INFO:root:Epoch[14] Batch [1460]	Speed: 661.22 samples/sec	acc=0.826758
+INFO:root:Epoch[14] Batch [1480]	Speed: 661.20 samples/sec	acc=0.829785
+INFO:root:Epoch[14] Batch [1500]	Speed: 661.27 samples/sec	acc=0.835156
+INFO:root:Epoch[14] Batch [1520]	Speed: 661.80 samples/sec	acc=0.823145
+INFO:root:Epoch[14] Batch [1540]	Speed: 661.56 samples/sec	acc=0.820215
+INFO:root:Epoch[14] Batch [1560]	Speed: 661.70 samples/sec	acc=0.827832
+INFO:root:Epoch[14] Batch [1580]	Speed: 660.74 samples/sec	acc=0.824121
+INFO:root:Epoch[14] Batch [1600]	Speed: 661.24 samples/sec	acc=0.824414
+INFO:root:Epoch[14] Batch [1620]	Speed: 662.76 samples/sec	acc=0.828418
+INFO:root:Epoch[14] Batch [1640]	Speed: 660.93 samples/sec	acc=0.825879
+INFO:root:Epoch[14] Batch [1660]	Speed: 658.56 samples/sec	acc=0.821582
+INFO:root:Epoch[14] Batch [1680]	Speed: 662.03 samples/sec	acc=0.833594
+INFO:root:Epoch[14] Batch [1700]	Speed: 660.66 samples/sec	acc=0.831836
+INFO:root:Epoch[14] Batch [1720]	Speed: 662.66 samples/sec	acc=0.819922
+lr-batch-epoch: 0.0001 1733 14
+INFO:root:Epoch[14] Batch [1740]	Speed: 662.22 samples/sec	acc=0.830176
+INFO:root:Epoch[14] Batch [1760]	Speed: 657.78 samples/sec	acc=0.823438
+INFO:root:Epoch[14] Batch [1780]	Speed: 665.25 samples/sec	acc=0.824902
+INFO:root:Epoch[14] Batch [1800]	Speed: 661.40 samples/sec	acc=0.826855
+INFO:root:Epoch[14] Batch [1820]	Speed: 661.17 samples/sec	acc=0.820410
+INFO:root:Epoch[14] Batch [1840]	Speed: 662.40 samples/sec	acc=0.829980
+INFO:root:Epoch[14] Batch [1860]	Speed: 650.31 samples/sec	acc=0.828418
+INFO:root:Epoch[14] Batch [1880]	Speed: 658.06 samples/sec	acc=0.822754
+INFO:root:Epoch[14] Batch [1900]	Speed: 661.98 samples/sec	acc=0.828125
+INFO:root:Epoch[14] Batch [1920]	Speed: 649.79 samples/sec	acc=0.826758
+INFO:root:Epoch[14] Batch [1940]	Speed: 671.72 samples/sec	acc=0.825195
+INFO:root:Epoch[14] Batch [1960]	Speed: 661.99 samples/sec	acc=0.827246
+INFO:root:Epoch[14] Batch [1980]	Speed: 662.11 samples/sec	acc=0.821094
+INFO:root:Epoch[14] Batch [2000]	Speed: 661.49 samples/sec	acc=0.825098
+INFO:root:Epoch[14] Batch [2020]	Speed: 661.18 samples/sec	acc=0.821484
+INFO:root:Epoch[14] Batch [2040]	Speed: 658.83 samples/sec	acc=0.822559
+INFO:root:Epoch[14] Batch [2060]	Speed: 660.67 samples/sec	acc=0.823633
+INFO:root:Epoch[14] Batch [2080]	Speed: 662.51 samples/sec	acc=0.821191
+INFO:root:Epoch[14] Batch [2100]	Speed: 662.02 samples/sec	acc=0.835840
+INFO:root:Epoch[14] Batch [2120]	Speed: 661.42 samples/sec	acc=0.827246
+INFO:root:Epoch[14] Batch [2140]	Speed: 661.52 samples/sec	acc=0.826562
+INFO:root:Epoch[14] Batch [2160]	Speed: 661.76 samples/sec	acc=0.824805
+INFO:root:Epoch[14] Batch [2180]	Speed: 661.83 samples/sec	acc=0.822363
+INFO:root:Epoch[14] Batch [2200]	Speed: 661.76 samples/sec	acc=0.825293
+INFO:root:Epoch[14] Batch [2220]	Speed: 661.52 samples/sec	acc=0.829395
+INFO:root:Epoch[14] Batch [2240]	Speed: 654.39 samples/sec	acc=0.828223
+INFO:root:Epoch[14] Batch [2260]	Speed: 662.36 samples/sec	acc=0.828027
+INFO:root:Epoch[14] Batch [2280]	Speed: 661.45 samples/sec	acc=0.822070
+INFO:root:Epoch[14] Batch [2300]	Speed: 660.74 samples/sec	acc=0.826660
+INFO:root:Epoch[14] Batch [2320]	Speed: 661.76 samples/sec	acc=0.823730
+INFO:root:Epoch[14] Batch [2340]	Speed: 662.79 samples/sec	acc=0.827148
+INFO:root:Epoch[14] Batch [2360]	Speed: 659.67 samples/sec	acc=0.822656
+INFO:root:Epoch[14] Batch [2380]	Speed: 665.46 samples/sec	acc=0.832324
+INFO:root:Epoch[14] Batch [2400]	Speed: 656.47 samples/sec	acc=0.821191
+INFO:root:Epoch[14] Batch [2420]	Speed: 661.48 samples/sec	acc=0.824316
+INFO:root:Epoch[14] Batch [2440]	Speed: 657.61 samples/sec	acc=0.821680
+INFO:root:Epoch[14] Batch [2460]	Speed: 662.31 samples/sec	acc=0.831055
+INFO:root:Epoch[14] Batch [2480]	Speed: 661.65 samples/sec	acc=0.815820
+INFO:root:Epoch[14] Batch [2500]	Speed: 660.07 samples/sec	acc=0.825391
+INFO:root:Epoch[14] Batch [2520]	Speed: 662.18 samples/sec	acc=0.822070
+INFO:root:Epoch[14] Batch [2540]	Speed: 661.96 samples/sec	acc=0.826270
+INFO:root:Epoch[14] Batch [2560]	Speed: 661.68 samples/sec	acc=0.826074
+INFO:root:Epoch[14] Batch [2580]	Speed: 661.56 samples/sec	acc=0.832910
+INFO:root:Epoch[14] Batch [2600]	Speed: 660.74 samples/sec	acc=0.829395
+INFO:root:Epoch[14] Batch [2620]	Speed: 662.02 samples/sec	acc=0.827344
+INFO:root:Epoch[14] Batch [2640]	Speed: 662.85 samples/sec	acc=0.826367
+INFO:root:Epoch[14] Batch [2660]	Speed: 662.04 samples/sec	acc=0.827051
+INFO:root:Epoch[14] Batch [2680]	Speed: 661.47 samples/sec	acc=0.825781
+INFO:root:Epoch[14] Batch [2700]	Speed: 662.31 samples/sec	acc=0.830957
+INFO:root:Epoch[14] Batch [2720]	Speed: 661.83 samples/sec	acc=0.830566
+[108000][MARGIN]0.801206,0.419432
+lr-batch-epoch: 0.0001 2733 14
+testing verification..
+(14000, 512)
+infer time 23.730797
+[cfp_ff][108000]XNorm: 21.432662
+[cfp_ff][108000]Accuracy-Flip: 0.99814+-0.00222
+testing verification..
+(14000, 512)
+infer time 23.797014
+[cfp_fp][108000]XNorm: 19.782774
+[cfp_fp][108000]Accuracy-Flip: 0.92800+-0.01416
+testing verification..
+(12000, 512)
+infer time 21.415407
+[agedb_30][108000]XNorm: 22.670755
+[agedb_30][108000]Accuracy-Flip: 0.97733+-0.00684
+testing verification..
+(12000, 512)
+infer time 18.815919
+[lfw][108000]XNorm: 22.223260
+[lfw][108000]Accuracy-Flip: 0.99767+-0.00281
+[108000]Accuracy-Highest: 0.99817
+INFO:root:Epoch[14] Batch [2740]	Speed: 82.77 samples/sec	acc=0.829395
+INFO:root:Epoch[14] Batch [2760]	Speed: 662.12 samples/sec	acc=0.826172
+INFO:root:Epoch[14] Batch [2780]	Speed: 661.64 samples/sec	acc=0.824512
+INFO:root:Epoch[14] Batch [2800]	Speed: 661.87 samples/sec	acc=0.823633
+INFO:root:Epoch[14] Batch [2820]	Speed: 662.40 samples/sec	acc=0.825293
+INFO:root:Epoch[14] Batch [2840]	Speed: 662.03 samples/sec	acc=0.825684
+INFO:root:Epoch[14] Batch [2860]	Speed: 661.77 samples/sec	acc=0.829785
+INFO:root:Epoch[14] Batch [2880]	Speed: 662.46 samples/sec	acc=0.822656
+INFO:root:Epoch[14] Batch [2900]	Speed: 661.64 samples/sec	acc=0.830273
+INFO:root:Epoch[14] Batch [2920]	Speed: 662.56 samples/sec	acc=0.822754
+INFO:root:Epoch[14] Batch [2940]	Speed: 662.49 samples/sec	acc=0.822266
+INFO:root:Epoch[14] Batch [2960]	Speed: 661.77 samples/sec	acc=0.827246
+INFO:root:Epoch[14] Batch [2980]	Speed: 661.46 samples/sec	acc=0.818164
+INFO:root:Epoch[14] Batch [3000]	Speed: 662.43 samples/sec	acc=0.829102
+INFO:root:Epoch[14] Batch [3020]	Speed: 662.06 samples/sec	acc=0.824121
+INFO:root:Epoch[14] Batch [3040]	Speed: 662.07 samples/sec	acc=0.834082
+INFO:root:Epoch[14] Batch [3060]	Speed: 656.21 samples/sec	acc=0.827246
+INFO:root:Epoch[14] Batch [3080]	Speed: 665.62 samples/sec	acc=0.826562
+INFO:root:Epoch[14] Batch [3100]	Speed: 669.57 samples/sec	acc=0.820801
+INFO:root:Epoch[14] Batch [3120]	Speed: 662.90 samples/sec	acc=0.829395
+INFO:root:Epoch[14] Batch [3140]	Speed: 661.12 samples/sec	acc=0.832520
+INFO:root:Epoch[14] Batch [3160]	Speed: 661.70 samples/sec	acc=0.829883
+INFO:root:Epoch[14] Batch [3180]	Speed: 658.36 samples/sec	acc=0.820801
+INFO:root:Epoch[14] Batch [3200]	Speed: 661.11 samples/sec	acc=0.832422
+INFO:root:Epoch[14] Batch [3220]	Speed: 662.54 samples/sec	acc=0.825000
+INFO:root:Epoch[14] Batch [3240]	Speed: 661.36 samples/sec	acc=0.830469
+INFO:root:Epoch[14] Batch [3260]	Speed: 662.41 samples/sec	acc=0.825000
+INFO:root:Epoch[14] Batch [3280]	Speed: 662.02 samples/sec	acc=0.831152
+INFO:root:Epoch[14] Batch [3300]	Speed: 661.66 samples/sec	acc=0.819824
+INFO:root:Epoch[14] Batch [3320]	Speed: 639.28 samples/sec	acc=0.821875
+INFO:root:Epoch[14] Batch [3340]	Speed: 661.48 samples/sec	acc=0.821777
+INFO:root:Epoch[14] Batch [3360]	Speed: 653.23 samples/sec	acc=0.826074
+INFO:root:Epoch[14] Batch [3380]	Speed: 662.14 samples/sec	acc=0.824902
+INFO:root:Epoch[14] Batch [3400]	Speed: 660.75 samples/sec	acc=0.829883
+INFO:root:Epoch[14] Batch [3420]	Speed: 661.99 samples/sec	acc=0.828125
+INFO:root:Epoch[14] Batch [3440]	Speed: 661.46 samples/sec	acc=0.827441
+INFO:root:Epoch[14] Batch [3460]	Speed: 661.62 samples/sec	acc=0.828223
+INFO:root:Epoch[14] Batch [3480]	Speed: 662.35 samples/sec	acc=0.830273
+INFO:root:Epoch[14] Batch [3500]	Speed: 661.64 samples/sec	acc=0.827539
+INFO:root:Epoch[14] Batch [3520]	Speed: 658.45 samples/sec	acc=0.819531
+INFO:root:Epoch[14] Batch [3540]	Speed: 662.64 samples/sec	acc=0.819629
+INFO:root:Epoch[14] Batch [3560]	Speed: 661.48 samples/sec	acc=0.829688
+INFO:root:Epoch[14] Batch [3580]	Speed: 658.42 samples/sec	acc=0.825879
+INFO:root:Epoch[14] Batch [3600]	Speed: 662.84 samples/sec	acc=0.822852
+INFO:root:Epoch[14] Batch [3620]	Speed: 661.61 samples/sec	acc=0.822266
+INFO:root:Epoch[14] Batch [3640]	Speed: 662.68 samples/sec	acc=0.821191
+INFO:root:Epoch[14] Batch [3660]	Speed: 662.34 samples/sec	acc=0.824805
+INFO:root:Epoch[14] Batch [3680]	Speed: 660.37 samples/sec	acc=0.822168
+INFO:root:Epoch[14] Batch [3700]	Speed: 661.91 samples/sec	acc=0.824707
+INFO:root:Epoch[14] Batch [3720]	Speed: 662.31 samples/sec	acc=0.828613
+lr-batch-epoch: 0.0001 3733 14
+INFO:root:Epoch[14] Batch [3740]	Speed: 661.37 samples/sec	acc=0.825195
+INFO:root:Epoch[14] Batch [3760]	Speed: 662.44 samples/sec	acc=0.822656
+INFO:root:Epoch[14] Batch [3780]	Speed: 662.46 samples/sec	acc=0.830371
+INFO:root:Epoch[14] Batch [3800]	Speed: 661.51 samples/sec	acc=0.830664
+INFO:root:Epoch[14] Batch [3820]	Speed: 648.82 samples/sec	acc=0.826172
+INFO:root:Epoch[14] Batch [3840]	Speed: 661.87 samples/sec	acc=0.824316
+INFO:root:Epoch[14] Batch [3860]	Speed: 662.37 samples/sec	acc=0.828711
+INFO:root:Epoch[14] Batch [3880]	Speed: 662.14 samples/sec	acc=0.813281
+INFO:root:Epoch[14] Batch [3900]	Speed: 663.17 samples/sec	acc=0.822754
+INFO:root:Epoch[14] Batch [3920]	Speed: 659.26 samples/sec	acc=0.823340
+INFO:root:Epoch[14] Batch [3940]	Speed: 674.53 samples/sec	acc=0.824121
+INFO:root:Epoch[14] Batch [3960]	Speed: 662.77 samples/sec	acc=0.827051
+INFO:root:Epoch[14] Batch [3980]	Speed: 656.89 samples/sec	acc=0.827832
+INFO:root:Epoch[14] Batch [4000]	Speed: 661.75 samples/sec	acc=0.823242
+INFO:root:Epoch[14] Batch [4020]	Speed: 662.66 samples/sec	acc=0.828125
+INFO:root:Epoch[14] Batch [4040]	Speed: 661.36 samples/sec	acc=0.820605
+INFO:root:Epoch[14] Batch [4060]	Speed: 662.78 samples/sec	acc=0.822363
+INFO:root:Epoch[14] Batch [4080]	Speed: 662.15 samples/sec	acc=0.830566
+INFO:root:Epoch[14] Batch [4100]	Speed: 654.88 samples/sec	acc=0.825391
+INFO:root:Epoch[14] Batch [4120]	Speed: 658.99 samples/sec	acc=0.827637
+INFO:root:Epoch[14] Batch [4140]	Speed: 662.43 samples/sec	acc=0.828613
+INFO:root:Epoch[14] Batch [4160]	Speed: 660.42 samples/sec	acc=0.826855
+INFO:root:Epoch[14] Batch [4180]	Speed: 661.61 samples/sec	acc=0.820996
+INFO:root:Epoch[14] Batch [4200]	Speed: 662.48 samples/sec	acc=0.831152
+INFO:root:Epoch[14] Batch [4220]	Speed: 652.26 samples/sec	acc=0.817773
+INFO:root:Epoch[14] Batch [4240]	Speed: 662.03 samples/sec	acc=0.822656
+INFO:root:Epoch[14] Batch [4260]	Speed: 662.47 samples/sec	acc=0.825293
+INFO:root:Epoch[14] Batch [4280]	Speed: 660.87 samples/sec	acc=0.826562
+INFO:root:Epoch[14] Batch [4300]	Speed: 656.11 samples/sec	acc=0.828418
+INFO:root:Epoch[14] Batch [4320]	Speed: 660.25 samples/sec	acc=0.827148
+INFO:root:Epoch[14] Batch [4340]	Speed: 657.03 samples/sec	acc=0.828125
+INFO:root:Epoch[14] Batch [4360]	Speed: 661.99 samples/sec	acc=0.828125
+INFO:root:Epoch[14] Batch [4380]	Speed: 662.87 samples/sec	acc=0.828320
+INFO:root:Epoch[14] Batch [4400]	Speed: 661.96 samples/sec	acc=0.832227
+INFO:root:Epoch[14] Batch [4420]	Speed: 661.72 samples/sec	acc=0.824512
+INFO:root:Epoch[14] Batch [4440]	Speed: 662.68 samples/sec	acc=0.830762
+INFO:root:Epoch[14] Batch [4460]	Speed: 661.46 samples/sec	acc=0.833203
+INFO:root:Epoch[14] Batch [4480]	Speed: 661.01 samples/sec	acc=0.827148
+INFO:root:Epoch[14] Batch [4500]	Speed: 661.29 samples/sec	acc=0.830566
+INFO:root:Epoch[14] Batch [4520]	Speed: 660.56 samples/sec	acc=0.826367
+INFO:root:Epoch[14] Batch [4540]	Speed: 662.60 samples/sec	acc=0.826660
+INFO:root:Epoch[14] Batch [4560]	Speed: 661.86 samples/sec	acc=0.821973
+INFO:root:Epoch[14] Batch [4580]	Speed: 661.89 samples/sec	acc=0.827051
+INFO:root:Epoch[14] Batch [4600]	Speed: 660.37 samples/sec	acc=0.824707
+INFO:root:Epoch[14] Batch [4620]	Speed: 635.72 samples/sec	acc=0.825586
+INFO:root:Epoch[14] Batch [4640]	Speed: 659.14 samples/sec	acc=0.825293
+INFO:root:Epoch[14] Batch [4660]	Speed: 662.39 samples/sec	acc=0.833008
+INFO:root:Epoch[14] Batch [4680]	Speed: 662.01 samples/sec	acc=0.819141
+INFO:root:Epoch[14] Batch [4700]	Speed: 660.20 samples/sec	acc=0.827637
+INFO:root:Epoch[14] Batch [4720]	Speed: 662.65 samples/sec	acc=0.827441
+[110000][MARGIN]0.798437,0.416107
+lr-batch-epoch: 0.0001 4733 14
+testing verification..
+(14000, 512)
+infer time 24.934157
+[cfp_ff][110000]XNorm: 21.432704
+[cfp_ff][110000]Accuracy-Flip: 0.99814+-0.00222
+testing verification..
+(14000, 512)
+infer time 27.715869
+[cfp_fp][110000]XNorm: 19.793554
+[cfp_fp][110000]Accuracy-Flip: 0.92643+-0.01393
+testing verification..
+(12000, 512)
+infer time 20.559592
+[agedb_30][110000]XNorm: 22.695860
+[agedb_30][110000]Accuracy-Flip: 0.97817+-0.00660
+testing verification..
+(12000, 512)
+infer time 20.800936
+[lfw][110000]XNorm: 22.211984
+[lfw][110000]Accuracy-Flip: 0.99800+-0.00267
+[110000]Accuracy-Highest: 0.99817
+INFO:root:Epoch[14] Batch [4740]	Speed: 79.01 samples/sec	acc=0.826953
+INFO:root:Epoch[14] Batch [4760]	Speed: 677.98 samples/sec	acc=0.823145
+INFO:root:Epoch[14] Batch [4780]	Speed: 662.13 samples/sec	acc=0.829590
+INFO:root:Epoch[14] Batch [4800]	Speed: 658.48 samples/sec	acc=0.828711
+INFO:root:Epoch[14] Batch [4820]	Speed: 662.27 samples/sec	acc=0.831152
+INFO:root:Epoch[14] Batch [4840]	Speed: 662.42 samples/sec	acc=0.819238
+INFO:root:Epoch[14] Batch [4860]	Speed: 662.90 samples/sec	acc=0.826074
+INFO:root:Epoch[14] Batch [4880]	Speed: 661.51 samples/sec	acc=0.822852
+INFO:root:Epoch[14] Batch [4900]	Speed: 662.17 samples/sec	acc=0.818555
+INFO:root:Epoch[14] Batch [4920]	Speed: 662.37 samples/sec	acc=0.829297
+INFO:root:Epoch[14] Batch [4940]	Speed: 666.07 samples/sec	acc=0.825781
+INFO:root:Epoch[14] Batch [4960]	Speed: 658.13 samples/sec	acc=0.821680
+INFO:root:Epoch[14] Batch [4980]	Speed: 662.73 samples/sec	acc=0.831836
+INFO:root:Epoch[14] Batch [5000]	Speed: 661.86 samples/sec	acc=0.823340
+INFO:root:Epoch[14] Batch [5020]	Speed: 661.95 samples/sec	acc=0.831836
+INFO:root:Epoch[14] Batch [5040]	Speed: 662.04 samples/sec	acc=0.827832
+INFO:root:Epoch[14] Batch [5060]	Speed: 661.67 samples/sec	acc=0.825879
+INFO:root:Epoch[14] Batch [5080]	Speed: 662.13 samples/sec	acc=0.826074
+INFO:root:Epoch[14] Batch [5100]	Speed: 662.56 samples/sec	acc=0.824023
+INFO:root:Epoch[14] Batch [5120]	Speed: 661.20 samples/sec	acc=0.833008
+INFO:root:Epoch[14] Batch [5140]	Speed: 662.29 samples/sec	acc=0.825098
+INFO:root:Epoch[14] Batch [5160]	Speed: 662.65 samples/sec	acc=0.828711
+INFO:root:Epoch[14] Batch [5180]	Speed: 662.05 samples/sec	acc=0.829395
+INFO:root:Epoch[14] Batch [5200]	Speed: 661.62 samples/sec	acc=0.823438
+INFO:root:Epoch[14] Batch [5220]	Speed: 653.48 samples/sec	acc=0.833496
+INFO:root:Epoch[14] Batch [5240]	Speed: 661.86 samples/sec	acc=0.826465
+INFO:root:Epoch[14] Batch [5260]	Speed: 662.07 samples/sec	acc=0.825098
+INFO:root:Epoch[14] Batch [5280]	Speed: 661.21 samples/sec	acc=0.832129
+INFO:root:Epoch[14] Batch [5300]	Speed: 655.14 samples/sec	acc=0.822266
+INFO:root:Epoch[14] Batch [5320]	Speed: 662.01 samples/sec	acc=0.827051
+INFO:root:Epoch[14] Batch [5340]	Speed: 662.20 samples/sec	acc=0.831445
+INFO:root:Epoch[14] Batch [5360]	Speed: 660.98 samples/sec	acc=0.829297
+INFO:root:Epoch[14] Batch [5380]	Speed: 662.53 samples/sec	acc=0.822754
+INFO:root:Epoch[14] Batch [5400]	Speed: 662.88 samples/sec	acc=0.828613
+INFO:root:Epoch[14] Batch [5420]	Speed: 661.42 samples/sec	acc=0.831738
+INFO:root:Epoch[14] Batch [5440]	Speed: 662.12 samples/sec	acc=0.824902
+INFO:root:Epoch[14] Batch [5460]	Speed: 663.33 samples/sec	acc=0.824023
+INFO:root:Epoch[14] Batch [5480]	Speed: 661.12 samples/sec	acc=0.826660
+INFO:root:Epoch[14] Batch [5500]	Speed: 662.12 samples/sec	acc=0.825977
+INFO:root:Epoch[14] Batch [5520]	Speed: 661.94 samples/sec	acc=0.824512
+INFO:root:Epoch[14] Batch [5540]	Speed: 661.93 samples/sec	acc=0.825879
+INFO:root:Epoch[14] Batch [5560]	Speed: 658.71 samples/sec	acc=0.822168
+INFO:root:Epoch[14] Batch [5580]	Speed: 666.17 samples/sec	acc=0.829199
+INFO:root:Epoch[14] Batch [5600]	Speed: 661.33 samples/sec	acc=0.826367
+INFO:root:Epoch[14] Batch [5620]	Speed: 661.48 samples/sec	acc=0.827246
+INFO:root:Epoch[14] Batch [5640]	Speed: 662.55 samples/sec	acc=0.822559
+INFO:root:Epoch[14] Batch [5660]	Speed: 656.27 samples/sec	acc=0.821680
+INFO:root:Epoch[14] Batch [5680]	Speed: 665.94 samples/sec	acc=0.822754
+INFO:root:Epoch[14] Batch [5700]	Speed: 661.87 samples/sec	acc=0.832422
+INFO:root:Epoch[14] Batch [5720]	Speed: 666.73 samples/sec	acc=0.827930
+lr-batch-epoch: 0.0001 5733 14
+INFO:root:Epoch[14] Batch [5740]	Speed: 661.00 samples/sec	acc=0.826660
+INFO:root:Epoch[14] Batch [5760]	Speed: 661.96 samples/sec	acc=0.825098
+INFO:root:Epoch[14] Batch [5780]	Speed: 661.59 samples/sec	acc=0.818555
+INFO:root:Epoch[14] Batch [5800]	Speed: 661.89 samples/sec	acc=0.819434
+INFO:root:Epoch[14] Batch [5820]	Speed: 662.83 samples/sec	acc=0.816113
+INFO:root:Epoch[14] Batch [5840]	Speed: 661.75 samples/sec	acc=0.821191
+INFO:root:Epoch[14] Batch [5860]	Speed: 662.32 samples/sec	acc=0.829102
+INFO:root:Epoch[14] Batch [5880]	Speed: 663.81 samples/sec	acc=0.829883
+INFO:root:Epoch[14] Batch [5900]	Speed: 671.15 samples/sec	acc=0.825391
+INFO:root:Epoch[14] Batch [5920]	Speed: 662.40 samples/sec	acc=0.825586
+INFO:root:Epoch[14] Batch [5940]	Speed: 661.38 samples/sec	acc=0.831152
+INFO:root:Epoch[14] Batch [5960]	Speed: 654.04 samples/sec	acc=0.826660
+INFO:root:Epoch[14] Batch [5980]	Speed: 654.60 samples/sec	acc=0.827637
+INFO:root:Epoch[14] Batch [6000]	Speed: 662.83 samples/sec	acc=0.828223
+INFO:root:Epoch[14] Batch [6020]	Speed: 661.70 samples/sec	acc=0.827930
+INFO:root:Epoch[14] Batch [6040]	Speed: 662.05 samples/sec	acc=0.822070
+INFO:root:Epoch[14] Batch [6060]	Speed: 655.41 samples/sec	acc=0.828320
+INFO:root:Epoch[14] Batch [6080]	Speed: 661.75 samples/sec	acc=0.823633
+INFO:root:Epoch[14] Batch [6100]	Speed: 661.93 samples/sec	acc=0.821582
+INFO:root:Epoch[14] Batch [6120]	Speed: 661.66 samples/sec	acc=0.827051
+INFO:root:Epoch[14] Batch [6140]	Speed: 662.04 samples/sec	acc=0.826270
+INFO:root:Epoch[14] Batch [6160]	Speed: 661.70 samples/sec	acc=0.825586
+INFO:root:Epoch[14] Batch [6180]	Speed: 660.46 samples/sec	acc=0.829688
+INFO:root:Epoch[14] Batch [6200]	Speed: 661.75 samples/sec	acc=0.829492
+INFO:root:Epoch[14] Batch [6220]	Speed: 661.62 samples/sec	acc=0.817383
+INFO:root:Epoch[14] Batch [6240]	Speed: 662.25 samples/sec	acc=0.821484
+INFO:root:Epoch[14] Batch [6260]	Speed: 661.50 samples/sec	acc=0.820605
+INFO:root:Epoch[14] Batch [6280]	Speed: 661.93 samples/sec	acc=0.833398
+INFO:root:Epoch[14] Batch [6300]	Speed: 662.60 samples/sec	acc=0.828320
+INFO:root:Epoch[14] Batch [6320]	Speed: 661.61 samples/sec	acc=0.820312
+INFO:root:Epoch[14] Batch [6340]	Speed: 659.62 samples/sec	acc=0.825098
+INFO:root:Epoch[14] Batch [6360]	Speed: 661.72 samples/sec	acc=0.832520
+INFO:root:Epoch[14] Batch [6380]	Speed: 660.62 samples/sec	acc=0.821094
+INFO:root:Epoch[14] Batch [6400]	Speed: 655.83 samples/sec	acc=0.823828
+INFO:root:Epoch[14] Batch [6420]	Speed: 663.71 samples/sec	acc=0.826270
+INFO:root:Epoch[14] Batch [6440]	Speed: 662.10 samples/sec	acc=0.820996
+INFO:root:Epoch[14] Batch [6460]	Speed: 668.68 samples/sec	acc=0.823633
+INFO:root:Epoch[14] Batch [6480]	Speed: 662.27 samples/sec	acc=0.828711
+INFO:root:Epoch[14] Batch [6500]	Speed: 665.57 samples/sec	acc=0.822070
+INFO:root:Epoch[14] Batch [6520]	Speed: 669.04 samples/sec	acc=0.823047
+INFO:root:Epoch[14] Batch [6540]	Speed: 662.86 samples/sec	acc=0.824609
+INFO:root:Epoch[14] Batch [6560]	Speed: 661.04 samples/sec	acc=0.818457
+INFO:root:Epoch[14] Batch [6580]	Speed: 662.56 samples/sec	acc=0.829004
+INFO:root:Epoch[14] Batch [6600]	Speed: 662.78 samples/sec	acc=0.824902
+INFO:root:Epoch[14] Batch [6620]	Speed: 662.58 samples/sec	acc=0.826172
+INFO:root:Epoch[14] Batch [6640]	Speed: 662.55 samples/sec	acc=0.832715
+INFO:root:Epoch[14] Batch [6660]	Speed: 662.67 samples/sec	acc=0.835742
+INFO:root:Epoch[14] Batch [6680]	Speed: 666.32 samples/sec	acc=0.828125
+INFO:root:Epoch[14] Batch [6700]	Speed: 663.35 samples/sec	acc=0.825293
+INFO:root:Epoch[14] Batch [6720]	Speed: 658.66 samples/sec	acc=0.826855
+[112000][MARGIN]0.805804,0.425627
+lr-batch-epoch: 0.0001 6733 14
+testing verification..
+(14000, 512)
+infer time 17.271577
+[cfp_ff][112000]XNorm: 21.469734
+[cfp_ff][112000]Accuracy-Flip: 0.99829+-0.00229
+testing verification..
+(14000, 512)
+infer time 18.281102
+[cfp_fp][112000]XNorm: 19.813611
+[cfp_fp][112000]Accuracy-Flip: 0.92657+-0.01354
+testing verification..
+(12000, 512)
+infer time 16.304625
+[agedb_30][112000]XNorm: 22.723814
+[agedb_30][112000]Accuracy-Flip: 0.97817+-0.00598
+testing verification..
