
# ##################################################################################
# # 数据清洗
# # 作者: Xinyu Ou (http://ouxinyu.cn)
# # 数据集名称：十二生肖数据集Zodiac
# # 本程序功能:
# # 对图像坏样本，进行索引，并保存到bad.txt中，扫描文件夹时，自动跳过文件夹'.DS_Store'和'.ipynb_checkpoints'
# ###################################################################################

# import os
# import cv2
# import codecs

# # 本地运行时，需要修改数据集的名称和绝对路径，注意和文件夹名称一致
# dataset_name = 'Zodiac'
# dataset_path = 'D:\\Workspace\\ExpDatasets\\'
# dataset_root_path = os.path.join(dataset_path, dataset_name)
# excluded_folder = ['.DS_Store', '.ipynb_checkpoints']    # 被排除的文件夹
# class_prefix = ['train', 'valid', 'test']
# num_bad = 0
# num_good = 0
# num_folder = 0

# # 检测坏文件列表是否存在，如果存在则先删除。
# bad_list = os.path.join(dataset_root_path, 'bad.txt')
# if os.path.exists(bad_list):
#     os.remove(bad_list)

# # 执行数据清洗
# with codecs.open(bad_list, 'a', 'utf-8') as f_bad:
#     for prefix in class_prefix:
#         class_name_list = os.listdir(os.path.join(dataset_root_path, prefix))
#         for class_name in class_name_list:
#             if class_name not in excluded_folder:    # 跳过排除文件夹
#                 images = os.listdir(os.path.join(dataset_root_path, prefix, class_name))
#                 for image in images:
#                     if image not in excluded_folder: # 跳过排除文件夹                
#                         img_path = os.path.join(dataset_root_path, prefix, class_name, image)
#                         try: # 通过尝试读取并显示图像维度来判断样本是否损坏
#                             img = cv2.imread(img_path, 1)
#                             x = img.shape
#                             num_good += 1
#                             pass
#                         except:                                     
#                             bad_file = os.path.join(prefix, class_name, image)
#                             f_bad.write("{}\n".format(bad_file))
#                             num_bad += 1
#             num_folder += 1
#             print('\r 当前清洗进度:{}/{}'.format(num_folder, 3*len(class_name_list)), end='')
 
# print('数据集清洗完成, 损坏文件{}个, 正常文件{}.'.format(num_bad, num_good))


##################################################################################
# 数据集预处理
# 作者: Xinyu Ou (http://ouxinyu.cn)
# 数据集名称：十二生肖数据集Zodiac
# 数据集简介: 数据集包含12个类别，其中训练集样本7199个, 验证集样本650个, 测试集样本660个, 共计8509个，其中包含9个损坏的文件。
# 本程序功能:
# 1. 将数据集由官方进行划分，大体上训练集、验证集、测试集的比例为: 85:7.5:7.5
# 2. 代码将生成4个列表文件：训练集列表train.txt, 验证集列表val.txt, 测试集列表test.txt, 训练验证集trainval.txt
# 3. 数据集基本信息：数据集的基本信息使用json格式进行输出，包括数据库名称、数据样本的数量、类别数以及类别标签。
###################################################################################

import os
import cv2
import json
import codecs

# 初始化参数
num_trainval = 0
num_train = 0
num_val = 0
num_test = 0
class_dim = 0
dataset_info = {
    'dataset_name': '',
    'num_trainval': -1,
    'num_train': -1,
    'num_val': -1,
    'num_test': -1,
    'class_dim': -1,
    'label_dict': {}
}

# 本地运行时，需要修改数据集的名称和绝对路径，注意和文件夹名称一致
dataset_name = 'Zodiac'
dataset_path = 'D:\\Workspace\\ExpDatasets\\'
dataset_root_path = os.path.join(dataset_path, dataset_name)
class_prefix = ['train', 'valid', 'test']
excluded_folder = ['.DS_Store', '.ipynb_checkpoints']    # 被排除的文件或文件夹

# 定义生成文件的路径
# data_path = os.path.join(dataset_root_path,=) # 该数据的样本分别保存在train,valid和test文件夹中，因此不需要统一指定路径
trainval_list = os.path.join(dataset_root_path, 'trainval.txt')
train_list = os.path.join(dataset_root_path, 'train.txt')
val_list = os.path.join(dataset_root_path, 'val.txt')
test_list = os.path.join(dataset_root_path, 'test.txt')
dataset_info_list = os.path.join(dataset_root_path, 'dataset_info.json')

# 读取数据清洗获得的坏样本列表
bad_list = os.path.join(dataset_root_path, 'bad.txt')     
with codecs.open(bad_list, 'r', 'utf-8') as f_bad:
    bad_file = f_bad.read().splitlines()

# 检测数据集列表是否存在，如果存在则先删除。其中测试集列表是一次写入，因此可以通过'w'参数进行覆盖写入，而不用进行手动删除。
if os.path.exists(trainval_list):
    os.remove(trainval_list)
if os.path.exists(train_list):
    os.remove(train_list)
if os.path.exists(val_list):
    os.remove(val_list)
if os.path.exists(test_list):
    os.remove(test_list)

# 获取类别的名称，因为train,valid,test的类别是相同的，因此只需要从train中获取即可
class_name_list = os.listdir(os.path.join(dataset_root_path, 'train'))

# 分别从train,valid和test文件夹中去索引图像，并写入列表文件中
with codecs.open(trainval_list, 'a', 'utf-8') as f_trainval:
    with codecs.open(train_list, 'a', 'utf-8') as f_train:
        with codecs.open(val_list, 'a', 'utf-8') as f_val:
            with codecs.open(test_list, 'a', 'utf-8') as f_test:
                for prefix in class_prefix:
                    class_name_dir = os.listdir(os.path.join(dataset_root_path, prefix))
                    for i in range(len(class_name_list)):
                        class_name = class_name_list[i]                        
                        dataset_info['label_dict'][i] = class_name_list[i]
                        images = os.listdir(os.path.join(dataset_root_path, prefix, class_name))
                        for image in images:
                            if image not in excluded_folder and os.path.join(prefix, class_name, image) not in bad_file:   # 判断文件是否是坏样本
                                if prefix == 'train':
                                    f_train.write("{}\t{}\n".format(os.path.join(dataset_root_path, prefix, class_name, image), str(i)))
                                    f_trainval.write("{}\t{}\n".format(os.path.join(dataset_root_path, prefix, class_name, image), str(i)))
                                    num_train += 1
                                    num_trainval += 1
                                elif prefix == 'valid':
                                    f_val.write("{}\t{}\n".format(os.path.join(dataset_root_path, prefix, class_name, image), str(i)))
                                    f_trainval.write("{}\t{}\n".format(os.path.join(dataset_root_path, prefix, class_name, image), str(i)))
                                    num_val += 1
                                    num_trainval += 1
                                elif prefix == 'test':
                                    f_test.write("{}\t{}\n".format(os.path.join(dataset_root_path, prefix, class_name, image), str(i)))
                                    num_test += 1
                    
# 将数据集信息保存到json文件中供训练时使用
dataset_info['dataset_name'] = dataset_name
dataset_info['num_trainval'] = num_trainval
dataset_info['num_train'] = num_train
dataset_info['num_val'] = num_val
dataset_info['num_test'] = num_test
dataset_info['class_dim'] = len(class_name_list)

# 输出数据集信息json和统计情况
with codecs.open(dataset_info_list, 'w', encoding='utf-8') as f_dataset_info:  
    json.dump(dataset_info, f_dataset_info, ensure_ascii=False, indent=4, separators=(',', ':')) # 格式化字典格式的参数列表
    
print("图像列表已生成, 其中训练验证集样本{}个，训练集样本{}个, 验证集样本{}个, 测试集样本{}个, 共计{}个。".format(num_trainval, num_train, num_val, num_test, num_train+num_val+num_test))

图像列表已生成, 其中训练验证集样本7840个，训练集样本7190个, 验证集样本650个, 测试集样本660个, 共计8500个。


###### 展示数据集列表信息 ###################3
from pprint import pprint
with open(dataset_info_list, 'r') as f_info:
    dataset_info = json.load(f_info)
pprint(dataset_info)

{'class_dim': 12,
 'dataset_name': 'Zodiac',
 'label_dict': {'0': 'dog',
                '1': 'dragon',
                '10': 'snake',
                '11': 'tiger',
                '2': 'goat',
                '3': 'horse',
                '4': 'monkey',
                '5': 'ox',
                '6': 'pig',
                '7': 'rabbit',
                '8': 'ratt',
                '9': 'rooster'},
 'num_test': 660,
 'num_train': 7190,
 'num_trainval': 7840,
 'num_val': 650}


#################导入依赖库################################################## 
import os
import json
import codecs
import numpy as np
import time                        # 载入time时间库,用于计算训练时间
import paddle
import matplotlib.pyplot as plt    # 载入python的第三方图像处理库
from pprint import pprint