+(12000, 512)
+infer time 19.865969
+[lfw][112000]XNorm: 22.262860
+[lfw][112000]Accuracy-Flip: 0.99767+-0.00271
+[112000]Accuracy-Highest: 0.99817
+INFO:root:Epoch[14] Batch [6740]	Speed: 96.46 samples/sec	acc=0.823535
+INFO:root:Epoch[14] Batch [6760]	Speed: 669.92 samples/sec	acc=0.822363
+INFO:root:Epoch[14] Batch [6780]	Speed: 662.33 samples/sec	acc=0.829004
+INFO:root:Epoch[14] Batch [6800]	Speed: 661.67 samples/sec	acc=0.832129
+INFO:root:Epoch[14] Batch [6820]	Speed: 661.24 samples/sec	acc=0.826660
+INFO:root:Epoch[14] Batch [6840]	Speed: 662.76 samples/sec	acc=0.820508
+INFO:root:Epoch[14] Batch [6860]	Speed: 661.49 samples/sec	acc=0.823730
+INFO:root:Epoch[14] Batch [6880]	Speed: 661.79 samples/sec	acc=0.827441
diff --git a/model/model-0000.params b/model/model-0000.params
new file mode 100644
index 0000000..35118ad
Binary files /dev/null and b/model/model-0000.params differ
diff --git a/model/model-symbol.json b/model/model-symbol.json
new file mode 100644
index 0000000..cea9abc
--- /dev/null
+++ b/model/model-symbol.json
@@ -0,0 +1,2399 @@
+{
+  "nodes": [
+    {
+      "op": "null", 
+      "name": "data", 
+      "inputs": []
+    }, 
+    {
+      "op": "_minus_scalar", 
+      "name": "_minusscalar0", 
+      "attrs": {"scalar": "127.5"}, 
+      "inputs": [[0, 0, 0]]
+    }, 
+    {
+      "op": "_mul_scalar", 
+      "name": "_mulscalar0", 
+      "attrs": {"scalar": "0.0078125"}, 
+      "inputs": [[1, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_1_conv2d_weight", 
+      "attrs": {
+        "kernel": "(3, 3)", 
+        "no_bias": "True", 
+        "num_filter": "8", 
+        "num_group": "1", 
+        "pad": "(1, 1)", 
+        "stride": "(1, 1)"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "Convolution", 
+      "name": "conv_1_conv2d", 
+      "attrs": {
+        "kernel": "(3, 3)", 
+        "no_bias": "True", 
+        "num_filter": "8", 
+        "num_group": "1", 
+        "pad": "(1, 1)", 
+        "stride": "(1, 1)"
+      }, 
+      "inputs": [[2, 0, 0], [3, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_1_batchnorm_gamma", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_1_batchnorm_beta", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_1_batchnorm_moving_mean", 
+      "attrs": {
+        "__init__": "[\"zero\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_1_batchnorm_moving_var", 
+      "attrs": {
+        "__init__": "[\"one\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "BatchNorm", 
+      "name": "conv_1_batchnorm", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": [[4, 0, 0], [5, 0, 0], [6, 0, 0], [7, 0, 1], [8, 0, 1]]
+    }, 
+    {
+      "op": "Activation", 
+      "name": "conv_1_relu", 
+      "attrs": {"act_type": "relu"}, 
+      "inputs": [[9, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_2_dw_conv2d_weight", 
+      "attrs": {
+        "kernel": "(3, 3)", 
+        "no_bias": "True", 
+        "num_filter": "8", 
+        "num_group": "8", 
+        "pad": "(1, 1)", 
+        "stride": "(1, 1)"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "Convolution", 
+      "name": "conv_2_dw_conv2d", 
+      "attrs": {
+        "kernel": "(3, 3)", 
+        "no_bias": "True", 
+        "num_filter": "8", 
+        "num_group": "8", 
+        "pad": "(1, 1)", 
+        "stride": "(1, 1)"
+      }, 
+      "inputs": [[10, 0, 0], [11, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_2_dw_batchnorm_gamma", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_2_dw_batchnorm_beta", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_2_dw_batchnorm_moving_mean", 
+      "attrs": {
+        "__init__": "[\"zero\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_2_dw_batchnorm_moving_var", 
+      "attrs": {
+        "__init__": "[\"one\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "BatchNorm", 
+      "name": "conv_2_dw_batchnorm", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": [[12, 0, 0], [13, 0, 0], [14, 0, 0], [15, 0, 1], [16, 0, 1]]
+    }, 
+    {
+      "op": "Activation", 
+      "name": "conv_2_dw_relu", 
+      "attrs": {"act_type": "relu"}, 
+      "inputs": [[17, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_2_conv2d_weight", 
+      "attrs": {
+        "kernel": "(1, 1)", 
+        "no_bias": "True", 
+        "num_filter": "16", 
+        "num_group": "1", 
+        "pad": "(0, 0)", 
+        "stride": "(1, 1)"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "Convolution", 
+      "name": "conv_2_conv2d", 
+      "attrs": {
+        "kernel": "(1, 1)", 
+        "no_bias": "True", 
+        "num_filter": "16", 
+        "num_group": "1", 
+        "pad": "(0, 0)", 
+        "stride": "(1, 1)"
+      }, 
+      "inputs": [[18, 0, 0], [19, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_2_batchnorm_gamma", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_2_batchnorm_beta", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_2_batchnorm_moving_mean", 
+      "attrs": {
+        "__init__": "[\"zero\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_2_batchnorm_moving_var", 
+      "attrs": {
+        "__init__": "[\"one\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "BatchNorm", 
+      "name": "conv_2_batchnorm", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": [[20, 0, 0], [21, 0, 0], [22, 0, 0], [23, 0, 1], [24, 0, 1]]
+    }, 
+    {
+      "op": "Activation", 
+      "name": "conv_2_relu", 
+      "attrs": {"act_type": "relu"}, 
+      "inputs": [[25, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_3_dw_conv2d_weight", 
+      "attrs": {
+        "kernel": "(3, 3)", 
+        "no_bias": "True", 
+        "num_filter": "16", 
+        "num_group": "16", 
+        "pad": "(1, 1)", 
+        "stride": "(2, 2)"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "Convolution", 
+      "name": "conv_3_dw_conv2d", 
+      "attrs": {
+        "kernel": "(3, 3)", 
+        "no_bias": "True", 
+        "num_filter": "16", 
+        "num_group": "16", 
+        "pad": "(1, 1)", 
+        "stride": "(2, 2)"
+      }, 
+      "inputs": [[26, 0, 0], [27, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_3_dw_batchnorm_gamma", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_3_dw_batchnorm_beta", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_3_dw_batchnorm_moving_mean", 
+      "attrs": {
+        "__init__": "[\"zero\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_3_dw_batchnorm_moving_var", 
+      "attrs": {
+        "__init__": "[\"one\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "BatchNorm", 
+      "name": "conv_3_dw_batchnorm", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": [[28, 0, 0], [29, 0, 0], [30, 0, 0], [31, 0, 1], [32, 0, 1]]
+    }, 
+    {
+      "op": "Activation", 
+      "name": "conv_3_dw_relu", 
+      "attrs": {"act_type": "relu"}, 
+      "inputs": [[33, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_3_conv2d_weight", 
+      "attrs": {
+        "kernel": "(1, 1)", 
+        "no_bias": "True", 
+        "num_filter": "32", 
+        "num_group": "1", 
+        "pad": "(0, 0)", 
+        "stride": "(1, 1)"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "Convolution", 
+      "name": "conv_3_conv2d", 
+      "attrs": {
+        "kernel": "(1, 1)", 
+        "no_bias": "True", 
+        "num_filter": "32", 
+        "num_group": "1", 
+        "pad": "(0, 0)", 
+        "stride": "(1, 1)"
+      }, 
+      "inputs": [[34, 0, 0], [35, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_3_batchnorm_gamma", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_3_batchnorm_beta", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_3_batchnorm_moving_mean", 
+      "attrs": {
+        "__init__": "[\"zero\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_3_batchnorm_moving_var", 
+      "attrs": {
+        "__init__": "[\"one\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "BatchNorm", 
+      "name": "conv_3_batchnorm", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": [[36, 0, 0], [37, 0, 0], [38, 0, 0], [39, 0, 1], [40, 0, 1]]
+    }, 
+    {
+      "op": "Activation", 
+      "name": "conv_3_relu", 
+      "attrs": {"act_type": "relu"}, 
+      "inputs": [[41, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_4_dw_conv2d_weight", 
+      "attrs": {
+        "kernel": "(3, 3)", 
+        "no_bias": "True", 
+        "num_filter": "32", 
+        "num_group": "32", 
+        "pad": "(1, 1)", 
+        "stride": "(1, 1)"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "Convolution", 
+      "name": "conv_4_dw_conv2d", 
+      "attrs": {
+        "kernel": "(3, 3)", 
+        "no_bias": "True", 
+        "num_filter": "32", 
+        "num_group": "32", 
+        "pad": "(1, 1)", 
+        "stride": "(1, 1)"
+      }, 
+      "inputs": [[42, 0, 0], [43, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_4_dw_batchnorm_gamma", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_4_dw_batchnorm_beta", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_4_dw_batchnorm_moving_mean", 
+      "attrs": {
+        "__init__": "[\"zero\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_4_dw_batchnorm_moving_var", 
+      "attrs": {
+        "__init__": "[\"one\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "BatchNorm", 
+      "name": "conv_4_dw_batchnorm", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": [[44, 0, 0], [45, 0, 0], [46, 0, 0], [47, 0, 1], [48, 0, 1]]
+    }, 
+    {
+      "op": "Activation", 
+      "name": "conv_4_dw_relu", 
+      "attrs": {"act_type": "relu"}, 
+      "inputs": [[49, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_4_conv2d_weight", 
+      "attrs": {
+        "kernel": "(1, 1)", 
+        "no_bias": "True", 
+        "num_filter": "32", 
+        "num_group": "1", 
+        "pad": "(0, 0)", 
+        "stride": "(1, 1)"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "Convolution", 
+      "name": "conv_4_conv2d", 
+      "attrs": {
+        "kernel": "(1, 1)", 
+        "no_bias": "True", 
+        "num_filter": "32", 
+        "num_group": "1", 
+        "pad": "(0, 0)", 
+        "stride": "(1, 1)"
+      }, 
+      "inputs": [[50, 0, 0], [51, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_4_batchnorm_gamma", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_4_batchnorm_beta", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_4_batchnorm_moving_mean", 
+      "attrs": {
+        "__init__": "[\"zero\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_4_batchnorm_moving_var", 
+      "attrs": {
+        "__init__": "[\"one\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "BatchNorm", 
+      "name": "conv_4_batchnorm", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": [[52, 0, 0], [53, 0, 0], [54, 0, 0], [55, 0, 1], [56, 0, 1]]
+    }, 
+    {
+      "op": "Activation", 
+      "name": "conv_4_relu", 
+      "attrs": {"act_type": "relu"}, 
+      "inputs": [[57, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_5_dw_conv2d_weight", 
+      "attrs": {
+        "kernel": "(3, 3)", 
+        "no_bias": "True", 
+        "num_filter": "32", 
+        "num_group": "32", 
+        "pad": "(1, 1)", 
+        "stride": "(2, 2)"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "Convolution", 
+      "name": "conv_5_dw_conv2d", 
+      "attrs": {
+        "kernel": "(3, 3)", 
+        "no_bias": "True", 
+        "num_filter": "32", 
+        "num_group": "32", 
+        "pad": "(1, 1)", 
+        "stride": "(2, 2)"
+      }, 
+      "inputs": [[58, 0, 0], [59, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_5_dw_batchnorm_gamma", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_5_dw_batchnorm_beta", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_5_dw_batchnorm_moving_mean", 
+      "attrs": {
+        "__init__": "[\"zero\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_5_dw_batchnorm_moving_var", 
+      "attrs": {
+        "__init__": "[\"one\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "BatchNorm", 
+      "name": "conv_5_dw_batchnorm", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": [[60, 0, 0], [61, 0, 0], [62, 0, 0], [63, 0, 1], [64, 0, 1]]
+    }, 
+    {
+      "op": "Activation", 
+      "name": "conv_5_dw_relu", 
+      "attrs": {"act_type": "relu"}, 
+      "inputs": [[65, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_5_conv2d_weight", 
+      "attrs": {
+        "kernel": "(1, 1)", 
+        "no_bias": "True", 
+        "num_filter": "64", 
+        "num_group": "1", 
+        "pad": "(0, 0)", 
+        "stride": "(1, 1)"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "Convolution", 
+      "name": "conv_5_conv2d", 
+      "attrs": {
+        "kernel": "(1, 1)", 
+        "no_bias": "True", 
+        "num_filter": "64", 
+        "num_group": "1", 
+        "pad": "(0, 0)", 
+        "stride": "(1, 1)"
+      }, 
+      "inputs": [[66, 0, 0], [67, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_5_batchnorm_gamma", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_5_batchnorm_beta", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_5_batchnorm_moving_mean", 
+      "attrs": {
+        "__init__": "[\"zero\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_5_batchnorm_moving_var", 
+      "attrs": {
+        "__init__": "[\"one\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "BatchNorm", 
+      "name": "conv_5_batchnorm", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": [[68, 0, 0], [69, 0, 0], [70, 0, 0], [71, 0, 1], [72, 0, 1]]
+    }, 
+    {
+      "op": "Activation", 
+      "name": "conv_5_relu", 
+      "attrs": {"act_type": "relu"}, 
+      "inputs": [[73, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_6_dw_conv2d_weight", 
+      "attrs": {
+        "kernel": "(3, 3)", 
+        "no_bias": "True", 
+        "num_filter": "64", 
+        "num_group": "64", 
+        "pad": "(1, 1)", 
+        "stride": "(1, 1)"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "Convolution", 
+      "name": "conv_6_dw_conv2d", 
+      "attrs": {
+        "kernel": "(3, 3)", 
+        "no_bias": "True", 
+        "num_filter": "64", 
+        "num_group": "64", 
+        "pad": "(1, 1)", 
+        "stride": "(1, 1)"
+      }, 
+      "inputs": [[74, 0, 0], [75, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_6_dw_batchnorm_gamma", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_6_dw_batchnorm_beta", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_6_dw_batchnorm_moving_mean", 
+      "attrs": {
+        "__init__": "[\"zero\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_6_dw_batchnorm_moving_var", 
+      "attrs": {
+        "__init__": "[\"one\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "BatchNorm", 
+      "name": "conv_6_dw_batchnorm", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": [[76, 0, 0], [77, 0, 0], [78, 0, 0], [79, 0, 1], [80, 0, 1]]
+    }, 
+    {
+      "op": "Activation", 
+      "name": "conv_6_dw_relu", 
+      "attrs": {"act_type": "relu"}, 
+      "inputs": [[81, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_6_conv2d_weight", 
+      "attrs": {
+        "kernel": "(1, 1)", 
+        "no_bias": "True", 
+        "num_filter": "64", 
+        "num_group": "1", 
+        "pad": "(0, 0)", 
+        "stride": "(1, 1)"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "Convolution", 
+      "name": "conv_6_conv2d", 
+      "attrs": {
+        "kernel": "(1, 1)", 
+        "no_bias": "True", 
+        "num_filter": "64", 
+        "num_group": "1", 
+        "pad": "(0, 0)", 
+        "stride": "(1, 1)"
+      }, 
+      "inputs": [[82, 0, 0], [83, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_6_batchnorm_gamma", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_6_batchnorm_beta", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_6_batchnorm_moving_mean", 
+      "attrs": {
+        "__init__": "[\"zero\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_6_batchnorm_moving_var", 
+      "attrs": {
+        "__init__": "[\"one\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "BatchNorm", 
+      "name": "conv_6_batchnorm", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": [[84, 0, 0], [85, 0, 0], [86, 0, 0], [87, 0, 1], [88, 0, 1]]
+    }, 
+    {
+      "op": "Activation", 
+      "name": "conv_6_relu", 
+      "attrs": {"act_type": "relu"}, 
+      "inputs": [[89, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_7_dw_conv2d_weight", 
+      "attrs": {
+        "kernel": "(3, 3)", 
+        "no_bias": "True", 
+        "num_filter": "64", 
+        "num_group": "64", 
+        "pad": "(1, 1)", 
+        "stride": "(2, 2)"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "Convolution", 
+      "name": "conv_7_dw_conv2d", 
+      "attrs": {
+        "kernel": "(3, 3)", 
+        "no_bias": "True", 
+        "num_filter": "64", 
+        "num_group": "64", 
+        "pad": "(1, 1)", 
+        "stride": "(2, 2)"
+      }, 
+      "inputs": [[90, 0, 0], [91, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_7_dw_batchnorm_gamma", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_7_dw_batchnorm_beta", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_7_dw_batchnorm_moving_mean", 
+      "attrs": {
+        "__init__": "[\"zero\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_7_dw_batchnorm_moving_var", 
+      "attrs": {
+        "__init__": "[\"one\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "BatchNorm", 
+      "name": "conv_7_dw_batchnorm", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": [[92, 0, 0], [93, 0, 0], [94, 0, 0], [95, 0, 1], [96, 0, 1]]
+    }, 
+    {
+      "op": "Activation", 
+      "name": "conv_7_dw_relu", 
+      "attrs": {"act_type": "relu"}, 
+      "inputs": [[97, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_7_conv2d_weight", 
+      "attrs": {
+        "kernel": "(1, 1)", 
+        "no_bias": "True", 
+        "num_filter": "128", 
+        "num_group": "1", 
+        "pad": "(0, 0)", 
+        "stride": "(1, 1)"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "Convolution", 
+      "name": "conv_7_conv2d", 
+      "attrs": {
+        "kernel": "(1, 1)", 
+        "no_bias": "True", 
+        "num_filter": "128", 
+        "num_group": "1", 
+        "pad": "(0, 0)", 
+        "stride": "(1, 1)"
+      }, 
+      "inputs": [[98, 0, 0], [99, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_7_batchnorm_gamma", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_7_batchnorm_beta", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_7_batchnorm_moving_mean", 
+      "attrs": {
+        "__init__": "[\"zero\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_7_batchnorm_moving_var", 
+      "attrs": {
+        "__init__": "[\"one\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "BatchNorm", 
+      "name": "conv_7_batchnorm", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": [[100, 0, 0], [101, 0, 0], [102, 0, 0], [103, 0, 1], [104, 0, 1]]
+    }, 
+    {
+      "op": "Activation", 
+      "name": "conv_7_relu", 
+      "attrs": {"act_type": "relu"}, 
+      "inputs": [[105, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_8_dw_conv2d_weight", 
+      "attrs": {
+        "kernel": "(3, 3)", 
+        "no_bias": "True", 
+        "num_filter": "128", 
+        "num_group": "128", 
+        "pad": "(1, 1)", 
+        "stride": "(1, 1)"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "Convolution", 
+      "name": "conv_8_dw_conv2d", 
+      "attrs": {
+        "kernel": "(3, 3)", 
+        "no_bias": "True", 
+        "num_filter": "128", 
+        "num_group": "128", 
+        "pad": "(1, 1)", 
+        "stride": "(1, 1)"
+      }, 
+      "inputs": [[106, 0, 0], [107, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_8_dw_batchnorm_gamma", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_8_dw_batchnorm_beta", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_8_dw_batchnorm_moving_mean", 
+      "attrs": {
+        "__init__": "[\"zero\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_8_dw_batchnorm_moving_var", 
+      "attrs": {
+        "__init__": "[\"one\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "BatchNorm", 
+      "name": "conv_8_dw_batchnorm", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": [[108, 0, 0], [109, 0, 0], [110, 0, 0], [111, 0, 1], [112, 0, 1]]
+    }, 
+    {
+      "op": "Activation", 
+      "name": "conv_8_dw_relu", 
+      "attrs": {"act_type": "relu"}, 
+      "inputs": [[113, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_8_conv2d_weight", 
+      "attrs": {
+        "kernel": "(1, 1)", 
+        "no_bias": "True", 
+        "num_filter": "128", 
+        "num_group": "1", 
+        "pad": "(0, 0)", 
+        "stride": "(1, 1)"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "Convolution", 
+      "name": "conv_8_conv2d", 
+      "attrs": {
+        "kernel": "(1, 1)", 
+        "no_bias": "True", 
+        "num_filter": "128", 
+        "num_group": "1", 
+        "pad": "(0, 0)", 
+        "stride": "(1, 1)"
+      }, 
+      "inputs": [[114, 0, 0], [115, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_8_batchnorm_gamma", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_8_batchnorm_beta", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_8_batchnorm_moving_mean", 
+      "attrs": {
+        "__init__": "[\"zero\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_8_batchnorm_moving_var", 
+      "attrs": {
+        "__init__": "[\"one\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "BatchNorm", 
+      "name": "conv_8_batchnorm", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": [[116, 0, 0], [117, 0, 0], [118, 0, 0], [119, 0, 1], [120, 0, 1]]
+    }, 
+    {
+      "op": "Activation", 
+      "name": "conv_8_relu", 
+      "attrs": {"act_type": "relu"}, 
+      "inputs": [[121, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_9_dw_conv2d_weight", 
+      "attrs": {
+        "kernel": "(3, 3)", 
+        "no_bias": "True", 
+        "num_filter": "128", 
+        "num_group": "128", 
+        "pad": "(1, 1)", 
+        "stride": "(1, 1)"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "Convolution", 
+      "name": "conv_9_dw_conv2d", 
+      "attrs": {
+        "kernel": "(3, 3)", 
+        "no_bias": "True", 
+        "num_filter": "128", 
+        "num_group": "128", 
+        "pad": "(1, 1)", 
+        "stride": "(1, 1)"
+      }, 
+      "inputs": [[122, 0, 0], [123, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_9_dw_batchnorm_gamma", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_9_dw_batchnorm_beta", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_9_dw_batchnorm_moving_mean", 
+      "attrs": {
+        "__init__": "[\"zero\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_9_dw_batchnorm_moving_var", 
+      "attrs": {
+        "__init__": "[\"one\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "BatchNorm", 
+      "name": "conv_9_dw_batchnorm", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": [[124, 0, 0], [125, 0, 0], [126, 0, 0], [127, 0, 1], [128, 0, 1]]
+    }, 
+    {
+      "op": "Activation", 
+      "name": "conv_9_dw_relu", 
+      "attrs": {"act_type": "relu"}, 
+      "inputs": [[129, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_9_conv2d_weight", 
+      "attrs": {
+        "kernel": "(1, 1)", 
+        "no_bias": "True", 
+        "num_filter": "128", 
+        "num_group": "1", 
+        "pad": "(0, 0)", 
+        "stride": "(1, 1)"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "Convolution", 
+      "name": "conv_9_conv2d", 
+      "attrs": {
+        "kernel": "(1, 1)", 
+        "no_bias": "True", 
+        "num_filter": "128", 
+        "num_group": "1", 
+        "pad": "(0, 0)", 
+        "stride": "(1, 1)"
+      }, 
+      "inputs": [[130, 0, 0], [131, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_9_batchnorm_gamma", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_9_batchnorm_beta", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_9_batchnorm_moving_mean", 
+      "attrs": {
+        "__init__": "[\"zero\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_9_batchnorm_moving_var", 
+      "attrs": {
+        "__init__": "[\"one\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "BatchNorm", 
+      "name": "conv_9_batchnorm", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": [[132, 0, 0], [133, 0, 0], [134, 0, 0], [135, 0, 1], [136, 0, 1]]
+    }, 
+    {
+      "op": "Activation", 
+      "name": "conv_9_relu", 
+      "attrs": {"act_type": "relu"}, 
+      "inputs": [[137, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_10_dw_conv2d_weight", 
+      "attrs": {
+        "kernel": "(3, 3)", 
+        "no_bias": "True", 
+        "num_filter": "128", 
+        "num_group": "128", 
+        "pad": "(1, 1)", 
+        "stride": "(1, 1)"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "Convolution", 
+      "name": "conv_10_dw_conv2d", 
+      "attrs": {
+        "kernel": "(3, 3)", 
+        "no_bias": "True", 
+        "num_filter": "128", 
+        "num_group": "128", 
+        "pad": "(1, 1)", 
+        "stride": "(1, 1)"
+      }, 
+      "inputs": [[138, 0, 0], [139, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_10_dw_batchnorm_gamma", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_10_dw_batchnorm_beta", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_10_dw_batchnorm_moving_mean", 
+      "attrs": {
+        "__init__": "[\"zero\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_10_dw_batchnorm_moving_var", 
+      "attrs": {
+        "__init__": "[\"one\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "BatchNorm", 
+      "name": "conv_10_dw_batchnorm", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": [[140, 0, 0], [141, 0, 0], [142, 0, 0], [143, 0, 1], [144, 0, 1]]
+    }, 
+    {
+      "op": "Activation", 
+      "name": "conv_10_dw_relu", 
+      "attrs": {"act_type": "relu"}, 
+      "inputs": [[145, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_10_conv2d_weight", 
+      "attrs": {
+        "kernel": "(1, 1)", 
+        "no_bias": "True", 
+        "num_filter": "128", 
+        "num_group": "1", 
+        "pad": "(0, 0)", 
+        "stride": "(1, 1)"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "Convolution", 
+      "name": "conv_10_conv2d", 
+      "attrs": {
+        "kernel": "(1, 1)", 
+        "no_bias": "True", 
+        "num_filter": "128", 
+        "num_group": "1", 
+        "pad": "(0, 0)", 
+        "stride": "(1, 1)"
+      }, 
+      "inputs": [[146, 0, 0], [147, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_10_batchnorm_gamma", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_10_batchnorm_beta", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_10_batchnorm_moving_mean", 
+      "attrs": {
+        "__init__": "[\"zero\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_10_batchnorm_moving_var", 
+      "attrs": {
+        "__init__": "[\"one\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "BatchNorm", 
+      "name": "conv_10_batchnorm", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": [[148, 0, 0], [149, 0, 0], [150, 0, 0], [151, 0, 1], [152, 0, 1]]
+    }, 
+    {
+      "op": "Activation", 
+      "name": "conv_10_relu", 
+      "attrs": {"act_type": "relu"}, 
+      "inputs": [[153, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_11_dw_conv2d_weight", 
+      "attrs": {
+        "kernel": "(3, 3)", 
+        "no_bias": "True", 
+        "num_filter": "128", 
+        "num_group": "128", 
+        "pad": "(1, 1)", 
+        "stride": "(1, 1)"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "Convolution", 
+      "name": "conv_11_dw_conv2d", 
+      "attrs": {
+        "kernel": "(3, 3)", 
+        "no_bias": "True", 
+        "num_filter": "128", 
+        "num_group": "128", 
+        "pad": "(1, 1)", 
+        "stride": "(1, 1)"
+      }, 
+      "inputs": [[154, 0, 0], [155, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_11_dw_batchnorm_gamma", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_11_dw_batchnorm_beta", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_11_dw_batchnorm_moving_mean", 
+      "attrs": {
+        "__init__": "[\"zero\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_11_dw_batchnorm_moving_var", 
+      "attrs": {
+        "__init__": "[\"one\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "BatchNorm", 
+      "name": "conv_11_dw_batchnorm", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": [[156, 0, 0], [157, 0, 0], [158, 0, 0], [159, 0, 1], [160, 0, 1]]
+    }, 
+    {
+      "op": "Activation", 
+      "name": "conv_11_dw_relu", 
+      "attrs": {"act_type": "relu"}, 
+      "inputs": [[161, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_11_conv2d_weight", 
+      "attrs": {
+        "kernel": "(1, 1)", 
+        "no_bias": "True", 
+        "num_filter": "128", 
+        "num_group": "1", 
+        "pad": "(0, 0)", 
+        "stride": "(1, 1)"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "Convolution", 
+      "name": "conv_11_conv2d", 
+      "attrs": {
+        "kernel": "(1, 1)", 
+        "no_bias": "True", 
+        "num_filter": "128", 
+        "num_group": "1", 
+        "pad": "(0, 0)", 
+        "stride": "(1, 1)"
+      }, 
+      "inputs": [[162, 0, 0], [163, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_11_batchnorm_gamma", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_11_batchnorm_beta", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_11_batchnorm_moving_mean", 
+      "attrs": {
+        "__init__": "[\"zero\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_11_batchnorm_moving_var", 
+      "attrs": {
+        "__init__": "[\"one\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "BatchNorm", 
+      "name": "conv_11_batchnorm", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": [[164, 0, 0], [165, 0, 0], [166, 0, 0], [167, 0, 1], [168, 0, 1]]
+    }, 
+    {
+      "op": "Activation", 
+      "name": "conv_11_relu", 
+      "attrs": {"act_type": "relu"}, 
+      "inputs": [[169, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_12_dw_conv2d_weight", 
+      "attrs": {
+        "kernel": "(3, 3)", 
+        "no_bias": "True", 
+        "num_filter": "128", 
+        "num_group": "128", 
+        "pad": "(1, 1)", 
+        "stride": "(1, 1)"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "Convolution", 
+      "name": "conv_12_dw_conv2d", 
+      "attrs": {
+        "kernel": "(3, 3)", 
+        "no_bias": "True", 
+        "num_filter": "128", 
+        "num_group": "128", 
+        "pad": "(1, 1)", 
+        "stride": "(1, 1)"
+      }, 
+      "inputs": [[170, 0, 0], [171, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_12_dw_batchnorm_gamma", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_12_dw_batchnorm_beta", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_12_dw_batchnorm_moving_mean", 
+      "attrs": {
+        "__init__": "[\"zero\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_12_dw_batchnorm_moving_var", 
+      "attrs": {
+        "__init__": "[\"one\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "BatchNorm", 
+      "name": "conv_12_dw_batchnorm", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": [[172, 0, 0], [173, 0, 0], [174, 0, 0], [175, 0, 1], [176, 0, 1]]
+    }, 
+    {
+      "op": "Activation", 
+      "name": "conv_12_dw_relu", 
+      "attrs": {"act_type": "relu"}, 
+      "inputs": [[177, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_12_conv2d_weight", 
+      "attrs": {
+        "kernel": "(1, 1)", 
+        "no_bias": "True", 
+        "num_filter": "128", 
+        "num_group": "1", 
+        "pad": "(0, 0)", 
+        "stride": "(1, 1)"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "Convolution", 
+      "name": "conv_12_conv2d", 
+      "attrs": {
+        "kernel": "(1, 1)", 
+        "no_bias": "True", 
+        "num_filter": "128", 
+        "num_group": "1", 
+        "pad": "(0, 0)", 
+        "stride": "(1, 1)"
+      }, 
+      "inputs": [[178, 0, 0], [179, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_12_batchnorm_gamma", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_12_batchnorm_beta", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_12_batchnorm_moving_mean", 
+      "attrs": {
+        "__init__": "[\"zero\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_12_batchnorm_moving_var", 
+      "attrs": {
+        "__init__": "[\"one\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "BatchNorm", 
+      "name": "conv_12_batchnorm", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": [[180, 0, 0], [181, 0, 0], [182, 0, 0], [183, 0, 1], [184, 0, 1]]
+    }, 
+    {
+      "op": "Activation", 
+      "name": "conv_12_relu", 
+      "attrs": {"act_type": "relu"}, 
+      "inputs": [[185, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_13_dw_conv2d_weight", 
+      "attrs": {
+        "kernel": "(3, 3)", 
+        "no_bias": "True", 
+        "num_filter": "128", 
+        "num_group": "128", 
+        "pad": "(1, 1)", 
+        "stride": "(2, 2)"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "Convolution", 
+      "name": "conv_13_dw_conv2d", 
+      "attrs": {
+        "kernel": "(3, 3)", 
+        "no_bias": "True", 
+        "num_filter": "128", 
+        "num_group": "128", 
+        "pad": "(1, 1)", 
+        "stride": "(2, 2)"
+      }, 
+      "inputs": [[186, 0, 0], [187, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_13_dw_batchnorm_gamma", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_13_dw_batchnorm_beta", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_13_dw_batchnorm_moving_mean", 
+      "attrs": {
+        "__init__": "[\"zero\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_13_dw_batchnorm_moving_var", 
+      "attrs": {
+        "__init__": "[\"one\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "BatchNorm", 
+      "name": "conv_13_dw_batchnorm", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": [[188, 0, 0], [189, 0, 0], [190, 0, 0], [191, 0, 1], [192, 0, 1]]
+    }, 
+    {
+      "op": "Activation", 
+      "name": "conv_13_dw_relu", 
+      "attrs": {"act_type": "relu"}, 
+      "inputs": [[193, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_13_conv2d_weight", 
+      "attrs": {
+        "kernel": "(1, 1)", 
+        "no_bias": "True", 
+        "num_filter": "256", 
+        "num_group": "1", 
+        "pad": "(0, 0)", 
+        "stride": "(1, 1)"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "Convolution", 
+      "name": "conv_13_conv2d", 
+      "attrs": {
+        "kernel": "(1, 1)", 
+        "no_bias": "True", 
+        "num_filter": "256", 
+        "num_group": "1", 
+        "pad": "(0, 0)", 
+        "stride": "(1, 1)"
+      }, 
+      "inputs": [[194, 0, 0], [195, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_13_batchnorm_gamma", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_13_batchnorm_beta", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_13_batchnorm_moving_mean", 
+      "attrs": {
+        "__init__": "[\"zero\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_13_batchnorm_moving_var", 
+      "attrs": {
+        "__init__": "[\"one\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "BatchNorm", 
+      "name": "conv_13_batchnorm", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": [[196, 0, 0], [197, 0, 0], [198, 0, 0], [199, 0, 1], [200, 0, 1]]
+    }, 
+    {
+      "op": "Activation", 
+      "name": "conv_13_relu", 
+      "attrs": {"act_type": "relu"}, 
+      "inputs": [[201, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_14_dw_conv2d_weight", 
+      "attrs": {
+        "kernel": "(3, 3)", 
+        "no_bias": "True", 
+        "num_filter": "256", 
+        "num_group": "256", 
+        "pad": "(1, 1)", 
+        "stride": "(1, 1)"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "Convolution", 
+      "name": "conv_14_dw_conv2d", 
+      "attrs": {
+        "kernel": "(3, 3)", 
+        "no_bias": "True", 
+        "num_filter": "256", 
+        "num_group": "256", 
+        "pad": "(1, 1)", 
+        "stride": "(1, 1)"
+      }, 
+      "inputs": [[202, 0, 0], [203, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_14_dw_batchnorm_gamma", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_14_dw_batchnorm_beta", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_14_dw_batchnorm_moving_mean", 
+      "attrs": {
+        "__init__": "[\"zero\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_14_dw_batchnorm_moving_var", 
+      "attrs": {
+        "__init__": "[\"one\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "BatchNorm", 
+      "name": "conv_14_dw_batchnorm", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": [[204, 0, 0], [205, 0, 0], [206, 0, 0], [207, 0, 1], [208, 0, 1]]
+    }, 
+    {
+      "op": "Activation", 
+      "name": "conv_14_dw_relu", 
+      "attrs": {"act_type": "relu"}, 
+      "inputs": [[209, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_14_conv2d_weight", 
+      "attrs": {
+        "kernel": "(1, 1)", 
+        "no_bias": "True", 
+        "num_filter": "256", 
+        "num_group": "1", 
+        "pad": "(0, 0)", 
+        "stride": "(1, 1)"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "Convolution", 
+      "name": "conv_14_conv2d", 
+      "attrs": {
+        "kernel": "(1, 1)", 
+        "no_bias": "True", 
+        "num_filter": "256", 
+        "num_group": "1", 
+        "pad": "(0, 0)", 
+        "stride": "(1, 1)"
+      }, 
+      "inputs": [[210, 0, 0], [211, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_14_batchnorm_gamma", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_14_batchnorm_beta", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_14_batchnorm_moving_mean", 
+      "attrs": {
+        "__init__": "[\"zero\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "conv_14_batchnorm_moving_var", 
+      "attrs": {
+        "__init__": "[\"one\", {}]", 
+        "fix_gamma": "True"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "BatchNorm", 
+      "name": "conv_14_batchnorm", 
+      "attrs": {"fix_gamma": "True"}, 
+      "inputs": [[212, 0, 0], [213, 0, 0], [214, 0, 0], [215, 0, 1], [216, 0, 1]]
+    }, 
+    {
+      "op": "Activation", 
+      "name": "conv_14_relu", 
+      "attrs": {"act_type": "relu"}, 
+      "inputs": [[217, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "bn1_gamma", 
+      "attrs": {
+        "eps": "2e-05", 
+        "fix_gamma": "False", 
+        "momentum": "0.9"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "bn1_beta", 
+      "attrs": {
+        "eps": "2e-05", 
+        "fix_gamma": "False", 
+        "momentum": "0.9"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "bn1_moving_mean", 
+      "attrs": {
+        "__init__": "[\"zero\", {}]", 
+        "eps": "2e-05", 
+        "fix_gamma": "False", 
+        "momentum": "0.9"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "bn1_moving_var", 
+      "attrs": {
+        "__init__": "[\"one\", {}]", 
+        "eps": "2e-05", 
+        "fix_gamma": "False", 
+        "momentum": "0.9"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "BatchNorm", 
+      "name": "bn1", 
+      "attrs": {
+        "eps": "2e-05", 
+        "fix_gamma": "False", 
+        "momentum": "0.9"
+      }, 
+      "inputs": [[218, 0, 0], [219, 0, 0], [220, 0, 0], [221, 0, 1], [222, 0, 1]]
+    }, 
+    {
+      "op": "null", 
+      "name": "relu1_gamma", 
+      "attrs": {
+        "__init__": "[\"Constant\", {\"value\": 0.25}]", 
+        "act_type": "prelu"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "LeakyReLU", 
+      "name": "relu1", 
+      "attrs": {"act_type": "prelu"}, 
+      "inputs": [[223, 0, 0], [224, 0, 0]]
+    }, 
+    {
+      "op": "Pooling", 
+      "name": "pool1", 
+      "attrs": {
+        "global_pool": "True", 
+        "kernel": "(7, 7)", 
+        "pool_type": "avg"
+      }, 
+      "inputs": [[225, 0, 0]]
+    }, 
+    {
+      "op": "Flatten", 
+      "name": "flatten0", 
+      "inputs": [[226, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "pre_fc1_weight", 
+      "attrs": {"num_hidden": "202"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "pre_fc1_bias", 
+      "attrs": {"num_hidden": "202"}, 
+      "inputs": []
+    }, 
+    {
+      "op": "FullyConnected", 
+      "name": "pre_fc1", 
+      "attrs": {"num_hidden": "202"}, 
+      "inputs": [[227, 0, 0], [228, 0, 0], [229, 0, 0]]
+    }, 
+    {
+      "op": "null", 
+      "name": "fc1_gamma", 
+      "attrs": {
+        "eps": "2e-05", 
+        "fix_gamma": "True", 
+        "momentum": "0.9"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "fc1_beta", 
+      "attrs": {
+        "eps": "2e-05", 
+        "fix_gamma": "True", 
+        "momentum": "0.9"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "fc1_moving_mean", 
+      "attrs": {
+        "__init__": "[\"zero\", {}]", 
+        "eps": "2e-05", 
+        "fix_gamma": "True", 
+        "momentum": "0.9"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "null", 
+      "name": "fc1_moving_var", 
+      "attrs": {
+        "__init__": "[\"one\", {}]", 
+        "eps": "2e-05", 
+        "fix_gamma": "True", 
+        "momentum": "0.9"
+      }, 
+      "inputs": []
+    }, 
+    {
+      "op": "BatchNorm", 
+      "name": "fc1", 
+      "attrs": {
+        "eps": "2e-05", 
+        "fix_gamma": "True", 
+        "momentum": "0.9"
+      }, 
+      "inputs": [[230, 0, 0], [231, 0, 0], [232, 0, 0], [233, 0, 1], [234, 0, 1]]
+    }
+  ], 
+  "arg_nodes": [
+    0, 
+    3, 
+    5, 
+    6, 
+    7, 
+    8, 
+    11, 
+    13, 
+    14, 
+    15, 
+    16, 
+    19, 
+    21, 
+    22, 
+    23, 
+    24, 
+    27, 
+    29, 
+    30, 
+    31, 
+    32, 
+    35, 
+    37, 
+    38, 
+    39, 
+    40, 
+    43, 
+    45, 
+    46, 
+    47, 
+    48, 
+    51, 
+    53, 
+    54, 
+    55, 
+    56, 
+    59, 
+    61, 
+    62, 
+    63, 
+    64, 
+    67, 
+    69, 
+    70, 
+    71, 
+    72, 
+    75, 
+    77, 
+    78, 
+    79, 
+    80, 
+    83, 
+    85, 
+    86, 
+    87, 
+    88, 
+    91, 
+    93, 
+    94, 
+    95, 
+    96, 
+    99, 
+    101, 
+    102, 
+    103, 
+    104, 
+    107, 
+    109, 
+    110, 
+    111, 
+    112, 
+    115, 
+    117, 
+    118, 
+    119, 
+    120, 
+    123, 
+    125, 
+    126, 
+    127, 
+    128, 
+    131, 
+    133, 
+    134, 
+    135, 
+    136, 
+    139, 
+    141, 
+    142, 
+    143, 
+    144, 
+    147, 
+    149, 
+    150, 
+    151, 
+    152, 
+    155, 
+    157, 
+    158, 
+    159, 
+    160, 
+    163, 
+    165, 
+    166, 
+    167, 
+    168, 
+    171, 
+    173, 
+    174, 
+    175, 
+    176, 
+    179, 
+    181, 
+    182, 
+    183, 
+    184, 
+    187, 
+    189, 
+    190, 
+    191, 
+    192, 
+    195, 
+    197, 
+    198, 
+    199, 
+    200, 
+    203, 
+    205, 
+    206, 
+    207, 
+    208, 
+    211, 
+    213, 
+    214, 
+    215, 
+    216, 
+    219, 
+    220, 
+    221, 
+    222, 
+    224, 
+    228, 
+    229, 
+    231, 
+    232, 
+    233, 
+    234
+  ], 
+  "node_row_ptr": [
+    0, 
+    1, 
+    2, 
+    3, 
+    4, 
+    5, 
+    6, 
+    7, 
+    8, 
+    9, 
+    12, 
+    13, 
+    14, 
+    15, 
+    16, 
+    17, 
+    18, 
+    19, 
+    22, 
+    23, 
+    24, 
+    25, 
+    26, 
+    27, 
+    28, 
+    29, 
+    32, 
+    33, 
+    34, 
+    35, 
+    36, 
+    37, 
+    38, 
+    39, 
+    42, 
+    43, 
+    44, 
+    45, 
+    46, 
+    47, 
+    48, 
+    49, 
+    52, 
+    53, 
+    54, 
+    55, 
+    56, 
+    57, 
+    58, 
+    59, 
+    62, 
+    63, 
+    64, 
+    65, 
+    66, 
+    67, 
+    68, 
+    69, 
+    72, 
+    73, 
+    74, 
+    75, 
+    76, 
+    77, 
+    78, 
+    79, 
+    82, 
+    83, 
+    84, 
+    85, 
+    86, 
+    87, 
+    88, 
+    89, 
+    92, 
+    93, 
+    94, 
+    95, 
+    96, 
+    97, 
+    98, 
+    99, 
+    102, 
+    103, 
+    104, 
+    105, 
+    106, 
+    107, 
+    108, 
+    109, 
+    112, 
+    113, 
+    114, 
+    115, 
+    116, 
+    117, 
+    118, 
+    119, 
+    122, 
+    123, 
+    124, 
+    125, 
+    126, 
+    127, 
+    128, 
+    129, 
+    132, 
+    133, 
+    134, 
+    135, 
+    136, 
+    137, 
+    138, 
+    139, 
+    142, 
+    143, 
+    144, 
+    145, 
+    146, 
+    147, 
+    148, 
+    149, 
+    152, 
+    153, 
+    154, 
+    155, 
+    156, 
+    157, 
+    158, 
+    159, 
+    162, 
+    163, 
+    164, 
+    165, 
+    166, 
+    167, 
+    168, 
+    169, 
+    172, 
+    173, 
+    174, 
+    175, 
+    176, 
+    177, 
+    178, 
+    179, 
+    182, 
+    183, 
+    184, 
+    185, 
+    186, 
+    187, 
+    188, 
+    189, 
+    192, 
+    193, 
+    194, 
+    195, 
+    196, 
+    197, 
+    198, 
+    199, 
+    202, 
+    203, 
+    204, 
+    205, 
+    206, 
+    207, 
+    208, 
+    209, 
+    212, 
+    213, 
+    214, 
+    215, 
+    216, 
+    217, 
+    218, 
+    219, 
+    222, 
+    223, 
+    224, 
+    225, 
+    226, 
+    227, 
+    228, 
+    229, 
+    232, 
+    233, 
+    234, 
+    235, 
+    236, 
+    237, 
+    238, 
+    239, 
+    242, 
+    243, 
+    244, 
+    245, 
+    246, 
+    247, 
+    248, 
+    249, 
+    252, 
+    253, 
+    254, 
+    255, 
+    256, 
+    257, 
+    258, 
+    259, 
+    262, 
+    263, 
+    264, 
+    265, 
+    266, 
+    267, 
+    268, 
+    269, 
+    272, 
+    273, 
+    274, 
+    275, 
+    276, 
+    277, 
+    280, 
+    281, 
+    282, 
+    283, 
+    284, 
+    285, 
+    286, 
+    287, 
+    288, 
+    289, 
+    290, 
+    291, 
+    294
+  ], 
+  "heads": [[235, 0, 0]], 
+  "attrs": {"mxnet_version": ["int", 10300]}
+}
\ No newline at end of file
diff --git a/model/onnx/centerface.onnx b/model/onnx/centerface.onnx
new file mode 100644
index 0000000..1487d5f
Binary files /dev/null and b/model/onnx/centerface.onnx differ
diff --git a/model/onnx/centerface_bnmerged.onnx b/model/onnx/centerface_bnmerged.onnx
new file mode 100644
index 0000000..e334bcc
Binary files /dev/null and b/model/onnx/centerface_bnmerged.onnx differ
diff --git a/models/__init__.py b/models/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/models/__pycache__/__init__.cpython-38.pyc b/models/__pycache__/__init__.cpython-38.pyc
new file mode 100644
index 0000000..c7188aa
Binary files /dev/null and b/models/__pycache__/__init__.cpython-38.pyc differ
diff --git a/models/__pycache__/net.cpython-38.pyc b/models/__pycache__/net.cpython-38.pyc
new file mode 100644
index 0000000..42cbfbf
Binary files /dev/null and b/models/__pycache__/net.cpython-38.pyc differ
diff --git a/models/__pycache__/retinaface.cpython-38.pyc b/models/__pycache__/retinaface.cpython-38.pyc
new file mode 100644
index 0000000..2fcb8a3
Binary files /dev/null and b/models/__pycache__/retinaface.cpython-38.pyc differ
diff --git a/models/net.py b/models/net.py
new file mode 100644
index 0000000..beb6040
--- /dev/null
+++ b/models/net.py
@@ -0,0 +1,137 @@
+import time
+import torch
+import torch.nn as nn
+import torchvision.models._utils as _utils
+import torchvision.models as models
+import torch.nn.functional as F
+from torch.autograd import Variable
+
+def conv_bn(inp, oup, stride = 1, leaky = 0):
+    return nn.Sequential(
+        nn.Conv2d(inp, oup, 3, stride, 1, bias=False),
+        nn.BatchNorm2d(oup),
+        nn.LeakyReLU(negative_slope=leaky, inplace=True)
+    )
+
+def conv_bn_no_relu(inp, oup, stride):
+    return nn.Sequential(
+        nn.Conv2d(inp, oup, 3, stride, 1, bias=False),
+        nn.BatchNorm2d(oup),
+    )
+
+def conv_bn1X1(inp, oup, stride, leaky=0):
+    return nn.Sequential(
+        nn.Conv2d(inp, oup, 1, stride, padding=0, bias=False),
+        nn.BatchNorm2d(oup),
+        nn.LeakyReLU(negative_slope=leaky, inplace=True)
+    )
+
+def conv_dw(inp, oup, stride, leaky=0.1):
+    return nn.Sequential(
+        nn.Conv2d(inp, inp, 3, stride, 1, groups=inp, bias=False),
+        nn.BatchNorm2d(inp),
+        nn.LeakyReLU(negative_slope= leaky,inplace=True),
+
+        nn.Conv2d(inp, oup, 1, 1, 0, bias=False),
+        nn.BatchNorm2d(oup),
+        nn.LeakyReLU(negative_slope= leaky,inplace=True),
+    )
+
+class SSH(nn.Module):
+    def __init__(self, in_channel, out_channel):
+        super(SSH, self).__init__()
+        assert out_channel % 4 == 0
+        leaky = 0
+        if (out_channel <= 64):
+            leaky = 0.1
+        self.conv3X3 = conv_bn_no_relu(in_channel, out_channel//2, stride=1)
+
+        self.conv5X5_1 = conv_bn(in_channel, out_channel//4, stride=1, leaky = leaky)
+        self.conv5X5_2 = conv_bn_no_relu(out_channel//4, out_channel//4, stride=1)
+
+        self.conv7X7_2 = conv_bn(out_channel//4, out_channel//4, stride=1, leaky = leaky)
+        self.conv7x7_3 = conv_bn_no_relu(out_channel//4, out_channel//4, stride=1)
+
+    def forward(self, input):
+        conv3X3 = self.conv3X3(input)
+
+        conv5X5_1 = self.conv5X5_1(input)
+        conv5X5 = self.conv5X5_2(conv5X5_1)
+
+        conv7X7_2 = self.conv7X7_2(conv5X5_1)
+        conv7X7 = self.conv7x7_3(conv7X7_2)
+
+        out = torch.cat([conv3X3, conv5X5, conv7X7], dim=1)
+        out = F.relu(out)
+        return out
+
+class FPN(nn.Module):
+    def __init__(self,in_channels_list,out_channels):
+        super(FPN,self).__init__()
+        leaky = 0
+        if (out_channels <= 64):
+            leaky = 0.1
+        self.output1 = conv_bn1X1(in_channels_list[0], out_channels, stride = 1, leaky = leaky)
+        self.output2 = conv_bn1X1(in_channels_list[1], out_channels, stride = 1, leaky = leaky)
+        self.output3 = conv_bn1X1(in_channels_list[2], out_channels, stride = 1, leaky = leaky)
+
+        self.merge1 = conv_bn(out_channels, out_channels, leaky = leaky)
+        self.merge2 = conv_bn(out_channels, out_channels, leaky = leaky)
+
+    def forward(self, input):
+        # names = list(input.keys())
+        input = list(input.values())
+
+        output1 = self.output1(input[0])
+        output2 = self.output2(input[1])
+        output3 = self.output3(input[2])
+
+        up3 = F.interpolate(output3, size=[output2.size(2), output2.size(3)], mode="nearest")
+        output2 = output2 + up3
+        output2 = self.merge2(output2)
+
+        up2 = F.interpolate(output2, size=[output1.size(2), output1.size(3)], mode="nearest")
+        output1 = output1 + up2
+        output1 = self.merge1(output1)
+
+        out = [output1, output2, output3]
+        return out
+
+
+
+class MobileNetV1(nn.Module):
+    def __init__(self):
+        super(MobileNetV1, self).__init__()
+        self.stage1 = nn.Sequential(
+            conv_bn(3, 8, 2, leaky = 0.1),    # 3
+            conv_dw(8, 16, 1),   # 7
+            conv_dw(16, 32, 2),  # 11
+            conv_dw(32, 32, 1),  # 19
+            conv_dw(32, 64, 2),  # 27
+            conv_dw(64, 64, 1),  # 43
+        )
+        self.stage2 = nn.Sequential(
+            conv_dw(64, 128, 2),  # 43 + 16 = 59
+            conv_dw(128, 128, 1), # 59 + 32 = 91
+            conv_dw(128, 128, 1), # 91 + 32 = 123
+            conv_dw(128, 128, 1), # 123 + 32 = 155
+            conv_dw(128, 128, 1), # 155 + 32 = 187
+            conv_dw(128, 128, 1), # 187 + 32 = 219
+        )
+        self.stage3 = nn.Sequential(
+            conv_dw(128, 256, 2), # 219 +3 2 = 241
+            conv_dw(256, 256, 1), # 241 + 64 = 301
+        )
+        self.avg = nn.AdaptiveAvgPool2d((1,1))
+        self.fc = nn.Linear(256, 1000)
+
+    def forward(self, x):
+        x = self.stage1(x)
+        x = self.stage2(x)
+        x = self.stage3(x)
+        x = self.avg(x)
+        # x = self.model(x)
+        x = x.view(-1, 256)
+        x = self.fc(x)
+        return x
+
diff --git a/models/retinaface.py b/models/retinaface.py
new file mode 100644
index 0000000..d530bd8
--- /dev/null
+++ b/models/retinaface.py
@@ -0,0 +1,127 @@
+import torch
+import torch.nn as nn
+import torchvision.models.detection.backbone_utils as backbone_utils
+import torchvision.models._utils as _utils
+import torch.nn.functional as F
+from collections import OrderedDict
+
+from models.net import MobileNetV1 as MobileNetV1
+from models.net import FPN as FPN
+from models.net import SSH as SSH
+
+
+
+class ClassHead(nn.Module):
+    def __init__(self,inchannels=512,num_anchors=3):
+        super(ClassHead,self).__init__()
+        self.num_anchors = num_anchors
+        self.conv1x1 = nn.Conv2d(inchannels,self.num_anchors*2,kernel_size=(1,1),stride=1,padding=0)
+
+    def forward(self,x):
+        out = self.conv1x1(x)
+        out = out.permute(0,2,3,1).contiguous()
+        
+        return out.view(out.shape[0], -1, 2)
+
+class BboxHead(nn.Module):
+    def __init__(self,inchannels=512,num_anchors=3):
+        super(BboxHead,self).__init__()
+        self.conv1x1 = nn.Conv2d(inchannels,num_anchors*4,kernel_size=(1,1),stride=1,padding=0)
+
+    def forward(self,x):
+        out = self.conv1x1(x)
+        out = out.permute(0,2,3,1).contiguous()
+
+        return out.view(out.shape[0], -1, 4)
+
+class LandmarkHead(nn.Module):
+    def __init__(self,inchannels=512,num_anchors=3):
+        super(LandmarkHead,self).__init__()
+        self.conv1x1 = nn.Conv2d(inchannels,num_anchors*10,kernel_size=(1,1),stride=1,padding=0)
+
+    def forward(self,x):
+        out = self.conv1x1(x)
+        out = out.permute(0,2,3,1).contiguous()
+
+        return out.view(out.shape[0], -1, 10)
+
+class RetinaFace(nn.Module):
+    def __init__(self, cfg = None, phase = 'train'):
+        """
+        :param cfg:  Network related settings.