################全局参数配置################################################### 
#### 1. 训练超参数定义
train_parameters = {
    'course_name': 'DeepLearning',
    'project_name': 'HW06TransferLearning',
    'dataset_name': 'Zodiac',
    'architecture': 'Mobilenetv2',
    'training_data': 'train',
    'starting_time': time.strftime("%Y%m%d%H%M", time.localtime()),          # 全局启动时间
    'input_size': [3, 227, 227],             # 输入样本的尺度
    'mean_value': [0.485, 0.456, 0.406],     # Imagenet均值
    'std_value': [0.229, 0.224, 0.225],      # Imagenet标准差
    'num_trainval': -1,  
    'num_train': -1,
    'num_val': -1,
    'num_test': -1,
    'class_dim': -1,
    'label_dict': {},    
    'total_epoch': 10,                # 总迭代次数, 代码调试好后考虑
    'batch_size': 64,                 # 设置每个批次的数据大小，同时对训练提供器和测试
    'log_interval': 10,                # 设置训练过程中，每隔多少个batch显示一次
    'eval_interval': 1,               # 设置每个多少个epoch测试一次
    'checkpointed': False,               # 是否保存checkpoint模型
    'checkpoint_train': False,          # 是否接着上一次保存的参数接着训练，优先级高于预训练模型
    'checkpoint_model':'Butterfly_Mobilenetv2',            # 设置恢复训练时载入的模型参数
    'checkpoint_time': '202102182058',   #  恢复训练时所指向的指定时间戳
    'pretrained': True,              # 是否使用预训练的模型    
    'pretrained_model':'API',              # 设置预训练模型, API|Butterflies_AlexNet_final
    'dataset_root_path': 'D:\\Workspace\\ExpDatasets\\',
    'result_root_path': 'D:\\Workspace\\ExpResults\\',
    'deployment_root_path': 'D:\\Workspace\\ExpDeployments\\',
    'useGPU': True,                   # True | Flase
    'learning_strategy': {            # 学习率和优化器相关参数
        'optimizer_strategy': 'Momentum',                   # 优化器：Momentum, RMS, SGD, Adam
        'learning_rate_strategy': 'CosineAnnealingDecay',   # 学习率策略: 固定fixed, 分段衰减PiecewiseDecay, 余弦退火CosineAnnealingDecay, 指数ExponentialDecay, 多项式PolynomialDecay
        'learning_rate': 0.001,                              # 固定学习率
        'momentum': 0.9,                                    # 动量
        'Piecewise_boundaries': [60, 80, 90],               # 分段衰减：变换边界，每当运行到epoch时调整一次
        'Piecewise_values': [0.01, 0.001, 0.0001, 0.00001], # 分段衰减：步进学习率，每次调节的具体值
        'Exponential_gamma': 0.9,                           # 指数衰减：衰减指数
        'Polynomial_decay_steps': 10,                       # 多项式衰减：衰减周期，每个多少个epoch衰减一次
        'verbose': False
    },
    'augmentation_strategy': { 
        'withAugmentation': True,            # 数据扩展相关参数
        'augmentation_prob': 0.5,            # 设置数据增广的概率
        'rotate_angle': 15,                  # 随机旋转的角度
        'Hflip_prob': 0.5,                   # 随机翻转的概率
        'brightness': 0.4, 
        'contrast': 0.4,
        'saturation': 0.4,
        'hue': 0.4,
    },
}

#### 2. 设置简化参数名
args = train_parameters  
argsAS = args['augmentation_strategy']
argsLS = train_parameters['learning_strategy']

if not args['pretrained']:
    model_name = args['dataset_name'] + '_' + args['architecture'] + '_withoutPretrained'
else:
    model_name = args['dataset_name'] + '_' + args['architecture']


#### 3. 定义设备工作模式 [GPU|CPU]
# 定义使用CPU还是GPU，使用CPU时use_cuda = False,使用GPU时use_cuda = True
def init_device(useGPU=args['useGPU']):
    paddle.set_device('gpu:0') if useGPU else paddle.set_device('cpu')
init_device()


#### 4.定义各种路径：模型、训练、日志结果图
# 4.1 数据集路径
dataset_root_path = os.path.join(args['dataset_root_path'], args['dataset_name'])
json_dataset_info = os.path.join(dataset_root_path, 'dataset_info.json')

# 4.2 训练过程涉及的相关路径
result_root_path = os.path.join(args['result_root_path'], args['project_name'], model_name)
checkpoint_models_path = os.path.join(result_root_path, 'checkpoint_models')               # 迭代训练模型保存路径
final_figures_path = os.path.join(result_root_path, 'final_figures')                       # 训练过程曲线图
final_models_path = os.path.join(result_root_path, 'final_models')                         # 最终用于部署和推理的模型
logs_path = os.path.join(result_root_path, 'logs')                                         # 训练过程日志

# 4.3 checkpoint_ 路径用于定义恢复训练所用的模型保存
checkpoint_model = os.path.join(checkpoint_models_path, args['checkpoint_model'])



# 4.4 验证和测试时的相关路径(文件)
deployment_root_path = os.path.join(args['deployment_root_path'], args['course_name'], args['project_name'], model_name)
deployment_checkpoint_model = os.path.join(deployment_root_path, 'checkpoint_models', model_name + '_final')
deployment_final_model = os.path.join(deployment_root_path, 'final_models', model_name + '_final')
deployment_final_figures_path = os.path.join(deployment_root_path, 'final_figures')
deployment_logs_path = os.path.join(deployment_root_path, 'logs')
deployment_pretrained_model = os.path.join(deployment_root_path, 'pretrained_dir', args['pretrained_model'])

# 4.5 初始化结果目录
def init_result_path():
    if not os.path.exists(final_models_path):
        os.makedirs(final_models_path)
    if not os.path.exists(final_figures_path):
        os.makedirs(final_figures_path)
    if not os.path.exists(logs_path):
        os.makedirs(logs_path)
    if not os.path.exists(checkpoint_models_path):
        os.makedirs(checkpoint_models_path)
init_result_path()


#### 5. 初始化参数
def init_train_parameters():
    dataset_info = json.loads(open(json_dataset_info, 'r', encoding='utf-8').read())    
    train_parameters['num_trainval'] = dataset_info['num_trainval']
    train_parameters['num_train'] = dataset_info['num_train']
    train_parameters['num_val'] = dataset_info['num_val']
    train_parameters['num_test'] = dataset_info['num_test']
    train_parameters['class_dim'] = dataset_info['class_dim']
    train_parameters['label_dict'] = dataset_info['label_dict']
init_train_parameters()

# 输出训练参数 train_parameters
# if __name__ == '__main__':
#     pprint(args)


import os
import sys
import cv2
import numpy as np
import paddle
import paddle.vision.transforms as T
from paddle.io import DataLoader
paddle.vision.set_image_backend('cv2')


input_size = (args['input_size'][1], args['input_size'][2])

# 1. 数据集的定义
class Dataset(paddle.io.Dataset):
    def __init__(self, dataset_root_path, mode='test', withAugmentation=argsAS['withAugmentation']):
        assert mode in ['train', 'val', 'test', 'trainval']
        self.data = []
        self.withAugmentation = withAugmentation
        
        with open(os.path.join(dataset_root_path, mode+'.txt')) as f:
            for line in f.readlines():
                info = line.strip().split('\t')
                image_path = os.path.join(dataset_root_path, 'Data', info[0].strip())
                
                if len(info) == 2:
                    self.data.append([image_path, info[1].strip()])
                elif len(info) == 1:
                    self.data.append([image_path, -1])
        
        prob = np.random.random()
        if mode in ['train', 'trainval'] and prob >= argsAS['augmentation_prob']:
            self.transforms = T.Compose([
                T.RandomResizedCrop(input_size),
                T.RandomHorizontalFlip(argsAS['Hflip_prob']),
                T.RandomRotation(argsAS['rotate_angle']),
                T.ColorJitter(brightness=argsAS['brightness'], contrast=argsAS['contrast'], saturation=argsAS['saturation'], hue=argsAS['hue']),
                T.ToTensor(),
                T.Normalize(mean=args['mean_value'], std=args['std_value'])
            ])
        elif mode in ['val', 'test'] or prob < argsAS['augmentation_prob']:
            self.transforms = T.Compose([
                T.Resize(input_size),
                T.ToTensor(),
                T.Normalize(mean=args['mean_value'], std=args['std_value'])                
            ])
    
    # 根据索引获取单个样本
    def __getitem__(self, index):
        image_path, label = self.data[index]
        image = cv2.imread(image_path, 1) # 使用cv2进行数据读取可以强制将的图像转化为彩色模式，其中0为灰度模式，1为彩色模式
        
        if self.withAugmentation == True:
            image = self.transforms(image)
        
        return image, np.array(label, dtype='int64')
                
    # 获取样本总数    
    def __len__(self):        
        return len(self.data)


###############################################################    
# 测试输入数据类：分别输出进行预处理和未进行预处理的数据形态和例图
if __name__ == "__main__":
    import random

    # 1. 载入数据
    dataset_val = Dataset(dataset_root_path, mode='val')
    i = random.randrange(0, len(dataset_val))
    img1 = dataset_val[i][0]
    print('验证数据预处理前的数据形态(进行预处理后): {}'.format(img1.shape))  
  
    dataset_test = Dataset(dataset_root_path, mode='test', withAugmentation=False)
    j = random.randrange(0, len(dataset_test))
    img2 = dataset_test[j][0]
    print('测试数据预处理前的数据形态(未进行预处理): {}'.format(img2.shape))

C:\Users\Administrator\anaconda3\lib\site-packages\ipykernel\ipkernel.py:287: DeprecationWarning: `should_run_async` will not call `transform_cell` automatically in the future. Please pass the result to `transformed_cell` argument and any exception that happen during thetransform in `preprocessing_exc_tuple` in IPython 7.17 and above.
  and should_run_async(code)

验证数据预处理前的数据形态(进行预处理后): [3, 227, 227]
测试数据预处理前的数据形态(未进行预处理): (433, 770, 3)


import os
import sys
from paddle.io import DataLoader

# 1. 从数据集库中获取数据
dataset_trainval = Dataset(dataset_root_path, mode='trainval')
dataset_train = Dataset(dataset_root_path, mode='train')
dataset_val = Dataset(dataset_root_path, mode='val')
dataset_test = Dataset(dataset_root_path, mode='test')

# 2. 创建读取器
trainval_reader = DataLoader(dataset_trainval, batch_size=args['batch_size'], shuffle=True, drop_last=True)
train_reader = DataLoader(dataset_train, batch_size=args['batch_size'], shuffle=True, drop_last=True)
val_reader = DataLoader(dataset_val, batch_size=args['batch_size'], shuffle=False, drop_last=False)
test_reader = DataLoader(dataset_test, batch_size=args['batch_size'], shuffle=False, drop_last=False)

# 测试读取器
if __name__ == "__main__":
    for i, (image, label) in enumerate(val_reader()):
        print('验证集batch_{}的图像形态:{}, 标签形态:{}'.format(i, image.shape, label.shape))
        break

验证集batch_0的图像形态:[64, 3, 227, 227], 标签形态:[64]


# 绘制训练batch精度和平均loss
def draw_process(title, loss_label, accuracy_label, iters, losses, accuracies, figurename=None, isShow=False):

    # 第一组坐标轴 Loss
    _, ax1 = plt.subplots()   # plt.subplots(figsize=(14,6))
    ax1.plot(iters, losses, color='red', label=loss_label)
    ax1.set_xlabel('Iters', fontsize=20)
    ax1.set_ylabel(loss_label, fontsize=20)
    max_loss = max(losses)
    ax1.set_ylim(0, max_loss*1.2)

    # 第二组坐标轴 accuracy
    ax2 = ax1.twinx()
    ax2.plot(iters, accuracies, color='blue', label=accuracy_label)
    ax2.set_ylabel(accuracy_label, fontsize=20)
    max_acc = max(accuracies)
    ax2.set_ylim(0, max_acc*1.2)

    plt.title(title, fontsize=24)
    # 图例
    handles1, labels1 = ax1.get_legend_handles_labels()
    handles2, labels2 = ax2.get_legend_handles_labels()
    plt.legend(handles1+handles2, labels1+labels2, loc='best')
    plt.grid()

    # 将绘图结果保存到final_figures目录
    if figurename != None:
        if not os.path.exists(final_figures_path):
            os.makedirs(final_figures_path)
        plt.savefig(os.path.join(final_figures_path, figurename + '.png'))
        # 将绘图数据保存到final_figures目录
        figure_data = np.array([iters, losses, accuracies])
        np.save(os.path.join(final_figures_path, figurename + '.npy'), figure_data)

    # 显示绘图结果
    if isShow is True:
        plt.show()
        
        
### 绘图测试 ###################################################
if __name__ == '__main__':
    root_path = deployment_final_figures_path
    try:
        train = np.load(os.path.join(root_path, 'train.npy'))
        draw_process('train', 'loss', 'accuracy', train[0], train[1], train[2], figurename=None, isShow=True)
    except:
        print('数据不存在，无法进行绘制')


#################################################
# 修改者: Xinyu Ou (http://ouxinyu.cn)
# 功能: 输出日志, 并保存到日志文件
# 格式: 2021-02-03 23:03:07,354 - INFO: [Messages]
# 调用方法：
# from utils.logger import logger
# logger.info('Good morning')
#################################################

import os
import sys
import logging

def init_log_config():
    """
    初始化日志相关配置
    :return:
    """
    global logger
    logger = logging.getLogger()
    logger.setLevel(logging.INFO)
    if not os.path.exists(logs_path):
        os.makedirs(logs_path)
    log_name = os.path.join(logs_path, model_name + '.logs')
    sh = logging.StreamHandler()                                      # 打印到屏幕控制多台
    fh = logging.FileHandler(log_name, mode='w', encoding='utf8')     # 打印到文件
    fh.setLevel(logging.DEBUG)
    formatter = logging.Formatter("%(asctime)s - %(levelname)s: %(message)s")
    fh.setFormatter(formatter)
    sh.setFormatter(formatter)
    logger.addHandler(fh)  # 输出到文件
    logger.addHandler(sh)  # 输出到控制台
    return logger

logger = init_log_config()

# 测试
if __name__ == "__main__":   
    logger.info('测试一下, 模型名称: {}'.format(model_name))

C:\Users\Administrator\anaconda3\lib\site-packages\ipykernel\ipkernel.py:287: DeprecationWarning: `should_run_async` will not call `transform_cell` automatically in the future. Please pass the result to `transformed_cell` argument and any exception that happen during thetransform in `preprocessing_exc_tuple` in IPython 7.17 and above.
  and should_run_async(code)
2021-08-18 20:38:27,227 - INFO: 测试一下, 模型名称: Zodiac_Mobilenetv2


# import paddle
# from paddle.static import InputSpec

# # 设置输入样本的维度
# input_spec = InputSpec(shape=[None] + train_parameters['input_size'], dtype='float32', name='image')
# label_spec = InputSpec(shape=[None, 1], dtype='int64', name='label')

# # 载入官方标准模型，若不存在则会自动进行下载，pretrained=True|False控制是否使用Imagenet预训练参数
# network = paddle.vision.models.mobilenet_v2(num_classes=args['class_dim'], pretrained=args['pretrained'])

# model = paddle.Model(network, input_spec, label_spec) 
# logger.info('模型参数信息：')
# logger.info(model.summary()) # 是否显示神经网络的具体信息


import sys
import os
import paddle
import paddle.optimizer as optimizer

def learning_rate_setting(verbose=argsLS['verbose']):
    if argsLS['learning_rate_strategy'] == 'PiecewiseDecay':
        lr = optimizer.lr.PiecewiseDecay(boundaries=argsLS['Piecewise_boundaries'], values=argsLS['Piecewise_values'], verbose=verbose)
    elif argsLS['learning_rate_strategy'] == 'CosineAnnealingDecay':
        step_each_epoch = args['num_train'] // (args['batch_size'] * 2)
        T_max = step_each_epoch * args['total_epoch']
        lr = optimizer.lr.CosineAnnealingDecay(learning_rate=argsLS['learning_rate'], T_max=T_max, verbose=verbose)
    elif argsLS['learning_rate_strategy'] == 'ExponentialDecay':
        lr = optimizer.lr.ExponentialDecay(learning_rate=argsLS['learning_rate'], gamma=argsLS['Exponential_gamma'], verbose=verbose)
    elif argsLS['learning_rate_strategy'] == 'PolynomialDecay':
        lr = optimizer.lr.PolynomialDecay(learning_rate=argsLS['learning_rate'], decay_steps=argsLS['Polynomial_decay_steps'], verbose=verbose)
    else:
        lr = argsLS['learning_rate']
    return lr

def optimizer_setting(model, lr):
    if argsLS['optimizer_strategy'] == 'Momentum':
        # 阶梯型的学习率适合比较大规模的训练数据
        opt = optimizer.Momentum(learning_rate=lr, momentum=argsLS['momentum'], parameters=model.parameters())
    elif argsLS['optimizer_strategy'] == 'RMS':
        # 阶梯型的学习率适合比较大规模的训练数据
        opt = optimizer.RMSProp(learning_rate=lr, parameters=model.parameters())
    elif argsLS['optimizer_strategy'] == 'SGD':
        # loss下降相对较慢，但是最终效果不错，阶梯型的学习率适合比较大规模的训练数据
        opt = optimizer.SGD(learning_rate=lr, parameters=model.parameters())
    elif argsLS['optimizer_strategy'] == 'Adam':
        # 能够比较快速的降低 loss，但是相对后期乏力
        opt = optimizer.Adam(learning_rate=lr, parameters=model.parameters())
    else:
        print('学习率设置错误, 请重新设置。')

    return opt


# 学习率输出测试
# if __name__ == '__main__':
#     print('当前学习率策略为: {}'.format(argsLS['learning_rate_strategy']))
#     linear = paddle.nn.Linear(10, 10)
#     lr = learning_rate_setting(verbose=True)
#     opt = optimizer_setting(linear, lr)
#     if argsLS['learning_rate_strategy'] == 'fixed':
#         print('learning = {}'.format(argsLS['learning_rate']))
#     else:
#         for epoch in range(20):
#             for batch_id in range(10):
#                 x = paddle.uniform([10, 10])
#                 out = linear(x)
#                 loss = paddle.mean(out)
#                 loss.backward()
#                 opt.step()
#                 opt.clear_gradients()
#                 # lr.step()    # If you update learning rate each step
#             lr.step()        # If you update learning rate each epoch


# 载入项目文件夹
import sys
import numpy as np
import paddle
import paddle.nn.functional as F
from paddle.static import InputSpec

__all__ = ['eval']

def eval(model, data_reader, verbose=0):
    accuracies_top1 = []
    accuracies_top5 = []
    losses = []
    n_total = 0
    
    for batch_id, (image, label) in enumerate(data_reader):
        n_batch = len(label)
        n_total = n_total + n_batch

        label = paddle.unsqueeze(label, axis=1)

        loss, acc = model.eval_batch([image], [label])
        losses.append(loss[0])                    
        accuracies_top1.append(acc[0][0]*n_batch)
        accuracies_top5.append(acc[0][1]*n_batch) 

        if verbose == 1:
            print('\r Batch:{}/{}, acc_top1:[{:.5f}], acc_top5:[{:.5f}]'.format(batch_id+1, len(data_reader), acc[0][0], acc[0][1]), end='') 
                
    avg_loss = np.sum(losses)/n_total                 # loss 记录的是当前batch的累积值
    avg_acc_top1 = np.sum(accuracies_top1)/n_total    # metric 是当前batch的平均值
    avg_acc_top5 = np.sum(accuracies_top5)/n_total

    return avg_loss, avg_acc_top1, avg_acc_top5

##############################################################     
if __name__ == '__main__':
    try:
        # 设置输入样本的维度    
        input_spec = InputSpec(shape=[None] + args['input_size'], dtype='float32', name='image')
        label_spec = InputSpec(shape=[None, 1], dtype='int64', name='label')

        # 载入模型

        network = paddle.vision.models.mobilenet_v2(num_classes=args['class_dim'])
        model = paddle.Model(network, input_spec, label_spec)                   # 模型实例化
        model.load(deployment_checkpoint_path)                                # 载入调优模型的参数
        model.prepare(loss=paddle.nn.CrossEntropyLoss(),                        # 设置loss
                      metrics=paddle.metric.Accuracy(topk=(1,5)))               # 设置评价指标
    