+        :param phase: train or test.
+        """
+        super(RetinaFace,self).__init__()
+        self.phase = phase
+        backbone = None
+        if cfg['name'] == 'mobilenet0.25':
+            backbone = MobileNetV1()
+            if cfg['pretrain']:
+                checkpoint = torch.load("./weights/mobilenetV1X0.25_pretrain.tar", map_location=torch.device('cpu'))
+                from collections import OrderedDict
+                new_state_dict = OrderedDict()
+                for k, v in checkpoint['state_dict'].items():
+                    name = k[7:]  # remove module.
+                    new_state_dict[name] = v
+                # load params
+                backbone.load_state_dict(new_state_dict)
+        elif cfg['name'] == 'Resnet50':
+            import torchvision.models as models
+            backbone = models.resnet50(pretrained=cfg['pretrain'])
+
+        self.body = _utils.IntermediateLayerGetter(backbone, cfg['return_layers'])
+        in_channels_stage2 = cfg['in_channel']
+        in_channels_list = [
+            in_channels_stage2 * 2,
+            in_channels_stage2 * 4,
+            in_channels_stage2 * 8,
+        ]
+        out_channels = cfg['out_channel']
+        self.fpn = FPN(in_channels_list,out_channels)
+        self.ssh1 = SSH(out_channels, out_channels)
+        self.ssh2 = SSH(out_channels, out_channels)
+        self.ssh3 = SSH(out_channels, out_channels)
+
+        self.ClassHead = self._make_class_head(fpn_num=3, inchannels=cfg['out_channel'])
+        self.BboxHead = self._make_bbox_head(fpn_num=3, inchannels=cfg['out_channel'])
+        self.LandmarkHead = self._make_landmark_head(fpn_num=3, inchannels=cfg['out_channel'])
+
+    def _make_class_head(self,fpn_num=3,inchannels=64,anchor_num=2):
+        classhead = nn.ModuleList()
+        for i in range(fpn_num):
+            classhead.append(ClassHead(inchannels,anchor_num))
+        return classhead
+    
+    def _make_bbox_head(self,fpn_num=3,inchannels=64,anchor_num=2):
+        bboxhead = nn.ModuleList()
+        for i in range(fpn_num):
+            bboxhead.append(BboxHead(inchannels,anchor_num))
+        return bboxhead
+
+    def _make_landmark_head(self,fpn_num=3,inchannels=64,anchor_num=2):
+        landmarkhead = nn.ModuleList()
+        for i in range(fpn_num):
+            landmarkhead.append(LandmarkHead(inchannels,anchor_num))
+        return landmarkhead
+
+    def forward(self,inputs):
+        out = self.body(inputs)
+
+        # FPN
+        fpn = self.fpn(out)
+
+        # SSH
+        feature1 = self.ssh1(fpn[0])
+        feature2 = self.ssh2(fpn[1])
+        feature3 = self.ssh3(fpn[2])
+        features = [feature1, feature2, feature3]
+
+        bbox_regressions = torch.cat([self.BboxHead[i](feature) for i, feature in enumerate(features)], dim=1)
+        classifications = torch.cat([self.ClassHead[i](feature) for i, feature in enumerate(features)],dim=1)
+        ldm_regressions = torch.cat([self.LandmarkHead[i](feature) for i, feature in enumerate(features)], dim=1)
+
+        if self.phase == 'train':
+            output = (bbox_regressions, classifications, ldm_regressions)
+        else:
+            output = (bbox_regressions, F.softmax(classifications, dim=-1), ldm_regressions)
+        return output
\ No newline at end of file
diff --git a/partial_fc.py b/partial_fc.py
new file mode 100644
index 0000000..9325b19
--- /dev/null
+++ b/partial_fc.py
@@ -0,0 +1,161 @@
+import logging
+import os
+
+import torch
+import torch.distributed as dist
+from torch.nn import Module
+from torch.nn.functional import normalize, linear
+from torch.nn.parameter import Parameter
+
+
+class PartialFC(Module):
+    """
+    Author: {Xiang An, Yang Xiao, XuHan Zhu} in DeepGlint,
+    Partial FC: Training 10 Million Identities on a Single Machine
+    See the original paper:
+    https://arxiv.org/abs/2010.05222
+    """
+
+    @torch.no_grad()
+    def __init__(self, rank, local_rank, world_size, batch_size, resume,
+                 margin_softmax, num_classes, sample_rate=1.0, embedding_size=512, prefix="./"):
+        super(PartialFC, self).__init__()
+        #
+        self.num_classes: int = num_classes
+        self.rank: int = rank
+        self.local_rank: int = local_rank
+        self.device: torch.device = torch.device("cuda:{}".format(self.local_rank))
+        self.world_size: int = world_size
+        self.batch_size: int = batch_size
+        self.margin_softmax: callable = margin_softmax
+        self.sample_rate: float = sample_rate
+        self.embedding_size: int = embedding_size
+        self.prefix: str = prefix
+        self.num_local: int = num_classes // world_size + int(rank < num_classes % world_size)
+        self.class_start: int = num_classes // world_size * rank + min(rank, num_classes % world_size)
+        self.num_sample: int = int(self.sample_rate * self.num_local)
+
+        self.weight_name = os.path.join(self.prefix, "rank:{}_softmax_weight.pt".format(self.rank))
+        self.weight_mom_name = os.path.join(self.prefix, "rank:{}_softmax_weight_mom.pt".format(self.rank))
+
+        if resume:
+            try:
+                self.weight: torch.Tensor = torch.load(self.weight_name)
+                logging.info("softmax weight resume successfully!")
+            except (FileNotFoundError, KeyError, IndexError):
+                self.weight = torch.normal(0, 0.01, (self.num_local, self.embedding_size), device=self.device)
+                logging.info("softmax weight resume fail!")
+
+            try:
+                self.weight_mom: torch.Tensor = torch.load(self.weight_mom_name)
+                logging.info("softmax weight mom resume successfully!")
+            except (FileNotFoundError, KeyError, IndexError):
+                self.weight_mom: torch.Tensor = torch.zeros_like(self.weight)
+                logging.info("softmax weight mom resume fail!")
+        else:
+            self.weight = torch.normal(0, 0.01, (self.num_local, self.embedding_size), device=self.device)
+            self.weight_mom: torch.Tensor = torch.zeros_like(self.weight)
+            logging.info("softmax weight init successfully!")
+            logging.info("softmax weight mom init successfully!")
+        self.stream: torch.cuda.Stream = torch.cuda.Stream(local_rank)
+
+        self.index = None
+        if int(self.sample_rate) == 1:
+            self.update = lambda: 0
+            self.sub_weight = Parameter(self.weight)
+            self.sub_weight_mom = self.weight_mom
+        else:
+            self.sub_weight = Parameter(torch.empty((0, 0)).cuda(local_rank))
+
+    def save_params(self):
+        torch.save(self.weight.data, self.weight_name)
+        torch.save(self.weight_mom, self.weight_mom_name)
+
+    @torch.no_grad()
+    def sample(self, total_label):
+        index_positive = (self.class_start <= total_label) & (total_label < self.class_start + self.num_local)
+        total_label[~index_positive] = -1
+        total_label[index_positive] -= self.class_start
+        if int(self.sample_rate) != 1:
+            positive = torch.unique(total_label[index_positive], sorted=True)
+            if self.num_sample - positive.size(0) >= 0:
+                perm = torch.rand(size=[self.num_local], device=self.device)
+                perm[positive] = 2.0
+                index = torch.topk(perm, k=self.num_sample)[1]
+                index = index.sort()[0]
+            else:
+                index = positive
+            self.index = index
+            total_label[index_positive] = torch.searchsorted(index, total_label[index_positive])
+            self.sub_weight = Parameter(self.weight[index])
+            self.sub_weight_mom = self.weight_mom[index]
+
+    def forward(self, total_features, norm_weight):
+        torch.cuda.current_stream().wait_stream(self.stream)
+        logits = linear(total_features, norm_weight)
+        return logits
+
+    @torch.no_grad()
+    def update(self):
+        self.weight_mom[self.index] = self.sub_weight_mom
+        self.weight[self.index] = self.sub_weight
+
+    def prepare(self, label, optimizer):
+        with torch.cuda.stream(self.stream):
+            total_label = torch.zeros(
+                size=[self.batch_size * self.world_size], device=self.device, dtype=torch.long)
+            dist.all_gather(list(total_label.chunk(self.world_size, dim=0)), label)
+            self.sample(total_label)
+            optimizer.state.pop(optimizer.param_groups[-1]['params'][0], None)
+            optimizer.param_groups[-1]['params'][0] = self.sub_weight
+            optimizer.state[self.sub_weight]['momentum_buffer'] = self.sub_weight_mom
+            norm_weight = normalize(self.sub_weight)
+            return total_label, norm_weight
+
+    def forward_backward(self, label, features, optimizer):
+        total_label, norm_weight = self.prepare(label, optimizer)
+        total_features = torch.zeros(
+            size=[self.batch_size * self.world_size, self.embedding_size], device=self.device)
+        dist.all_gather(list(total_features.chunk(self.world_size, dim=0)), features.data)
+        total_features.requires_grad = True
+
+        logits = self.forward(total_features, norm_weight)
+        logits = self.margin_softmax(logits, total_label)
+
+        with torch.no_grad():
+            max_fc = torch.max(logits, dim=1, keepdim=True)[0]
+            dist.all_reduce(max_fc, dist.ReduceOp.MAX)
+
+            # calculate exp(logits) and all-reduce
+            logits_exp = torch.exp(logits - max_fc)
+            logits_sum_exp = logits_exp.sum(dim=1, keepdims=True)
+            dist.all_reduce(logits_sum_exp, dist.ReduceOp.SUM)
+
+            # calculate prob
+            logits_exp.div_(logits_sum_exp)
+
+            # get one-hot
+            grad = logits_exp
+            index = torch.where(total_label != -1)[0]
+            one_hot = torch.zeros(size=[index.size()[0], grad.size()[1]], device=grad.device)
+            one_hot.scatter_(1, total_label[index, None], 1)
+
+            # calculate loss
+            loss = torch.zeros(grad.size()[0], 1, device=grad.device)
+            loss[index] = grad[index].gather(1, total_label[index, None])
+            dist.all_reduce(loss, dist.ReduceOp.SUM)
+            loss_v = loss.clamp_min_(1e-30).log_().mean() * (-1)
+
+            # calculate grad
+            grad[index] -= one_hot
+            grad.div_(self.batch_size * self.world_size)
+
+        logits.backward(grad)
+        if total_features.grad is not None:
+            total_features.grad.detach_()
+        x_grad: torch.Tensor = torch.zeros_like(features, requires_grad=True)
+        # feature gradient all-reduce
+        dist.reduce_scatter(x_grad, list(total_features.grad.chunk(self.world_size, dim=0)))
+        x_grad = x_grad * self.world_size
+        # backward backbone
+        return x_grad, loss_v
diff --git a/play.py b/play.py
new file mode 100644
index 0000000..f8e3d51
--- /dev/null
+++ b/play.py
@@ -0,0 +1,14 @@
+import cv2
+cap = cv2.VideoCapture("rtsp://admin:2020@uestc@192.168.30.83:554/h264")
+ret, frame = cap.read()
+h, w = frame.shape[:2]
+print("hight:"+str(h)+"with:"+str(w))
+fps = cap.get(cv2.CAP_PROP_FPS)
+print(fps)
+# while ret:
+#     cv2.imshow('out', frame)
+#     if cv2.waitKey(1) & 0xFF == ord('q'):
+#         break
+#     ret, frame = cap.read()
+cap.release()
+cv2.destroyAllWindows()
\ No newline at end of file
diff --git a/realtime_detect.py b/realtime_detect.py
new file mode 100644
index 0000000..b830415
--- /dev/null
+++ b/realtime_detect.py
@@ -0,0 +1,282 @@
+import argparse
+import subprocess
+import time
+import cv2
+import torch
+import numpy as np
+from skimage import transform as trans
+from PIL import Image, ImageDraw, ImageFont
+from data import cfg_mnet, cfg_re50
+from face_api import load_arcface_model, load_npy
+from layers.functions.prior_box import PriorBox
+from retinaface_detect import set_retinaface_conf, load_retinaface_model, findAll
+from utils.nms.py_cpu_nms import py_cpu_nms
+from utils.box_utils import decode, decode_landm
+import faiss
+
+ppi = 1280
+ppi2 = 1100
+step = 3
+
+def detect_rtsp(rtsp, out_rtsp, net, arcface_model, index ,database_name_list, k_v, args):
+    tic_total = time.time()
+    cfg = None
+    if args.network == "mobile0.25":
+        cfg = cfg_mnet
+    elif args.network == "resnet50":
+        cfg = cfg_re50
+    device = torch.device("cpu" if args.cpu else "cuda")
+    resize = 1
+
+    # testing begin
+    cap = cv2.VideoCapture(rtsp)
+    ret, frame = cap.read()
+    h, w = frame.shape[:2]
+
+    factor = 0
+    if (w > ppi):
+        factor = h / w
+        frame = cv2.resize(frame, (ppi, int(ppi * factor)))
+        h, w = frame.shape[:2]
+    arf = 1
+    detect_h, detect_w = frame.shape[:2]
+    frame_detect = frame
+    factor2 = 0
+    if (w > ppi2):
+        factor2 = h / w
+        frame_detect = cv2.resize(frame, (ppi2, int(ppi2 * factor2)))
+        detect_h, detect_w = frame_detect.shape[:2]
+        arf = w/detect_w
+    print(w,h)
+    print(detect_w,detect_h)
+
+    fps = cap.get(cv2.CAP_PROP_FPS)
+    #print(fps)
+    size = (w, h)
+    sizeStr = str(size[0]) + 'x' + str(size[1])
+    if(out_rtsp.startswith("rtsp")):
+        command = ['ffmpeg',
+                   '-y', '-an',
+                   '-f', 'rawvideo',
+                   '-vcodec', 'rawvideo',
+                   '-pix_fmt', 'bgr24',
+                   '-s', sizeStr,
+                   '-r', "25",
+                   '-i', '-',
+                   '-c:v', 'libx265',
+                   '-b:v', '3000k',
+                   '-pix_fmt', 'yuv420p',
+                   '-preset', 'ultrafast',
+                   '-f', 'rtsp',
+                   out_rtsp]
+    pipe = subprocess.Popen(command, shell=False, stdin=subprocess.PIPE)
+    #out = cv2.VideoWriter("output.avi", cv2.VideoWriter_fourcc(*'XVID'), fps, size)
+    number = step
+    dets = []
+    name_list = []
+    font = ImageFont.truetype("font.ttf", 22)
+    priorbox = PriorBox(cfg, image_size=(detect_h, detect_w))
+    priors = priorbox.forward()
+    priors = priors.to(device)
+    prior_data = priors.data
+
+    scale = torch.Tensor([detect_w, detect_h, detect_w, detect_h])
+    scale = scale.to(device)
+    scale1 = torch.Tensor([detect_w, detect_h, detect_w, detect_h,
+                           detect_w, detect_h, detect_w, detect_h,
+                           detect_w, detect_h])
+    scale1 = scale1.to(device)
+
+    src1 = np.array([
+        [38.3814, 51.6963],
+        [73.6186, 51.5014],
+        [56.1120, 71.7366],
+        [41.6361, 92.3655],
+        [70.8167, 92.2041]], dtype=np.float32)
+    tform = trans.SimilarityTransform()
+
+    while ret:
+        tic_all = time.time()
+        if number == step:
+            tic = time.time()
+            img = np.float32(frame_detect)
+            img -= (104, 117, 123)
+            img = img.transpose(2, 0, 1)
+            img = torch.from_numpy(img).unsqueeze(0)
+            img = img.to(device)
+
+            loc, conf, landms = net(img)  # forward pass
+
+            boxes = decode(loc.data.squeeze(0), prior_data, cfg['variance'])
+            boxes = boxes * scale / resize
+            boxes = boxes.cpu().numpy()
+            scores = conf.squeeze(0).data.cpu().numpy()[:, 1]
+            landms = decode_landm(landms.data.squeeze(0), prior_data, cfg['variance'])
+
+            landms = landms * scale1 / resize
+            landms = landms.cpu().numpy()
+
+            # ignore low scores
+            inds = np.where(scores > args.confidence_threshold)[0]
+            boxes = boxes[inds]
+            landms = landms[inds]
+            scores = scores[inds]
+
+            # keep top-K before NMS
+            order = scores.argsort()[::-1][:args.top_k]
+            boxes = boxes[order]
+            landms = landms[order]
+            scores = scores[order]
+
+            # do NMS
+            dets = np.hstack((boxes, scores[:, np.newaxis])).astype(np.float32, copy=False)
+            keep = py_cpu_nms(dets, args.nms_threshold)
+            # keep = nms(dets, args.nms_threshold,force_cpu=args.cpu)
+            dets = dets[keep, :]
+            landms = landms[keep]
+
+            # keep top-K faster NMS
+            dets = dets[:args.keep_top_k, :]
+            landms = landms[:args.keep_top_k, :]
+
+            dets = np.concatenate((dets, landms), axis=1)
+            face_list = []
+            name_list = []
+            print('net forward time: {:.4f}'.format(time.time() - tic))
+            start_time_findall = time.time()
+            for i, det in enumerate(dets[:4]):
+                if det[4] < args.vis_thres:
+                    continue
+                #boxes, score = det[:4], det[4]
+                dst = np.reshape(landms[i], (5, 2))
+                dst = dst * arf
+
+                tform.estimate(dst, src1)
+                M = tform.params[0:2, :]
+                frame2 = cv2.warpAffine(frame, M, (w, h), borderValue=0.0)
+                img112 = frame2[0:112, 0:112, :]
+                face_list.append(img112)
+
+            if len(face_list) != 0:
+                face_list = np.array(face_list)
+                face_list = face_list.transpose((0, 3, 1, 2))
+                face_list = np.array(face_list, dtype=np.float32)
+                face_list -= 127.5
+                face_list /= 127.5
+                print(face_list.shape)
+                print("warpALL time: " + str(time.time() - start_time_findall ))
+                #start_time = time.time()
+                name_list = findAll(face_list, arcface_model, index ,database_name_list, k_v, "cpu" if args.cpu else "cuda")
+                #print(name_list)
+
+            #print("findOneframe time: " + str(time.time() - start_time_findall))
+            # start_time = time.time()
+            # if (len(dets) != 0):
+            #     for i, det in enumerate(dets[:]):
+            #         if det[4] < args.vis_thres:
+            #             continue
+            #         boxes, score = det[:4], det[4]
+            #         boxes = boxes * arf
+            #         name = name_list[i]
+            #         cv2.rectangle(frame, (int(boxes[0]), int(boxes[1])), (int(boxes[2]), int(boxes[3])), (255, 0, 0), 2)
+            #         cv2.putText(frame, name, (int(boxes[0]), int(boxes[1])), cv2.FONT_HERSHEY_COMPLEX, 0.4,(0, 225, 255), 1)
+            start_time = time.time()
+            if(len(dets) != 0):
+                img_PIL = Image.fromarray(cv2.cvtColor(frame, cv2.COLOR_BGR2RGB))
+                draw = ImageDraw.Draw(img_PIL)
+                for i, det in enumerate(dets[:4]):
+                    if det[4] < args.vis_thres:
+                        continue
+                    boxes, score = det[:4], det[4]
+                    boxes = boxes * arf
+                    name = name_list[i]
+                    if not isinstance(name, np.unicode):
+                        name = name.decode('utf8')
+                    draw.text((int(boxes[0]), int(boxes[1])), name, fill=(255, 0, 0), font=font)
+                    draw.rectangle((int(boxes[0]), int(boxes[1]), int(boxes[2]), int(boxes[3])), outline="green", width=3)
+                frame = cv2.cvtColor(np.asarray(img_PIL), cv2.COLOR_RGB2BGR)
+            pipe.stdin.write(frame.tostring())
+            #out.write(frame)
+            print("drawOneframe time: " + str(time.time() - start_time))
+            start_time = time.time()
+            ret, frame = cap.read()
+            frame_detect = frame
+            number = 0
+            if (ret != 0 and factor != 0):
+                frame = cv2.resize(frame, (ppi, int(ppi * factor)))
+            if (ret != 0 and factor2 != 0):
+                frame_detect = cv2.resize(frame, (ppi2, int(ppi2 * factor2)))
+            print("readframe time: " + str(time.time() - start_time))
+        else:
+            number += 1
+            # if (len(dets) != 0):
+            #     for i, det in enumerate(dets[:4]):
+            #         if det[4] < args.vis_thres:
+            #             continue
+            #         boxes, score = det[:4], det[4]
+            #         cv2.rectangle(frame, (int(boxes[0]), int(boxes[1])), (int(boxes[2]), int(boxes[3])), (2, 255, 0), 1)
+            if (len(dets) != 0):
+                img_PIL = Image.fromarray(cv2.cvtColor(frame, cv2.COLOR_BGR2RGB))
+                draw = ImageDraw.Draw(img_PIL)
+                for i, det in enumerate(dets[:4]):
+                    if det[4] < args.vis_thres:
+                        continue
+                    boxes, score = det[:4], det[4]
+                    boxes = boxes * arf
+                    name = name_list[i]
+                    if not isinstance(name, np.unicode):
+                        name = name.decode('utf8')
+                    draw.text((int(boxes[0]), int(boxes[1])), name, fill=(255, 0, 0), font=font)
+                    draw.rectangle((int(boxes[0]), int(boxes[1]), int(boxes[2]), int(boxes[3])), outline="green",
+                                   width=3)
+                frame = cv2.cvtColor(np.asarray(img_PIL), cv2.COLOR_RGB2BGR)
+            start_time = time.time()
+            pipe.stdin.write(frame.tostring())
+            #out.write(frame)
+            print("writeframe time: " + str(time.time() - start_time))
+            start_time = time.time()
+            ret, frame = cap.read()
+            frame_detect = frame
+            if (ret != 0 and factor != 0):
+                frame = cv2.resize(frame, (ppi, int(ppi * factor)))
+            if (ret != 0 and factor2 != 0):
+                frame_detect = cv2.resize(frame, (ppi2, int(ppi2 * factor2)))
+            print("readframe time: " + str(time.time() - start_time))
+        print('all time: {:.4f}'.format(time.time() - tic_all))
+    cap.release()
+    #out.release()
+    pipe.terminate()
+    print('total time: {:.4f}'.format(time.time() - tic_total))
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "--rtsp",
+        type=str,
+        default="",
+        dest="rtsp_path"
+    )
+    args = parser.parse_args()
+    cpu_or_cuda = "cuda" if torch.cuda.is_available() else "cpu"
+    # 加载人脸识别模型
+    arcface_model = load_arcface_model("./model/backbone100.pth", cpu_or_cuda="cuda")
+    # 加载人脸检测模型
+    retinaface_args = set_retinaface_conf(cpu_or_cuda=cpu_or_cuda)
+    retinaface_model = load_retinaface_model(retinaface_args)
+    k_v = load_npy("./Database/student.npy")
+    #print(list(k_v.keys()))
+    database_name_list = list(k_v.keys())
+    vector_list = np.array(list(k_v.values()))
+    print(vector_list.shape)
+    nlist = 10
+    quantizer = faiss.IndexFlatL2(512)  # the other index
+    index = faiss.IndexIVFFlat(quantizer, 512, nlist, faiss.METRIC_L2)
+    index.train(vector_list)
+    #index = faiss.IndexFlatL2(512)
+    index.add(vector_list)
+    index.nprobe=10
+
+    detect_rtsp(args.rtsp_path, 'rtsp://localhost:5001/test2', retinaface_model, arcface_model, index ,database_name_list, k_v, retinaface_args)
+
+    #detect_rtsp("rtsp://admin:2020@uestc@192.168.14.32:8557/h264", 'rtsp://localhost:5001/test2', retinaface_model, arcface_model, index ,database_name_list, k_v, retinaface_args)
+    #detect_rtsp("cut.mp4", 'rtsp://localhost:5001/test2', retinaface_model, arcface_model, k_v, retinaface_args)
diff --git a/recognition_video.py b/recognition_video.py
new file mode 100644
index 0000000..cc45a75
--- /dev/null
+++ b/recognition_video.py
@@ -0,0 +1,283 @@
+import time
+from centerface import CenterFace
+from skimage import transform as trans
+import numpy as np
+import torch
+import cv2
+from backbones import iresnet100, iresnet18
+from create_database import findOne, load_npy,findAll
+from PIL import Image, ImageDraw,ImageFont
+
+def show():
+    cap = cv2.VideoCapture("test.mp4")
+    ret, frame = cap.read()
+    h, w = frame.shape[:2]
+    centerface = CenterFace()
+    size = (int(cap.get(cv2.CAP_PROP_FRAME_WIDTH)),
+            int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT)))
+    out = cv2.VideoWriter('ccvt6.mp4', cv2.VideoWriter_fourcc('m', 'p', '4', 'v'), 30, size)
+    while ret:
+        start_time = time.time()
+        dets, lms = centerface(frame, h, w, threshold=0.35)
+        end_time = time.time()
+        print("findOne time: " + str(end_time - start_time))
+        for det in dets:
+            boxes, score = det[:4], det[4]
+            cv2.rectangle(frame, (int(boxes[0]), int(boxes[1])), (int(boxes[2]), int(boxes[3])), (2, 255, 0), 1)
+        for lm in lms:
+            for i in range(0, 5):
+                cv2.circle(frame, (int(lm[i * 2]), int(lm[i * 2 + 1])), 2, (0, 0, 255), -1)
+        cv2.imshow('out', frame)
+        out.write(frame)
+        # Press Q on keyboard to stop recording
+        if cv2.waitKey(1) & 0xFF == ord('q'):
+            break
+        ret, frame = cap.read()
+    cap.release()
+    out.release()
+    cv2.destroyAllWindows()
+
+def video():
+    model = iresnet100()
+    model.load_state_dict(torch.load("./model/backbone100.pth", map_location="cpu"))
+    model.eval()
+    k_v = load_npy("student.npy")
+    count = 0
+    #cap = cv2.VideoCapture("http://ivi.bupt.edu.cn/hls/cctv6hd.m3u8")
+    cap = cv2.VideoCapture("software.mp4")
+    ret, frame = cap.read()
+    h, w = frame.shape[:2]
+    size = (int(cap.get(cv2.CAP_PROP_FRAME_WIDTH)),
+            int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT)))
+    fps = cap.get(cv2.CAP_PROP_FPS)
+    out = cv2.VideoWriter('ttt.mp4', cv2.VideoWriter_fourcc('m', 'p', '4', 'v'), fps, size)
+    centerface = CenterFace()
+    while ret:
+        start_time = time.time()
+        dets, lms = centerface(frame, h, w, threshold=0.35)
+        end_time = time.time()
+        print("detectOneframe time: " + str(end_time - start_time))
+        face_list = []
+        name_list = []
+        for i,det in enumerate(dets):
+            boxes, score = det[:4], det[4]
+            img_w = int(boxes[2] - boxes[0])
+            img_h = int(boxes[3] - boxes[1])
+            distace = int(abs(img_w - img_h) / 2)
+            img_w1 = int(boxes[0]) - distace
+            img_w2 = int(boxes[2]) + distace
+            # print(img_w,img_h,distace,max_hw)
+            if img_w <= img_h and img_w1 >= 0 and img_w2 <= frame.shape[1]:
+                img112 = frame[int(boxes[1]):int(boxes[3]), img_w1:img_w2, :]
+                img112 = cv2.resize(img112, (112, 112))
+                # cv2.imwrite("./img/man"+str(count)+".jpg", img112)
+                # count += 1
+                face_list.append(img112)
+            else:
+                img112 = frame[int(boxes[1]):int(boxes[3]), int(boxes[0]):int(boxes[2]), :]
+                img112 = cv2.resize(img112, (112, 112))
+                face_list.append(img112)
+        if len(face_list) != 0:
+            face_list = np.array(face_list)
+            face_list = face_list.transpose((0,3,1,2))
+            face_list = np.array(face_list, dtype=np.float32)
+            face_list -= 127.5
+            face_list /= 127.5
+            print(face_list.shape)
+            face_list = torch.from_numpy(face_list)
+            start_time = time.time()
+
+            for face in face_list:
+                face = face[np.newaxis, :, :, :]
+
+                name_list.append(findOne(face,model,k_v))
+            end_time = time.time()
+            print("findOneframe time: "+str(end_time-start_time))
+        img_PIL = Image.fromarray(cv2.cvtColor(frame, cv2.COLOR_BGR2RGB))
+        draw = ImageDraw.Draw(img_PIL)
+        font = ImageFont.truetype("font.ttf",12)
+        for i,det in enumerate(dets):
+            boxes, score = det[:4], det[4]
+            # cv2.rectangle(frame, (int(boxes[0]), int(boxes[1])), (int(boxes[2]), int(boxes[3])), (2, 255, 0), 1)
+            # cv2.putText(frame, name_list[i], (int(boxes[0]), int(boxes[1])), cv2.FONT_HERSHEY_COMPLEX, 0.4,
+            #             (0, 225, 255), 1)
+            name = name_list[i][:3]
+            if not isinstance(name, np.unicode):
+                name = name.decode('utf8')
+            draw.text((int(boxes[0]), int(boxes[1])),name,fill=(0, 225, 255),font=font)
+            draw.rectangle((int(boxes[0]), int(boxes[1]),int(boxes[2]), int(boxes[3])),outline="green",width=1)
+        frame = cv2.cvtColor(np.asarray(img_PIL),cv2.COLOR_RGB2BGR)
+        cv2.imshow('out', frame)
+        out.write(frame)
+        # Press Q on keyboard to stop recording
+        if cv2.waitKey(1) & 0xFF == ord('q'):
+            break
+        ret, frame = cap.read()
+    cap.release()
+    out.release()
+    cv2.destroyAllWindows()
+def video_GPU():
+    model = iresnet100()
+    model.load_state_dict(torch.load("./model/backbone100.pth", map_location="cpu"))
+    model.eval()
+    k_v = load_npy("student.npy")
+    count = 0
+    #cap = cv2.VideoCapture("http://ivi.bupt.edu.cn/hls/cctv6hd.m3u8")
+    cap = cv2.VideoCapture("software.mp4")
+    ret, frame = cap.read()
+    h, w = frame.shape[:2]
+    size = (int(cap.get(cv2.CAP_PROP_FRAME_WIDTH)),
+            int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT)))
+    fps = cap.get(cv2.CAP_PROP_FPS)
+    out = cv2.VideoWriter('ttt.mp4', cv2.VideoWriter_fourcc('m', 'p', '4', 'v'), fps, size)
+    centerface = CenterFace()
+    while ret:
+        start_time = time.time()
+        dets, lms = centerface(frame, h, w, threshold=0.35)
+        end_time = time.time()
+        print("detectOneframe time: " + str(end_time - start_time))
+        face_list = []
+        name_list = []
+        for i,det in enumerate(dets):
+            boxes, score = det[:4], det[4]
+            img_w = int(boxes[2] - boxes[0])
+            img_h = int(boxes[3] - boxes[1])
+            distace = int(abs(img_w - img_h) / 2)
+            img_w1 = int(boxes[0]) - distace
+            img_w2 = int(boxes[2]) + distace
+            # print(img_w,img_h,distace,max_hw)
+            if img_w <= img_h and img_w1 >= 0 and img_w2 <= frame.shape[1]:
+                img112 = frame[int(boxes[1]):int(boxes[3]), img_w1:img_w2, :]
+                img112 = cv2.resize(img112, (112, 112))
+                # cv2.imwrite("./img/man"+str(count)+".jpg", img112)
+                # count += 1
+                face_list.append(img112)
+            else:
+                img112 = frame[int(boxes[1]):int(boxes[3]), int(boxes[0]):int(boxes[2]), :]
+                img112 = cv2.resize(img112, (112, 112))
+                face_list.append(img112)
+        if len(face_list) != 0:
+            face_list = np.array(face_list)
+            face_list = face_list.transpose((0,3,1,2))
+            face_list = np.array(face_list, dtype=np.float32)
+            face_list -= 127.5
+            face_list /= 127.5
+            print(face_list.shape)
+            face_list = torch.from_numpy(face_list)
+            start_time = time.time()
+            name_list = findAll(face_list, model, k_v)
+            # for face in face_list:
+            #     face = face[np.newaxis, :, :, :]
+            #
+            #     name_list.append(findOne(face,model,k_v))
+            end_time = time.time()
+            print("findOneframe time: "+str(end_time-start_time))
+        img_PIL = Image.fromarray(cv2.cvtColor(frame, cv2.COLOR_BGR2RGB))
+        draw = ImageDraw.Draw(img_PIL)
+        font = ImageFont.truetype("font.ttf",18)
+        for i,det in enumerate(dets):
+            boxes, score = det[:4], det[4]
+            # cv2.rectangle(frame, (int(boxes[0]), int(boxes[1])), (int(boxes[2]), int(boxes[3])), (2, 255, 0), 1)
+            # cv2.putText(frame, name_list[i], (int(boxes[0]), int(boxes[1])), cv2.FONT_HERSHEY_COMPLEX, 0.4,
+            #             (0, 225, 255), 1)
+            name = name_list[i][:3]
+            if not isinstance(name, np.unicode):
+                name = name.decode('utf8')
+            draw.text((int(boxes[0]), int(boxes[1])),name,fill=(255, 0, 0),font=font)
+            draw.rectangle((int(boxes[0]), int(boxes[1]),int(boxes[2]), int(boxes[3])),outline="green",width=2)
+        frame = cv2.cvtColor(np.asarray(img_PIL),cv2.COLOR_RGB2BGR)
+        cv2.imshow('out', frame)
+        out.write(frame)
+        # Press Q on keyboard to stop recording
+        if cv2.waitKey(1) & 0xFF == ord('q'):
+            break
+        ret, frame = cap.read()
+    cap.release()
+    out.release()
+    cv2.destroyAllWindows()
+
+def video_GPU_retinaface():
+    model = iresnet100()
+    model.load_state_dict(torch.load("./model/backbone100.pth", map_location="cpu"))
+    model.eval()
+    k_v = load_npy("student.npy")
+    count = 0
+    #cap = cv2.VideoCapture("http://ivi.bupt.edu.cn/hls/cctv6hd.m3u8")
+    cap = cv2.VideoCapture("software.mp4")
+    ret, frame = cap.read()
+    h, w = frame.shape[:2]
+    size = (int(cap.get(cv2.CAP_PROP_FRAME_WIDTH)),
+            int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT)))
+    fps = cap.get(cv2.CAP_PROP_FPS)
+    out = cv2.VideoWriter('ttt.mp4', cv2.VideoWriter_fourcc('m', 'p', '4', 'v'), fps, size)
+    centerface = CenterFace()
+    while ret:
+        start_time = time.time()
+        dets, lms = centerface(frame, h, w, threshold=0.35)
+        end_time = time.time()
+        print("detectOneframe time: " + str(end_time - start_time))
+        face_list = []
+        name_list = []
+        print(dets.shape)
+        for i,det in enumerate(dets):
+            boxes, score = det[:4], det[4]
+            img_w = int(boxes[2] - boxes[0])
+            img_h = int(boxes[3] - boxes[1])
+            distace = int(abs(img_w - img_h) / 2)
+            img_w1 = int(boxes[0]) - distace
+            img_w2 = int(boxes[2]) + distace
+            # print(img_w,img_h,distace,max_hw)
+            if img_w <= img_h and img_w1 >= 0 and img_w2 <= frame.shape[1]:
+                img112 = frame[int(boxes[1]):int(boxes[3]), img_w1:img_w2, :]
+                img112 = cv2.resize(img112, (112, 112))
+                # cv2.imwrite("./img/man"+str(count)+".jpg", img112)
+                # count += 1
+                face_list.append(img112)
+            else:
+                img112 = frame[int(boxes[1]):int(boxes[3]), int(boxes[0]):int(boxes[2]), :]
+                img112 = cv2.resize(img112, (112, 112))
+                face_list.append(img112)
+        if len(face_list) != 0:
+            face_list = np.array(face_list)
+            face_list = face_list.transpose((0,3,1,2))
+            face_list = np.array(face_list, dtype=np.float32)
+            face_list -= 127.5
+            face_list /= 127.5
+            print(face_list.shape)
+            face_list = torch.from_numpy(face_list)
+            start_time = time.time()
+            name_list = findAll(face_list, model, k_v)
+            # for face in face_list:
+            #     face = face[np.newaxis, :, :, :]
+            #
+            #     name_list.append(findOne(face,model,k_v))
+            end_time = time.time()
+            print("findOneframe time: "+str(end_time-start_time))
+        img_PIL = Image.fromarray(cv2.cvtColor(frame, cv2.COLOR_BGR2RGB))
+        draw = ImageDraw.Draw(img_PIL)
+        font = ImageFont.truetype("font.ttf",18)
+        for i,det in enumerate(dets):
+            boxes, score = det[:4], det[4]
+            # cv2.rectangle(frame, (int(boxes[0]), int(boxes[1])), (int(boxes[2]), int(boxes[3])), (2, 255, 0), 1)
+            # cv2.putText(frame, name_list[i], (int(boxes[0]), int(boxes[1])), cv2.FONT_HERSHEY_COMPLEX, 0.4,
+            #             (0, 225, 255), 1)
+            name = name_list[i][:3]
+            if not isinstance(name, np.unicode):
+                name = name.decode('utf8')
+            draw.text((int(boxes[0]), int(boxes[1])),name,fill=(255, 0, 0),font=font)
+            draw.rectangle((int(boxes[0]), int(boxes[1]),int(boxes[2]), int(boxes[3])),outline="green",width=2)
+        frame = cv2.cvtColor(np.asarray(img_PIL),cv2.COLOR_RGB2BGR)
+        cv2.imshow('out', frame)
+        out.write(frame)
+        # Press Q on keyboard to stop recording
+        if cv2.waitKey(1) & 0xFF == ord('q'):
+            break
+        ret, frame = cap.read()
+    cap.release()
+    out.release()
+    cv2.destroyAllWindows()
+
+
+video_GPU_retinaface()
+#video_GPU()
+#show()
diff --git a/requirements.txt b/requirements.txt
new file mode 100644
index 0000000..fbb1f29
--- /dev/null
+++ b/requirements.txt
@@ -0,0 +1,111 @@
+Package                Version    
+---------------------- -----------
+appdirs                1.4.4      
+attrs                  21.2.0     
+backcall               0.2.0      
+beautifulsoup4         4.9.3      
+certifi                2021.5.30  
+cffi                   1.14.0     
+chardet                4.0.0      
+click                  8.0.1      
+conda                  4.9.1      
+conda-build            3.20.5     
+conda-package-handling 1.7.0      
+cryptography           2.9.2      
+cycler                 0.10.0     
+dataclasses            0.6        
+decorator              4.4.2      
+dnspython              2.0.0      
+faiss-cpu              1.7.1      
+filelock               3.0.12     
+fire                   0.4.0      
+Flask                  1.1.2      
+future                 0.18.2     
+glob2                  0.7        
+graphsurgeon           0.4.5      
+graphviz               0.8.4      
+h5py                   3.3.0      
+idna                   2.10       
+imageio                2.9.0      
+iniconfig              1.1.1      
+ipython                7.18.1     
+ipython-genutils       0.2.0      
+itsdangerous           2.0.1      
+jedi                   0.17.2     
+Jinja2                 3.0.1      
+joblib                 1.0.1      
+kiwisolver             1.3.1      
+libarchive-c           2.9        
+Mako                   1.1.4      
+MarkupSafe             2.0.1      
+matplotlib             3.4.1      
+mkl-fft                1.2.0      
+mkl-random             1.1.1      
+mkl-service            2.3.0      
+mxnet                  1.8.0.post0
+networkx               2.5.1      
+nltk                   3.6        
+numpy                  1.20.3     
+olefile                0.46       
+opencv-python          4.5.1.48   
+packaging              21.0       
+pandas                 1.2.4      
+parso                  0.7.0      
+pexpect                4.8.0      
+pickleshare            0.7.5      
+Pillow                 8.0.0      
+pip                    20.0.2     
+pkginfo                1.6.0      
+pluggy                 1.0.0      
+prefetch-generator     1.0.1      
+prompt-toolkit         3.0.8      
+protobuf               3.15.8     
+psutil                 5.7.2      
+ptyprocess             0.6.0      
+py                     1.9.0      
+pycosat                0.6.3      
+pycparser              2.20       
+pycuda                 2021.1     
+Pygments               2.7.1      
+pyOpenSSL              19.1.0     
+pyparsing              2.4.7      
+PySocks                1.7.1      
+pytest                 6.2.5      
+python-dateutil        2.8.1      
+python-etcd            0.4.5      
+pytools                2021.2.6   
+pytz                   2020.1     
+PyWavelets             1.1.1      
+PyYAML                 5.3.1      
+pyzmq                  22.1.0     
+regex                  2021.8.3   
+requests               2.25.1     
+ruamel-yaml            0.15.87    
+scikit-image           0.18.1     
+scipy                  1.6.3      
+seaborn                0.11.1     
+setuptools             57.1.0     
+six                    1.14.0     
+soupsieve              2.0.1      
+tensorboard-logger     0.1.0      
+tensorrt               7.2.3.4    
+termcolor              1.1.0      
+tifffile               2021.4.8   
+toml                   0.10.2     
+torch                  1.7.1      
+torch2trt              0.2.0      
+torchelastic           0.2.1      
+torchfile              0.1.0      
+torchtext              0.8.0      
+torchvision            0.8.2      
+tornado                6.1        
+tqdm                   4.46.0     
+traitlets              5.0.5      
+typing-extensions      3.7.4.3    
+uff                    0.6.9      
+urllib3                1.26.5     
+visdom                 0.1.8      
+wcwidth                0.2.5      
+websocket-client       1.1.0      
+Werkzeug               2.0.1      
+wheel                  0.34.2     
diff --git a/retinaface_arcface.py b/retinaface_arcface.py
new file mode 100644
index 0000000..2660192
--- /dev/null
+++ b/retinaface_arcface.py
@@ -0,0 +1,762 @@
+from __future__ import print_function
+import os
+import argparse
+import re
+
+import faiss
+import torch
+import torch.backends.cudnn as cudnn
+import numpy as np
+from data import cfg_mnet, cfg_re50
+from face_api import create_database_from_img, load_arcface_model, findAll
+from layers.functions.prior_box import PriorBox
+from utils.nms.py_cpu_nms import py_cpu_nms
+import cv2
+from models.retinaface import RetinaFace
+from utils.box_utils import decode, decode_landm
+import time
+from face_api import load_arcface_model, load_npy
+from skimage import transform as trans
+from backbones import iresnet100, iresnet18
+#from create_database import findOne, load_npy,findAll
+from PIL import Image, ImageDraw,ImageFont
+
+parser = argparse.ArgumentParser(description='Retinaface')
+
+parser.add_argument('-m', '--trained_model', default='./weights/mobilenet0.25_Final.pth',
+                    type=str, help='Trained state_dict file path to open')
+parser.add_argument('--network', default='mobile0.25', help='Backbone network mobile0.25 or resnet50')
+parser.add_argument('--cpu', action="store_true", default=False if torch.cuda.is_available() else True, help='Use cpu inference')
+parser.add_argument('--confidence_threshold', default=0.02, type=float, help='confidence_threshold')
+parser.add_argument('--top_k', default=5000, type=int, help='top_k')
+parser.add_argument('--nms_threshold', default=0.4, type=float, help='nms_threshold')
+parser.add_argument('--keep_top_k', default=750, type=int, help='keep_top_k')
+parser.add_argument('-s', '--save_image', action="store_true", default=True, help='show detection results')
+parser.add_argument('--vis_thres', default=0.6, type=float, help='visualization_threshold')
+args = parser.parse_args()
+
+
+def check_keys(model, pretrained_state_dict):
+    ckpt_keys = set(pretrained_state_dict.keys())
+    model_keys = set(model.state_dict().keys())
+    used_pretrained_keys = model_keys & ckpt_keys
+    unused_pretrained_keys = ckpt_keys - model_keys
+    missing_keys = model_keys - ckpt_keys
+    print('Missing keys:{}'.format(len(missing_keys)))
+    print('Unused checkpoint keys:{}'.format(len(unused_pretrained_keys)))
+    print('Used keys:{}'.format(len(used_pretrained_keys)))
+    assert len(used_pretrained_keys) > 0, 'load NONE from pretrained checkpoint'
+    return True
+
+
+def remove_prefix(state_dict, prefix):
+    ''' Old style model is stored with all names of parameters sharing common prefix 'module.' '''
+    print('remove prefix \'{}\''.format(prefix))
+    f = lambda x: x.split(prefix, 1)[-1] if x.startswith(prefix) else x
+    return {f(key): value for key, value in state_dict.items()}
+
+
+def load_model(model, pretrained_path, load_to_cpu):
+    print('Loading pretrained model from {}'.format(pretrained_path))
+    if load_to_cpu:
+        pretrained_dict = torch.load(pretrained_path, map_location=lambda storage, loc: storage)
+    else:
+        device = torch.cuda.current_device()
+        pretrained_dict = torch.load(pretrained_path, map_location=lambda storage, loc: storage.cuda(device))
+    if "state_dict" in pretrained_dict.keys():
+        pretrained_dict = remove_prefix(pretrained_dict['state_dict'], 'module.')
+    else:
+        pretrained_dict = remove_prefix(pretrained_dict, 'module.')
+    check_keys(model, pretrained_dict)
+    model.load_state_dict(pretrained_dict, strict=False)
+    return model
+
+def image_to112x112_retinaface():
+    torch.set_grad_enabled(False)
+    cfg = None
+    if args.network == "mobile0.25":
+        cfg = cfg_mnet
+    elif args.network == "resnet50":
+        cfg = cfg_re50
+    # net and model
+    net = RetinaFace(cfg=cfg, phase = 'test')
+    net = load_model(net, args.trained_model, args.cpu)
+    net.eval()
+    print('Finished loading model!')
+    #print(net)
+    cudnn.benchmark = True
+    device = torch.device("cpu" if args.cpu else "cuda")
+    net = net.to(device)
+
+    resize = 1
+    input_path = r"D:\Download\out\cfp"
+    output_path = "D:\Download\out\cfp_align"
+    folder1 = os.listdir(input_path)
+    count = 0
+    count2 =0
+    for f in folder1:
+        output_name_path = os.path.join(output_path, f)
+        if os.path.exists(output_name_path) == 0:
+            os.makedirs(output_name_path)
+        img_name_path = os.path.join(input_path, f)
+        img_list = os.listdir(img_name_path)
+
+        for img in img_list:
+            count2 +=1
+            print(count2)
+            path = os.path.join(img_name_path, img)
+            align_img_path = os.path.join(output_name_path, img)
+            # print(path)
+            frame = cv2.imread(path)
+            h, w = frame.shape[:2]
+            img = np.float32(frame)
+            im_height, im_width, _ = img.shape
+            scale = torch.Tensor([img.shape[1], img.shape[0], img.shape[1], img.shape[0]])
+            img -= (104, 117, 123)
+            img = img.transpose(2, 0, 1)
+            img = torch.from_numpy(img).unsqueeze(0)
+            img = img.to(device)
+            scale = scale.to(device)
+
+            tic = time.time()
+            loc, conf, landms = net(img)  # forward pass
+            print('net forward time: {:.4f}'.format(time.time() - tic))
+
+            priorbox = PriorBox(cfg, image_size=(im_height, im_width))
+            priors = priorbox.forward()
+            priors = priors.to(device)
+            prior_data = priors.data
+            boxes = decode(loc.data.squeeze(0), prior_data, cfg['variance'])
+            boxes = boxes * scale / resize
+            boxes = boxes.cpu().numpy()
+            scores = conf.squeeze(0).data.cpu().numpy()[:, 1]
+            landms = decode_landm(landms.data.squeeze(0), prior_data, cfg['variance'])
+            scale1 = torch.Tensor([img.shape[3], img.shape[2], img.shape[3], img.shape[2],
+                                   img.shape[3], img.shape[2], img.shape[3], img.shape[2],
+                                   img.shape[3], img.shape[2]])
+            scale1 = scale1.to(device)
+            landms = landms * scale1 / resize
+            landms = landms.cpu().numpy()
+
+            # ignore low scores
+            inds = np.where(scores > args.confidence_threshold)[0]
+            boxes = boxes[inds]
+            landms = landms[inds]
+            scores = scores[inds]
+
+            # keep top-K before NMS
+            order = scores.argsort()[::-1][:args.top_k]
+            boxes = boxes[order]
+            landms = landms[order]
+            scores = scores[order]
+
+            # do NMS
+            dets = np.hstack((boxes, scores[:, np.newaxis])).astype(np.float32, copy=False)
+            keep = py_cpu_nms(dets, args.nms_threshold)
+            # keep = nms(dets, args.nms_threshold,force_cpu=args.cpu)
+            dets = dets[keep, :]
+            landms = landms[keep]
+
+            # keep top-K faster NMS
+            dets = dets[:args.keep_top_k, :]
+            landms = landms[:args.keep_top_k, :]
+
+            dets = np.concatenate((dets, landms), axis=1)
+            score = 500
+            # show image
+            if args.save_image:
+                dst = []
+                for i, det in enumerate(dets):
+                    if det[4] < args.vis_thres:
+                        continue
+                    center_x = (det[2] + det[0]) / 2
+                    center_y = (det[3] + det[1]) / 2
+                    if abs(center_x - 125) + abs(center_y - 125) < score:
+                        score = abs(center_x - 125) + abs(center_y - 125)
+                        dst = np.reshape(landms[i], (5, 2))
+                if len(dst) > 0:
+                    src1 = np.array([
+                        [38.3814, 51.6963],
+                        [73.6186, 51.5014],
+                        [56.1120, 71.7366],
+                        [41.6361, 92.3655],
+                        [70.8167, 92.2041]], dtype=np.float32)
+                    tform = trans.SimilarityTransform()
+                    tform.estimate(dst, src1)
+                    M = tform.params[0:2, :]
+
+                    if w < 112 or h < 112:
+                        count += 1
+                        #print(align_img_path)
+                        continue
+                    frame = cv2.warpAffine(frame, M, (w, h), borderValue=0.0)
+                    img112 = frame[0:112, 0:112, :]
+                    cv2.imwrite(align_img_path, img112)
+    print(">112 number"+str(count))
+
+def sfz_to112x112_retinaface(arcface_model,cpu_or_cuda):
+    torch.set_grad_enabled(False)
+    cfg = None
+    if args.network == "mobile0.25":
+        cfg = cfg_mnet
+    elif args.network == "resnet50":
+        cfg = cfg_re50
+    # net and model
+    net = RetinaFace(cfg=cfg, phase = 'test')
+    net = load_model(net, args.trained_model, args.cpu)
+    net.eval()
+    print('Finished loading model!')
+    #print(net)
+    cudnn.benchmark = True
+    device = torch.device("cpu" if args.cpu else "cuda")
+    net = net.to(device)
+
+    resize = 1
+    input_path = r"D:\Download\out\alig_students_all"
+    output_path = r"D:\Download\out\alig_students_all"
+    folder1 = os.listdir(input_path)
+    count = 0
+    count2 =0
+    print(len(folder1))
+    # print(folder1[0][:-4])
+    # return 0
+    order_img = []
+    order_name = []
+    tic = time.time()
+    for img_name in folder1[:2500]:
+        # output_name_path = os.path.join(output_path, img_name)
+        # if os.path.exists(output_name_path) == 0:
+        #     os.makedirs(output_name_path)
+        img_name_path = os.path.join(input_path, img_name)
+        #img_list = os.listdir(img_name_path)
+        count2 += 1
+        if (count2 % 1000 == 0):
+            print('net forward time: {:.4f}'.format(time.time() - tic))
+            print(count2)
+            if len(order_img) > 0:
+                order_img = np.array(order_img)
+                order_img = order_img.transpose((0, 3, 1, 2))
+                order_img = np.array(order_img, dtype=np.float32)
+                order_img -= 127.5
+                order_img /= 127.5
+                # order_img = np.array(order_img)
+                # print(order_img.shape)
+                # print(len(order_name))
+                create_database_from_img(order_name, order_img, arcface_model, "./Database/sfz_test.npy", cpu_or_cuda)
+                order_img = []
+                order_name = []
+                tic = time.time()
+
+        # if img_name[19] != "1":
+        #     continue
+
+        #path = os.path.join(img_name_path, img)
+        align_img_path = os.path.join(output_path, img_name)
+        # print(path)
+        #frame = cv2.imdecode(np.fromfile(img_name_path, dtype=np.uint8), cv2.IMREAD_COLOR)
+        try:
+            frame = cv2.imdecode(np.fromfile(img_name_path, dtype=np.uint8), cv2.IMREAD_COLOR)
+            h, w, d = frame.shape
+        except AttributeError:
+            print(img_name)
+            continue
+        if d == 1:
+            continue
+        factor = h / w
+        if (w > 1000):
+            frame = cv2.resize(frame, (600, int(600 * factor)))
+        h, w = frame.shape[:2]
+        img = np.float32(frame)
+        im_height, im_width, _ = img.shape
+        scale = torch.Tensor([img.shape[1], img.shape[0], img.shape[1], img.shape[0]])
+        img -= (104, 117, 123)
+        img = img.transpose(2, 0, 1)
+        img = torch.from_numpy(img).unsqueeze(0)
+        img = img.to(device)
+        scale = scale.to(device)
+
+        #tic = time.time()
+        loc, conf, landms = net(img)  # forward pass
+        #print('net forward time: {:.4f}'.format(time.time() - tic))
+
+        priorbox = PriorBox(cfg, image_size=(im_height, im_width))
+        priors = priorbox.forward()
+        priors = priors.to(device)
+        prior_data = priors.data
+        boxes = decode(loc.data.squeeze(0), prior_data, cfg['variance'])
+        boxes = boxes * scale / resize
+        boxes = boxes.cpu().numpy()
+        scores = conf.squeeze(0).data.cpu().numpy()[:, 1]
+        landms = decode_landm(landms.data.squeeze(0), prior_data, cfg['variance'])
+        scale1 = torch.Tensor([img.shape[3], img.shape[2], img.shape[3], img.shape[2],
+                               img.shape[3], img.shape[2], img.shape[3], img.shape[2],
+                               img.shape[3], img.shape[2]])
+        scale1 = scale1.to(device)
+        landms = landms * scale1 / resize
+        landms = landms.cpu().numpy()
+
+        # ignore low scores
+        inds = np.where(scores > args.confidence_threshold)[0]
+        boxes = boxes[inds]
+        landms = landms[inds]
+        scores = scores[inds]
+
+        # keep top-K before NMS
+        order = scores.argsort()[::-1][:args.top_k]
+        boxes = boxes[order]
+        landms = landms[order]
+        scores = scores[order]
+
+        # do NMS
+        dets = np.hstack((boxes, scores[:, np.newaxis])).astype(np.float32, copy=False)
+        keep = py_cpu_nms(dets, args.nms_threshold)
+        # keep = nms(dets, args.nms_threshold,force_cpu=args.cpu)
+        dets = dets[keep, :]
+        landms = landms[keep]
+
+        # keep top-K faster NMS
+        dets = dets[:args.keep_top_k, :]
+        landms = landms[:args.keep_top_k, :]
+
+        dets = np.concatenate((dets, landms), axis=1)
+        score = 500
+        # show image
+        if args.save_image:
+            dst = []
+            for i, det in enumerate(dets):
+                if det[4] < args.vis_thres:
+                    continue
+                # center_x = (det[2] + det[0]) / 2
+                # center_y = (det[3] + det[1]) / 2
+                # if abs(center_x - 125) + abs(center_y - 125) < score:
+                #     score = abs(center_x - 125) + abs(center_y - 125)
+                dst = np.reshape(landms[i], (5, 2))
+            if len(dst) > 0:
+                src1 = np.array([
+                    [38.3814, 51.6963],
+                    [73.6186, 51.5014],
+                    [56.1120, 71.7366],
+                    [41.6361, 92.3655],
+                    [70.8167, 92.2041]], dtype=np.float32)
+                tform = trans.SimilarityTransform()
+                tform.estimate(dst, src1)
+                M = tform.params[0:2, :]
+
+                if w < 112 or h < 112:
+                    count += 1
+                    print(img_name_path)
+                    continue
+                frame = cv2.warpAffine(frame, M, (w, h), borderValue=0.0)
+                img112 = frame[0:112, 0:112, :]
+                order_img.append(img112)
+                order_name.append(img_name[:-6])
+                #cv2.imencode('.jpg', img112)[1].tofile(align_img_path)
+                #cv2.imwrite(align_img_path, img112)
+
+    print(">112 number"+str(count))
+    if len(order_img) > 0:
+        order_img = np.array(order_img)
+        order_img = order_img.transpose((0, 3, 1, 2))
+        order_img = np.array(order_img, dtype=np.float32)
+        order_img -= 127.5
+        order_img /= 127.5
+        #order_img = np.array(order_img)
+        # print(order_img.shape)
+        # print(len(order_name))
+        create_database_from_img(order_name, order_img, arcface_model, "./Database/sfz_test.npy", cpu_or_cuda)
+
+def count_accuracy(arcface_model,cpu_or_cuda,index ,database_name_list):
+    torch.set_grad_enabled(False)
+    cfg = None
+    if args.network == "mobile0.25":
+        cfg = cfg_mnet
+    elif args.network == "resnet50":
+        cfg = cfg_re50
+    # net and model
+    net = RetinaFace(cfg=cfg, phase = 'test')
+    net = load_model(net, args.trained_model, args.cpu)
+    net.eval()
+    print('Finished loading model!')
+    #print(net)
+    cudnn.benchmark = True
+    device = torch.device("cpu" if args.cpu else "cuda")
+    net = net.to(device)
+
+    resize = 1
+    input_path = r"../face/czrkzp2"
+    folder1 = os.listdir(input_path)
+    count = 0
+    count2 =0
+    print(len(folder1))
+    # print(folder1[0][:-4])
+    # return 0
+    order_img = []
+    order_name = []
+    tic = time.time()
+    for img_name in folder1[:15000]:
+        # output_name_path = os.path.join(output_path, img_name)
+        # if os.path.exists(output_name_path) == 0:
+        #     os.makedirs(output_name_path)
+        img_name_path = os.path.join(input_path, img_name)
+        #img_list = os.listdir(img_name_path)
+        count2 += 1
+        if (count2 % 5000 == 0):
+            print('net forward time: {:.4f}'.format(time.time() - tic))
+            print(count2)
+            # if len(order_img) > 0:
+            #     order_img = np.array(order_img)
+            #     order_img = order_img.transpose((0, 3, 1, 2))
+            #     order_img = np.array(order_img, dtype=np.float32)
+            #     order_img -= 127.5
+            #     order_img /= 127.5
+            #     # order_img = np.array(order_img)
+            #     # print(order_img.shape)
+            #     # print(len(order_name))
+            #     create_database_from_img(order_name, order_img, arcface_model, "./Database/sfz_test.npy", cpu_or_cuda)
+            #     order_img = []
+            #     order_name = []
+            #     tic = time.time()
+
+        if img_name[19] == "1":
+            continue
+
+        #path = os.path.join(img_name_path, img)
+        #align_img_path = os.path.join(output_path, img_name)
+        # print(path)
+        #frame = cv2.imdecode(np.fromfile(img_name_path, dtype=np.uint8), cv2.IMREAD_COLOR)
+        try:
+            frame = cv2.imread(img_name_path)
+            h, w, d = frame.shape
+        except AttributeError:
+            print(img_name)
+            continue
+        if d == 1:
+            continue
+        factor = h / w
+        if (w > 1000):
+            frame = cv2.resize(frame, (600, int(600 * factor)))
+        h, w = frame.shape[:2]
+        img = np.float32(frame)
+        im_height, im_width, _ = img.shape
+        scale = torch.Tensor([img.shape[1], img.shape[0], img.shape[1], img.shape[0]])
+        img -= (104, 117, 123)
+        img = img.transpose(2, 0, 1)
+        img = torch.from_numpy(img).unsqueeze(0)
+        img = img.to(device)
+        scale = scale.to(device)
+
+        #tic = time.time()
+        loc, conf, landms = net(img)  # forward pass
+        #print('net forward time: {:.4f}'.format(time.time() - tic))
+
+        priorbox = PriorBox(cfg, image_size=(im_height, im_width))
+        priors = priorbox.forward()
+        priors = priors.to(device)
+        prior_data = priors.data
+        boxes = decode(loc.data.squeeze(0), prior_data, cfg['variance'])
+        boxes = boxes * scale / resize
+        boxes = boxes.cpu().numpy()
+        scores = conf.squeeze(0).data.cpu().numpy()[:, 1]
+        landms = decode_landm(landms.data.squeeze(0), prior_data, cfg['variance'])
+        scale1 = torch.Tensor([img.shape[3], img.shape[2], img.shape[3], img.shape[2],
+                               img.shape[3], img.shape[2], img.shape[3], img.shape[2],
+                               img.shape[3], img.shape[2]])
+        scale1 = scale1.to(device)
+        landms = landms * scale1 / resize
+        landms = landms.cpu().numpy()
+
+        # ignore low scores
+        inds = np.where(scores > args.confidence_threshold)[0]
+        boxes = boxes[inds]
+        landms = landms[inds]
+        scores = scores[inds]
+
+        # keep top-K before NMS
+        order = scores.argsort()[::-1][:args.top_k]
+        boxes = boxes[order]
+        landms = landms[order]
+        scores = scores[order]
+
+        # do NMS
+        dets = np.hstack((boxes, scores[:, np.newaxis])).astype(np.float32, copy=False)
+        keep = py_cpu_nms(dets, args.nms_threshold)
+        # keep = nms(dets, args.nms_threshold,force_cpu=args.cpu)
+        dets = dets[keep, :]
+        landms = landms[keep]
+
+        # keep top-K faster NMS
+        dets = dets[:args.keep_top_k, :]
+        landms = landms[:args.keep_top_k, :]
+
+        dets = np.concatenate((dets, landms), axis=1)
+        score = 500
+        # show image
+        if args.save_image:
+            dst = []
+            for i, det in enumerate(dets):
+                if det[4] < args.vis_thres:
+                    continue
+                # center_x = (det[2] + det[0]) / 2
+                # center_y = (det[3] + det[1]) / 2
+                # if abs(center_x - 125) + abs(center_y - 125) < score:
+                #     score = abs(center_x - 125) + abs(center_y - 125)
+                dst = np.reshape(landms[i], (5, 2))
+            if len(dst) > 0:
+                src1 = np.array([
+                    [38.3814, 51.6963],
+                    [73.6186, 51.5014],
+                    [56.1120, 71.7366],
+                    [41.6361, 92.3655],
+                    [70.8167, 92.2041]], dtype=np.float32)
+                tform = trans.SimilarityTransform()
+                tform.estimate(dst, src1)
+                M = tform.params[0:2, :]
+
+                if w < 112 or h < 112:
+                    count += 1
+                    print(img_name_path)
+                    continue
+                frame = cv2.warpAffine(frame, M, (w, h), borderValue=0.0)
+                img112 = frame[0:112, 0:112, :]
+                order_img.append(img112)
+                order_name.append(img_name)
+                #cv2.imencode('.jpg', img112)[1].tofile(align_img_path)
+                #cv2.imwrite(align_img_path, img112)
+
+    print(">112 number"+str(count))
+    if len(order_img) > 0:
+        order_img = np.array(order_img)
+        order_img = order_img.transpose((0, 3, 1, 2))
+        order_img = np.array(order_img, dtype=np.float32)
+        order_img -= 127.5
+        order_img /= 127.5
+        #order_img = np.array(order_img)
+        # print(order_img.shape)
+        # print(len(order_name))
+        count_acc(order_name,order_img,arcface_model,index ,database_name_list,cpu_or_cuda)
+
+def count_acc(order_name,order_img,model,index ,database_name_list,cpu_or_cuda):
+    pred_name = []
+    unknown = []
+    print(order_img.shape)
+
+    start_time = time.time()
+    # order_img = torch.from_numpy(order_img)
+    # order_img = order_img.to(torch.device("cuda" if cpu_or_cuda == "cuda" else "cpu"))
+    batch = 256
+    now = 0
+    number = len(order_img)
+    # number = 1400
+    for i in range(number):
+        unknown.append("unknown")
+
+    while now < number:
+        if now + batch < number:
+            name = findAll(order_img[now:now + batch], model, index ,database_name_list, cpu_or_cuda)
+        else:
+            name = findAll(order_img[now:number], model, index ,database_name_list, cpu_or_cuda)
+        now = now + batch
+        for na in name:
+            pred_name.append(na)
+        print("batch" + str(now))
+    end_time = time.time()
+    print("findAll time: " + str(end_time - start_time))
+    # print(len(pred_name))
+    right = 0
+    for i, name in enumerate(pred_name):
+        if pred_name[i] == order_name[i][:-6]:
+            right += 1
+    filed = 0
+    for i, name in enumerate(pred_name):
+        if pred_name[i] == unknown[i]:
+            filed += 1
+            #print(order_name[i])
+    error = 0
+    print("----------------")
+    for i, name in enumerate(pred_name):
+        if pred_name[i] != order_name[i][:-6]:
+            error += 1
+            #print(order_name[i] + " " + pred_name[i] + " ")
+    #print(order_name)
+    #print(pred_name)
+    print("total:" + str(number))
+    print("right:" + str(right+filed) + " rate:" + str((filed+right) / number))
+    #print("filed:" + str(filed) + " rate:" + str(filed / number))
+    print("error:" + str(error - filed) + " rate:" + str((error - filed) / number))
+# if __name__ == '__main__':
+#     torch.set_grad_enabled(False)
+#     cfg = None
+#     if args.network == "mobile0.25":
+#         cfg = cfg_mnet
+#     elif args.network == "resnet50":
+#         cfg = cfg_re50
+#     # net and model
+#     net = RetinaFace(cfg=cfg, phase = 'test')
+#     net = load_model(net, args.trained_model, args.cpu)
+#     net.eval()
+#     print('Finished loading model!')