        # 执行评估函数，并输出验证集样本的损失和精度
        print('开始评估...')
        avg_loss, avg_acc_top1, avg_acc_top5 = eval(model, val_reader(), verbose=1)
        print('\r [验证集] 损失: {:.5f}, top1精度:{:.5f}, top5精度为:{:.5f} \n'.format(avg_loss, avg_acc_top1, avg_acc_top5), end='')    
        avg_loss, avg_acc_top1, avg_acc_top5 = eval(model, test_reader(), verbose=1)
        print('\r [测试集] 损失: {:.5f}, top1精度:{:.5f}, top5精度为:{:.5f}'.format(avg_loss, avg_acc_top1, avg_acc_top5), end='')
    except:
        print('数据不存在, 跳过测试')

数据不存在, 跳过测试


import os
import time
import json
import paddle
from paddle.static import InputSpec

# 初始配置变量
total_epoch = train_parameters['total_epoch']

# 初始化绘图列表
all_train_iters = []
all_train_losses = []
all_train_accs_top1 = []
all_train_accs_top5 = []
all_test_losses = []
all_test_iters = []
all_test_accs_top1 = []
all_test_accs_top5 = []

def train(model):  
    # 初始化临时变量
    num_batch = 0
    best_result = 0
    best_result_id = 0
    elapsed = 0


    # 根据config文件设置训练数据来源
    if train_parameters['training_data'] == 'trainval':
        data_reader = trainval_reader
    elif train_parameters['training_data'] == 'train':
        data_reader = train_reader

    for epoch in range(1, total_epoch+1):
        for batch_id, (image, label) in enumerate(data_reader):
            num_batch += 1
            
            label = paddle.unsqueeze(label, axis=1)                        
            loss, acc = model.train_batch([image], [label])
       
            if num_batch % train_parameters['log_interval'] == 0: # 每10个batch显示一次日志，适合大数据集  
                avg_loss = loss[0][0]
                acc_top1 = acc[0][0]
                acc_top5 = acc[0][1]
                
                elapsed_step = time.perf_counter() - elapsed - start
                elapsed = time.perf_counter() - start
                logger.info('Epoch:{}/{}, batch:{}, train_loss:[{:.5f}], acc_top1:[{:.5f}], acc_top5:[{:.5f}]({:.2f}s)'
                            .format(epoch, total_epoch, num_batch, loss[0][0], acc[0][0], acc[0][1], elapsed_step))

                # 记录训练过程，用于可视化训练过程中的loss和accuracy
                all_train_iters.append(num_batch)
                all_train_losses.append(avg_loss)
                all_train_accs_top1.append(acc_top1)
                all_train_accs_top5.append(acc_top5)

        # 每隔一定周期进行一次测试
        if epoch % train_parameters['eval_interval'] == 0 or epoch == total_epoch:
            # 模型校验
            avg_loss, avg_acc_top1, avg_acc_top5 = eval(model, val_reader())            
            logger.info('[validation] Epoch:{}/{}, val_loss:[{:.5f}], val_top1:[{:.5f}], val_top5:[{:.5f}]'.format(epoch, total_epoch, avg_loss, avg_acc_top1, avg_acc_top5))

            # 记录测试过程，用于可视化训练过程中的loss和accuracy
            all_test_iters.append(epoch)
            all_test_losses.append(avg_loss)
            all_test_accs_top1.append(avg_acc_top1)            
            all_test_accs_top5.append(avg_acc_top5)
            

            # 将性能最好的模型保存为final模型
            if avg_acc_top1 > best_result:
                best_result = avg_acc_top1
                best_result_id = epoch
                
                # finetune model 用于调优和恢复训练
                model.save(os.path.join(checkpoint_models_path, model_name + '_final')) 
                # inference model 用于部署和预测 
                model.save(os.path.join(final_models_path, model_name + '_final'), training=False)
                logger.info('已保存当前测试模型(epoch={})为最优模型:{}_final'.format(best_result_id, model_name)) 
            logger.info('最优top1测试精度:{:.5f} (epoch={})'.format(best_result, best_result_id))
            
            # 根据需要决定是否需要将每次测试的模型都进行保存(if needed)，保存模型需要耗费一定的运算时间和大量的存储资源
            # 建议在训练大型模型时，开启该选项，方便训练中断时能够及时恢复训练
            # 训练小型模型（训练时间短）时，可以关闭该选项，以进一步提高训练速度
            if train_parameters['checkpointed']:
                model.save(os.path.join(checkpoint_models_path, model_name + '_' + str(epoch)))


    logger.info('训练完成，最终性能accuracy={:.5f}(epoch={}), 总耗时{:.2f}s, 已将其保存为：{}_final'.format(best_result, best_result_id, time.perf_counter() - start, model_name))


#### 训练主函数 ########################################################3
if __name__ == '__main__':
#     model = MyNet(num_classes=10)
#     train(model)
    # 将此次训练的超参数进行保存
    data = json.dumps(train_parameters, indent=4, ensure_ascii=False, sort_keys=False, separators=(',', ':'))   # 格式化字典格式的参数列表
    logger.info(data)
    # 启动训练过程
    logger.info('训练参数保存完毕，使用{}模型, 训练{}数据, 训练集{}, 启动训练...'.format(train_parameters['architecture'],train_parameters['dataset_name'],train_parameters['training_data']))
    logger.info('当前模型目录为：{}'.format(model_name + '_' + train_parameters['starting_time']))

    # 设置输入样本的维度
    input_spec = InputSpec(shape=[None] + train_parameters['input_size'], dtype='float32', name='image')
    label_spec = InputSpec(shape=[None, 1], dtype='int64', name='label')
  
    # 载入官方标准模型，若不存在则会自动进行下载，pretrained=True|False控制是否使用Imagenet预训练参数
    network = paddle.vision.models.mobilenet_v2(num_classes=args['class_dim'], pretrained=args['pretrained'])
 
    model = paddle.Model(network, input_spec, label_spec) 
    logger.info('模型参数信息：')
    logger.info(model.summary()) # 是否显示神经网络的具体信息

    if train_parameters['checkpoint_train'] == True:
        model.load(checkpoint_load_model)
        logger.info('载入{}中断模型和参数完毕，开始从checkpoint恢复训练'.format(train_parameters['architecture']))
        logger.info('checkpoint模型:{}'.format(checkpoint_load_model))
    else:
        if train_parameters['pretrained'] == False:
            logger.info('载入{}模型完毕，从初始状态开始训练'.format(train_parameters['architecture']))
        elif train_parameters['pretrained_model'] == 'API':
            logger.info('载入Imagenet-{}预训练模型完毕，开始微调训练(fine-tune)'.format(train_parameters['architecture']))
        else:
            model.load(project_pretrained_model)
            logger.info('载入自定义预训练{}模型完毕，开始微调训练(fine-tune)'.format(train_parameters['architecture']))
            logger.info('预训练模型:{}'.format(project_pretrained_model))
          
    # 设置学习率、优化器、损失函数和评价指标
    lr = learning_rate_setting()
    optimizer = optimizer_setting(model, lr)    
    model.prepare(optimizer,
                  paddle.nn.CrossEntropyLoss(),
                  paddle.metric.Accuracy(topk=(1,5)))   

    # 启动训练过程
    start = time.perf_counter()
    train(model)
    logger.info('训练完毕，结果路径{}.'.format(result_root_path))

    # 输出训练过程图
    logger.info('Done.')
    draw_process("Training Process", 'Train Loss', 'Train Accuracy(top1)', all_train_iters, all_train_losses, all_train_accs_top1, 'train')
    draw_process("Validation Results", 'Validation Loss', 'Validation Accuracy(top1)', all_test_iters, all_test_losses, all_test_accs_top1, 'val')

2021-08-18 20:38:27,387 - INFO: {
    "course_name":"DeepLearning",
    "project_name":"HW06TransferLearning",
    "dataset_name":"Zodiac",
    "architecture":"Mobilenetv2",
    "training_data":"train",
    "starting_time":"202108182038",
    "input_size":[
        3,
        227,
        227
    ],
    "mean_value":[
        0.485,
        0.456,
        0.406
    ],
    "std_value":[
        0.229,
        0.224,
        0.225
    ],
    "num_trainval":7840,
    "num_train":7190,
    "num_val":650,
    "num_test":660,
    "class_dim":12,
    "label_dict":{
        "0":"dog",
        "1":"dragon",
        "2":"goat",
        "3":"horse",
        "4":"monkey",
        "5":"ox",
        "6":"pig",
        "7":"rabbit",
        "8":"ratt",
        "9":"rooster",
        "10":"snake",
        "11":"tiger"
    },
    "total_epoch":10,
    "batch_size":64,
    "log_interval":10,
    "eval_interval":1,
    "checkpointed":false,
    "checkpoint_train":false,
    "checkpoint_model":"Butterfly_Mobilenetv2",
    "checkpoint_time":"202102182058",
    "pretrained":true,
    "pretrained_model":"API",
    "dataset_root_path":"D:\\Workspace\\ExpDatasets\\",
    "result_root_path":"D:\\Workspace\\ExpResults\\",
    "deployment_root_path":"D:\\Workspace\\ExpDeployments\\",
    "useGPU":true,
    "learning_strategy":{
        "optimizer_strategy":"Momentum",
        "learning_rate_strategy":"CosineAnnealingDecay",
        "learning_rate":0.001,
        "momentum":0.9,
        "Piecewise_boundaries":[
            60,
            80,
            90
        ],
        "Piecewise_values":[
            0.01,
            0.001,
            0.0001,
            1e-05
        ],
        "Exponential_gamma":0.9,
        "Polynomial_decay_steps":10,
        "verbose":false
    },
    "augmentation_strategy":{
        "withAugmentation":true,
        "augmentation_prob":0.5,
        "rotate_angle":15,
        "Hflip_prob":0.5,
        "brightness":0.4,
        "contrast":0.4,
        "saturation":0.4,
        "hue":0.4
    }
}
2021-08-18 20:38:27,389 - INFO: 训练参数保存完毕，使用Mobilenetv2模型, 训练Zodiac数据, 训练集train, 启动训练...
2021-08-18 20:38:27,390 - INFO: 当前模型目录为：Zodiac_Mobilenetv2_202108182038
2021-08-18 20:38:27,458 - INFO: unique_endpoints {''}
2021-08-18 20:38:27,459 - INFO: File C:\Users\Administrator/.cache/paddle/hapi/weights\mobilenet_v2_x1.0.pdparams md5 checking...
2021-08-18 20:38:27,501 - INFO: Found C:\Users\Administrator/.cache/paddle/hapi/weights\mobilenet_v2_x1.0.pdparams
C:\Users\Administrator\anaconda3\lib\site-packages\paddle\fluid\dygraph\layers.py:1301: UserWarning: Skip loading for classifier.1.weight. classifier.1.weight receives a shape [1280, 1000], but the expected shape is [1280, 12].
  warnings.warn(("Skip loading for {}. ".format(key) + str(err)))
C:\Users\Administrator\anaconda3\lib\site-packages\paddle\fluid\dygraph\layers.py:1301: UserWarning: Skip loading for classifier.1.bias. classifier.1.bias receives a shape [1000], but the expected shape is [12].
  warnings.warn(("Skip loading for {}. ".format(key) + str(err)))
2021-08-18 20:38:27,643 - INFO: 模型参数信息：
2021-08-18 20:38:27,682 - INFO: {'total_params': 2273356, 'trainable_params': 2205132}
2021-08-18 20:38:27,682 - INFO: 载入Imagenet-Mobilenetv2预训练模型完毕，开始微调训练(fine-tune)