+#     #print(net)
+#     cudnn.benchmark = True
+#     device = torch.device("cpu" if args.cpu else "cuda")
+#     net = net.to(device)
+#
+#     resize = 1
+#
+#     # testing begin
+#     cap = cv2.VideoCapture("rtsp://47.108.74.82:8557/h264")
+#     ret, frame = cap.read()
+#     h, w = frame.shape[:2]
+#     fps = cap.get(cv2.CAP_PROP_FPS)
+#     size = (int(cap.get(cv2.CAP_PROP_FRAME_WIDTH)),
+#             int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT)))
+#     #out = cv2.VideoWriter('out.mp4', cv2.VideoWriter_fourcc('m', 'p', '4', 'v'), fps, size)
+#     out = cv2.VideoWriter('ttttttt.avi', cv2.VideoWriter_fourcc(*'XVID'), fps, size)
+#     number = 0
+#
+#     model = iresnet100()
+#     model.load_state_dict(torch.load("./model/backbone100.pth", map_location="cpu"))
+#     model.eval()
+#     k_v = load_npy("./Database/student.npy")
+#
+#     while ret:
+#         tic = time.time()
+#         img = np.float32(frame)
+#         im_height, im_width, _ = img.shape
+#         scale = torch.Tensor([img.shape[1], img.shape[0], img.shape[1], img.shape[0]])
+#         img -= (104, 117, 123)
+#         img = img.transpose(2, 0, 1)
+#         img = torch.from_numpy(img).unsqueeze(0)
+#         img = img.to(device)
+#         scale = scale.to(device)
+#
+#         loc, conf, landms = net(img)  # forward pass
+#
+#
+#         priorbox = PriorBox(cfg, image_size=(im_height, im_width))
+#         priors = priorbox.forward()
+#         priors = priors.to(device)
+#         prior_data = priors.data
+#         boxes = decode(loc.data.squeeze(0), prior_data, cfg['variance'])
+#         boxes = boxes * scale / resize
+#         boxes = boxes.cpu().numpy()
+#         scores = conf.squeeze(0).data.cpu().numpy()[:, 1]
+#         landms = decode_landm(landms.data.squeeze(0), prior_data, cfg['variance'])
+#         scale1 = torch.Tensor([img.shape[3], img.shape[2], img.shape[3], img.shape[2],
+#                                img.shape[3], img.shape[2], img.shape[3], img.shape[2],
+#                                img.shape[3], img.shape[2]])
+#         scale1 = scale1.to(device)
+#         landms = landms * scale1 / resize
+#         landms = landms.cpu().numpy()
+#
+#         # ignore low scores
+#         inds = np.where(scores > args.confidence_threshold)[0]
+#         boxes = boxes[inds]
+#         landms = landms[inds]
+#         scores = scores[inds]
+#
+#         # keep top-K before NMS
+#         order = scores.argsort()[::-1][:args.top_k]
+#         boxes = boxes[order]
+#         landms = landms[order]
+#         scores = scores[order]
+#
+#         # do NMS
+#         dets = np.hstack((boxes, scores[:, np.newaxis])).astype(np.float32, copy=False)
+#         keep = py_cpu_nms(dets, args.nms_threshold)
+#         # keep = nms(dets, args.nms_threshold,force_cpu=args.cpu)
+#         dets = dets[keep, :]
+#         landms = landms[keep]
+#
+#         # keep top-K faster NMS
+#         dets = dets[:args.keep_top_k, :]
+#         landms = landms[:args.keep_top_k, :]
+#
+#         dets = np.concatenate((dets, landms), axis=1)
+#         face_list = []
+#         name_list = []
+#         #print(dets[:4])
+#         print('net forward time: {:.4f}'.format(time.time() - tic))
+#         start_time = time.time()
+#         for i, det in enumerate(dets):
+#             if det[4] < args.vis_thres:
+#                         continue
+#             boxes, score = det[:4], det[4]
+#             dst = np.reshape(landms[i],(5,2))
+#             #print(dst.shape)
+#             src1 = np.array([
+#                 [38.3814, 51.6963],
+#                 [73.6186, 51.5014],
+#                 [56.1120, 71.7366],
+#                 [41.6361, 92.3655],
+#                 [70.8167, 92.2041]], dtype=np.float32)
+#             #print(src1.shape)
+#             tform = trans.SimilarityTransform()
+#             tform.estimate(dst, src1)
+#             M = tform.params[0:2, :]
+#             frame2 = cv2.warpAffine(frame, M, (w, h), borderValue=0.0)
+#             img112 = frame2[0:112, 0:112, :]
+#             # cv2.imwrite("./img/man"+str(count)+".jpg", img112)
+#             # count += 1
+#             face_list.append(img112)
+#
+#         if len(face_list) != 0:
+#             face_list = np.array(face_list)
+#             face_list = face_list.transpose((0, 3, 1, 2))
+#             face_list = np.array(face_list, dtype=np.float32)
+#             face_list -= 127.5
+#             face_list /= 127.5
+#             print(face_list.shape)
+#             face_list = torch.from_numpy(face_list)
+#
+#             name_list = findAll(face_list, model, k_v)
+#             end_time = time.time()
+#             print("findOneframe time: " + str(end_time - start_time))
+#         start_time = time.time()
+#         img_PIL = Image.fromarray(cv2.cvtColor(frame, cv2.COLOR_BGR2RGB))
+#         draw = ImageDraw.Draw(img_PIL)
+#         font = ImageFont.truetype("font.ttf", 22)
+#         for i, det in enumerate(dets):
+#             if det[4] < args.vis_thres:
+#                         continue
+#             boxes, score = det[:4], det[4]
+#             #print(name_list)
+#             name = name_list[i]
+#             mo = r'[\u4e00-\u9fa5]*'
+#             name = re.match(mo, name).group(0)
+#             if not isinstance(name, np.unicode):
+#                 name = name.decode('utf8')
+#             draw.text((int(boxes[0]), int(boxes[1])), name, fill=(255, 0, 0), font=font)
+#             draw.rectangle((int(boxes[0]), int(boxes[1]), int(boxes[2]), int(boxes[3])), outline="green", width=3)
+#         frame = cv2.cvtColor(np.asarray(img_PIL), cv2.COLOR_RGB2BGR)
+#         cv2.imshow('out', frame)
+#         out.write(frame)
+#         end_time = time.time()
+#         print("drawOneframe time: " + str(end_time - start_time))
+#         # Press Q on keyboard to stop recording
+#         if cv2.waitKey(1) & 0xFF == ord('q'):
+#             break
+#         ret, frame = cap.read()
+#     cap.release()
+#     out.release()
+#     cv2.destroyAllWindows()
+if __name__ == '__main__':
+    cpu_or_cuda = "cuda" if torch.cuda.is_available() else "cpu"
+    arcface_model = load_arcface_model("./model/backbone100.pth", cpu_or_cuda=cpu_or_cuda)
+
+    k_v = load_npy("./Database/sfz_test.npy")
+    database_name_list = list(k_v.keys())
+    vector_list = np.array(list(k_v.values()))
+    print(vector_list.shape)
+    # print(database_name_list)
+    nlist = 500
+    quantizer = faiss.IndexFlatL2(512)  # the other index
+    index = faiss.IndexIVFFlat(quantizer, 512, nlist, faiss.METRIC_L2)
+    index.train(vector_list)
+    # index = faiss.IndexFlatL2(512)
+    index.add(vector_list)
+    index.nprobe = 50
+
+    count_accuracy(arcface_model, cpu_or_cuda, index, database_name_list)
+    # sfz_to112x112_retinaface(arcface_model,cpu_or_cuda)
+
+
diff --git a/retinaface_detect.py b/retinaface_detect.py
new file mode 100644
index 0000000..cc41137
--- /dev/null
+++ b/retinaface_detect.py
@@ -0,0 +1,483 @@
+from __future__ import print_function
+import re
+import time
+import cv2
+import torch
+import torch.backends.cudnn as cudnn
+import numpy as np
+from skimage import transform as trans
+from PIL import Image, ImageDraw, ImageFont
+from data import cfg_mnet, cfg_re50
+from layers.functions.prior_box import PriorBox
+from utils.nms.py_cpu_nms import py_cpu_nms
+from models.retinaface import RetinaFace
+from utils.box_utils import decode, decode_landm
+
+threshold = 1.05
+ppi = 1280
+step = 3
+
+class ConfRetinaface(object):
+    def __init__(self, trained_model, network, cpu, confidence_threshold, top_k, nms_threshold, keep_top_k, vis_thres):
+        self.trained_model = trained_model
+        self.network = network
+        self.cpu = cpu
+        self.confidence_threshold = confidence_threshold
+        self.top_k = top_k
+        self.nms_threshold = nms_threshold
+        self.keep_top_k = keep_top_k
+        self.vis_thres = vis_thres
+
+
+def set_retinaface_conf(cpu_or_cuda):
+    args = ConfRetinaface(trained_model='./weights/mobilenet0.25_Final.pth',
+                          network='mobile0.25',
+                          cpu=True if cpu_or_cuda == 'cpu' else False,
+                          confidence_threshold=0.02,
+                          top_k=5000,
+                          nms_threshold=0.4,
+                          keep_top_k=750,
+                          vis_thres=0.6)
+    return args
+
+
+def check_keys(model, pretrained_state_dict):
+    ckpt_keys = set(pretrained_state_dict.keys())
+    model_keys = set(model.state_dict().keys())
+    used_pretrained_keys = model_keys & ckpt_keys
+    unused_pretrained_keys = ckpt_keys - model_keys
+    missing_keys = model_keys - ckpt_keys
+    print('Missing keys:{}'.format(len(missing_keys)))
+    print('Unused checkpoint keys:{}'.format(len(unused_pretrained_keys)))
+    print('Used keys:{}'.format(len(used_pretrained_keys)))
+    assert len(used_pretrained_keys) > 0, 'load NONE from pretrained checkpoint'
+    return True
+
+
+def remove_prefix(state_dict, prefix):
+    ''' Old style model is stored with all names of parameters sharing common prefix 'module.' '''
+    print('remove prefix \'{}\''.format(prefix))
+    f = lambda x: x.split(prefix, 1)[-1] if x.startswith(prefix) else x
+    return {f(key): value for key, value in state_dict.items()}
+
+
+def load_model(model, pretrained_path, load_to_cpu):
+    print('Loading pretrained model from {}'.format(pretrained_path))
+    if load_to_cpu:
+        pretrained_dict = torch.load(pretrained_path, map_location=lambda storage, loc: storage)
+    else:
+        device = torch.cuda.current_device()
+        pretrained_dict = torch.load(pretrained_path, map_location=lambda storage, loc: storage.cuda(device))
+    if "state_dict" in pretrained_dict.keys():
+        pretrained_dict = remove_prefix(pretrained_dict['state_dict'], 'module.')
+    else:
+        pretrained_dict = remove_prefix(pretrained_dict, 'module.')
+    check_keys(model, pretrained_dict)
+    model.load_state_dict(pretrained_dict, strict=False)
+    return model
+
+
+# 加载retinaface模型
+def load_retinaface_model(args):
+    torch.set_grad_enabled(False)
+    cfg = None
+    if args.network == "mobile0.25":
+        cfg = cfg_mnet
+    elif args.network == "resnet50":
+        cfg = cfg_re50
+    # net and model
+    net = RetinaFace(cfg=cfg, phase='test')
+    net = load_model(net, args.trained_model, args.cpu)
+    net.eval()
+    cudnn.benchmark = True
+    device = torch.device("cpu" if args.cpu else "cuda")
+    net = net.to(device)
+    print('Finished loading model!')
+    return net
+
+
+# 计算两个特征向量的欧式距离
+def findEuclideanDistance(source_representation, test_representation):
+    euclidean_distance = source_representation - test_representation
+    euclidean_distance = np.sum(np.multiply(euclidean_distance, euclidean_distance))
+    euclidean_distance = np.sqrt(euclidean_distance)
+    return euclidean_distance
+
+
+# 归一化欧氏距离
+def l2_normalize(x):
+    return x / np.sqrt(np.sum(np.multiply(x, x)))
+
+
+# 计算此特征向量与人脸库中的哪个人脸特征向量距离最近
+def findmindistance(pred, threshold, k_v):
+    distance = 10
+    most_like = ""
+    for name in k_v.keys():
+        tmp = findEuclideanDistance(k_v[name], pred)
+        if distance > tmp:
+            distance = tmp
+            most_like = name
+    if distance < threshold:
+        return most_like
+    else:
+        return "unknown"
+
+#
+def faiss_find_face(pred,index ,database_name_list):
+    #print(len(database_name_list))
+    start_time = time.time()
+    D, I = index.search(pred, 1)
+    name_list = []
+    end_time = time.time()
+    print("faiss cost %fs" % (end_time - start_time))
+    print(D, I)
+    # if D[0][0] < threshold:
+    #     print(database_name_list[I[0][0]])
+    #     return database_name_list[I[0][0]]
+    # else:
+    #     return "unknown"
+    for i,index in enumerate(I):
+        if D[i][0] < threshold:
+            #print(database_name_list[I[0][0]])
+            name_list.append(database_name_list[index[0]])
+        else:
+            name_list.append("unknown")
+    return name_list
+
+# 从人脸库中找到传入的人脸列表中的所有人脸
+def findAll(imglist, model, index ,database_name_list, k_v, cpu_or_cuda):
+    start_time = time.time()
+    imglist = torch.from_numpy(imglist)
+    imglist = imglist.to(torch.device("cuda" if cpu_or_cuda == "cuda" else "cpu"))
+    with torch.no_grad():
+        name_list = []
+        pred = model(imglist)
+        pred = pred.cpu().numpy()
+        print("predOne time: " + str(time.time() - start_time))
+        #print(pred.shape)
+        start_time = time.time()
+        #name_list = faiss_find_face(l2_normalize(pred), index, database_name_list)
+        for pr in pred:
+            name = findmindistance(l2_normalize(pr), threshold=threshold, k_v=k_v)
+            print(name)
+            # print(l2_normalize(pr).shape)
+            #pr = np.expand_dims(l2_normalize(pr), 0)
+            #print(pr.shape)
+            #name = faiss_find_face(pr,index ,database_name_list)
+            if name != "unknown":
+                mo = r'[\u4e00-\u9fa5_a-zA-Z]*'
+                name = re.match(mo, name)
+                name_list.append(name.group(0))
+            else:
+                name_list.append("unknown")
+            #name_list.append(name)
+        print("findOne time: " + str(time.time() - start_time))
+        return name_list
+
+
+# 检测单张人脸，返回1x3x112x112的数组
+def detect_one(path, net, args):
+    cfg = None
+    if args.network == "mobile0.25":
+        cfg = cfg_mnet
+    elif args.network == "resnet50":
+        cfg = cfg_re50
+
+    device = torch.device("cpu" if args.cpu else "cuda")
+    resize = 1
+
+    # testing begin
+    frame = cv2.imdecode(np.fromfile(path, dtype=np.uint8), cv2.IMREAD_COLOR)
+    h, w = frame.shape[:2]
+    factor = h / w
+    if (w > 1000):
+        frame = cv2.resize(frame, (600, int(600 * factor)))
+    h, w = frame.shape[:2]
+
+    tic = time.time()
+    img = np.float32(frame)
+    im_height, im_width, _ = img.shape
+    scale = torch.Tensor([img.shape[1], img.shape[0], img.shape[1], img.shape[0]])
+    img -= (104, 117, 123)
+    img = img.transpose(2, 0, 1)
+    img = torch.from_numpy(img).unsqueeze(0)
+    img = img.to(device)
+    scale = scale.to(device)
+
+    loc, conf, landms = net(img)  # forward pass
+    #print(loc.shape,landms.shape,conf.shape)
+    priorbox = PriorBox(cfg, image_size=(im_height, im_width))
+    priors = priorbox.forward()
+    priors = priors.to(device)
+    prior_data = priors.data
+    boxes = decode(loc.data.squeeze(0), prior_data, cfg['variance'])
+    boxes = boxes * scale / resize
+    boxes = boxes.cpu().numpy()
+    scores = conf.squeeze(0).data.cpu().numpy()[:, 1]
+    landms = decode_landm(landms.data.squeeze(0), prior_data, cfg['variance'])
+    scale1 = torch.Tensor([img.shape[3], img.shape[2], img.shape[3], img.shape[2],
+                           img.shape[3], img.shape[2], img.shape[3], img.shape[2],
+                           img.shape[3], img.shape[2]])
+    scale1 = scale1.to(device)
+    landms = landms * scale1 / resize
+    landms = landms.cpu().numpy()
+
+    # ignore low scores
+    inds = np.where(scores > args.confidence_threshold)[0]
+    boxes = boxes[inds]
+    landms = landms[inds]
+    scores = scores[inds]
+
+    # keep top-K before NMS
+    order = scores.argsort()[::-1][:args.top_k]
+    boxes = boxes[order]
+    landms = landms[order]
+    scores = scores[order]
+
+    # do NMS
+    dets = np.hstack((boxes, scores[:, np.newaxis])).astype(np.float32, copy=False)
+    keep = py_cpu_nms(dets, args.nms_threshold)
+    # keep = nms(dets, args.nms_threshold,force_cpu=args.cpu)
+    dets = dets[keep, :]
+    landms = landms[keep]
+
+    # keep top-K faster NMS
+    dets = dets[:args.keep_top_k, :]
+    landms = landms[:args.keep_top_k, :]
+
+    dets = np.concatenate((dets, landms), axis=1)
+    face_list = []
+    box_and_point = []
+    # print(dets[:4])
+    # print('net forward time: {:.4f}'.format(time.time() - tic))
+    print(len(dets))
+    for i, det in enumerate(dets):
+
+        if det[4] < args.vis_thres:
+            continue
+        box_and_point.append(det)
+        dst = np.reshape(landms[i], (5, 2))
+        # print(dst.shape)
+        src1 = np.array([
+            [38.3814, 51.6963],
+            [73.6186, 51.5014],
+            [56.1120, 71.7366],
+            [41.6361, 92.3655],
+            [70.8167, 92.2041]], dtype=np.float32)
+        # print(src1.shape)
+        tform = trans.SimilarityTransform()
+        tform.estimate(dst, src1)
+        M = tform.params[0:2, :]
+        frame2 = cv2.warpAffine(frame, M, (w, h), borderValue=0.0)
+        img112 = frame2[0:112, 0:112, :]
+        # cv2.imshow('out', img112)
+        # cv2.waitKey(0)
+        face_list.append(img112)
+    if len(face_list) > 0:
+        face_list = np.array(face_list)
+        face_list = face_list.transpose((0, 3, 1, 2))
+        face_list = np.array(face_list, dtype=np.float32)
+        face_list -= 127.5
+        face_list /= 127.5
+        box_and_point = np.array(box_and_point)
+        # face_list = torch.from_numpy(face_list)
+        # cv2.imshow('out', img112)
+        # cv2.waitKey(0)
+    return face_list, box_and_point
+
+
+# 检测视频中的人脸并人脸识别
+def detect_video(video_path, output_path, net, arcface_model, k_v, args):
+    tic_total = time.time()
+    cfg = None
+    if args.network == "mobile0.25":
+        cfg = cfg_mnet
+    elif args.network == "resnet50":
+        cfg = cfg_re50
+    device = torch.device("cpu" if args.cpu else "cuda")
+    resize = 1
+
+    # testing begin
+    cap = cv2.VideoCapture(video_path)
+    ret, frame = cap.read()
+    h, w = frame.shape[:2]
+    factor = 0
+    if (w > ppi):
+        factor = h / w
+        frame = cv2.resize(frame, (ppi, int(ppi * factor)))
+        h, w = frame.shape[:2]
+
+    fps = cap.get(cv2.CAP_PROP_FPS)
+    size = (w, h)
+    # size = (int(cap.get(cv2.CAP_PROP_FRAME_WIDTH)),
+    #         int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT)))
+    # out = cv2.VideoWriter('out.mp4', cv2.VideoWriter_fourcc('m', 'p', '4', 'v'), fps, size)
+    out = cv2.VideoWriter(output_path, cv2.VideoWriter_fourcc(*'XVID'), fps, size)
+    number = step
+    dets = []
+    name_list = []
+    font = ImageFont.truetype("font.ttf", 22)
+    priorbox = PriorBox(cfg, image_size=(h, w))
+    priors = priorbox.forward()
+    priors = priors.to(device)
+    prior_data = priors.data
+
+    scale = torch.Tensor([w, h, w, h])
+    scale = scale.to(device)
+    scale1 = torch.Tensor([w, h, w, h,
+                           w, h, w, h,
+                           w, h])
+    scale1 = scale1.to(device)
+
+    src1 = np.array([
+        [38.3814, 51.6963],
+        [73.6186, 51.5014],
+        [56.1120, 71.7366],
+        [41.6361, 92.3655],
+        [70.8167, 92.2041]], dtype=np.float32)
+    # print(src1.shape)
+    tform = trans.SimilarityTransform()
+
+    while ret:
+        tic_all = time.time()
+        if number == step:
+            tic = time.time()
+            img = np.float32(frame)
+            img -= (104, 117, 123)
+            img = img.transpose(2, 0, 1)
+            img = torch.from_numpy(img).unsqueeze(0)
+            img = img.to(device)
+
+            loc, conf, landms = net(img)  # forward pass
+
+            boxes = decode(loc.data.squeeze(0), prior_data, cfg['variance'])
+            boxes = boxes * scale / resize
+            boxes = boxes.cpu().numpy()
+            scores = conf.squeeze(0).data.cpu().numpy()[:, 1]
+            landms = decode_landm(landms.data.squeeze(0), prior_data, cfg['variance'])
+
+            landms = landms * scale1 / resize
+            landms = landms.cpu().numpy()
+
+            # ignore low scores
+            inds = np.where(scores > args.confidence_threshold)[0]
+            boxes = boxes[inds]
+            landms = landms[inds]
+            scores = scores[inds]
+
+            # keep top-K before NMS
+            order = scores.argsort()[::-1][:args.top_k]
+            boxes = boxes[order]
+            landms = landms[order]
+            scores = scores[order]
+
+            # do NMS
+            dets = np.hstack((boxes, scores[:, np.newaxis])).astype(np.float32, copy=False)
+            keep = py_cpu_nms(dets, args.nms_threshold)
+            # keep = nms(dets, args.nms_threshold,force_cpu=args.cpu)
+            dets = dets[keep, :]
+            landms = landms[keep]
+
+            # keep top-K faster NMS
+            dets = dets[:args.keep_top_k, :]
+            landms = landms[:args.keep_top_k, :]
+
+            dets = np.concatenate((dets, landms), axis=1)
+            face_list = []
+            name_list = []
+            # print(dets[:4])
+            print('net forward time: {:.4f}'.format(time.time() - tic))
+            start_time = time.time()
+            for i, det in enumerate(dets[:4]):
+                if det[4] < args.vis_thres:
+                    continue
+                boxes, score = det[:4], det[4]
+                dst = np.reshape(landms[i], (5, 2))
+                # print(dst.shape)
+
+                tform.estimate(dst, src1)
+                M = tform.params[0:2, :]
+                frame2 = cv2.warpAffine(frame, M, (w, h), borderValue=0.0)
+                img112 = frame2[0:112, 0:112, :]
+                face_list.append(img112)
+
+            if len(face_list) != 0:
+                face_list = np.array(face_list)
+                face_list = face_list.transpose((0, 3, 1, 2))
+                face_list = np.array(face_list, dtype=np.float32)
+                face_list -= 127.5
+                face_list /= 127.5
+                print(face_list.shape)
+                # face_list = torch.from_numpy(face_list)
+                name_list = findAll(face_list, arcface_model, k_v, "cpu" if args.cpu else "cuda")
+            end_time = time.time()
+            print("findOneframe time: " + str(end_time - start_time))
+            start_time = time.time()
+            if (len(dets) != 0):
+                for i, det in enumerate(dets[:4]):
+                    if det[4] < args.vis_thres:
+                        continue
+                    boxes, score = det[:4], det[4]
+                    cv2.rectangle(frame, (int(boxes[0]), int(boxes[1])), (int(boxes[2]), int(boxes[3])), (2, 255, 0), 1)
+
+            # if (len(dets) != 0):
+            #     img_PIL = Image.fromarray(cv2.cvtColor(frame, cv2.COLOR_BGR2RGB))
+            #     draw = ImageDraw.Draw(img_PIL)
+            #
+            #     for i, det in enumerate(dets[:4]):
+            #         if det[4] < args.vis_thres:
+            #             continue
+            #         boxes, score = det[:4], det[4]
+            #         # print(name_list)
+            #         name = name_list[i]
+            #         if not isinstance(name, np.unicode):
+            #             name = name.decode('utf8')
+            #         draw.text((int(boxes[0]), int(boxes[1])), name, fill=(255, 0, 0), font=font)
+            #         draw.rectangle((int(boxes[0]), int(boxes[1]), int(boxes[2]), int(boxes[3])), outline="green", width=3)
+            #     frame = cv2.cvtColor(np.asarray(img_PIL), cv2.COLOR_RGB2BGR)
+            #cv2.imshow('out', frame)
+            #cv2.waitKey(0)
+            out.write(frame)
+            end_time = time.time()
+            print("drawOneframe time: " + str(end_time - start_time))
+            # Press Q on keyboard to stop recording
+            # if cv2.waitKey(1) & 0xFF == ord('q'):
+            #     break
+            ret, frame = cap.read()
+            number = 0
+            if (ret != 0 and factor != 0):
+                frame = cv2.resize(frame, (ppi, int(ppi * factor)))
+        else:
+            number += 1
+            if (len(dets) != 0):
+                img_PIL = Image.fromarray(cv2.cvtColor(frame, cv2.COLOR_BGR2RGB))
+                draw = ImageDraw.Draw(img_PIL)
+                for i, det in enumerate(dets[:4]):
+                    if det[4] < args.vis_thres:
+                        continue
+                    boxes, score = det[:4], det[4]
+                    # print(name_list)
+                    name = name_list[i]
+                    if not isinstance(name, np.unicode):
+                        name = name.decode('utf8')
+                    draw.text((int(boxes[0]), int(boxes[1])), name, fill=(255, 0, 0), font=font)
+                    draw.rectangle((int(boxes[0]), int(boxes[1]), int(boxes[2]), int(boxes[3])), outline="green",
+                                   width=3)
+                frame = cv2.cvtColor(np.asarray(img_PIL), cv2.COLOR_RGB2BGR)
+            out.write(frame)
+            start_time = time.time()
+            ret, frame = cap.read()
+            if (ret != 0 and factor != 0):
+                frame = cv2.resize(frame, (ppi, int(ppi * factor)))
+            print("readframe time: " + str(time.time() - start_time))
+        print('all time: {:.4f}'.format(time.time() - tic_all))
+    cap.release()
+    out.release()
+    print('total time: {:.4f}'.format(time.time() - tic_total))
+    #cv2.destroyAllWindows()
+
+
+if __name__ == "__main__":
+    args = set_retinaface_conf()
+    print(args.cpu)
diff --git a/src/__pycache__/generate_patches.cpython-38.pyc b/src/__pycache__/generate_patches.cpython-38.pyc
new file mode 100644
index 0000000..134581e
Binary files /dev/null and b/src/__pycache__/generate_patches.cpython-38.pyc differ
diff --git a/src/__pycache__/utility.cpython-38.pyc b/src/__pycache__/utility.cpython-38.pyc
new file mode 100644
index 0000000..5dbf5db
Binary files /dev/null and b/src/__pycache__/utility.cpython-38.pyc differ
diff --git a/src/data_io/__pycache__/functional.cpython-38.pyc b/src/data_io/__pycache__/functional.cpython-38.pyc
new file mode 100644
index 0000000..7457343
Binary files /dev/null and b/src/data_io/__pycache__/functional.cpython-38.pyc differ
diff --git a/src/data_io/__pycache__/transform.cpython-38.pyc b/src/data_io/__pycache__/transform.cpython-38.pyc
new file mode 100644
index 0000000..f7a288d
Binary files /dev/null and b/src/data_io/__pycache__/transform.cpython-38.pyc differ
diff --git a/src/data_io/dataset_folder.py b/src/data_io/dataset_folder.py
new file mode 100644
index 0000000..c651f41
--- /dev/null
+++ b/src/data_io/dataset_folder.py
@@ -0,0 +1,65 @@
+# -*- coding: utf-8 -*-
+# @Time : 20-6-4 下午4:04
+# @Author : zhuying
+# @Company : Minivision
+# @File : dataset_folder.py
+# @Software : PyCharm
+
+import cv2
+import torch
+from torchvision import datasets
+import numpy as np
+
+
+def opencv_loader(path):
+    img = cv2.imread(path)
+    return img
+
+
+class DatasetFolderFT(datasets.ImageFolder):
+    def __init__(self, root, transform=None, target_transform=None,
+                 ft_width=10, ft_height=10, loader=opencv_loader):
+        super(DatasetFolderFT, self).__init__(root, transform, target_transform, loader)
+        self.root = root
+        self.ft_width = ft_width
+        self.ft_height = ft_height
+
+    def __getitem__(self, index):
+        path, target = self.samples[index]
+        sample = self.loader(path)
+        # generate the FT picture of the sample
+        ft_sample = generate_FT(sample)
+        if sample is None:
+            print('image is None --> ', path)
+        if ft_sample is None:
+            print('FT image is None -->', path)
+        assert sample is not None
+
+        ft_sample = cv2.resize(ft_sample, (self.ft_width, self.ft_height))
+        ft_sample = torch.from_numpy(ft_sample).float()
+        ft_sample = torch.unsqueeze(ft_sample, 0)
+
+        if self.transform is not None:
+            try:
+                sample = self.transform(sample)
+            except Exception as err:
+                print('Error Occured: %s' % err, path)
+        if self.target_transform is not None:
+            target = self.target_transform(target)
+        return sample, ft_sample, target
+
+
+def generate_FT(image):
+    image = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
+    f = np.fft.fft2(image)
+    fshift = np.fft.fftshift(f)
+    fimg = np.log(np.abs(fshift)+1)
+    maxx = -1
+    minn = 100000
+    for i in range(len(fimg)):
+        if maxx < max(fimg[i]):
+            maxx = max(fimg[i])
+        if minn > min(fimg[i]):
+            minn = min(fimg[i])
+    fimg = (fimg - minn+1) / (maxx - minn+1)
+    return fimg
\ No newline at end of file
diff --git a/src/data_io/dataset_loader.py b/src/data_io/dataset_loader.py
new file mode 100644
index 0000000..0bfdb60
--- /dev/null
+++ b/src/data_io/dataset_loader.py
@@ -0,0 +1,33 @@
+# -*- coding: utf-8 -*-
+# @Time : 20-6-4 下午3:40
+# @Author : zhuying
+# @Company : Minivision
+# @File : dataset_loader.py
+# @Software : PyCharm
+
+from torch.utils.data import DataLoader
+from src.data_io.dataset_folder import DatasetFolderFT
+from src.data_io import transform as trans
+
+
+def get_train_loader(conf):
+    train_transform = trans.Compose([
+        trans.ToPILImage(),
+        trans.RandomResizedCrop(size=tuple(conf.input_size),
+                                scale=(0.9, 1.1)),
+        trans.ColorJitter(brightness=0.4,
+                          contrast=0.4, saturation=0.4, hue=0.1),
+        trans.RandomRotation(10),
+        trans.RandomHorizontalFlip(),
+        trans.ToTensor()
+    ])
+    root_path = '{}/{}'.format(conf.train_root_path, conf.patch_info)
+    trainset = DatasetFolderFT(root_path, train_transform,
+                               None, conf.ft_width, conf.ft_height)
+    train_loader = DataLoader(
+        trainset,
+        batch_size=conf.batch_size,
+        shuffle=True,
+        pin_memory=True,
+        num_workers=16)
+    return train_loader
diff --git a/src/data_io/functional.py b/src/data_io/functional.py
new file mode 100644
index 0000000..a448bc2
--- /dev/null
+++ b/src/data_io/functional.py
@@ -0,0 +1,589 @@
+# -*- coding: utf-8 -*-
+# @Time : 20-6-4 下午6:18
+# @Author : zhuying
+# @Company : Minivision
+# @File : functional.py
+# @Software : PyCharm
+
+from __future__ import division
+import torch
+from PIL import Image, ImageOps, ImageEnhance
+try:
+    import accimage
+except ImportError:
+    accimage = None
+import numpy as np
+import numbers
+import types
+import collections
+import warnings
+
+
+def _is_pil_image(img):
+    if accimage is not None:
+        return isinstance(img, (Image.Image, accimage.Image))
+    else:
+        return isinstance(img, Image.Image)
+
+
+def _is_tensor_image(img):
+    return torch.is_tensor(img) and img.ndimension() == 3
+
+
+def _is_numpy_image(img):
+    return isinstance(img, np.ndarray) and (img.ndim in {2, 3})
+
+
+def to_tensor(pic):
+    """Convert a ``PIL Image`` or ``numpy.ndarray`` to tensor.
+
+    See ``ToTensor`` for more details.
+
+    Args:
+        pic (PIL Image or numpy.ndarray): Image to be converted to tensor.
+
+    Returns:
+        Tensor: Converted image.
+    """
+    if not(_is_pil_image(pic) or _is_numpy_image(pic)):
+        raise TypeError('pic should be PIL Image or ndarray. Got {}'.format(type(pic)))
+
+    if isinstance(pic, np.ndarray):
+        # handle numpy array
+        # IR image channel=1: modify by lzc --> 20190730
+        if pic.ndim == 2:
+            pic = pic.reshape((pic.shape[0], pic.shape[1], 1))
+
+        img = torch.from_numpy(pic.transpose((2, 0, 1)))
+        # backward compatibility
+        # return img.float().div(255)  modify by zkx
+        return img.float()
+    if accimage is not None and isinstance(pic, accimage.Image):
+        nppic = np.zeros([pic.channels, pic.height, pic.width], dtype=np.float32)
+        pic.copyto(nppic)
+        return torch.from_numpy(nppic)
+
+    # handle PIL Image
+    if pic.mode == 'I':
+        img = torch.from_numpy(np.array(pic, np.int32, copy=False))
+    elif pic.mode == 'I;16':
+        img = torch.from_numpy(np.array(pic, np.int16, copy=False))
+    else:
+        img = torch.ByteTensor(torch.ByteStorage.from_buffer(pic.tobytes()))
+    # PIL image mode: 1, L, P, I, F, RGB, YCbCr, RGBA, CMYK
+    if pic.mode == 'YCbCr':
+        nchannel = 3
+    elif pic.mode == 'I;16':
+        nchannel = 1
+    else:
+        nchannel = len(pic.mode)
+    img = img.view(pic.size[1], pic.size[0], nchannel)
+    # put it from HWC to CHW format
+    # yikes, this transpose takes 80% of the loading time/CPU
+    img = img.transpose(0, 1).transpose(0, 2).contiguous()
+    if isinstance(img, torch.ByteTensor):
+        # return img.float().div(255)  #modified by zkx
+        return img.float()
+    else:
+        return img
+
+
+def to_pil_image(pic, mode=None):
+    """Convert a tensor or an ndarray to PIL Image.
+
+    See :class:`~torchvision.transforms.ToPIlImage` for more details.
+
+    Args:
+        pic (Tensor or numpy.ndarray): Image to be converted to PIL Image.
+        mode (`PIL.Image mode`_): color space and pixel depth of input data (optional).
+
+    .. _PIL.Image mode: http://pillow.readthedocs.io/en/3.4.x/handbook/concepts.html#modes
+
+    Returns:
+        PIL Image: Image converted to PIL Image.
+    """
+    if not(_is_numpy_image(pic) or _is_tensor_image(pic)):
+        raise TypeError('pic should be Tensor or ndarray. Got {}.'.format(type(pic)))
+
+    npimg = pic
+    if isinstance(pic, torch.FloatTensor):
+        pic = pic.mul(255).byte()
+    if torch.is_tensor(pic):
+        npimg = np.transpose(pic.numpy(), (1, 2, 0))
+
+    if not isinstance(npimg, np.ndarray):
+        raise TypeError('Input pic must be a torch.Tensor or NumPy ndarray, ' +
+                        'not {}'.format(type(npimg)))
+
+    if npimg.shape[2] == 1:
+        expected_mode = None
+        npimg = npimg[:, :, 0]
+        if npimg.dtype == np.uint8:
+            expected_mode = 'L'
+        if npimg.dtype == np.int16:
+            expected_mode = 'I;16'
+        if npimg.dtype == np.int32:
+            expected_mode = 'I'
+        elif npimg.dtype == np.float32:
+            expected_mode = 'F'
+        if mode is not None and mode != expected_mode:
+            raise ValueError("Incorrect mode ({}) supplied for input type {}. Should be {}"
+                             .format(mode, np.dtype, expected_mode))
+        mode = expected_mode
+
+    elif npimg.shape[2] == 4:
+        permitted_4_channel_modes = ['RGBA', 'CMYK']
+        if mode is not None and mode not in permitted_4_channel_modes:
+            raise ValueError("Only modes {} are supported for 4D inputs".format(permitted_4_channel_modes))
+
+        if mode is None and npimg.dtype == np.uint8:
+            mode = 'RGBA'
+    else:
+        permitted_3_channel_modes = ['RGB', 'YCbCr', 'HSV']
+        if mode is not None and mode not in permitted_3_channel_modes:
+            raise ValueError("Only modes {} are supported for 3D inputs".format(permitted_3_channel_modes))
+        if mode is None and npimg.dtype == np.uint8:
+            mode = 'RGB'
+
+    if mode is None:
+        raise TypeError('Input type {} is not supported'.format(npimg.dtype))
+
+    return Image.fromarray(npimg, mode=mode)
+
+
+def normalize(tensor, mean, std):
+    """Normalize a tensor image with mean and standard deviation.
+
+    See ``Normalize`` for more details.
+
+    Args:
+        tensor (Tensor): Tensor image of size (C, H, W) to be normalized.
+        mean (sequence): Sequence of means for each channel.
+        std (sequence): Sequence of standard deviations for each channely.
+
+    Returns:
+        Tensor: Normalized Tensor image.
+    """
+    if not _is_tensor_image(tensor):
+        raise TypeError('tensor is not a torch image.')
+
+    for t, m, s in zip(tensor, mean, std):
+        t.sub_(m).div_(s)
+    return tensor
+
+
+def resize(img, size, interpolation=Image.BILINEAR):
+    """Resize the input PIL Image to the given size.
+
+    Args:
+        img (PIL Image): Image to be resized.
+        size (sequence or int): Desired output size. If size is a sequence like
+            (h, w), the output size will be matched to this. If size is an int,
+            the smaller edge of the image will be matched to this number maintaing
+            the aspect ratio. i.e, if height > width, then image will be rescaled to
+            (size * height / width, size)
+        interpolation (int, optional): Desired interpolation. Default is
+            ``PIL.Image.BILINEAR``
+
+    Returns:
+        PIL Image: Resized image.
+    """
+    if not _is_pil_image(img):
+        raise TypeError('img should be PIL Image. Got {}'.format(type(img)))
+    if not (isinstance(size, int) or (isinstance(size, collections.Iterable) and len(size) == 2)):
+        raise TypeError('Got inappropriate size arg: {}'.format(size))
+
+    if isinstance(size, int):
+        w, h = img.size
+        if (w <= h and w == size) or (h <= w and h == size):
+            return img
+        if w < h:
+            ow = size
+            oh = int(size * h / w)
+            return img.resize((ow, oh), interpolation)
+        else:
+            oh = size
+            ow = int(size * w / h)
+            return img.resize((ow, oh), interpolation)
+    else:
+        return img.resize(size[::-1], interpolation)
+
+
+def scale(*args, **kwargs):
+    warnings.warn("The use of the transforms.Scale transform is deprecated, " +
+                  "please use transforms.Resize instead.")
+    return resize(*args, **kwargs)
+
+
+def pad(img, padding, fill=0):
+    """Pad the given PIL Image on all sides with the given "pad" value.
+
+    Args:
+        img (PIL Image): Image to be padded.
+        padding (int or tuple): Padding on each border. If a single int is provided this
+            is used to pad all borders. If tuple of length 2 is provided this is the padding
+            on left/right and top/bottom respectively. If a tuple of length 4 is provided
+            this is the padding for the left, top, right and bottom borders
+            respectively.
+        fill: Pixel fill value. Default is 0. If a tuple of
+            length 3, it is used to fill R, G, B channels respectively.
+
+    Returns:
+        PIL Image: Padded image.
+    """
+    if not _is_pil_image(img):
+        raise TypeError('img should be PIL Image. Got {}'.format(type(img)))
+
+    if not isinstance(padding, (numbers.Number, tuple)):
+        raise TypeError('Got inappropriate padding arg')
+    if not isinstance(fill, (numbers.Number, str, tuple)):
+        raise TypeError('Got inappropriate fill arg')
+
+    if isinstance(padding, collections.Sequence) and len(padding) not in [2, 4]:
+        raise ValueError("Padding must be an int or a 2, or 4 element tuple, not a " +
+                         "{} element tuple".format(len(padding)))
+
+    return ImageOps.expand(img, border=padding, fill=fill)
+
+
+def crop(img, i, j, h, w):
+    """Crop the given PIL Image.
+
+    Args:
+        img (PIL Image): Image to be cropped.
+        i: Upper pixel coordinate.
+        j: Left pixel coordinate.
+        h: Height of the cropped image.
+        w: Width of the cropped image.
+
+    Returns:
+        PIL Image: Cropped image.
+    """
+    if not _is_pil_image(img):
+        raise TypeError('img should be PIL Image. Got {}'.format(type(img)))
+
+    return img.crop((j, i, j + w, i + h))
+
+
+def center_crop(img, output_size):
+    if isinstance(output_size, numbers.Number):
+        output_size = (int(output_size), int(output_size))
+    w, h = img.size
+    th, tw = output_size
+    i = int(round((h - th) / 2.))
+    j = int(round((w - tw) / 2.))
+    return crop(img, i, j, th, tw)
+
+
+def resized_crop(img, i, j, h, w, size, interpolation=Image.BILINEAR):
+    """Crop the given PIL Image and resize it to desired size.
+
+    Notably used in RandomResizedCrop.
+
+    Args:
+        img (PIL Image): Image to be cropped.
+        i: Upper pixel coordinate.
+        j: Left pixel coordinate.
+        h: Height of the cropped image.
+        w: Width of the cropped image.
+        size (sequence or int): Desired output size. Same semantics as ``scale``.
+        interpolation (int, optional): Desired interpolation. Default is
+            ``PIL.Image.BILINEAR``.
+    Returns:
+        PIL Image: Cropped image.
+    """
+    assert _is_pil_image(img), 'img should be PIL Image'
+    img = crop(img, i, j, h, w)
+    img = resize(img, size, interpolation)
+    return img
+
+
+def hflip(img):
+    """Horizontally flip the given PIL Image.
+
+    Args:
+        img (PIL Image): Image to be flipped.
+
+    Returns:
+        PIL Image:  Horizontall flipped image.
+    """
+    if not _is_pil_image(img):
+        raise TypeError('img should be PIL Image. Got {}'.format(type(img)))
+
+    return img.transpose(Image.FLIP_LEFT_RIGHT)
+
+
+def vflip(img):
+    """Vertically flip the given PIL Image.
+
+    Args:
+        img (PIL Image): Image to be flipped.
+
+    Returns:
+        PIL Image:  Vertically flipped image.
+    """
+    if not _is_pil_image(img):
+        raise TypeError('img should be PIL Image. Got {}'.format(type(img)))
+
+    return img.transpose(Image.FLIP_TOP_BOTTOM)
+
+
+def five_crop(img, size):
+    """Crop the given PIL Image into four corners and the central crop.
+
+    .. Note::
+        This transform returns a tuple of images and there may be a
+        mismatch in the number of inputs and targets your ``Dataset`` returns.
+
+    Args:
+       size (sequence or int): Desired output size of the crop. If size is an
+           int instead of sequence like (h, w), a square crop (size, size) is
+           made.
+    Returns:
+        tuple: tuple (tl, tr, bl, br, center) corresponding top left,
+            top right, bottom left, bottom right and center crop.
+    """
+    if isinstance(size, numbers.Number):
+        size = (int(size), int(size))
+    else:
+        assert len(size) == 2, "Please provide only two dimensions (h, w) for size."
+
+    w, h = img.size
+    crop_h, crop_w = size
+    if crop_w > w or crop_h > h:
+        raise ValueError("Requested crop size {} is bigger than input size {}".format(size,
+                                                                                      (h, w)))
+    tl = img.crop((0, 0, crop_w, crop_h))
+    tr = img.crop((w - crop_w, 0, w, crop_h))
+    bl = img.crop((0, h - crop_h, crop_w, h))
+    br = img.crop((w - crop_w, h - crop_h, w, h))
+    center = center_crop(img, (crop_h, crop_w))
+    return (tl, tr, bl, br, center)
+
+
+def ten_crop(img, size, vertical_flip=False):
+    """Crop the given PIL Image into four corners and the central crop plus the
+       flipped version of these (horizontal flipping is used by default).
+
+    .. Note::
+        This transform returns a tuple of images and there may be a
+        mismatch in the number of inputs and targets your ``Dataset`` returns.
+
+       Args:
+           size (sequence or int): Desired output size of the crop. If size is an
+               int instead of sequence like (h, w), a square crop (size, size) is
+               made.
+           vertical_flip (bool): Use vertical flipping instead of horizontal
+
+        Returns:
+            tuple: tuple (tl, tr, bl, br, center, tl_flip, tr_flip, bl_flip,
+                br_flip, center_flip) corresponding top left, top right,
+                bottom left, bottom right and center crop and same for the
+                flipped image.
+    """
+    if isinstance(size, numbers.Number):
+        size = (int(size), int(size))
+    else:
+        assert len(size) == 2, "Please provide only two dimensions (h, w) for size."
+
+    first_five = five_crop(img, size)
+
+    if vertical_flip:
+        img = vflip(img)
+    else:
+        img = hflip(img)
+
+    second_five = five_crop(img, size)
+    return first_five + second_five
+
+
+def adjust_brightness(img, brightness_factor):
+    """Adjust brightness of an Image.
+
+    Args:
+        img (PIL Image): PIL Image to be adjusted.
+        brightness_factor (float):  How much to adjust the brightness. Can be
+            any non negative number. 0 gives a black image, 1 gives the
+            original image while 2 increases the brightness by a factor of 2.
+
+    Returns:
+        PIL Image: Brightness adjusted image.
+    """
+    if not _is_pil_image(img):
+        raise TypeError('img should be PIL Image. Got {}'.format(type(img)))
+
+    enhancer = ImageEnhance.Brightness(img)
+    img = enhancer.enhance(brightness_factor)
+    return img
+
+
+def adjust_contrast(img, contrast_factor):
+    """Adjust contrast of an Image.
+
+    Args:
+        img (PIL Image): PIL Image to be adjusted.
+        contrast_factor (float): How much to adjust the contrast. Can be any
+            non negative number. 0 gives a solid gray image, 1 gives the
+            original image while 2 increases the contrast by a factor of 2.
+
+    Returns:
+        PIL Image: Contrast adjusted image.
+    """
+    if not _is_pil_image(img):
+        raise TypeError('img should be PIL Image. Got {}'.format(type(img)))
+
+    enhancer = ImageEnhance.Contrast(img)
+    img = enhancer.enhance(contrast_factor)
+    return img
+
+
+def adjust_saturation(img, saturation_factor):
+    """Adjust color saturation of an image.
+
+    Args:
+        img (PIL Image): PIL Image to be adjusted.
+        saturation_factor (float):  How much to adjust the saturation. 0 will
+            give a black and white image, 1 will give the original image while
+            2 will enhance the saturation by a factor of 2.
+
+    Returns:
+        PIL Image: Saturation adjusted image.
+    """
+    if not _is_pil_image(img):
+        raise TypeError('img should be PIL Image. Got {}'.format(type(img)))
+
+    enhancer = ImageEnhance.Color(img)
+    img = enhancer.enhance(saturation_factor)
+    return img
+
+
+def adjust_hue(img, hue_factor):
+    """Adjust hue of an image.
+
+    The image hue is adjusted by converting the image to HSV and
+    cyclically shifting the intensities in the hue channel (H).
+    The image is then converted back to original image mode.
+
+    `hue_factor` is the amount of shift in H channel and must be in the
+    interval `[-0.5, 0.5]`.
+
+    See https://en.wikipedia.org/wiki/Hue for more details on Hue.
+
+    Args:
+        img (PIL Image): PIL Image to be adjusted.
+        hue_factor (float):  How much to shift the hue channel. Should be in
+            [-0.5, 0.5]. 0.5 and -0.5 give complete reversal of hue channel in
+            HSV space in positive and negative direction respectively.
+            0 means no shift. Therefore, both -0.5 and 0.5 will give an image
+            with complementary colors while 0 gives the original image.
+
+    Returns:
+        PIL Image: Hue adjusted image.
+    """
+    if not(-0.5 <= hue_factor <= 0.5):
+        raise ValueError('hue_factor is not in [-0.5, 0.5].'.format(hue_factor))
+
+    if not _is_pil_image(img):
+        raise TypeError('img should be PIL Image. Got {}'.format(type(img)))
+
+    input_mode = img.mode
+    if input_mode in {'L', '1', 'I', 'F'}:
+        return img
+
+    h, s, v = img.convert('HSV').split()
+
+    np_h = np.array(h, dtype=np.uint8)
+    # uint8 addition take cares of rotation across boundaries
+    with np.errstate(over='ignore'):
+        np_h += np.uint8(hue_factor * 255)
+    h = Image.fromarray(np_h, 'L')
+
+    img = Image.merge('HSV', (h, s, v)).convert(input_mode)
+    return img
+
+
+def adjust_gamma(img, gamma, gain=1):
+    """Perform gamma correction on an image.
+
+    Also known as Power Law Transform. Intensities in RGB mode are adjusted
+    based on the following equation:
+
+        I_out = 255 * gain * ((I_in / 255) ** gamma)
+
+    See https://en.wikipedia.org/wiki/Gamma_correction for more details.
+
+    Args:
+        img (PIL Image): PIL Image to be adjusted.
+        gamma (float): Non negative real number. gamma larger than 1 make the
+            shadows darker, while gamma smaller than 1 make dark regions
+            lighter.
+        gain (float): The constant multiplier.
+    """
+    if not _is_pil_image(img):
+        raise TypeError('img should be PIL Image. Got {}'.format(type(img)))
+
+    if gamma < 0:
+        raise ValueError('Gamma should be a non-negative real number')
+
+    input_mode = img.mode
+    img = img.convert('RGB')
+
+    np_img = np.array(img, dtype=np.float32)
+    np_img = 255 * gain * ((np_img / 255) ** gamma)
+    np_img = np.uint8(np.clip(np_img, 0, 255))
+
+    img = Image.fromarray(np_img, 'RGB').convert(input_mode)
+    return img
+
+
+def rotate(img, angle, resample=False, expand=False, center=None):
+    """Rotate the image by angle and then (optionally) translate it by (n_columns, n_rows)
+
+
+    Args:
+        img (PIL Image): PIL Image to be rotated.
+        angle ({float, int}): In degrees degrees counter clockwise order.
+        resample ({PIL.Image.NEAREST, PIL.Image.BILINEAR, PIL.Image.BICUBIC}, optional):
+            An optional resampling filter.
+            See http://pillow.readthedocs.io/en/3.4.x/handbook/concepts.html#filters
+            If omitted, or if the image has mode "1" or "P", it is set to PIL.Image.NEAREST.
+        expand (bool, optional): Optional expansion flag.
+            If true, expands the output image to make it large enough to hold the entire rotated image.
+            If false or omitted, make the output image the same size as the input image.
+            Note that the expand flag assumes rotation around the center and no translation.
+        center (2-tuple, optional): Optional center of rotation.
+            Origin is the upper left corner.
+            Default is the center of the image.
+    """
+
+    if not _is_pil_image(img):
+        raise TypeError('img should be PIL Image. Got {}'.format(type(img)))
+
+    return img.rotate(angle, resample, expand, center)
+
+
+def to_grayscale(img, num_output_channels=1):
+    """Convert image to grayscale version of image.
+
+    Args:
+        img (PIL Image): Image to be converted to grayscale.
+
+    Returns:
+        PIL Image:  Grayscale version of the image.
+                    if num_output_channels == 1 : returned image is single channel
+                    if num_output_channels == 3 : returned image is 3 channel with r == g == b
+    """
+    if not _is_pil_image(img):
+        raise TypeError('img should be PIL Image. Got {}'.format(type(img)))
+
+    if num_output_channels == 1:
+        img = img.convert('L')
+    elif num_output_channels == 3:
+        img = img.convert('L')
+        np_img = np.array(img, dtype=np.uint8)
+        np_img = np.dstack([np_img, np_img, np_img])
+        img = Image.fromarray(np_img, 'RGB')
+    else:
+        raise ValueError('num_output_channels should be either 1 or 3')
+
+    return img
diff --git a/src/data_io/transform.py b/src/data_io/transform.py
new file mode 100644
index 0000000..5aa1654
--- /dev/null
+++ b/src/data_io/transform.py
@@ -0,0 +1,347 @@
+# -*- coding: utf-8 -*-
+# @Time : 20-6-4 下午4:19
+# @Author : zhuying
+# @Company : Minivision
+# @File : transform.py
+# @Software : PyCharm
+
+from __future__ import division
+import math
+import random
+from PIL import Image
+try:
+    import accimage
+except ImportError:
+    accimage = None
+import numpy as np
+import numbers
+import types
+
+from src.data_io import functional as F
+
+__all__ = ["Compose", "ToTensor", "ToPILImage", "Normalize", "RandomHorizontalFlip",
+           "Lambda", "RandomResizedCrop", "ColorJitter", "RandomRotation"]
+
+
+class Compose(object):
+    """Composes several transforms together.
+
+    Args:
+        transforms (list of ``Transform`` objects): list of transforms to compose.
+
+    Example:
+        >>> transforms.Compose([
+        >>>     transforms.CenterCrop(10),
+        >>>     transforms.ToTensor(),
+        >>> ])
+    """
+
+    def __init__(self, transforms):
+        self.transforms = transforms
+
+    def __call__(self, img):
+        for t in self.transforms:
+            img = t(img)
+        return img
+
+
+class ToTensor(object):
+
+    """Convert a ``PIL Image`` or ``numpy.ndarray`` to tensor.
+
+    Converts a PIL Image or numpy.ndarray (H x W x C) in the range
+    [0, 255] to a torch.FloatTensor of shape (C x H x W) in the range [0.0, 1.0].
+    """
+
+    def __call__(self, pic):
+        """
+        Args:
+            pic (PIL Image or numpy.ndarray): Image to be converted to tensor.
+
+        Returns:
+            Tensor: Converted image.
+        """
+        return F.to_tensor(pic)
+
+
+class Lambda(object):
+    """Apply a user-defined lambda as a transform.
+
+    Args:
+        lambd (function): Lambda/function to be used for transform.
+    """
+
+    def __init__(self, lambd):
+        assert isinstance(lambd, types.LambdaType)
+        self.lambd = lambd
+
+    def __call__(self, img):
+        return self.lambd(img)
+
+
+class ToPILImage(object):
+    """Convert a tensor or an ndarray to PIL Image.
+
+    Converts a torch.*Tensor of shape C x H x W or a numpy ndarray of shape
+    H x W x C to a PIL Image while preserving the value range.
+
+    Args:
+        mode (`PIL.Image mode`_): color space and pixel depth of input data (optional).
+            If ``mode`` is ``None`` (default) there are some assumptions made about the input data:
+            1. If the input has 3 channels, the ``mode`` is assumed to be ``RGB``.
+            2. If the input has 4 channels, the ``mode`` is assumed to be ``RGBA``.
+            3. If the input has 1 channel, the ``mode`` is determined by the data type (i,e,
+            ``int``, ``float``, ``short``).
+
+    .. _PIL.Image mode: http://pillow.readthedocs.io/en/3.4.x/handbook/concepts.html#modes
+    """
+    def __init__(self, mode=None):
+        self.mode = mode
+
+    def __call__(self, pic):
+        """
+        Args:
+            pic (Tensor or numpy.ndarray): Image to be converted to PIL Image.
+
+        Returns:
+            PIL Image: Image converted to PIL Image.
+
+        """
+        return F.to_pil_image(pic, self.mode)
+
+
+class Normalize(object):
+    """Normalize an tensor image with mean and standard deviation.
+    Given mean: ``(M1,...,Mn)`` and std: ``(S1,..,Sn)`` for ``n`` channels, this transform
+    will normalize each channel of the input ``torch.*Tensor`` i.e.
+    ``input[channel] = (input[channel] - mean[channel]) / std[channel]``
+
+    Args:
+        mean (sequence): Sequence of means for each channel.
+        std (sequence): Sequence of standard deviations for each channel.
+    """
+
+    def __init__(self, mean, std):
+        self.mean = mean
+        self.std = std
+
+    def __call__(self, tensor):
+        """
+        Args:
+            tensor (Tensor): Tensor image of size (C, H, W) to be normalized.
+
+        Returns:
+            Tensor: Normalized Tensor image.
+        """
+        return F.normalize(tensor, self.mean, self.std)
+
+
+class RandomHorizontalFlip(object):
+    """Horizontally flip the given PIL Image randomly with a probability of 0.5."""
+
+    def __call__(self, img):
+        """
+        Args:
+            img (PIL Image): Image to be flipped.
+
+        Returns:
+            PIL Image: Randomly flipped image.
+        """
+        if random.random() < 0.5:
+            return F.hflip(img)
+        return img
+
+
+class RandomResizedCrop(object):
+    """Crop the given PIL Image to random size and aspect ratio.
+
+    A crop of random size (default: of 0.08 to 1.0) of the original size and a random
+    aspect ratio (default: of 3/4 to 4/3) of the original aspect ratio is made. This crop
+    is finally resized to given size.
+    This is popularly used to train the Inception networks.
+
+    Args:
+        size: expected output size of each edge
+        scale: range of size of the origin size cropped
+        ratio: range of aspect ratio of the origin aspect ratio cropped
+        interpolation: Default: PIL.Image.BILINEAR
+    """
+
+    def __init__(self, size, scale=(0.08, 1.0), ratio=(3. / 4., 4. / 3.), interpolation=Image.BILINEAR):
+        if isinstance(size, tuple):
+            self.size = size
+        else:
+            self.size = (size, size)
+        self.interpolation = interpolation
+        self.scale = scale
+        self.ratio = ratio
+
+    @staticmethod
+    def get_params(img, scale, ratio):
+        """Get parameters for ``crop`` for a random sized crop.
+
+        Args:
+            img (PIL Image): Image to be cropped.
+            scale (tuple): range of size of the origin size cropped
+            ratio (tuple): range of aspect ratio of the origin aspect ratio cropped
+
+        Returns:
+            tuple: params (i, j, h, w) to be passed to ``crop`` for a random
+                sized crop.
+        """
+        for attempt in range(10):
+            area = img.size[0] * img.size[1]
+            target_area = random.uniform(*scale) * area
+            aspect_ratio = random.uniform(*ratio)
+
+            w = int(round(math.sqrt(target_area * aspect_ratio)))
+            h = int(round(math.sqrt(target_area / aspect_ratio)))
+
+            if random.random() < 0.5:
+                w, h = h, w
+
+            if w <= img.size[0] and h <= img.size[1]:
+                i = random.randint(0, img.size[1] - h)
+                j = random.randint(0, img.size[0] - w)
+                return i, j, h, w
+
+        # Fallback
+        w = min(img.size[0], img.size[1])
+        i = (img.size[1] - w) // 2
+        j = (img.size[0] - w) // 2
+        return i, j, w, w
+
+    def __call__(self, img):
+        """
+        Args:
+            img (PIL Image): Image to be flipped.
+
+        Returns:
+            PIL Image: Randomly cropped and resize image.
+        """
+        i, j, h, w = self.get_params(img, self.scale, self.ratio)
+        return F.resized_crop(img, i, j, h, w, self.size, self.interpolation)
+
+
+class ColorJitter(object):
+    """Randomly change the brightness, contrast and saturation of an image.
+
+    Args:
+        brightness (float): How much to jitter brightness. brightness_factor
+            is chosen uniformly from [max(0, 1 - brightness), 1 + brightness].
+        contrast (float): How much to jitter contrast. contrast_factor
+            is chosen uniformly from [max(0, 1 - contrast), 1 + contrast].
+        saturation (float): How much to jitter saturation. saturation_factor
+            is chosen uniformly from [max(0, 1 - saturation), 1 + saturation].
+        hue(float): How much to jitter hue. hue_factor is chosen uniformly from
+            [-hue, hue]. Should be >=0 and <= 0.5.
+    """
+    def __init__(self, brightness=0, contrast=0, saturation=0, hue=0):
+        self.brightness = brightness
+        self.contrast = contrast
+        self.saturation = saturation
+        self.hue = hue
+
+    @staticmethod
+    def get_params(brightness, contrast, saturation, hue):
+        """Get a randomized transform to be applied on image.
+
+        Arguments are same as that of __init__.
+
+        Returns:
+            Transform which randomly adjusts brightness, contrast and
+            saturation in a random order.
+        """
+        transforms = []
+        if brightness > 0:
+            brightness_factor = np.random.uniform(max(0, 1 - brightness), 1 + brightness)
+            transforms.append(Lambda(lambda img: F.adjust_brightness(img, brightness_factor)))
+
+        if contrast > 0:
+            contrast_factor = np.random.uniform(max(0, 1 - contrast), 1 + contrast)
+            transforms.append(Lambda(lambda img: F.adjust_contrast(img, contrast_factor)))
+
+        if saturation > 0:
+            saturation_factor = np.random.uniform(max(0, 1 - saturation), 1 + saturation)
+            transforms.append(Lambda(lambda img: F.adjust_saturation(img, saturation_factor)))
+
+        if hue > 0:
+            hue_factor = np.random.uniform(-hue, hue)
+            transforms.append(Lambda(lambda img: F.adjust_hue(img, hue_factor)))
+
+        np.random.shuffle(transforms)
+        transform = Compose(transforms)
+
+        return transform
+
+    def __call__(self, img):
+        """
+        Args:
+            img (PIL Image): Input image.
+
+        Returns:
+            PIL Image: Color jittered image.
+        """
+        transform = self.get_params(self.brightness, self.contrast,
+                                    self.saturation, self.hue)
+        return transform(img)
+
+
+class RandomRotation(object):
+    """Rotate the image by angle.
+
+    Args:
+        degrees (sequence or float or int): Range of degrees to select from.
+            If degrees is a number instead of sequence like (min, max), the range of degrees
+            will be (-degrees, +degrees).
+        resample ({PIL.Image.NEAREST, PIL.Image.BILINEAR, PIL.Image.BICUBIC}, optional):
+            An optional resampling filter.
+            See http://pillow.readthedocs.io/en/3.4.x/handbook/concepts.html#filters
+            If omitted, or if the image has mode "1" or "P", it is set to PIL.Image.NEAREST.
+        expand (bool, optional): Optional expansion flag.
+            If true, expands the output to make it large enough to hold the entire rotated image.
+            If false or omitted, make the output image the same size as the input image.
+            Note that the expand flag assumes rotation around the center and no translation.
+        center (2-tuple, optional): Optional center of rotation.
+            Origin is the upper left corner.
+            Default is the center of the image.
+    """
+
+    def __init__(self, degrees, resample=False, expand=False, center=None):
+        if isinstance(degrees, numbers.Number):
+            if degrees < 0:
+                raise ValueError("If degrees is a single number, it must be positive.")
+            self.degrees = (-degrees, degrees)
+        else:
+            if len(degrees) != 2:
+                raise ValueError("If degrees is a sequence, it must be of len 2.")
+            self.degrees = degrees
+
+        self.resample = resample
+        self.expand = expand
+        self.center = center
+
+    @staticmethod
+    def get_params(degrees):
+        """Get parameters for ``rotate`` for a random rotation.
+
+        Returns:
+            sequence: params to be passed to ``rotate`` for random rotation.
+        """
+        angle = np.random.uniform(degrees[0], degrees[1])
+
+        return angle
+
+    def __call__(self, img):
+        """
+            img (PIL Image): Image to be rotated.
+
+        Returns:
+            PIL Image: Rotated image.