-------------------------------------------------------------------------------
   Layer (type)         Input Shape          Output Shape         Param #    
===============================================================================
     Conv2D-53       [[1, 3, 227, 227]]   [1, 32, 114, 114]         864      
  BatchNorm2D-53    [[1, 32, 114, 114]]   [1, 32, 114, 114]         128      
     ReLU6-36       [[1, 32, 114, 114]]   [1, 32, 114, 114]          0       
     Conv2D-54      [[1, 32, 114, 114]]   [1, 32, 114, 114]         288      
  BatchNorm2D-54    [[1, 32, 114, 114]]   [1, 32, 114, 114]         128      
     ReLU6-37       [[1, 32, 114, 114]]   [1, 32, 114, 114]          0       
     Conv2D-55      [[1, 32, 114, 114]]   [1, 16, 114, 114]         512      
  BatchNorm2D-55    [[1, 16, 114, 114]]   [1, 16, 114, 114]         64       
InvertedResidual-18 [[1, 32, 114, 114]]   [1, 16, 114, 114]          0       
     Conv2D-56      [[1, 16, 114, 114]]   [1, 96, 114, 114]        1,536     
  BatchNorm2D-56    [[1, 96, 114, 114]]   [1, 96, 114, 114]         384      
     ReLU6-38       [[1, 96, 114, 114]]   [1, 96, 114, 114]          0       
     Conv2D-57      [[1, 96, 114, 114]]    [1, 96, 57, 57]          864      
  BatchNorm2D-57     [[1, 96, 57, 57]]     [1, 96, 57, 57]          384      
     ReLU6-39        [[1, 96, 57, 57]]     [1, 96, 57, 57]           0       
     Conv2D-58       [[1, 96, 57, 57]]     [1, 24, 57, 57]         2,304     
  BatchNorm2D-58     [[1, 24, 57, 57]]     [1, 24, 57, 57]          96       
InvertedResidual-19 [[1, 16, 114, 114]]    [1, 24, 57, 57]           0       
     Conv2D-59       [[1, 24, 57, 57]]     [1, 144, 57, 57]        3,456     
  BatchNorm2D-59     [[1, 144, 57, 57]]    [1, 144, 57, 57]         576      
     ReLU6-40        [[1, 144, 57, 57]]    [1, 144, 57, 57]          0       
     Conv2D-60       [[1, 144, 57, 57]]    [1, 144, 57, 57]        1,296     
  BatchNorm2D-60     [[1, 144, 57, 57]]    [1, 144, 57, 57]         576      
     ReLU6-41        [[1, 144, 57, 57]]    [1, 144, 57, 57]          0       
     Conv2D-61       [[1, 144, 57, 57]]    [1, 24, 57, 57]         3,456     
  BatchNorm2D-61     [[1, 24, 57, 57]]     [1, 24, 57, 57]          96       
InvertedResidual-20  [[1, 24, 57, 57]]     [1, 24, 57, 57]           0       
     Conv2D-62       [[1, 24, 57, 57]]     [1, 144, 57, 57]        3,456     
  BatchNorm2D-62     [[1, 144, 57, 57]]    [1, 144, 57, 57]         576      
     ReLU6-42        [[1, 144, 57, 57]]    [1, 144, 57, 57]          0       
     Conv2D-63       [[1, 144, 57, 57]]    [1, 144, 29, 29]        1,296     
  BatchNorm2D-63     [[1, 144, 29, 29]]    [1, 144, 29, 29]         576      
     ReLU6-43        [[1, 144, 29, 29]]    [1, 144, 29, 29]          0       
     Conv2D-64       [[1, 144, 29, 29]]    [1, 32, 29, 29]         4,608     
  BatchNorm2D-64     [[1, 32, 29, 29]]     [1, 32, 29, 29]          128      
InvertedResidual-21  [[1, 24, 57, 57]]     [1, 32, 29, 29]           0       
     Conv2D-65       [[1, 32, 29, 29]]     [1, 192, 29, 29]        6,144     
  BatchNorm2D-65     [[1, 192, 29, 29]]    [1, 192, 29, 29]         768      
     ReLU6-44        [[1, 192, 29, 29]]    [1, 192, 29, 29]          0       
     Conv2D-66       [[1, 192, 29, 29]]    [1, 192, 29, 29]        1,728     
  BatchNorm2D-66     [[1, 192, 29, 29]]    [1, 192, 29, 29]         768      
     ReLU6-45        [[1, 192, 29, 29]]    [1, 192, 29, 29]          0       
     Conv2D-67       [[1, 192, 29, 29]]    [1, 32, 29, 29]         6,144     
  BatchNorm2D-67     [[1, 32, 29, 29]]     [1, 32, 29, 29]          128      
InvertedResidual-22  [[1, 32, 29, 29]]     [1, 32, 29, 29]           0       
     Conv2D-68       [[1, 32, 29, 29]]     [1, 192, 29, 29]        6,144     
  BatchNorm2D-68     [[1, 192, 29, 29]]    [1, 192, 29, 29]         768      
     ReLU6-46        [[1, 192, 29, 29]]    [1, 192, 29, 29]          0       
     Conv2D-69       [[1, 192, 29, 29]]    [1, 192, 29, 29]        1,728     
  BatchNorm2D-69     [[1, 192, 29, 29]]    [1, 192, 29, 29]         768      
     ReLU6-47        [[1, 192, 29, 29]]    [1, 192, 29, 29]          0       
     Conv2D-70       [[1, 192, 29, 29]]    [1, 32, 29, 29]         6,144     
  BatchNorm2D-70     [[1, 32, 29, 29]]     [1, 32, 29, 29]          128      
InvertedResidual-23  [[1, 32, 29, 29]]     [1, 32, 29, 29]           0       
     Conv2D-71       [[1, 32, 29, 29]]     [1, 192, 29, 29]        6,144     
  BatchNorm2D-71     [[1, 192, 29, 29]]    [1, 192, 29, 29]         768      
     ReLU6-48        [[1, 192, 29, 29]]    [1, 192, 29, 29]          0       
     Conv2D-72       [[1, 192, 29, 29]]    [1, 192, 15, 15]        1,728     
  BatchNorm2D-72     [[1, 192, 15, 15]]    [1, 192, 15, 15]         768      
     ReLU6-49        [[1, 192, 15, 15]]    [1, 192, 15, 15]          0       
     Conv2D-73       [[1, 192, 15, 15]]    [1, 64, 15, 15]        12,288     
  BatchNorm2D-73     [[1, 64, 15, 15]]     [1, 64, 15, 15]          256      
InvertedResidual-24  [[1, 32, 29, 29]]     [1, 64, 15, 15]           0       
     Conv2D-74       [[1, 64, 15, 15]]     [1, 384, 15, 15]       24,576     
  BatchNorm2D-74     [[1, 384, 15, 15]]    [1, 384, 15, 15]        1,536     
     ReLU6-50        [[1, 384, 15, 15]]    [1, 384, 15, 15]          0       
     Conv2D-75       [[1, 384, 15, 15]]    [1, 384, 15, 15]        3,456     
  BatchNorm2D-75     [[1, 384, 15, 15]]    [1, 384, 15, 15]        1,536     
     ReLU6-51        [[1, 384, 15, 15]]    [1, 384, 15, 15]          0       
     Conv2D-76       [[1, 384, 15, 15]]    [1, 64, 15, 15]        24,576     
  BatchNorm2D-76     [[1, 64, 15, 15]]     [1, 64, 15, 15]          256      
InvertedResidual-25  [[1, 64, 15, 15]]     [1, 64, 15, 15]           0       
     Conv2D-77       [[1, 64, 15, 15]]     [1, 384, 15, 15]       24,576     
  BatchNorm2D-77     [[1, 384, 15, 15]]    [1, 384, 15, 15]        1,536     
     ReLU6-52        [[1, 384, 15, 15]]    [1, 384, 15, 15]          0       
     Conv2D-78       [[1, 384, 15, 15]]    [1, 384, 15, 15]        3,456     
  BatchNorm2D-78     [[1, 384, 15, 15]]    [1, 384, 15, 15]        1,536     
     ReLU6-53        [[1, 384, 15, 15]]    [1, 384, 15, 15]          0       
     Conv2D-79       [[1, 384, 15, 15]]    [1, 64, 15, 15]        24,576     
  BatchNorm2D-79     [[1, 64, 15, 15]]     [1, 64, 15, 15]          256      
InvertedResidual-26  [[1, 64, 15, 15]]     [1, 64, 15, 15]           0       
     Conv2D-80       [[1, 64, 15, 15]]     [1, 384, 15, 15]       24,576     
  BatchNorm2D-80     [[1, 384, 15, 15]]    [1, 384, 15, 15]        1,536     
     ReLU6-54        [[1, 384, 15, 15]]    [1, 384, 15, 15]          0       
     Conv2D-81       [[1, 384, 15, 15]]    [1, 384, 15, 15]        3,456     
  BatchNorm2D-81     [[1, 384, 15, 15]]    [1, 384, 15, 15]        1,536     
     ReLU6-55        [[1, 384, 15, 15]]    [1, 384, 15, 15]          0       
     Conv2D-82       [[1, 384, 15, 15]]    [1, 64, 15, 15]        24,576     
  BatchNorm2D-82     [[1, 64, 15, 15]]     [1, 64, 15, 15]          256      
InvertedResidual-27  [[1, 64, 15, 15]]     [1, 64, 15, 15]           0       
     Conv2D-83       [[1, 64, 15, 15]]     [1, 384, 15, 15]       24,576     
  BatchNorm2D-83     [[1, 384, 15, 15]]    [1, 384, 15, 15]        1,536     
     ReLU6-56        [[1, 384, 15, 15]]    [1, 384, 15, 15]          0       
     Conv2D-84       [[1, 384, 15, 15]]    [1, 384, 15, 15]        3,456     
  BatchNorm2D-84     [[1, 384, 15, 15]]    [1, 384, 15, 15]        1,536     
     ReLU6-57        [[1, 384, 15, 15]]    [1, 384, 15, 15]          0       
     Conv2D-85       [[1, 384, 15, 15]]    [1, 96, 15, 15]        36,864     
  BatchNorm2D-85     [[1, 96, 15, 15]]     [1, 96, 15, 15]          384      
InvertedResidual-28  [[1, 64, 15, 15]]     [1, 96, 15, 15]           0       
     Conv2D-86       [[1, 96, 15, 15]]     [1, 576, 15, 15]       55,296     
  BatchNorm2D-86     [[1, 576, 15, 15]]    [1, 576, 15, 15]        2,304     
     ReLU6-58        [[1, 576, 15, 15]]    [1, 576, 15, 15]          0       
     Conv2D-87       [[1, 576, 15, 15]]    [1, 576, 15, 15]        5,184     
  BatchNorm2D-87     [[1, 576, 15, 15]]    [1, 576, 15, 15]        2,304     
     ReLU6-59        [[1, 576, 15, 15]]    [1, 576, 15, 15]          0       
     Conv2D-88       [[1, 576, 15, 15]]    [1, 96, 15, 15]        55,296     
  BatchNorm2D-88     [[1, 96, 15, 15]]     [1, 96, 15, 15]          384      
InvertedResidual-29  [[1, 96, 15, 15]]     [1, 96, 15, 15]           0       
     Conv2D-89       [[1, 96, 15, 15]]     [1, 576, 15, 15]       55,296     
  BatchNorm2D-89     [[1, 576, 15, 15]]    [1, 576, 15, 15]        2,304     
     ReLU6-60        [[1, 576, 15, 15]]    [1, 576, 15, 15]          0       
     Conv2D-90       [[1, 576, 15, 15]]    [1, 576, 15, 15]        5,184     
  BatchNorm2D-90     [[1, 576, 15, 15]]    [1, 576, 15, 15]        2,304     
     ReLU6-61        [[1, 576, 15, 15]]    [1, 576, 15, 15]          0       
     Conv2D-91       [[1, 576, 15, 15]]    [1, 96, 15, 15]        55,296     
  BatchNorm2D-91     [[1, 96, 15, 15]]     [1, 96, 15, 15]          384      
InvertedResidual-30  [[1, 96, 15, 15]]     [1, 96, 15, 15]           0       
     Conv2D-92       [[1, 96, 15, 15]]     [1, 576, 15, 15]       55,296     
  BatchNorm2D-92     [[1, 576, 15, 15]]    [1, 576, 15, 15]        2,304     
     ReLU6-62        [[1, 576, 15, 15]]    [1, 576, 15, 15]          0       
     Conv2D-93       [[1, 576, 15, 15]]     [1, 576, 8, 8]         5,184     
  BatchNorm2D-93      [[1, 576, 8, 8]]      [1, 576, 8, 8]         2,304     
     ReLU6-63         [[1, 576, 8, 8]]      [1, 576, 8, 8]           0       
     Conv2D-94        [[1, 576, 8, 8]]      [1, 160, 8, 8]        92,160     
  BatchNorm2D-94      [[1, 160, 8, 8]]      [1, 160, 8, 8]          640      
InvertedResidual-31  [[1, 96, 15, 15]]      [1, 160, 8, 8]           0       
     Conv2D-95        [[1, 160, 8, 8]]      [1, 960, 8, 8]        153,600    
  BatchNorm2D-95      [[1, 960, 8, 8]]      [1, 960, 8, 8]         3,840     
     ReLU6-64         [[1, 960, 8, 8]]      [1, 960, 8, 8]           0       
     Conv2D-96        [[1, 960, 8, 8]]      [1, 960, 8, 8]         8,640     
  BatchNorm2D-96      [[1, 960, 8, 8]]      [1, 960, 8, 8]         3,840     
     ReLU6-65         [[1, 960, 8, 8]]      [1, 960, 8, 8]           0       
     Conv2D-97        [[1, 960, 8, 8]]      [1, 160, 8, 8]        153,600    
  BatchNorm2D-97      [[1, 160, 8, 8]]      [1, 160, 8, 8]          640      
InvertedResidual-32   [[1, 160, 8, 8]]      [1, 160, 8, 8]           0       
     Conv2D-98        [[1, 160, 8, 8]]      [1, 960, 8, 8]        153,600    
  BatchNorm2D-98      [[1, 960, 8, 8]]      [1, 960, 8, 8]         3,840     
     ReLU6-66         [[1, 960, 8, 8]]      [1, 960, 8, 8]           0       
     Conv2D-99        [[1, 960, 8, 8]]      [1, 960, 8, 8]         8,640     
  BatchNorm2D-99      [[1, 960, 8, 8]]      [1, 960, 8, 8]         3,840     
     ReLU6-67         [[1, 960, 8, 8]]      [1, 960, 8, 8]           0       
    Conv2D-100        [[1, 960, 8, 8]]      [1, 160, 8, 8]        153,600    
  BatchNorm2D-100     [[1, 160, 8, 8]]      [1, 160, 8, 8]          640      
InvertedResidual-33   [[1, 160, 8, 8]]      [1, 160, 8, 8]           0       
    Conv2D-101        [[1, 160, 8, 8]]      [1, 960, 8, 8]        153,600    
  BatchNorm2D-101     [[1, 960, 8, 8]]      [1, 960, 8, 8]         3,840     
     ReLU6-68         [[1, 960, 8, 8]]      [1, 960, 8, 8]           0       
    Conv2D-102        [[1, 960, 8, 8]]      [1, 960, 8, 8]         8,640     
  BatchNorm2D-102     [[1, 960, 8, 8]]      [1, 960, 8, 8]         3,840     
     ReLU6-69         [[1, 960, 8, 8]]      [1, 960, 8, 8]           0       
    Conv2D-103        [[1, 960, 8, 8]]      [1, 320, 8, 8]        307,200    
  BatchNorm2D-103     [[1, 320, 8, 8]]      [1, 320, 8, 8]         1,280     
InvertedResidual-34   [[1, 160, 8, 8]]      [1, 320, 8, 8]           0       
    Conv2D-104        [[1, 320, 8, 8]]     [1, 1280, 8, 8]        409,600    
  BatchNorm2D-104    [[1, 1280, 8, 8]]     [1, 1280, 8, 8]         5,120     
     ReLU6-70        [[1, 1280, 8, 8]]     [1, 1280, 8, 8]           0       
AdaptiveAvgPool2D-2  [[1, 1280, 8, 8]]     [1, 1280, 1, 1]           0       
     Dropout-2          [[1, 1280]]           [1, 1280]              0       
     Linear-2           [[1, 1280]]            [1, 12]            15,372     
===============================================================================
Total params: 2,273,356
Trainable params: 2,205,132
Non-trainable params: 68,224
-------------------------------------------------------------------------------
Input size (MB): 0.59
Forward/backward pass size (MB): 165.07
Params size (MB): 8.67
Estimated Total Size (MB): 174.34
-------------------------------------------------------------------------------