+        """
+
+        angle = self.get_params(self.degrees)
+
+        return F.rotate(img, angle, self.resample, self.expand, self.center)
+
+
diff --git a/src/default_config.py b/src/default_config.py
new file mode 100644
index 0000000..d630c6e
--- /dev/null
+++ b/src/default_config.py
@@ -0,0 +1,73 @@
+# -*- coding: utf-8 -*-
+# @Time : 20-6-4 上午9:12
+# @Author : zhuying
+# @Company : Minivision
+# @File : default_config.py
+# @Software : PyCharm
+# --*-- coding: utf-8 --*--
+"""
+default config for training
+"""
+
+import torch
+from datetime import datetime
+from easydict import EasyDict
+from src.utility import make_if_not_exist, get_width_height, get_kernel
+
+
+def get_default_config():
+    conf = EasyDict()
+
+    # ----------------------training---------------
+    conf.lr = 1e-1
+    # [9, 13, 15]
+    conf.milestones = [10, 15, 22]  # down learing rate
+    conf.gamma = 0.1
+    conf.epochs = 25
+    conf.momentum = 0.9
+    conf.batch_size = 1024
+
+    # model
+    conf.num_classes = 3
+    conf.input_channel = 3
+    conf.embedding_size = 128
+
+    # dataset
+    conf.train_root_path = './datasets/rgb_image'
+
+    # save file path
+    conf.snapshot_dir_path = './saved_logs/snapshot'
+
+    # log path
+    conf.log_path = './saved_logs/jobs'
+    # tensorboard
+    conf.board_loss_every = 10
+    # save model/iter
+    conf.save_every = 30
+
+    return conf
+
+
+def update_config(args, conf):
+    conf.devices = args.devices
+    conf.patch_info = args.patch_info
+    w_input, h_input = get_width_height(args.patch_info)
+    conf.input_size = [h_input, w_input]
+    conf.kernel_size = get_kernel(h_input, w_input)
+    conf.device = "cuda:{}".format(conf.devices[0]) if torch.cuda.is_available() else "cpu"
+
+    # resize fourier image size
+    conf.ft_height = 2*conf.kernel_size[0]
+    conf.ft_width = 2*conf.kernel_size[1]
+    current_time = datetime.now().strftime('%b%d_%H-%M-%S')
+    job_name = 'Anti_Spoofing_{}'.format(args.patch_info)
+    log_path = '{}/{}/{} '.format(conf.log_path, job_name, current_time)
+    snapshot_dir = '{}/{}'.format(conf.snapshot_dir_path, job_name)
+
+    make_if_not_exist(snapshot_dir)
+    make_if_not_exist(log_path)
+
+    conf.model_path = snapshot_dir
+    conf.log_path = log_path
+    conf.job_name = job_name
+    return conf
diff --git a/src/generate_patches.py b/src/generate_patches.py
new file mode 100644
index 0000000..0056269
--- /dev/null
+++ b/src/generate_patches.py
@@ -0,0 +1,65 @@
+# -*- coding: utf-8 -*-
+# @Time : 20-6-9 下午3:06
+# @Author : zhuying
+# @Company : Minivision
+# @File : test.py
+# @Software : PyCharm
+"""
+Create patch from original input image by using bbox coordinate
+"""
+
+import cv2
+import numpy as np
+
+
+class CropImage:
+    @staticmethod
+    def _get_new_box(src_w, src_h, bbox, scale):
+        x = bbox[0]
+        y = bbox[1]
+        box_w = bbox[2]
+        box_h = bbox[3]
+
+        scale = min((src_h-1)/box_h, min((src_w-1)/box_w, scale))
+
+        new_width = box_w * scale
+        new_height = box_h * scale
+        center_x, center_y = box_w/2+x, box_h/2+y
+
+        left_top_x = center_x-new_width/2
+        left_top_y = center_y-new_height/2
+        right_bottom_x = center_x+new_width/2
+        right_bottom_y = center_y+new_height/2
+
+        if left_top_x < 0:
+            right_bottom_x -= left_top_x
+            left_top_x = 0
+
+        if left_top_y < 0:
+            right_bottom_y -= left_top_y
+            left_top_y = 0
+
+        if right_bottom_x > src_w-1:
+            left_top_x -= right_bottom_x-src_w+1
+            right_bottom_x = src_w-1
+
+        if right_bottom_y > src_h-1:
+            left_top_y -= right_bottom_y-src_h+1
+            right_bottom_y = src_h-1
+
+        return int(left_top_x), int(left_top_y),\
+               int(right_bottom_x), int(right_bottom_y)
+
+    def crop(self, org_img, bbox, scale, out_w, out_h, crop=True):
+
+        if not crop:
+            dst_img = cv2.resize(org_img, (out_w, out_h))
+        else:
+            src_h, src_w, _ = np.shape(org_img)
+            left_top_x, left_top_y, \
+                right_bottom_x, right_bottom_y = self._get_new_box(src_w, src_h, bbox, scale)
+
+            img = org_img[left_top_y: right_bottom_y+1,
+                          left_top_x: right_bottom_x+1]
+            dst_img = cv2.resize(img, (out_w, out_h))
+        return dst_img
diff --git a/src/model_lib/MiniFASNet.py b/src/model_lib/MiniFASNet.py
new file mode 100644
index 0000000..548f773
--- /dev/null
+++ b/src/model_lib/MiniFASNet.py
@@ -0,0 +1,296 @@
+# -*- coding: utf-8 -*-
+# @Time : 20-6-3 下午4:45
+# @Author : zhuying
+# @Company : Minivision
+# @File : MiniFASNet.py
+# @Software : PyCharm
+import torch
+import torch.nn.functional as F
+from torch.nn import Linear, Conv2d, BatchNorm1d, BatchNorm2d, PReLU, ReLU, Sigmoid, \
+    AdaptiveAvgPool2d, Sequential, Module
+
+
+class L2Norm(Module):
+    def forward(self, input):
+        return F.normalize(input)
+
+
+class Flatten(Module):
+    def forward(self, input):
+        return input.view(input.size(0), -1)
+
+
+class Conv_block(Module):
+    def __init__(self, in_c, out_c, kernel=(1, 1), stride=(1, 1), padding=(0, 0), groups=1):
+        super(Conv_block, self).__init__()
+        self.conv = Conv2d(in_c, out_c, kernel_size=kernel, groups=groups,
+                           stride=stride, padding=padding, bias=False)
+        self.bn = BatchNorm2d(out_c)
+        self.prelu = PReLU(out_c)
+
+    def forward(self, x):
+        x = self.conv(x)
+        x = self.bn(x)
+        x = self.prelu(x)
+        return x
+
+
+class Linear_block(Module):
+    def __init__(self, in_c, out_c, kernel=(1, 1), stride=(1, 1), padding=(0, 0), groups=1):
+        super(Linear_block, self).__init__()
+        self.conv = Conv2d(in_c, out_channels=out_c, kernel_size=kernel,
+                           groups=groups, stride=stride, padding=padding, bias=False)
+        self.bn = BatchNorm2d(out_c)
+
+    def forward(self, x):
+        x = self.conv(x)
+        x = self.bn(x)
+        return x
+
+
+class Depth_Wise(Module):
+     def __init__(self, c1, c2, c3, residual=False, kernel=(3, 3), stride=(2, 2), padding=(1, 1), groups=1):
+        super(Depth_Wise, self).__init__()
+        c1_in, c1_out = c1
+        c2_in, c2_out = c2
+        c3_in, c3_out = c3
+        self.conv = Conv_block(c1_in, out_c=c1_out, kernel=(1, 1), padding=(0, 0), stride=(1, 1))
+        self.conv_dw = Conv_block(c2_in, c2_out, groups=c2_in, kernel=kernel, padding=padding, stride=stride)
+        self.project = Linear_block(c3_in, c3_out, kernel=(1, 1), padding=(0, 0), stride=(1, 1))
+        self.residual = residual
+
+     def forward(self, x):
+        if self.residual:
+            short_cut = x
+        x = self.conv(x)
+        x = self.conv_dw(x)
+        x = self.project(x)
+        if self.residual:
+            output = short_cut + x
+        else:
+            output = x
+        return output
+
+
+class Residual(Module):
+    def __init__(self, c1, c2, c3, num_block, groups, kernel=(3, 3), stride=(1, 1), padding=(1, 1)):
+        super(Residual, self).__init__()
+        modules = []
+        for i in range(num_block):
+            c1_tuple = c1[i]
+            c2_tuple = c2[i]
+            c3_tuple = c3[i]
+            modules.append(Depth_Wise(c1_tuple, c2_tuple, c3_tuple, residual=True,
+                                      kernel=kernel, padding=padding, stride=stride, groups=groups))
+        self.model = Sequential(*modules)
+
+    def forward(self, x):
+        return self.model(x)
+
+
+class SEModule(Module):
+    def __init__(self, channels, reduction):
+        super(SEModule, self).__init__()
+        self.avg_pool = AdaptiveAvgPool2d(1)
+        self.fc1 = Conv2d(
+            channels, channels // reduction, kernel_size=1, padding=0, bias=False)
+        self.bn1 = BatchNorm2d(channels // reduction)
+        self.relu = ReLU(inplace=True)
+        self.fc2 = Conv2d(
+            channels // reduction, channels, kernel_size=1, padding=0, bias=False)
+        self.bn2 = BatchNorm2d(channels)
+        self.sigmoid = Sigmoid()
+
+    def forward(self, x):
+        module_input = x
+        x = self.avg_pool(x)
+        x = self.fc1(x)
+        x = self.bn1(x)
+        x = self.relu(x)
+        x = self.fc2(x)
+        x = self.bn2(x)
+        x = self.sigmoid(x)
+        return module_input * x
+
+
+class ResidualSE(Module):
+    def __init__(self, c1, c2, c3, num_block, groups, kernel=(3, 3), stride=(1, 1), padding=(1, 1), se_reduct=4):
+        super(ResidualSE, self).__init__()
+        modules = []
+        for i in range(num_block):
+            c1_tuple = c1[i]
+            c2_tuple = c2[i]
+            c3_tuple = c3[i]
+            if i == num_block-1:
+                modules.append(
+                    Depth_Wise_SE(c1_tuple, c2_tuple, c3_tuple, residual=True, kernel=kernel, padding=padding, stride=stride,
+                               groups=groups, se_reduct=se_reduct))
+            else:
+                modules.append(Depth_Wise(c1_tuple, c2_tuple, c3_tuple, residual=True, kernel=kernel, padding=padding,
+                                          stride=stride, groups=groups))
+        self.model = Sequential(*modules)
+
+    def forward(self, x):
+        return self.model(x)
+
+
+class Depth_Wise_SE(Module):
+    def __init__(self, c1, c2, c3, residual=False, kernel=(3, 3), stride=(2, 2), padding=(1, 1), groups=1, se_reduct=8):
+        super(Depth_Wise_SE, self).__init__()
+        c1_in, c1_out = c1
+        c2_in, c2_out = c2
+        c3_in, c3_out = c3
+        self.conv = Conv_block(c1_in, out_c=c1_out, kernel=(1, 1), padding=(0, 0), stride=(1, 1))
+        self.conv_dw = Conv_block(c2_in, c2_out, groups=c2_in, kernel=kernel, padding=padding, stride=stride)
+        self.project = Linear_block(c3_in, c3_out, kernel=(1, 1), padding=(0, 0), stride=(1, 1))
+        self.residual = residual
+        self.se_module = SEModule(c3_out, se_reduct)
+
+    def forward(self, x):
+        if self.residual:
+            short_cut = x
+        x = self.conv(x)
+        x = self.conv_dw(x)
+        x = self.project(x)
+        if self.residual:
+            x = self.se_module(x)
+            output = short_cut + x
+        else:
+            output = x
+        return output
+
+
+class MiniFASNet(Module):
+    def __init__(self, keep, embedding_size, conv6_kernel=(7, 7),
+                 drop_p=0.0, num_classes=3, img_channel=3):
+        super(MiniFASNet, self).__init__()
+        self.embedding_size = embedding_size
+
+        self.conv1 = Conv_block(img_channel, keep[0], kernel=(3, 3), stride=(2, 2), padding=(1, 1))
+        self.conv2_dw = Conv_block(keep[0], keep[1], kernel=(3, 3), stride=(1, 1), padding=(1, 1), groups=keep[1])
+
+        c1 = [(keep[1], keep[2])]
+        c2 = [(keep[2], keep[3])]
+        c3 = [(keep[3], keep[4])]
+
+        self.conv_23 = Depth_Wise(c1[0], c2[0], c3[0], kernel=(3, 3), stride=(2, 2), padding=(1, 1), groups=keep[3])
+
+        c1 = [(keep[4], keep[5]), (keep[7], keep[8]), (keep[10], keep[11]), (keep[13], keep[14])]
+        c2 = [(keep[5], keep[6]), (keep[8], keep[9]), (keep[11], keep[12]), (keep[14], keep[15])]
+        c3 = [(keep[6], keep[7]), (keep[9], keep[10]), (keep[12], keep[13]), (keep[15], keep[16])]
+
+        self.conv_3 = Residual(c1, c2, c3, num_block=4, groups=keep[4], kernel=(3, 3), stride=(1, 1), padding=(1, 1))
+
+        c1 = [(keep[16], keep[17])]
+        c2 = [(keep[17], keep[18])]
+        c3 = [(keep[18], keep[19])]
+
+        self.conv_34 = Depth_Wise(c1[0], c2[0], c3[0], kernel=(3, 3), stride=(2, 2), padding=(1, 1), groups=keep[19])
+
+        c1 = [(keep[19], keep[20]), (keep[22], keep[23]), (keep[25], keep[26]), (keep[28], keep[29]),
+              (keep[31], keep[32]), (keep[34], keep[35])]
+        c2 = [(keep[20], keep[21]), (keep[23], keep[24]), (keep[26], keep[27]), (keep[29], keep[30]),
+              (keep[32], keep[33]), (keep[35], keep[36])]
+        c3 = [(keep[21], keep[22]), (keep[24], keep[25]), (keep[27], keep[28]), (keep[30], keep[31]),
+              (keep[33], keep[34]), (keep[36], keep[37])]
+
+        self.conv_4 = Residual(c1, c2, c3, num_block=6, groups=keep[19], kernel=(3, 3), stride=(1, 1), padding=(1, 1))
+
+        c1 = [(keep[37], keep[38])]
+        c2 = [(keep[38], keep[39])]
+        c3 = [(keep[39], keep[40])]
+
+        self.conv_45 = Depth_Wise(c1[0], c2[0], c3[0], kernel=(3, 3), stride=(2, 2), padding=(1, 1), groups=keep[40])
+
+        c1 = [(keep[40], keep[41]), (keep[43], keep[44])]
+        c2 = [(keep[41], keep[42]), (keep[44], keep[45])]
+        c3 = [(keep[42], keep[43]), (keep[45], keep[46])]
+
+        self.conv_5 = Residual(c1, c2, c3, num_block=2, groups=keep[40], kernel=(3, 3), stride=(1, 1), padding=(1, 1))
+        self.conv_6_sep = Conv_block(keep[46], keep[47], kernel=(1, 1), stride=(1, 1), padding=(0, 0))
+        self.conv_6_dw = Linear_block(keep[47], keep[48], groups=keep[48], kernel=conv6_kernel, stride=(1, 1), padding=(0, 0))
+        self.conv_6_flatten = Flatten()
+        self.linear = Linear(512, embedding_size, bias=False)
+        self.bn = BatchNorm1d(embedding_size)
+        self.drop = torch.nn.Dropout(p=drop_p)
+        self.prob = Linear(embedding_size, num_classes, bias=False)
+
+    def forward(self, x):
+        out = self.conv1(x)
+        out = self.conv2_dw(out)
+        out = self.conv_23(out)
+        out = self.conv_3(out)
+        out = self.conv_34(out)
+        out = self.conv_4(out)
+        out = self.conv_45(out)
+        out = self.conv_5(out)
+        out = self.conv_6_sep(out)
+        out = self.conv_6_dw(out)
+        out = self.conv_6_flatten(out)
+        if self.embedding_size != 512:
+            out = self.linear(out)
+        out = self.bn(out)
+        out = self.drop(out)
+        out = self.prob(out)
+        return out
+
+
+class MiniFASNetSE(MiniFASNet):
+    def __init__(self, keep, embedding_size, conv6_kernel=(7, 7),drop_p=0.75, num_classes=4, img_channel=3):
+        super(MiniFASNetSE, self).__init__(keep=keep, embedding_size=embedding_size, conv6_kernel=conv6_kernel,
+                                               drop_p=drop_p, num_classes=num_classes, img_channel=img_channel)
+
+        c1 = [(keep[4], keep[5]), (keep[7], keep[8]), (keep[10], keep[11]), (keep[13], keep[14])]
+        c2 = [(keep[5], keep[6]), (keep[8], keep[9]), (keep[11], keep[12]), (keep[14], keep[15])]
+        c3 = [(keep[6], keep[7]), (keep[9], keep[10]), (keep[12], keep[13]), (keep[15], keep[16])]
+
+        self.conv_3 = ResidualSE(c1, c2, c3, num_block=4, groups=keep[4], kernel=(3, 3), stride=(1, 1), padding=(1, 1))
+
+        c1 = [(keep[19], keep[20]), (keep[22], keep[23]), (keep[25], keep[26]), (keep[28], keep[29]),
+              (keep[31], keep[32]), (keep[34], keep[35])]
+        c2 = [(keep[20], keep[21]), (keep[23], keep[24]), (keep[26], keep[27]), (keep[29], keep[30]),
+              (keep[32], keep[33]), (keep[35], keep[36])]
+        c3 = [(keep[21], keep[22]), (keep[24], keep[25]), (keep[27], keep[28]), (keep[30], keep[31]),
+              (keep[33], keep[34]), (keep[36], keep[37])]
+
+        self.conv_4 = ResidualSE(c1, c2, c3, num_block=6, groups=keep[19], kernel=(3, 3), stride=(1, 1), padding=(1, 1))
+
+        c1 = [(keep[40], keep[41]), (keep[43], keep[44])]
+        c2 = [(keep[41], keep[42]), (keep[44], keep[45])]
+        c3 = [(keep[42], keep[43]), (keep[45], keep[46])]
+        self.conv_5 = ResidualSE(c1, c2, c3, num_block=2, groups=keep[40], kernel=(3, 3), stride=(1, 1), padding=(1, 1))
+
+
+
+keep_dict = {'1.8M': [32, 32, 103, 103, 64, 13, 13, 64, 26, 26,
+                      64, 13, 13, 64, 52, 52, 64, 231, 231, 128,
+                      154, 154, 128, 52, 52, 128, 26, 26, 128, 52,
+                      52, 128, 26, 26, 128, 26, 26, 128, 308, 308,
+                      128, 26, 26, 128, 26, 26, 128, 512, 512],
+
+             '1.8M_': [32, 32, 103, 103, 64, 13, 13, 64, 13, 13, 64, 13,
+                       13, 64, 13, 13, 64, 231, 231, 128, 231, 231, 128, 52,
+                       52, 128, 26, 26, 128, 77, 77, 128, 26, 26, 128, 26, 26,
+                       128, 308, 308, 128, 26, 26, 128, 26, 26, 128, 512, 512]
+             }
+
+
+# (80x80) flops: 0.044, params: 0.41
+def MiniFASNetV1(embedding_size=128, conv6_kernel=(7, 7),
+                     drop_p=0.2, num_classes=3, img_channel=3):
+    return MiniFASNet(keep_dict['1.8M'], embedding_size, conv6_kernel, drop_p, num_classes, img_channel)
+
+
+# (80x80) flops: 0.044, params: 0.43
+def MiniFASNetV2(embedding_size=128, conv6_kernel=(7, 7),
+                     drop_p=0.2, num_classes=3, img_channel=3):
+    return MiniFASNet(keep_dict['1.8M_'], embedding_size, conv6_kernel, drop_p, num_classes, img_channel)
+
+def MiniFASNetV1SE(embedding_size=128, conv6_kernel=(7, 7),
+                   drop_p=0.75, num_classes=3, img_channel=3):
+    return MiniFASNetSE(keep_dict['1.8M'], embedding_size, conv6_kernel,drop_p, num_classes, img_channel)
+
+# (80x80) flops: 0.044, params: 0.43
+def MiniFASNetV2SE(embedding_size=128, conv6_kernel=(7, 7),
+                   drop_p=0.75, num_classes=4, img_channel=3):
+    return MiniFASNetSE(keep_dict['1.8M_'], embedding_size, conv6_kernel,drop_p, num_classes, img_channel)
diff --git a/src/model_lib/MultiFTNet.py b/src/model_lib/MultiFTNet.py
new file mode 100644
index 0000000..f934df4
--- /dev/null
+++ b/src/model_lib/MultiFTNet.py
@@ -0,0 +1,79 @@
+# -*- coding: utf-8 -*-
+# @Time : 20-6-3 下午5:14
+# @Author : zhuying
+# @Company : Minivision
+# @File : MultiFTNet.py
+# @Software : PyCharm
+from torch import nn
+import torch.nn.functional as F
+from src.model_lib.MiniFASNet import MiniFASNetV1,MiniFASNetV2,MiniFASNetV1SE,MiniFASNetV2SE
+
+
+class FTGenerator(nn.Module):
+    def __init__(self, in_channels=48, out_channels=1):
+        super(FTGenerator, self).__init__()
+
+        self.ft = nn.Sequential(
+            nn.Conv2d(in_channels, 128, kernel_size=(3, 3), padding=1),
+            nn.BatchNorm2d(128),
+            nn.ReLU(inplace=True),
+
+            nn.Conv2d(128, 64, kernel_size=(3, 3), padding=1),
+            nn.BatchNorm2d(64),
+            nn.ReLU(inplace=True),
+
+            nn.Conv2d(64, out_channels, kernel_size=(3, 3), padding=1),
+            nn.BatchNorm2d(out_channels),
+            nn.ReLU(inplace=True)
+        )
+
+    def forward(self, x):
+        return self.ft(x)
+
+
+class MultiFTNet(nn.Module):
+    def __init__(self, img_channel=3, num_classes=3, embedding_size=128, conv6_kernel=(5, 5)):
+        super(MultiFTNet, self).__init__()
+        self.img_channel = img_channel
+        self.num_classes = num_classes
+        self.model = MiniFASNetV2SE(embedding_size=embedding_size, conv6_kernel=conv6_kernel,
+                                      num_classes=num_classes, img_channel=img_channel)
+        self.FTGenerator = FTGenerator(in_channels=128)
+        self._initialize_weights()
+
+    def _initialize_weights(self):
+        for m in self.modules():
+            if isinstance(m, nn.Conv2d):
+                nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu')
+                if m.bias is not None:
+                    nn.init.constant_(m.bias, 0)
+            elif isinstance(m, (nn.BatchNorm2d, nn.BatchNorm1d, nn.GroupNorm)):
+                nn.init.constant_(m.weight, 1)
+                nn.init.constant_(m.bias, 0)
+            elif isinstance(m, nn.Linear):
+                nn.init.normal_(m.weight, std=0.001)
+                if m.bias is not None:
+                    nn.init.constant_(m.bias, 0)
+
+    def forward(self, x):
+        x = self.model.conv1(x)
+        x = self.model.conv2_dw(x)
+        x = self.model.conv_23(x)
+        x = self.model.conv_3(x)
+        x = self.model.conv_34(x)
+        x = self.model.conv_4(x)
+        x1 = self.model.conv_45(x)
+        x1 = self.model.conv_5(x1)
+        x1 = self.model.conv_6_sep(x1)
+        x1 = self.model.conv_6_dw(x1)
+        x1 = self.model.conv_6_flatten(x1)
+        x1 = self.model.linear(x1)
+        x1 = self.model.bn(x1)
+        x1 = self.model.drop(x1)
+        cls = self.model.prob(x1)
+
+        if self.training:
+            ft = self.FTGenerator(x)
+            return cls, ft
+        else:
+            return cls
diff --git a/src/model_lib/__pycache__/MiniFASNet.cpython-38.pyc b/src/model_lib/__pycache__/MiniFASNet.cpython-38.pyc
new file mode 100644
index 0000000..e92fce8
Binary files /dev/null and b/src/model_lib/__pycache__/MiniFASNet.cpython-38.pyc differ
diff --git a/src/train_main.py b/src/train_main.py
new file mode 100644
index 0000000..7f3be7d
--- /dev/null
+++ b/src/train_main.py
@@ -0,0 +1,146 @@
+# -*- coding: utf-8 -*-
+# @Time : 20-6-4 上午9:59
+# @Author : zhuying
+# @Company : Minivision
+# @File : train_main.py
+# @Software : PyCharm
+
+import torch
+from torch import optim
+from torch.nn import CrossEntropyLoss, MSELoss
+from tqdm import tqdm
+from tensorboardX import SummaryWriter
+
+from src.utility import get_time
+from src.model_lib.MultiFTNet import MultiFTNet
+from src.data_io.dataset_loader import get_train_loader
+
+
+class TrainMain:
+    def __init__(self, conf):
+        self.conf = conf
+        self.board_loss_every = conf.board_loss_every
+        self.save_every = conf.save_every
+        self.step = 0
+        self.start_epoch = 0
+        self.train_loader = get_train_loader(self.conf)
+
+    def train_model(self):
+        self._init_model_param()
+        self._train_stage()
+
+    def _init_model_param(self):
+        self.cls_criterion = CrossEntropyLoss()
+        self.ft_criterion = MSELoss()
+        self.model = self._define_network()
+        self.optimizer = optim.SGD(self.model.module.parameters(),
+                                   lr=self.conf.lr,
+                                   weight_decay=5e-4,
+                                   momentum=self.conf.momentum)
+
+        self.schedule_lr = optim.lr_scheduler.MultiStepLR(
+            self.optimizer, self.conf.milestones, self.conf.gamma, - 1)
+
+        print("lr: ", self.conf.lr)
+        print("epochs: ", self.conf.epochs)
+        print("milestones: ", self.conf.milestones)
+
+    def _train_stage(self):
+        self.model.train()
+        running_loss = 0.
+        running_acc = 0.
+        running_loss_cls = 0.
+        running_loss_ft = 0.
+        is_first = True
+        for e in range(self.start_epoch, self.conf.epochs):
+            if is_first:
+                self.writer = SummaryWriter(self.conf.log_path)
+                is_first = False
+            print('epoch {} started'.format(e))
+            print("lr: ", self.schedule_lr.get_lr())
+
+            for sample, ft_sample, target in tqdm(iter(self.train_loader)):
+                imgs = [sample, ft_sample]
+                labels = target
+
+                loss, acc, loss_cls, loss_ft = self._train_batch_data(imgs, labels)
+                running_loss_cls += loss_cls
+                running_loss_ft += loss_ft
+                running_loss += loss
+                running_acc += acc
+
+                self.step += 1
+
+                if self.step % self.board_loss_every == 0 and self.step != 0:
+                    loss_board = running_loss / self.board_loss_every
+                    self.writer.add_scalar(
+                        'Training/Loss', loss_board, self.step)
+                    acc_board = running_acc / self.board_loss_every
+                    self.writer.add_scalar(
+                        'Training/Acc', acc_board, self.step)
+                    lr = self.optimizer.param_groups[0]['lr']
+                    self.writer.add_scalar(
+                        'Training/Learning_rate', lr, self.step)
+                    loss_cls_board = running_loss_cls / self.board_loss_every
+                    self.writer.add_scalar(
+                        'Training/Loss_cls', loss_cls_board, self.step)
+                    loss_ft_board = running_loss_ft / self.board_loss_every
+                    self.writer.add_scalar(
+                        'Training/Loss_ft', loss_ft_board, self.step)
+
+                    running_loss = 0.
+                    running_acc = 0.
+                    running_loss_cls = 0.
+                    running_loss_ft = 0.
+                if self.step % self.save_every == 0 and self.step != 0:
+                    time_stamp = get_time()
+                    self._save_state(time_stamp, extra=self.conf.job_name)
+            self.schedule_lr.step()
+
+        time_stamp = get_time()
+        self._save_state(time_stamp, extra=self.conf.job_name)
+        self.writer.close()
+
+    def _train_batch_data(self, imgs, labels):
+        self.optimizer.zero_grad()
+        labels = labels.to(self.conf.device)
+        embeddings, feature_map = self.model.forward(imgs[0].to(self.conf.device))
+
+        loss_cls = self.cls_criterion(embeddings, labels)
+        loss_fea = self.ft_criterion(feature_map, imgs[1].to(self.conf.device))
+
+        loss = 0.5*loss_cls + 0.5*loss_fea
+        acc = self._get_accuracy(embeddings, labels)[0]
+        loss.backward()
+        self.optimizer.step()
+        return loss.item(), acc, loss_cls.item(), loss_fea.item()
+
+    def _define_network(self):
+        param = {
+            'num_classes': self.conf.num_classes,
+            'img_channel': self.conf.input_channel,
+            'embedding_size': self.conf.embedding_size,
+            'conv6_kernel': self.conf.kernel_size}
+
+        model = MultiFTNet(**param).to(self.conf.device)
+        model = torch.nn.DataParallel(model, self.conf.devices)
+        model.to(self.conf.device)
+        return model
+
+    def _get_accuracy(self, output, target, topk=(1,)):
+        maxk = max(topk)
+        batch_size = target.size(0)
+        _, pred = output.topk(maxk, 1, True, True)
+        pred = pred.t()
+        correct = pred.eq(target.view(1, -1).expand_as(pred))
+
+        ret = []
+        for k in topk:
+            correct_k = correct[:k].view(-1).float().sum(dim=0, keepdim=True)
+            ret.append(correct_k.mul_(1. / batch_size))
+        return ret
+
+    def _save_state(self, time_stamp, extra=None):
+        save_path = self.conf.model_path
+        torch.save(self.model.state_dict(), save_path + '/' +
+                   ('{}_{}_model_iter-{}.pth'.format(time_stamp, extra, self.step)))
diff --git a/src/utility.py b/src/utility.py
new file mode 100644
index 0000000..5615498
--- /dev/null
+++ b/src/utility.py
@@ -0,0 +1,41 @@
+# -*- coding: utf-8 -*-
+# @Time : 20-6-4 下午2:13
+# @Author : zhuying
+# @Company : Minivision
+# @File : utility.py
+# @Software : PyCharm
+
+from datetime import datetime
+import os
+
+
+def get_time():
+    return (str(datetime.now())[:-10]).replace(' ', '-').replace(':', '-')
+
+
+def get_kernel(height, width):
+    kernel_size = ((height + 15) // 16, (width + 15) // 16)
+    return kernel_size
+
+
+def get_width_height(patch_info):
+    w_input = int(patch_info.split('x')[-1])
+    h_input = int(patch_info.split('x')[0].split('_')[-1])
+    return w_input,h_input
+
+
+def parse_model_name(model_name):
+    info = model_name.split('_')[0:-1]
+    h_input, w_input = info[-1].split('x')
+    model_type = model_name.split('.pth')[0].split('_')[-1]
+
+    if info[0] == "org":
+        scale = None
+    else:
+        scale = float(info[0])
+    return int(h_input), int(w_input), model_type, scale
+
+
+def make_if_not_exist(folder_path):
+    if not os.path.exists(folder_path):
+        os.makedirs(folder_path)
diff --git a/stream.py b/stream.py
new file mode 100644
index 0000000..0647975
--- /dev/null
+++ b/stream.py
@@ -0,0 +1,36 @@
+import cv2
+import subprocess
+
+rtsp = "rtsp://admin:2020@uestc@192.168.30.83:554/h264"
+out_rtsp = 'rtsp://localhost:5001/test'
+
+# 读取视频并获取属性
+cap = cv2.VideoCapture(rtsp)
+size = (int(cap.get(cv2.CAP_PROP_FRAME_WIDTH)), int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT)))
+sizeStr = str(size[0]) + 'x' + str(size[1])
+fps = cap.get(cv2.CAP_PROP_FPS)
+command = ['ffmpeg',
+           '-y', '-an',
+           '-f', 'rawvideo',
+           '-vcodec', 'rawvideo',
+           '-pix_fmt', 'bgr24',
+           '-s', sizeStr,
+           '-r', str(fps),
+           '-i', '-',
+           '-c:v', 'libx264',
+           '-pix_fmt', 'yuv420p',
+           '-preset', 'ultrafast',
+           '-f', 'rtsp',
+           out_rtsp]
+
+pipe = subprocess.Popen(command, shell=False, stdin=subprocess.PIPE)
+
+while cap.isOpened():
+    success, frame = cap.read()
+    if success:
+        if cv2.waitKey(1) & 0xFF == ord('q'):
+            break
+        pipe.stdin.write(frame.tostring())
+
+cap.release()
+pipe.terminate()
diff --git a/test.py b/test.py
new file mode 100644
index 0000000..352e5a9
--- /dev/null
+++ b/test.py
@@ -0,0 +1,191 @@
+import itertools
+import os
+import time
+
+import torch
+import cv2
+import numpy as np
+from backbones import iresnet50,iresnet18,iresnet100
+
+# print(torch.cuda.is_available())
+def load_image(img_path):
+    img = cv2.imread(img_path)
+    #print(img)
+    #img2 = np.fliplr(img)
+    #print(img2.shape)
+    img = img.transpose((2, 0, 1))
+    #img2 = img2.transpose((2, 0, 1))
+    img = img[np.newaxis, :, :, :]
+    #img2 = img2[np.newaxis, :, :, :]
+    #img = np.concatenate((img,img2))
+    img = np.array(img, dtype=np.float32)
+    img -= 127.5
+    img /= 127.5
+    return img
+
+def findEuclideanDistance(source_representation, test_representation):
+    euclidean_distance = source_representation - test_representation
+    euclidean_distance = np.sum(np.multiply(euclidean_distance, euclidean_distance))
+    euclidean_distance = np.sqrt(euclidean_distance)
+    return euclidean_distance
+
+def findCosineDistance(source_representation, test_representation):
+    a = np.matmul(np.transpose(source_representation), test_representation)
+    b = np.sum(np.multiply(source_representation, source_representation))
+    c = np.sum(np.multiply(test_representation, test_representation))
+    return 1 - (a / (np.sqrt(b) * np.sqrt(c)))
+
+def l2_normalize(x):
+    return x / np.sqrt(np.sum(np.multiply(x, x)))
+
+def cosin_metric(x1, x2):
+    return np.dot(x1, x2) / (np.linalg.norm(x1) * np.linalg.norm(x2))
+
+def get_file_list(file_name):
+    file1 = []
+    file2 = []
+    code = []
+    with open(file_name,"r") as f:
+        file_list = f.readlines()
+    for i in file_list:
+        data = i.replace("\n","").split(" ")
+        file1.append(data[0])
+        file2.append(data[1])
+        code.append(data[2])
+    return file1,file2,code
+
+def load_list_images(file_list):
+    all = np.zeros((len(file_list),3,112,112),dtype=np.float32)
+    count = 0
+    for img_path in file_list:
+        img = cv2.imread(img_path)
+        img = img.transpose((2, 0, 1))
+        img = img[np.newaxis, :, :, :]
+        all[count] = img
+        count = count + 1
+    all -= 127.5
+    all /= 127.5
+    return all
+
+def threshold_acc(distance,code):
+    #thresholds = np.arange(24,36,0.1)
+    thresholds = np.arange(0.5,2, 0.1)
+    total = len(code)
+    max_acc = 0
+    threshold = 0
+    code_net = []
+    for i in thresholds:
+        for ds in distance:
+            if ds < i:
+                code_net.append(1)
+            else:
+                code_net.append(0)
+        right = np.sum(code==code_net)
+        #print(threshold, max_acc)
+        if right/total > max_acc:
+            max_acc = right / total
+            threshold = i
+        code_net = []
+    return threshold,max_acc
+
+def reject_and_error(distance,code):
+    total = len(code)
+    threshold = 1.20
+    code_net = []
+    reject = 0
+    error = 0
+    for ds in distance:
+        if ds < threshold:
+            code_net.append(1)
+        else:
+            code_net.append(0)
+    right = np.sum(code==code_net)
+    max_acc = right / total
+    #print(max_acc,threshold)
+    for i in range(total):
+        if code[i] == 1:
+            if code_net[i] == 0:
+                reject += 1
+        else:
+            if code_net[i] == 1:
+                error += 1
+    return reject,error,right
+# if __name__=='__main__':
+#     img1 = load_image("D:\Download\out\centerface_test\Adolfo_Rodriguez_Saa\Adolfo_Rodriguez_Saa_0002.jpg")
+#     img1 = torch.from_numpy(img1)
+#     #print(img1.shape)
+#     #print(img1)
+#     img2 = load_image("D:\Download\out\centerface_database\Dario_Franchitti\Dario_Franchitti_0001.jpg")
+#     img2 = torch.from_numpy(img2)
+#     model = iresnet100()
+#     model.load_state_dict(torch.load("./model/backbone100.pth",map_location="cpu"))
+#     #print(model)
+#     model.eval()
+#     with torch.no_grad():
+#         start_time = time.time()
+#         pred1 = model(img1)
+#         #print(pred1.shape)
+#         #print(pred1)
+#         pred2 = model(img2)
+#         pred1 = pred1.numpy()
+#         #print(pred1.shape)
+#         pred2 = pred2.numpy()
+#         distance = findEuclideanDistance(l2_normalize(pred1),l2_normalize(pred2))
+#         end_time = time.time()
+#         print("EuclideanDistance is :" + str(distance))
+#         if distance < 1.20:
+#             print("对比两张相同的人脸用时："+str(end_time - start_time))
+#         else:
+#             print("对比两张不同的人脸用时：" + str(end_time - start_time))
+
+
+if __name__=='__main__':
+    model = iresnet100()
+    model.load_state_dict(torch.load("./model/backbone100.pth", map_location="cpu"))
+    model.eval()
+    file1, file2, code = get_file_list("pairs.txt")
+    code = np.array(code,dtype=np.int)
+    img = load_list_images(file1)
+    print(img.shape)
+    img = torch.from_numpy(img)
+    img2 = load_list_images(file2)
+    img2 = torch.from_numpy(img2)
+    batch_size = 128
+    now = 0
+    number = len(code)
+    distance = []
+    with torch.no_grad():
+        start_time = time.time()
+        while now < number:
+            if now + batch_size < number:
+                pred1 = model(img[now:now + batch_size])
+                pred2 = model(img2[now:now + batch_size])
+            else:
+                pred1 = model(img[now:])
+                pred2 = model(img2[now:])
+            now = now + batch_size
+            pred1 = pred1.numpy()
+            pred2 = pred2.numpy()
+            print("batch"+str(now))
+            for i in range(len(pred1)):
+                #distance.append(findEuclideanDistance(pred1[i], pred2[i]))
+                distance.append(findEuclideanDistance(l2_normalize(pred1[i]), l2_normalize(pred2[i])))
+        end_time = time.time()
+    distance = np.array(distance)
+    print("对比两张人脸平均用时：" + str((end_time - start_time) / number))
+    #print(distance)
+    # threshold,max_acc = threshold_acc(distance,code)
+    # print(threshold,max_acc)
+    reject, error,right = reject_and_error(distance, code)
+    print("正确人脸对数量："+str(right)+" 正确率："+str(right/number))
+    print("拒识人脸对数量：" + str(reject) + " 拒识率：" + str(2*reject / number))
+    print("误识人脸对数量：" + str(error) + " 误识率：" + str(2 * error / number))
+
+# sublist = ['zdz','bbb','aaa']
+# for item in itertools.combinations(sublist, 2):
+#     for name in item:
+#         print(name)
+# path = "D:\Download\out\output2\Alan_Trammell\Alan_Trammell_0001.jpg"
+# path = path.split("\\")
+# print(path)
+# print(os.path.join(path[0],path[1],path[2],path[3],path[4]))
diff --git a/test2.py b/test2.py
new file mode 100644
index 0000000..3dab2a3
--- /dev/null
+++ b/test2.py
@@ -0,0 +1,52 @@
+# import numpy as np
+# a = [1,2,3,4,5,6,7,8,9,10]
+# print(a[0::2])
+# print(a[1::2])
+#
+# src = np.array([
+#             [30.2946, 51.6963],
+#             [65.5318, 51.5014],
+#             [48.0252, 71.7366],
+#             [33.5493, 92.3655],
+#             [62.7299, 92.2041]], dtype=np.float32)
+# src[:, 0] += 8.0
+# #print(src[:, 1])
+# input_blob = np.zeros((2, 3, 2, 2), dtype=np.uint8)
+# #print(input_blob[0])
+
+import numpy as np
+d = 512                           # dimension
+nb = 23800                   # database size
+nq = 1                      # nb of queries
+np.random.seed(1234)             # make reproducible
+xb = np.random.random((nb, d)).astype('float32')
+#print(xb)
+#print(np.arange(nb))
+#print(xb[:, 0])
+
+# xb[:, 0] += np.arange(nb) / 1000.
+# #print(xb)
+#xq = np.random.random((nq, d)).astype('float32')
+#xq[:, 0] += np.arange(nq) / 1000.
+import faiss   # make faiss available
+nlist = 100
+#quantizer = faiss.IndexFlatL2(d)  # the other index
+#index = faiss.IndexIVFFlat(quantizer, d, nlist, faiss.METRIC_L2)
+#index.train(xb)
+index = faiss.IndexFlatL2(d)   # build the index
+print(index.is_trained)
+#print(xq.shape)
+index.add(xb)                  # add vectors to the index
+print(index.ntotal)
+#index.nprobe=30
+k = 1                          # we want to see 4 nearest neighbors
+import time
+for i in range(10):
+    xq = np.random.random((nq, d)).astype('float32')
+    start_time = time.time()
+    D, I = index.search(xq, k)     # actual search
+    end_time = time.time()
+    print("faiss cost %fs"%(end_time - start_time))
+    print(D,I)
+
+
diff --git a/test_CPU.py b/test_CPU.py
new file mode 100644
index 0000000..a78cd7e
--- /dev/null
+++ b/test_CPU.py
@@ -0,0 +1,140 @@
+import torch
+import cv2
+import numpy as np
+from backbones import iresnet50,iresnet18,iresnet100
+
+# print(torch.cuda.is_available())
+def load_image(img_path):
+    img = cv2.imread(img_path)
+    #print(img)
+    #img2 = np.fliplr(img)
+    #print(img2.shape)
+    img = img.transpose((2, 0, 1))
+    #img2 = img2.transpose((2, 0, 1))
+    img = img[np.newaxis, :, :, :]
+    #img2 = img2[np.newaxis, :, :, :]
+    #img = np.concatenate((img,img2))
+    img = np.array(img, dtype=np.float32)
+    img -= 127.5
+    img /= 127.5
+    return img
+
+def findEuclideanDistance(source_representation, test_representation):
+    euclidean_distance = source_representation - test_representation
+    euclidean_distance = np.sum(np.multiply(euclidean_distance, euclidean_distance))
+    euclidean_distance = np.sqrt(euclidean_distance)
+    return euclidean_distance
+
+def findCosineDistance(source_representation, test_representation):
+    a = np.matmul(np.transpose(source_representation), test_representation)
+    b = np.sum(np.multiply(source_representation, source_representation))
+    c = np.sum(np.multiply(test_representation, test_representation))
+    return 1 - (a / (np.sqrt(b) * np.sqrt(c)))
+
+def l2_normalize(x):
+    return x / np.sqrt(np.sum(np.multiply(x, x))-1)
+
+def cosin_metric(x1, x2):
+    return np.dot(x1, x2) / (np.linalg.norm(x1) * np.linalg.norm(x2))
+
+def get_file_list(file_name):
+    file1 = []
+    file2 = []
+    code = []
+    with open(file_name,"r") as f:
+        file_list = f.readlines()
+    for i in file_list:
+        data = i.replace("\n","").split(" ")
+        file1.append(data[0])
+        file2.append(data[1])
+        code.append(data[2])
+    return file1,file2,code
+
+def load_list_images(file_list):
+    all = np.zeros((len(file_list),3,112,112),dtype=np.float32)
+    count = 0
+    for img_path in file_list:
+        img = cv2.imread(img_path)
+        img = img.transpose((2, 0, 1))
+        img = img[np.newaxis, :, :, :]
+        all[count] = img
+        count = count + 1
+    all -= 127.5
+    all /= 127.5
+    return all
+
+def threshold_acc(distance,code):
+    thresholds = np.arange(0.5,2,0.1)
+    total = len(code)
+    max_acc = 0
+    threshold = 0
+    code_net = []
+    for i in thresholds:
+        for ds in distance:
+            if ds < i:
+                code_net.append(1)
+            else:
+                code_net.append(0)
+        right = np.sum(code==code_net)
+        #print(threshold, max_acc)
+        if right/total > max_acc:
+            max_acc = right / total
+            threshold = i
+        code_net = []
+    return threshold,max_acc
+
+# if __name__=='__main__':
+#     img1 = load_image("D:\Download\out\output\Aaron_Peirsol\Aaron_Peirsol_0001.jpg")
+#     img1 = torch.from_numpy(img1)
+#     print(img1.shape)
+#     #print(img1)
+#     img2 = load_image("D:\Download\out\output\Aaron_Peirsol\Aaron_Peirsol_0002.jpg")
+#     img2 = torch.from_numpy(img2)
+#     model = iresnet100()
+#     model.load_state_dict(torch.load("./model/backbone100.pth",map_location="cpu"))
+#     #print(model)
+#     model.eval()
+#     with torch.no_grad():
+#         pred1 = model(img1)
+#         print(pred1.shape)
+#         #print(pred1)
+#         pred2 = model(img2)
+#         pred1 = pred1.numpy()
+#         print(pred1.shape)
+#         pred2 = pred2.numpy()
+#         print("EuclideanDistance is :"+str(findEuclideanDistance(l2_normalize(pred1),l2_normalize(pred2))))
+#         print(findCosineDistance(pred1[0],pred2[0]))
+
+if __name__=='__main__':
+    model = iresnet100()
+    model.load_state_dict(torch.load("./model/backbone100.pth", map_location="cpu"))
+    model.eval()
+    file1, file2, code = get_file_list("pairs.txt")
+    code = np.array(code,dtype=np.int)
+    img = load_list_images(file1)
+    print(img.shape)
+    img = torch.from_numpy(img)
+    img2 = load_list_images(file2)
+    img2 = torch.from_numpy(img2)
+    batch_size = 64
+    now = 0
+    number = len(code)
+    distance = []
+    with torch.no_grad():
+        while now < number:
+            if now + batch_size < number:
+                pred1 = model(img[now:now + batch_size])
+                pred2 = model(img2[now:now + batch_size])
+            else:
+                pred1 = model(img[now:])
+                pred2 = model(img2[now:])
+            now = now + batch_size
+            pred1 = pred1.numpy()
+            pred2 = pred2.numpy()
+            print(pred1.shape)
+            for i in range(len(pred1)):
+                distance.append(findEuclideanDistance(l2_normalize(pred1[i]), l2_normalize(pred2[i])))
+    distance = np.array(distance)
+    print(distance[:100])
+    threshold,max_acc = threshold_acc(distance,code)
+    print(threshold,max_acc)
diff --git a/train.py b/train.py
new file mode 100644
index 0000000..220c231
--- /dev/null
+++ b/train.py
@@ -0,0 +1,133 @@
+import argparse
+import logging
+import os
+import time
+
+import torch
+import torch.distributed as dist
+import torch.nn.functional as F
+import torch.utils.data.distributed
+from torch.nn.utils import clip_grad_norm_
+
+import backbones
+import losses
+from config import config as cfg
+from dataset import MXFaceDataset, DataLoaderX
+from partial_fc import PartialFC
+from utils.utils_callbacks import CallBackVerification, CallBackLogging, CallBackModelCheckpoint
+from utils.utils_logging import AverageMeter, init_logging
+from utils.utils_amp import MaxClipGradScaler
+
+torch.backends.cudnn.benchmark = True
+
+
+def main(args):
+
+    world_size = int(os.environ['WORLD_SIZE'])
+    rank = int(os.environ['RANK'])
+    dist_url = "tcp://{}:{}".format(os.environ["MASTER_ADDR"], os.environ["MASTER_PORT"])
+    dist.init_process_group(backend='nccl', init_method=dist_url, rank=rank, world_size=world_size)
+    local_rank = args.local_rank
+    torch.cuda.set_device(local_rank)
+
+    if not os.path.exists(cfg.output) and rank is 0:
+        os.makedirs(cfg.output)
+    else:
+        time.sleep(2)
+
+    log_root = logging.getLogger()
+    init_logging(log_root, rank, cfg.output)
+    trainset = MXFaceDataset(root_dir=cfg.rec, local_rank=local_rank)
+    train_sampler = torch.utils.data.distributed.DistributedSampler(
+        trainset, shuffle=True)
+    train_loader = DataLoaderX(
+        local_rank=local_rank, dataset=trainset, batch_size=cfg.batch_size,
+        sampler=train_sampler, num_workers=0, pin_memory=True, drop_last=True)
+
+    dropout = 0.4 if cfg.dataset is "webface" else 0
+    backbone = eval("backbones.{}".format(args.network))(False, dropout=dropout, fp16=cfg.fp16).to(local_rank)
+
+    if args.resume:
+        try:
+            backbone_pth = os.path.join(cfg.output, "backbone.pth")
+            backbone.load_state_dict(torch.load(backbone_pth, map_location=torch.device(local_rank)))
+            if rank is 0:
+                logging.info("backbone resume successfully!")