C:\Users\Administrator\anaconda3\lib\site-packages\paddle\nn\layer\norm.py:640: UserWarning: When training, we now always track global mean and variance.
  warnings.warn(
2021-08-18 20:38:47,444 - INFO: Epoch:1/10, batch:10, train_loss:[2.35868], acc_top1:[0.18750], acc_top5:[0.65625](19.73s)
2021-08-18 20:39:08,187 - INFO: Epoch:1/10, batch:20, train_loss:[1.53091], acc_top1:[0.53125], acc_top5:[0.92188](20.74s)
2021-08-18 20:39:28,241 - INFO: Epoch:1/10, batch:30, train_loss:[1.24399], acc_top1:[0.60938], acc_top5:[0.92188](20.05s)
2021-08-18 20:39:49,964 - INFO: Epoch:1/10, batch:40, train_loss:[1.11863], acc_top1:[0.60938], acc_top5:[0.96875](21.72s)
2021-08-18 20:40:12,133 - INFO: Epoch:1/10, batch:50, train_loss:[0.84734], acc_top1:[0.76562], acc_top5:[0.95312](22.17s)
2021-08-18 20:40:34,673 - INFO: Epoch:1/10, batch:60, train_loss:[0.77916], acc_top1:[0.76562], acc_top5:[0.95312](22.54s)
2021-08-18 20:40:54,459 - INFO: Epoch:1/10, batch:70, train_loss:[0.63126], acc_top1:[0.82812], acc_top5:[1.00000](19.79s)
2021-08-18 20:41:14,705 - INFO: Epoch:1/10, batch:80, train_loss:[0.51194], acc_top1:[0.81250], acc_top5:[1.00000](20.25s)
2021-08-18 20:41:34,630 - INFO: Epoch:1/10, batch:90, train_loss:[0.54417], acc_top1:[0.84375], acc_top5:[1.00000](19.92s)
2021-08-18 20:41:52,263 - INFO: Epoch:1/10, batch:100, train_loss:[0.54743], acc_top1:[0.81250], acc_top5:[0.98438](17.63s)
2021-08-18 20:42:10,089 - INFO: Epoch:1/10, batch:110, train_loss:[0.36209], acc_top1:[0.92188], acc_top5:[1.00000](17.83s)
2021-08-18 20:42:27,441 - INFO: [validation] Epoch:1/10, val_loss:[0.00699], val_top1:[0.86615], val_top5:[0.99538]
C:\Users\Administrator\anaconda3\lib\site-packages\paddle\fluid\layers\utils.py:77: DeprecationWarning: Using or importing the ABCs from 'collections' instead of from 'collections.abc' is deprecated since Python 3.3, and in 3.9 it will stop working
  return (isinstance(seq, collections.Sequence) and
C:\Users\Administrator\anaconda3\lib\site-packages\paddle\fluid\dygraph\dygraph_to_static\convert_call_func.py:89: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray
  func_in_dict = func == v
C:\Users\Administrator\anaconda3\lib\site-packages\paddle\fluid\layers\math_op_patch.py:317: UserWarning: C:\Users\Administrator\anaconda3\lib\site-packages\paddle\vision\models\mobilenetv2.py:99
The behavior of expression A + B has been unified with elementwise_add(X, Y, axis=-1) from Paddle 2.0. If your code works well in the older versions but crashes in this version, try to use elementwise_add(X, Y, axis=0) instead of A + B. This transitional warning will be dropped in the future.
  warnings.warn(
2021-08-18 20:42:28,376 - INFO: 已保存当前测试模型(epoch=1)为最优模型:Zodiac_Mobilenetv2_final
2021-08-18 20:42:28,377 - INFO: 最优top1测试精度:0.86615 (epoch=1)
2021-08-18 20:42:35,218 - INFO: Epoch:2/10, batch:120, train_loss:[0.32727], acc_top1:[0.89062], acc_top5:[1.00000](25.13s)
2021-08-18 20:42:44,745 - INFO: Epoch:2/10, batch:130, train_loss:[0.33888], acc_top1:[0.89062], acc_top5:[0.98438](9.53s)
2021-08-18 20:42:54,753 - INFO: Epoch:2/10, batch:140, train_loss:[0.24190], acc_top1:[0.93750], acc_top5:[0.98438](10.01s)
2021-08-18 20:43:04,487 - INFO: Epoch:2/10, batch:150, train_loss:[0.39876], acc_top1:[0.90625], acc_top5:[1.00000](9.73s)
2021-08-18 20:43:14,891 - INFO: Epoch:2/10, batch:160, train_loss:[0.38700], acc_top1:[0.90625], acc_top5:[0.98438](10.40s)
2021-08-18 20:43:25,636 - INFO: Epoch:2/10, batch:170, train_loss:[0.31630], acc_top1:[0.89062], acc_top5:[0.98438](10.74s)
2021-08-18 20:43:35,091 - INFO: Epoch:2/10, batch:180, train_loss:[0.36716], acc_top1:[0.87500], acc_top5:[1.00000](9.46s)
2021-08-18 20:43:44,375 - INFO: Epoch:2/10, batch:190, train_loss:[0.32833], acc_top1:[0.90625], acc_top5:[0.98438](9.28s)
2021-08-18 20:43:54,686 - INFO: Epoch:2/10, batch:200, train_loss:[0.38044], acc_top1:[0.87500], acc_top5:[0.98438](10.31s)
2021-08-18 20:44:03,645 - INFO: Epoch:2/10, batch:210, train_loss:[0.28205], acc_top1:[0.90625], acc_top5:[0.96875](8.96s)
2021-08-18 20:44:13,641 - INFO: Epoch:2/10, batch:220, train_loss:[0.17315], acc_top1:[0.96875], acc_top5:[1.00000](10.00s)
2021-08-18 20:44:27,035 - INFO: [validation] Epoch:2/10, val_loss:[0.00529], val_top1:[0.89846], val_top5:[0.99385]
2021-08-18 20:44:27,699 - INFO: 已保存当前测试模型(epoch=2)为最优模型:Zodiac_Mobilenetv2_final
2021-08-18 20:44:27,700 - INFO: 最优top1测试精度:0.89846 (epoch=2)
2021-08-18 20:44:33,548 - INFO: Epoch:3/10, batch:230, train_loss:[0.23524], acc_top1:[0.92188], acc_top5:[1.00000](19.91s)
2021-08-18 20:44:43,755 - INFO: Epoch:3/10, batch:240, train_loss:[0.13375], acc_top1:[0.96875], acc_top5:[1.00000](10.21s)
2021-08-18 20:44:53,204 - INFO: Epoch:3/10, batch:250, train_loss:[0.24205], acc_top1:[0.90625], acc_top5:[1.00000](9.45s)
2021-08-18 20:45:02,680 - INFO: Epoch:3/10, batch:260, train_loss:[0.30691], acc_top1:[0.93750], acc_top5:[1.00000](9.48s)
2021-08-18 20:45:11,859 - INFO: Epoch:3/10, batch:270, train_loss:[0.23953], acc_top1:[0.95312], acc_top5:[1.00000](9.18s)
2021-08-18 20:45:20,276 - INFO: Epoch:3/10, batch:280, train_loss:[0.26566], acc_top1:[0.90625], acc_top5:[1.00000](8.42s)
2021-08-18 20:45:29,860 - INFO: Epoch:3/10, batch:290, train_loss:[0.21597], acc_top1:[0.95312], acc_top5:[1.00000](9.58s)
2021-08-18 20:45:39,252 - INFO: Epoch:3/10, batch:300, train_loss:[0.34542], acc_top1:[0.92188], acc_top5:[0.98438](9.39s)
2021-08-18 20:45:49,498 - INFO: Epoch:3/10, batch:310, train_loss:[0.16125], acc_top1:[0.98438], acc_top5:[1.00000](10.25s)
2021-08-18 20:45:59,449 - INFO: Epoch:3/10, batch:320, train_loss:[0.18767], acc_top1:[0.95312], acc_top5:[1.00000](9.95s)
2021-08-18 20:46:09,352 - INFO: Epoch:3/10, batch:330, train_loss:[0.32362], acc_top1:[0.93750], acc_top5:[0.98438](9.90s)
2021-08-18 20:46:25,174 - INFO: [validation] Epoch:3/10, val_loss:[0.00449], val_top1:[0.90462], val_top5:[0.99846]
2021-08-18 20:46:25,848 - INFO: 已保存当前测试模型(epoch=3)为最优模型:Zodiac_Mobilenetv2_final
2021-08-18 20:46:25,849 - INFO: 最优top1测试精度:0.90462 (epoch=3)
2021-08-18 20:46:29,669 - INFO: Epoch:4/10, batch:340, train_loss:[0.15083], acc_top1:[0.95312], acc_top5:[1.00000](20.32s)
2021-08-18 20:46:39,399 - INFO: Epoch:4/10, batch:350, train_loss:[0.07786], acc_top1:[0.98438], acc_top5:[1.00000](9.73s)
2021-08-18 20:46:49,194 - INFO: Epoch:4/10, batch:360, train_loss:[0.18050], acc_top1:[0.95312], acc_top5:[1.00000](9.79s)
2021-08-18 20:46:59,333 - INFO: Epoch:4/10, batch:370, train_loss:[0.15963], acc_top1:[0.95312], acc_top5:[1.00000](10.14s)
2021-08-18 20:47:08,950 - INFO: Epoch:4/10, batch:380, train_loss:[0.13318], acc_top1:[0.96875], acc_top5:[1.00000](9.62s)
2021-08-18 20:47:18,739 - INFO: Epoch:4/10, batch:390, train_loss:[0.14684], acc_top1:[0.93750], acc_top5:[1.00000](9.79s)
2021-08-18 20:47:29,050 - INFO: Epoch:4/10, batch:400, train_loss:[0.17527], acc_top1:[0.96875], acc_top5:[0.98438](10.31s)