+        except (FileNotFoundError, KeyError, IndexError, RuntimeError):
+            logging.info("resume fail, backbone init successfully!")
+
+    for ps in backbone.parameters():
+        dist.broadcast(ps, 0)
+    backbone = torch.nn.parallel.DistributedDataParallel(
+        module=backbone, broadcast_buffers=False, device_ids=[local_rank])
+    backbone.train()
+
+    margin_softmax = eval("losses.{}".format(args.loss))()
+    module_partial_fc = PartialFC(
+        rank=rank, local_rank=local_rank, world_size=world_size, resume=args.resume,
+        batch_size=cfg.batch_size, margin_softmax=margin_softmax, num_classes=cfg.num_classes,
+        sample_rate=cfg.sample_rate, embedding_size=cfg.embedding_size, prefix=cfg.output)
+
+    opt_backbone = torch.optim.SGD(
+        params=[{'params': backbone.parameters()}],
+        lr=cfg.lr / 512 * cfg.batch_size * world_size,
+        momentum=0.9, weight_decay=cfg.weight_decay)
+    opt_pfc = torch.optim.SGD(
+        params=[{'params': module_partial_fc.parameters()}],
+        lr=cfg.lr / 512 * cfg.batch_size * world_size,
+        momentum=0.9, weight_decay=cfg.weight_decay)
+
+    scheduler_backbone = torch.optim.lr_scheduler.LambdaLR(
+        optimizer=opt_backbone, lr_lambda=cfg.lr_func)
+    scheduler_pfc = torch.optim.lr_scheduler.LambdaLR(
+        optimizer=opt_pfc, lr_lambda=cfg.lr_func)
+
+    start_epoch = 0
+    total_step = int(len(trainset) / cfg.batch_size / world_size * cfg.num_epoch)
+    if rank is 0: logging.info("Total Step is: %d" % total_step)
+
+    callback_verification = CallBackVerification(2000, rank, cfg.val_targets, cfg.rec)
+    callback_logging = CallBackLogging(50, rank, total_step, cfg.batch_size, world_size, None)
+    callback_checkpoint = CallBackModelCheckpoint(rank, cfg.output)
+
+    loss = AverageMeter()
+    global_step = 0
+    grad_scaler = MaxClipGradScaler(cfg.batch_size, 128 * cfg.batch_size, growth_interval=100) if cfg.fp16 else None
+    for epoch in range(start_epoch, cfg.num_epoch):
+        train_sampler.set_epoch(epoch)
+        for step, (img, label) in enumerate(train_loader):
+            global_step += 1
+            features = F.normalize(backbone(img))
+            x_grad, loss_v = module_partial_fc.forward_backward(label, features, opt_pfc)
+            if cfg.fp16:
+                features.backward(grad_scaler.scale(x_grad))
+                grad_scaler.unscale_(opt_backbone)
+                clip_grad_norm_(backbone.parameters(), max_norm=5, norm_type=2)
+                grad_scaler.step(opt_backbone)
+                grad_scaler.update()
+            else:
+                features.backward(x_grad)
+                clip_grad_norm_(backbone.parameters(), max_norm=5, norm_type=2)
+                opt_backbone.step()
+
+            opt_pfc.step()
+            module_partial_fc.update()
+            opt_backbone.zero_grad()
+            opt_pfc.zero_grad()
+            loss.update(loss_v, 1)
+            callback_logging(global_step, loss, epoch, cfg.fp16, grad_scaler)
+            callback_verification(global_step, backbone)
+        callback_checkpoint(global_step, backbone, module_partial_fc)
+        scheduler_backbone.step()
+        scheduler_pfc.step()
+    dist.destroy_process_group()
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser(description='PyTorch ArcFace Training')
+    parser.add_argument('--local_rank', type=int, default=0, help='local_rank')
+    parser.add_argument('--network', type=str, default='iresnet50', help='backbone network')
+    parser.add_argument('--loss', type=str, default='ArcFace', help='loss function')
+    parser.add_argument('--resume', type=int, default=0, help='model resuming')
+    args_ = parser.parse_args()
+    main(args_)
diff --git a/utils/__init__.py b/utils/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/utils/__pycache__/__init__.cpython-38.pyc b/utils/__pycache__/__init__.cpython-38.pyc
new file mode 100644
index 0000000..39c856b
Binary files /dev/null and b/utils/__pycache__/__init__.cpython-38.pyc differ
diff --git a/utils/__pycache__/box_utils.cpython-38.pyc b/utils/__pycache__/box_utils.cpython-38.pyc
new file mode 100644
index 0000000..0331a96
Binary files /dev/null and b/utils/__pycache__/box_utils.cpython-38.pyc differ
diff --git a/utils/box_utils.py b/utils/box_utils.py
new file mode 100644
index 0000000..c1d12bc
--- /dev/null
+++ b/utils/box_utils.py
@@ -0,0 +1,330 @@
+import torch
+import numpy as np
+
+
+def point_form(boxes):
+    """ Convert prior_boxes to (xmin, ymin, xmax, ymax)
+    representation for comparison to point form ground truth data.
+    Args:
+        boxes: (tensor) center-size default boxes from priorbox layers.
+    Return:
+        boxes: (tensor) Converted xmin, ymin, xmax, ymax form of boxes.
+    """
+    return torch.cat((boxes[:, :2] - boxes[:, 2:]/2,     # xmin, ymin
+                     boxes[:, :2] + boxes[:, 2:]/2), 1)  # xmax, ymax
+
+
+def center_size(boxes):
+    """ Convert prior_boxes to (cx, cy, w, h)
+    representation for comparison to center-size form ground truth data.
+    Args:
+        boxes: (tensor) point_form boxes
+    Return:
+        boxes: (tensor) Converted xmin, ymin, xmax, ymax form of boxes.
+    """
+    return torch.cat((boxes[:, 2:] + boxes[:, :2])/2,  # cx, cy
+                     boxes[:, 2:] - boxes[:, :2], 1)  # w, h
+
+
+def intersect(box_a, box_b):
+    """ We resize both tensors to [A,B,2] without new malloc:
+    [A,2] -> [A,1,2] -> [A,B,2]
+    [B,2] -> [1,B,2] -> [A,B,2]
+    Then we compute the area of intersect between box_a and box_b.
+    Args:
+      box_a: (tensor) bounding boxes, Shape: [A,4].
+      box_b: (tensor) bounding boxes, Shape: [B,4].
+    Return:
+      (tensor) intersection area, Shape: [A,B].
+    """
+    A = box_a.size(0)
+    B = box_b.size(0)
+    max_xy = torch.min(box_a[:, 2:].unsqueeze(1).expand(A, B, 2),
+                       box_b[:, 2:].unsqueeze(0).expand(A, B, 2))
+    min_xy = torch.max(box_a[:, :2].unsqueeze(1).expand(A, B, 2),
+                       box_b[:, :2].unsqueeze(0).expand(A, B, 2))
+    inter = torch.clamp((max_xy - min_xy), min=0)
+    return inter[:, :, 0] * inter[:, :, 1]
+
+
+def jaccard(box_a, box_b):
+    """Compute the jaccard overlap of two sets of boxes.  The jaccard overlap
+    is simply the intersection over union of two boxes.  Here we operate on
+    ground truth boxes and default boxes.
+    E.g.:
+        A ∩ B / A ∪ B = A ∩ B / (area(A) + area(B) - A ∩ B)
+    Args:
+        box_a: (tensor) Ground truth bounding boxes, Shape: [num_objects,4]
+        box_b: (tensor) Prior boxes from priorbox layers, Shape: [num_priors,4]
+    Return:
+        jaccard overlap: (tensor) Shape: [box_a.size(0), box_b.size(0)]
+    """
+    inter = intersect(box_a, box_b)
+    area_a = ((box_a[:, 2]-box_a[:, 0]) *
+              (box_a[:, 3]-box_a[:, 1])).unsqueeze(1).expand_as(inter)  # [A,B]
+    area_b = ((box_b[:, 2]-box_b[:, 0]) *
+              (box_b[:, 3]-box_b[:, 1])).unsqueeze(0).expand_as(inter)  # [A,B]
+    union = area_a + area_b - inter
+    return inter / union  # [A,B]
+
+
+def matrix_iou(a, b):
+    """
+    return iou of a and b, numpy version for data augenmentation
+    """
+    lt = np.maximum(a[:, np.newaxis, :2], b[:, :2])
+    rb = np.minimum(a[:, np.newaxis, 2:], b[:, 2:])
+
+    area_i = np.prod(rb - lt, axis=2) * (lt < rb).all(axis=2)
+    area_a = np.prod(a[:, 2:] - a[:, :2], axis=1)
+    area_b = np.prod(b[:, 2:] - b[:, :2], axis=1)
+    return area_i / (area_a[:, np.newaxis] + area_b - area_i)
+
+
+def matrix_iof(a, b):
+    """
+    return iof of a and b, numpy version for data augenmentation
+    """
+    lt = np.maximum(a[:, np.newaxis, :2], b[:, :2])
+    rb = np.minimum(a[:, np.newaxis, 2:], b[:, 2:])
+
+    area_i = np.prod(rb - lt, axis=2) * (lt < rb).all(axis=2)
+    area_a = np.prod(a[:, 2:] - a[:, :2], axis=1)
+    return area_i / np.maximum(area_a[:, np.newaxis], 1)
+
+
+def match(threshold, truths, priors, variances, labels, landms, loc_t, conf_t, landm_t, idx):
+    """Match each prior box with the ground truth box of the highest jaccard
+    overlap, encode the bounding boxes, then return the matched indices
+    corresponding to both confidence and location preds.
+    Args:
+        threshold: (float) The overlap threshold used when mathing boxes.
+        truths: (tensor) Ground truth boxes, Shape: [num_obj, 4].
+        priors: (tensor) Prior boxes from priorbox layers, Shape: [n_priors,4].
+        variances: (tensor) Variances corresponding to each prior coord,
+            Shape: [num_priors, 4].
+        labels: (tensor) All the class labels for the image, Shape: [num_obj].
+        landms: (tensor) Ground truth landms, Shape [num_obj, 10].
+        loc_t: (tensor) Tensor to be filled w/ endcoded location targets.
+        conf_t: (tensor) Tensor to be filled w/ matched indices for conf preds.
+        landm_t: (tensor) Tensor to be filled w/ endcoded landm targets.
+        idx: (int) current batch index
+    Return:
+        The matched indices corresponding to 1)location 2)confidence 3)landm preds.
+    """
+    # jaccard index
+    overlaps = jaccard(
+        truths,
+        point_form(priors)
+    )
+    # (Bipartite Matching)
+    # [1,num_objects] best prior for each ground truth
+    best_prior_overlap, best_prior_idx = overlaps.max(1, keepdim=True)
+
+    # ignore hard gt
+    valid_gt_idx = best_prior_overlap[:, 0] >= 0.2
+    best_prior_idx_filter = best_prior_idx[valid_gt_idx, :]
+    if best_prior_idx_filter.shape[0] <= 0:
+        loc_t[idx] = 0
+        conf_t[idx] = 0
+        return
+
+    # [1,num_priors] best ground truth for each prior
+    best_truth_overlap, best_truth_idx = overlaps.max(0, keepdim=True)
+    best_truth_idx.squeeze_(0)
+    best_truth_overlap.squeeze_(0)
+    best_prior_idx.squeeze_(1)
+    best_prior_idx_filter.squeeze_(1)
+    best_prior_overlap.squeeze_(1)
+    best_truth_overlap.index_fill_(0, best_prior_idx_filter, 2)  # ensure best prior
+    # TODO refactor: index  best_prior_idx with long tensor
+    # ensure every gt matches with its prior of max overlap
+    for j in range(best_prior_idx.size(0)):     # 判别此anchor是预测哪一个boxes
+        best_truth_idx[best_prior_idx[j]] = j
+    matches = truths[best_truth_idx]            # Shape: [num_priors,4] 此处为每一个anchor对应的bbox取出来
+    conf = labels[best_truth_idx]               # Shape: [num_priors]      此处为每一个anchor对应的label取出来
+    conf[best_truth_overlap < threshold] = 0    # label as background   overlap<0.35的全部作为负样本
+    loc = encode(matches, priors, variances)
+
+    matches_landm = landms[best_truth_idx]
+    landm = encode_landm(matches_landm, priors, variances)
+    loc_t[idx] = loc    # [num_priors,4] encoded offsets to learn
+    conf_t[idx] = conf  # [num_priors] top class label for each prior
+    landm_t[idx] = landm
+
+
+def encode(matched, priors, variances):
+    """Encode the variances from the priorbox layers into the ground truth boxes
+    we have matched (based on jaccard overlap) with the prior boxes.
+    Args:
+        matched: (tensor) Coords of ground truth for each prior in point-form
+            Shape: [num_priors, 4].
+        priors: (tensor) Prior boxes in center-offset form
+            Shape: [num_priors,4].
+        variances: (list[float]) Variances of priorboxes
+    Return:
+        encoded boxes (tensor), Shape: [num_priors, 4]
+    """
+
+    # dist b/t match center and prior's center
+    g_cxcy = (matched[:, :2] + matched[:, 2:])/2 - priors[:, :2]
+    # encode variance
+    g_cxcy /= (variances[0] * priors[:, 2:])
+    # match wh / prior wh
+    g_wh = (matched[:, 2:] - matched[:, :2]) / priors[:, 2:]
+    g_wh = torch.log(g_wh) / variances[1]
+    # return target for smooth_l1_loss
+    return torch.cat([g_cxcy, g_wh], 1)  # [num_priors,4]
+
+def encode_landm(matched, priors, variances):
+    """Encode the variances from the priorbox layers into the ground truth boxes
+    we have matched (based on jaccard overlap) with the prior boxes.
+    Args:
+        matched: (tensor) Coords of ground truth for each prior in point-form
+            Shape: [num_priors, 10].
+        priors: (tensor) Prior boxes in center-offset form
+            Shape: [num_priors,4].
+        variances: (list[float]) Variances of priorboxes
+    Return:
+        encoded landm (tensor), Shape: [num_priors, 10]
+    """
+
+    # dist b/t match center and prior's center
+    matched = torch.reshape(matched, (matched.size(0), 5, 2))
+    priors_cx = priors[:, 0].unsqueeze(1).expand(matched.size(0), 5).unsqueeze(2)
+    priors_cy = priors[:, 1].unsqueeze(1).expand(matched.size(0), 5).unsqueeze(2)
+    priors_w = priors[:, 2].unsqueeze(1).expand(matched.size(0), 5).unsqueeze(2)
+    priors_h = priors[:, 3].unsqueeze(1).expand(matched.size(0), 5).unsqueeze(2)
+    priors = torch.cat([priors_cx, priors_cy, priors_w, priors_h], dim=2)
+    g_cxcy = matched[:, :, :2] - priors[:, :, :2]
+    # encode variance
+    g_cxcy /= (variances[0] * priors[:, :, 2:])
+    # g_cxcy /= priors[:, :, 2:]
+    g_cxcy = g_cxcy.reshape(g_cxcy.size(0), -1)
+    # return target for smooth_l1_loss
+    return g_cxcy
+
+
+# Adapted from https://github.com/Hakuyume/chainer-ssd
+def decode(loc, priors, variances):
+    """Decode locations from predictions using priors to undo
+    the encoding we did for offset regression at train time.
+    Args:
+        loc (tensor): location predictions for loc layers,
+            Shape: [num_priors,4]
+        priors (tensor): Prior boxes in center-offset form.
+            Shape: [num_priors,4].
+        variances: (list[float]) Variances of priorboxes
+    Return:
+        decoded bounding box predictions
+    """
+
+    boxes = torch.cat((
+        priors[:, :2] + loc[:, :2] * variances[0] * priors[:, 2:],
+        priors[:, 2:] * torch.exp(loc[:, 2:] * variances[1])), 1)
+    boxes[:, :2] -= boxes[:, 2:] / 2
+    boxes[:, 2:] += boxes[:, :2]
+    return boxes
+
+def decode_landm(pre, priors, variances):
+    """Decode landm from predictions using priors to undo
+    the encoding we did for offset regression at train time.
+    Args:
+        pre (tensor): landm predictions for loc layers,
+            Shape: [num_priors,10]
+        priors (tensor): Prior boxes in center-offset form.
+            Shape: [num_priors,4].
+        variances: (list[float]) Variances of priorboxes
+    Return:
+        decoded landm predictions
+    """
+    landms = torch.cat((priors[:, :2] + pre[:, :2] * variances[0] * priors[:, 2:],
+                        priors[:, :2] + pre[:, 2:4] * variances[0] * priors[:, 2:],
+                        priors[:, :2] + pre[:, 4:6] * variances[0] * priors[:, 2:],
+                        priors[:, :2] + pre[:, 6:8] * variances[0] * priors[:, 2:],
+                        priors[:, :2] + pre[:, 8:10] * variances[0] * priors[:, 2:],
+                        ), dim=1)
+    return landms
+
+
+def log_sum_exp(x):
+    """Utility function for computing log_sum_exp while determining
+    This will be used to determine unaveraged confidence loss across
+    all examples in a batch.
+    Args:
+        x (Variable(tensor)): conf_preds from conf layers
+    """
+    x_max = x.data.max()
+    return torch.log(torch.sum(torch.exp(x-x_max), 1, keepdim=True)) + x_max
+
+
+# Original author: Francisco Massa:
+# https://github.com/fmassa/object-detection.torch
+# Ported to PyTorch by Max deGroot (02/01/2017)
+def nms(boxes, scores, overlap=0.5, top_k=200):
+    """Apply non-maximum suppression at test time to avoid detecting too many
+    overlapping bounding boxes for a given object.
+    Args:
+        boxes: (tensor) The location preds for the img, Shape: [num_priors,4].
+        scores: (tensor) The class predscores for the img, Shape:[num_priors].
+        overlap: (float) The overlap thresh for suppressing unnecessary boxes.
+        top_k: (int) The Maximum number of box preds to consider.
+    Return:
+        The indices of the kept boxes with respect to num_priors.
+    """
+
+    keep = torch.Tensor(scores.size(0)).fill_(0).long()
+    if boxes.numel() == 0:
+        return keep
+    x1 = boxes[:, 0]
+    y1 = boxes[:, 1]
+    x2 = boxes[:, 2]
+    y2 = boxes[:, 3]
+    area = torch.mul(x2 - x1, y2 - y1)
+    v, idx = scores.sort(0)  # sort in ascending order
+    # I = I[v >= 0.01]
+    idx = idx[-top_k:]  # indices of the top-k largest vals
+    xx1 = boxes.new()
+    yy1 = boxes.new()
+    xx2 = boxes.new()
+    yy2 = boxes.new()
+    w = boxes.new()
+    h = boxes.new()
+
+    # keep = torch.Tensor()
+    count = 0
+    while idx.numel() > 0:
+        i = idx[-1]  # index of current largest val
+        # keep.append(i)
+        keep[count] = i
+        count += 1
+        if idx.size(0) == 1:
+            break
+        idx = idx[:-1]  # remove kept element from view
+        # load bboxes of next highest vals
+        torch.index_select(x1, 0, idx, out=xx1)
+        torch.index_select(y1, 0, idx, out=yy1)
+        torch.index_select(x2, 0, idx, out=xx2)
+        torch.index_select(y2, 0, idx, out=yy2)
+        # store element-wise max with next highest score
+        xx1 = torch.clamp(xx1, min=x1[i])
+        yy1 = torch.clamp(yy1, min=y1[i])
+        xx2 = torch.clamp(xx2, max=x2[i])
+        yy2 = torch.clamp(yy2, max=y2[i])
+        w.resize_as_(xx2)
+        h.resize_as_(yy2)
+        w = xx2 - xx1
+        h = yy2 - yy1
+        # check sizes of xx1 and xx2.. after each iteration
+        w = torch.clamp(w, min=0.0)
+        h = torch.clamp(h, min=0.0)
+        inter = w*h
+        # IoU = i / (area(a) + area(b) - i)
+        rem_areas = torch.index_select(area, 0, idx)  # load remaining areas)
+        union = (rem_areas - inter) + area[i]
+        IoU = inter/union  # store result in iou
+        # keep only elements with an IoU <= overlap
+        idx = idx[IoU.le(overlap)]
+    return keep, count
+
+
diff --git a/utils/nms/__init__.py b/utils/nms/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/utils/nms/__pycache__/__init__.cpython-38.pyc b/utils/nms/__pycache__/__init__.cpython-38.pyc
new file mode 100644
index 0000000..b730b73
Binary files /dev/null and b/utils/nms/__pycache__/__init__.cpython-38.pyc differ
diff --git a/utils/nms/__pycache__/py_cpu_nms.cpython-38.pyc b/utils/nms/__pycache__/py_cpu_nms.cpython-38.pyc
new file mode 100644
index 0000000..1831740
Binary files /dev/null and b/utils/nms/__pycache__/py_cpu_nms.cpython-38.pyc differ
diff --git a/utils/nms/py_cpu_nms.py b/utils/nms/py_cpu_nms.py
new file mode 100644
index 0000000..54e7b25
--- /dev/null
+++ b/utils/nms/py_cpu_nms.py
@@ -0,0 +1,38 @@
+# --------------------------------------------------------
+# Fast R-CNN
+# Copyright (c) 2015 Microsoft
+# Licensed under The MIT License [see LICENSE for details]
+# Written by Ross Girshick
+# --------------------------------------------------------
+
+import numpy as np
+
+def py_cpu_nms(dets, thresh):
+    """Pure Python NMS baseline."""
+    x1 = dets[:, 0]
+    y1 = dets[:, 1]
+    x2 = dets[:, 2]
+    y2 = dets[:, 3]
+    scores = dets[:, 4]
+
+    areas = (x2 - x1 + 1) * (y2 - y1 + 1)
+    order = scores.argsort()[::-1]
+
+    keep = []
+    while order.size > 0:
+        i = order[0]
+        keep.append(i)
+        xx1 = np.maximum(x1[i], x1[order[1:]])
+        yy1 = np.maximum(y1[i], y1[order[1:]])
+        xx2 = np.minimum(x2[i], x2[order[1:]])
+        yy2 = np.minimum(y2[i], y2[order[1:]])
+
+        w = np.maximum(0.0, xx2 - xx1 + 1)
+        h = np.maximum(0.0, yy2 - yy1 + 1)
+        inter = w * h
+        ovr = inter / (areas[i] + areas[order[1:]] - inter)
+
+        inds = np.where(ovr <= thresh)[0]
+        order = order[inds + 1]
+
+    return keep
diff --git a/utils/plot.py b/utils/plot.py
new file mode 100644
index 0000000..ccc588e
--- /dev/null
+++ b/utils/plot.py
@@ -0,0 +1,72 @@
+# coding: utf-8
+
+import os
+from pathlib import Path
+
+import matplotlib.pyplot as plt
+import numpy as np
+import pandas as pd
+from menpo.visualize.viewmatplotlib import sample_colours_from_colourmap
+from prettytable import PrettyTable
+from sklearn.metrics import roc_curve, auc
+
+image_path = "/data/anxiang/IJB_release/IJBC"
+files = [
+        "./ms1mv3_arcface_r100/ms1mv3_arcface_r100/ijbc.npy"
+]
+
+
+def read_template_pair_list(path):
+    pairs = pd.read_csv(path, sep=' ', header=None).values
+    t1 = pairs[:, 0].astype(np.int)
+    t2 = pairs[:, 1].astype(np.int)
+    label = pairs[:, 2].astype(np.int)
+    return t1, t2, label
+
+
+p1, p2, label = read_template_pair_list(
+    os.path.join('%s/meta' % image_path,
+                 '%s_template_pair_label.txt' % 'ijbc'))
+
+methods = []
+scores = []
+for file in files:
+    methods.append(file.split('/')[-2])
+    scores.append(np.load(file))
+
+methods = np.array(methods)
+scores = dict(zip(methods, scores))
+colours = dict(
+    zip(methods, sample_colours_from_colourmap(methods.shape[0], 'Set2')))
+x_labels = [10 ** -6, 10 ** -5, 10 ** -4, 10 ** -3, 10 ** -2, 10 ** -1]
+tpr_fpr_table = PrettyTable(['Methods'] + [str(x) for x in x_labels])
+fig = plt.figure()
+for method in methods:
+    fpr, tpr, _ = roc_curve(label, scores[method])
+    roc_auc = auc(fpr, tpr)
+    fpr = np.flipud(fpr)
+    tpr = np.flipud(tpr)  # select largest tpr at same fpr
+    plt.plot(fpr,
+             tpr,
+             color=colours[method],
+             lw=1,
+             label=('[%s (AUC = %0.4f %%)]' %
+                    (method.split('-')[-1], roc_auc * 100)))
+    tpr_fpr_row = []
+    tpr_fpr_row.append("%s-%s" % (method, "IJBC"))
+    for fpr_iter in np.arange(len(x_labels)):
+        _, min_index = min(
+            list(zip(abs(fpr - x_labels[fpr_iter]), range(len(fpr)))))
+        tpr_fpr_row.append('%.2f' % (tpr[min_index] * 100))
+    tpr_fpr_table.add_row(tpr_fpr_row)
+plt.xlim([10 ** -6, 0.1])
+plt.ylim([0.3, 1.0])
+plt.grid(linestyle='--', linewidth=1)
+plt.xticks(x_labels)
+plt.yticks(np.linspace(0.3, 1.0, 8, endpoint=True))
+plt.xscale('log')
+plt.xlabel('False Positive Rate')
+plt.ylabel('True Positive Rate')
+plt.title('ROC on IJB')
+plt.legend(loc="lower right")
+print(tpr_fpr_table)
diff --git a/utils/timer.py b/utils/timer.py
new file mode 100644
index 0000000..e4b3b80
--- /dev/null
+++ b/utils/timer.py
@@ -0,0 +1,40 @@
+# --------------------------------------------------------
+# Fast R-CNN
+# Copyright (c) 2015 Microsoft
+# Licensed under The MIT License [see LICENSE for details]
+# Written by Ross Girshick
+# --------------------------------------------------------
+
+import time
+
+
+class Timer(object):
+    """A simple timer."""
+    def __init__(self):
+        self.total_time = 0.
+        self.calls = 0
+        self.start_time = 0.
+        self.diff = 0.
+        self.average_time = 0.
+
+    def tic(self):
+        # using time.time instead of time.clock because time time.clock
+        # does not normalize for multithreading
+        self.start_time = time.time()
+
+    def toc(self, average=True):
+        self.diff = time.time() - self.start_time
+        self.total_time += self.diff
+        self.calls += 1
+        self.average_time = self.total_time / self.calls
+        if average:
+            return self.average_time
+        else:
+            return self.diff
+
+    def clear(self):
+        self.total_time = 0.
+        self.calls = 0
+        self.start_time = 0.
+        self.diff = 0.
+        self.average_time = 0.
diff --git a/utils/utils_amp.py b/utils/utils_amp.py
new file mode 100644
index 0000000..c60067e
--- /dev/null
+++ b/utils/utils_amp.py
@@ -0,0 +1,81 @@
+from typing import Dict, List
+
+import torch
+from torch._six import container_abcs
+from torch.cuda.amp import GradScaler
+
+
+class _MultiDeviceReplicator(object):
+    """
+    Lazily serves copies of a tensor to requested devices.  Copies are cached per-device.
+    """
+
+    def __init__(self, master_tensor: torch.Tensor) -> None:
+        assert master_tensor.is_cuda
+        self.master = master_tensor
+        self._per_device_tensors: Dict[torch.device, torch.Tensor] = {}
+
+    def get(self, device) -> torch.Tensor:
+        retval = self._per_device_tensors.get(device, None)
+        if retval is None:
+            retval = self.master.to(device=device, non_blocking=True, copy=True)
+            self._per_device_tensors[device] = retval
+        return retval
+
+
+class MaxClipGradScaler(GradScaler):
+    def __init__(self, init_scale, max_scale: float, growth_interval=100):
+        GradScaler.__init__(self, init_scale=init_scale, growth_interval=growth_interval)
+        self.max_scale = max_scale
+
+    def scale_clip(self):
+        if self.get_scale() == self.max_scale:
+            self.set_growth_factor(1)
+        elif self.get_scale() < self.max_scale:
+            self.set_growth_factor(2)
+        elif self.get_scale() > self.max_scale:
+            self._scale.fill_(self.max_scale)
+            self.set_growth_factor(1)
+
+    def scale(self, outputs):
+        """
+        Multiplies ('scales') a tensor or list of tensors by the scale factor.
+
+        Returns scaled outputs.  If this instance of :class:`GradScaler` is not enabled, outputs are returned
+        unmodified.
+
+        Arguments:
+            outputs (Tensor or iterable of Tensors):  Outputs to scale.
+        """
+        if not self._enabled:
+            return outputs
+        self.scale_clip()
+        # Short-circuit for the common case.
+        if isinstance(outputs, torch.Tensor):
+            assert outputs.is_cuda
+            if self._scale is None:
+                self._lazy_init_scale_growth_tracker(outputs.device)
+            assert self._scale is not None
+            return outputs * self._scale.to(device=outputs.device, non_blocking=True)
+
+        # Invoke the more complex machinery only if we're treating multiple outputs.
+        stash: List[_MultiDeviceReplicator] = []  # holds a reference that can be overwritten by apply_scale
+
+        def apply_scale(val):
+            if isinstance(val, torch.Tensor):
+                assert val.is_cuda
+                if len(stash) == 0:
+                    if self._scale is None:
+                        self._lazy_init_scale_growth_tracker(val.device)
+                    assert self._scale is not None
+                    stash.append(_MultiDeviceReplicator(self._scale))
+                return val * stash[0].get(val.device)
+            elif isinstance(val, container_abcs.Iterable):
+                iterable = map(apply_scale, val)
+                if isinstance(val, list) or isinstance(val, tuple):
+                    return type(val)(iterable)
+                else:
+                    return iterable
+            else:
+                raise ValueError("outputs must be a Tensor or an iterable of Tensors")
+        return apply_scale(outputs)
diff --git a/utils/utils_callbacks.py b/utils/utils_callbacks.py
new file mode 100644
index 0000000..1114301
--- /dev/null
+++ b/utils/utils_callbacks.py
@@ -0,0 +1,106 @@
+import logging
+import os
+import time
+from typing import List
+
+import torch
+
+from eval import verification
+from utils.utils_logging import AverageMeter
+from partial_fc import PartialFC
+
+
+class CallBackVerification(object):
+    def __init__(self, frequent, rank, val_targets, rec_prefix, image_size=(112, 112)):
+        self.frequent: int = frequent
+        self.rank: int = rank
+        self.highest_acc: float = 0.0
+        self.highest_acc_list: List[float] = [0.0] * len(val_targets)
+        self.ver_list: List[object] = []
+        self.ver_name_list: List[str] = []
+        if self.rank is 0:
+            self.init_dataset(val_targets=val_targets, data_dir=rec_prefix, image_size=image_size)
+
+    def ver_test(self, backbone: torch.nn.Module, global_step: int):
+        results = []
+        for i in range(len(self.ver_list)):
+            acc1, std1, acc2, std2, xnorm, embeddings_list = verification.test(
+                self.ver_list[i], backbone, 10, 10)
+            logging.info('[%s][%d]XNorm: %f' % (self.ver_name_list[i], global_step, xnorm))
+            logging.info('[%s][%d]Accuracy-Flip: %1.5f+-%1.5f' % (self.ver_name_list[i], global_step, acc2, std2))
+            if acc2 > self.highest_acc_list[i]:
+                self.highest_acc_list[i] = acc2
+            logging.info(
+                '[%s][%d]Accuracy-Highest: %1.5f' % (self.ver_name_list[i], global_step, self.highest_acc_list[i]))
+            results.append(acc2)
+
+    def init_dataset(self, val_targets, data_dir, image_size):
+        for name in val_targets:
+            path = os.path.join(data_dir, name + ".bin")
+            if os.path.exists(path):
+                data_set = verification.load_bin(path, image_size)
+                self.ver_list.append(data_set)
+                self.ver_name_list.append(name)
+
+    def __call__(self, num_update, backbone: torch.nn.Module):
+        if self.rank is 0 and num_update > 0 and num_update % self.frequent == 0:
+            backbone.eval()
+            self.ver_test(backbone, num_update)
+            backbone.train()
+
+
+class CallBackLogging(object):
+    def __init__(self, frequent, rank, total_step, batch_size, world_size, writer=None):
+        self.frequent: int = frequent
+        self.rank: int = rank
+        self.time_start = time.time()
+        self.total_step: int = total_step
+        self.batch_size: int = batch_size
+        self.world_size: int = world_size
+        self.writer = writer
+
+        self.init = False
+        self.tic = 0
+
+    def __call__(self, global_step, loss: AverageMeter, epoch: int, fp16: bool, grad_scaler: torch.cuda.amp.GradScaler):
+        if self.rank is 0 and global_step > 0 and global_step % self.frequent == 0:
+            if self.init:
+                try:
+                    speed: float = self.frequent * self.batch_size / (time.time() - self.tic)
+                    speed_total = speed * self.world_size
+                except ZeroDivisionError:
+                    speed_total = float('inf')
+
+                time_now = (time.time() - self.time_start) / 3600
+                time_total = time_now / ((global_step + 1) / self.total_step)
+                time_for_end = time_total - time_now
+                if self.writer is not None:
+                    self.writer.add_scalar('time_for_end', time_for_end, global_step)
+                    self.writer.add_scalar('loss', loss.avg, global_step)
+                if fp16:
+                    msg = "Speed %.2f samples/sec   Loss %.4f   Epoch: %d   Global Step: %d   "\
+                          "Fp16 Grad Scale: %2.f   Required: %1.f hours" % (
+                        speed_total, loss.avg, epoch, global_step, grad_scaler.get_scale(), time_for_end
+                    )
+                else:
+                    msg = "Speed %.2f samples/sec   Loss %.4f   Epoch: %d   Global Step: %d   Required: %1.f hours" % (
+                        speed_total, loss.avg, epoch, global_step, time_for_end
+                    )
+                logging.info(msg)
+                loss.reset()
+                self.tic = time.time()
+            else:
+                self.init = True
+                self.tic = time.time()
+
+
+class CallBackModelCheckpoint(object):
+    def __init__(self, rank, output="./"):
+        self.rank: int = rank
+        self.output: str = output
+
+    def __call__(self, global_step, backbone: torch.nn.Module, partial_fc: PartialFC = None):
+        if global_step > 100 and self.rank is 0:
+            torch.save(backbone.module.state_dict(), os.path.join(self.output, "backbone.pth"))
+        if global_step > 100 and partial_fc is not None:
+            partial_fc.save_params()
diff --git a/utils/utils_logging.py b/utils/utils_logging.py
new file mode 100644
index 0000000..7d58012
--- /dev/null
+++ b/utils/utils_logging.py
@@ -0,0 +1,40 @@
+import logging
+import os
+import sys
+
+
+class AverageMeter(object):
+    """Computes and stores the average and current value
+    """
+
+    def __init__(self):
+        self.val = None
+        self.avg = None
+        self.sum = None
+        self.count = None
+        self.reset()
+
+    def reset(self):
+        self.val = 0
+        self.avg = 0
+        self.sum = 0
+        self.count = 0
+
+    def update(self, val, n=1):
+        self.val = val
+        self.sum += val * n
+        self.count += n
+        self.avg = self.sum / self.count
+
+
+def init_logging(log_root, rank, models_root):
+    if rank is 0:
+        log_root.setLevel(logging.INFO)
+        formatter = logging.Formatter("Training: %(asctime)s-%(message)s")
+        handler_file = logging.FileHandler(os.path.join(models_root, "training.log"))
+        handler_stream = logging.StreamHandler(sys.stdout)
+        handler_file.setFormatter(formatter)
+        handler_stream.setFormatter(formatter)
+        log_root.addHandler(handler_file)
+        log_root.addHandler(handler_stream)
+        log_root.info('rank_id: %d' % rank)
diff --git a/utils/utils_os.py b/utils/utils_os.py
new file mode 100644
index 0000000..e69de29
diff --git a/weights/Resnet50_Final.pth b/weights/Resnet50_Final.pth
new file mode 100644
index 0000000..3ac99e7
Binary files /dev/null and b/weights/Resnet50_Final.pth differ
diff --git a/weights/mobilenet0.25_Final.pth b/weights/mobilenet0.25_Final.pth
new file mode 100644
index 0000000..a02809f
Binary files /dev/null and b/weights/mobilenet0.25_Final.pth differ
diff --git a/weights/mobilenetV1X0.25_pretrain.tar b/weights/mobilenetV1X0.25_pretrain.tar
new file mode 100644
index 0000000..cb53f82
Binary files /dev/null and b/weights/mobilenetV1X0.25_pretrain.tar differ