2021-08-18 20:47:38,745 - INFO: Epoch:4/10, batch:410, train_loss:[0.10365], acc_top1:[0.96875], acc_top5:[1.00000](9.69s)
2021-08-18 20:47:48,626 - INFO: Epoch:4/10, batch:420, train_loss:[0.20553], acc_top1:[0.93750], acc_top5:[1.00000](9.88s)
2021-08-18 20:47:56,370 - INFO: Epoch:4/10, batch:430, train_loss:[0.06721], acc_top1:[1.00000], acc_top5:[1.00000](7.74s)
2021-08-18 20:48:06,305 - INFO: Epoch:4/10, batch:440, train_loss:[0.14027], acc_top1:[0.95312], acc_top5:[1.00000](9.93s)
2021-08-18 20:48:23,898 - INFO: [validation] Epoch:4/10, val_loss:[0.00420], val_top1:[0.91846], val_top5:[1.00000]
2021-08-18 20:48:24,665 - INFO: 已保存当前测试模型(epoch=4)为最优模型:Zodiac_Mobilenetv2_final
2021-08-18 20:48:24,666 - INFO: 最优top1测试精度:0.91846 (epoch=4)
2021-08-18 20:48:26,821 - INFO: Epoch:5/10, batch:450, train_loss:[0.09366], acc_top1:[0.98438], acc_top5:[1.00000](20.52s)
2021-08-18 20:48:36,958 - INFO: Epoch:5/10, batch:460, train_loss:[0.09438], acc_top1:[0.98438], acc_top5:[1.00000](10.14s)
2021-08-18 20:48:47,249 - INFO: Epoch:5/10, batch:470, train_loss:[0.19238], acc_top1:[0.96875], acc_top5:[0.98438](10.29s)
2021-08-18 20:48:57,046 - INFO: Epoch:5/10, batch:480, train_loss:[0.09879], acc_top1:[0.96875], acc_top5:[1.00000](9.80s)
2021-08-18 20:49:07,041 - INFO: Epoch:5/10, batch:490, train_loss:[0.06695], acc_top1:[0.98438], acc_top5:[1.00000](9.99s)
2021-08-18 20:49:15,346 - INFO: Epoch:5/10, batch:500, train_loss:[0.07161], acc_top1:[0.96875], acc_top5:[1.00000](8.31s)
2021-08-18 20:49:24,664 - INFO: Epoch:5/10, batch:510, train_loss:[0.09566], acc_top1:[0.98438], acc_top5:[1.00000](9.32s)
2021-08-18 20:49:33,218 - INFO: Epoch:5/10, batch:520, train_loss:[0.06367], acc_top1:[1.00000], acc_top5:[1.00000](8.55s)
2021-08-18 20:49:41,788 - INFO: Epoch:5/10, batch:530, train_loss:[0.07820], acc_top1:[1.00000], acc_top5:[1.00000](8.57s)
2021-08-18 20:49:51,577 - INFO: Epoch:5/10, batch:540, train_loss:[0.10651], acc_top1:[0.96875], acc_top5:[1.00000](9.79s)
2021-08-18 20:50:01,029 - INFO: Epoch:5/10, batch:550, train_loss:[0.12549], acc_top1:[0.96875], acc_top5:[1.00000](9.45s)
2021-08-18 20:50:11,980 - INFO: Epoch:5/10, batch:560, train_loss:[0.07814], acc_top1:[0.98438], acc_top5:[1.00000](10.95s)
2021-08-18 20:50:21,376 - INFO: [validation] Epoch:5/10, val_loss:[0.00400], val_top1:[0.92615], val_top5:[0.99846]
2021-08-18 20:50:22,029 - INFO: 已保存当前测试模型(epoch=5)为最优模型:Zodiac_Mobilenetv2_final
2021-08-18 20:50:22,029 - INFO: 最优top1测试精度:0.92615 (epoch=5)
2021-08-18 20:50:32,246 - INFO: Epoch:6/10, batch:570, train_loss:[0.06100], acc_top1:[1.00000], acc_top5:[1.00000](20.27s)
2021-08-18 20:50:40,916 - INFO: Epoch:6/10, batch:580, train_loss:[0.07342], acc_top1:[0.98438], acc_top5:[1.00000](8.67s)
2021-08-18 20:50:49,712 - INFO: Epoch:6/10, batch:590, train_loss:[0.07284], acc_top1:[1.00000], acc_top5:[1.00000](8.80s)
2021-08-18 20:50:59,671 - INFO: Epoch:6/10, batch:600, train_loss:[0.14909], acc_top1:[0.95312], acc_top5:[0.98438](9.96s)
2021-08-18 20:51:08,865 - INFO: Epoch:6/10, batch:610, train_loss:[0.06966], acc_top1:[0.98438], acc_top5:[1.00000](9.19s)
2021-08-18 20:51:17,472 - INFO: Epoch:6/10, batch:620, train_loss:[0.06740], acc_top1:[0.98438], acc_top5:[1.00000](8.61s)
2021-08-18 20:51:26,750 - INFO: Epoch:6/10, batch:630, train_loss:[0.10529], acc_top1:[0.98438], acc_top5:[0.98438](9.28s)
2021-08-18 20:51:36,055 - INFO: Epoch:6/10, batch:640, train_loss:[0.05156], acc_top1:[1.00000], acc_top5:[1.00000](9.30s)
2021-08-18 20:51:45,744 - INFO: Epoch:6/10, batch:650, train_loss:[0.08652], acc_top1:[0.98438], acc_top5:[1.00000](9.69s)
2021-08-18 20:51:56,040 - INFO: Epoch:6/10, batch:660, train_loss:[0.04166], acc_top1:[1.00000], acc_top5:[1.00000](10.30s)
2021-08-18 20:52:07,168 - INFO: Epoch:6/10, batch:670, train_loss:[0.04415], acc_top1:[1.00000], acc_top5:[1.00000](11.13s)
2021-08-18 20:52:18,479 - INFO: [validation] Epoch:6/10, val_loss:[0.00400], val_top1:[0.91846], val_top5:[1.00000]
2021-08-18 20:52:18,480 - INFO: 最优top1测试精度:0.92615 (epoch=5)
2021-08-18 20:52:25,105 - INFO: Epoch:7/10, batch:680, train_loss:[0.03270], acc_top1:[1.00000], acc_top5:[1.00000](17.94s)
2021-08-18 20:52:34,802 - INFO: Epoch:7/10, batch:690, train_loss:[0.04794], acc_top1:[1.00000], acc_top5:[1.00000](9.70s)
2021-08-18 20:52:45,562 - INFO: Epoch:7/10, batch:700, train_loss:[0.04604], acc_top1:[1.00000], acc_top5:[1.00000](10.76s)
2021-08-18 20:52:54,632 - INFO: Epoch:7/10, batch:710, train_loss:[0.04353], acc_top1:[1.00000], acc_top5:[1.00000](9.07s)
2021-08-18 20:53:03,736 - INFO: Epoch:7/10, batch:720, train_loss:[0.05784], acc_top1:[1.00000], acc_top5:[1.00000](9.10s)
2021-08-18 20:53:12,762 - INFO: Epoch:7/10, batch:730, train_loss:[0.10201], acc_top1:[0.98438], acc_top5:[1.00000](9.03s)
2021-08-18 20:53:21,900 - INFO: Epoch:7/10, batch:740, train_loss:[0.11172], acc_top1:[0.96875], acc_top5:[1.00000](9.14s)
2021-08-18 20:53:32,386 - INFO: Epoch:7/10, batch:750, train_loss:[0.03383], acc_top1:[1.00000], acc_top5:[1.00000](10.49s)
2021-08-18 20:53:41,702 - INFO: Epoch:7/10, batch:760, train_loss:[0.05390], acc_top1:[1.00000], acc_top5:[1.00000](9.32s)
2021-08-18 20:53:52,079 - INFO: Epoch:7/10, batch:770, train_loss:[0.04625], acc_top1:[0.98438], acc_top5:[1.00000](10.38s)
2021-08-18 20:54:00,986 - INFO: Epoch:7/10, batch:780, train_loss:[0.04487], acc_top1:[1.00000], acc_top5:[1.00000](8.91s)
2021-08-18 20:54:14,524 - INFO: [validation] Epoch:7/10, val_loss:[0.00421], val_top1:[0.91692], val_top5:[1.00000]
2021-08-18 20:54:14,525 - INFO: 最优top1测试精度:0.92615 (epoch=5)
2021-08-18 20:54:20,247 - INFO: Epoch:8/10, batch:790, train_loss:[0.02338], acc_top1:[1.00000], acc_top5:[1.00000](19.26s)
2021-08-18 20:54:30,393 - INFO: Epoch:8/10, batch:800, train_loss:[0.03130], acc_top1:[1.00000], acc_top5:[1.00000](10.15s)
2021-08-18 20:54:38,556 - INFO: Epoch:8/10, batch:810, train_loss:[0.02492], acc_top1:[1.00000], acc_top5:[1.00000](8.16s)
2021-08-18 20:54:47,180 - INFO: Epoch:8/10, batch:820, train_loss:[0.02855], acc_top1:[1.00000], acc_top5:[1.00000](8.62s)
2021-08-18 20:54:59,162 - INFO: Epoch:8/10, batch:830, train_loss:[0.04445], acc_top1:[1.00000], acc_top5:[1.00000](11.98s)
2021-08-18 20:55:09,225 - INFO: Epoch:8/10, batch:840, train_loss:[0.02702], acc_top1:[1.00000], acc_top5:[1.00000](10.06s)
2021-08-18 20:55:18,196 - INFO: Epoch:8/10, batch:850, train_loss:[0.04049], acc_top1:[1.00000], acc_top5:[1.00000](8.97s)
2021-08-18 20:55:27,605 - INFO: Epoch:8/10, batch:860, train_loss:[0.02455], acc_top1:[1.00000], acc_top5:[1.00000](9.41s)
2021-08-18 20:55:37,179 - INFO: Epoch:8/10, batch:870, train_loss:[0.02132], acc_top1:[1.00000], acc_top5:[1.00000](9.57s)
2021-08-18 20:55:46,563 - INFO: Epoch:8/10, batch:880, train_loss:[0.04612], acc_top1:[0.98438], acc_top5:[1.00000](9.38s)
2021-08-18 20:55:55,424 - INFO: Epoch:8/10, batch:890, train_loss:[0.03026], acc_top1:[1.00000], acc_top5:[1.00000](8.86s)
2021-08-18 20:56:10,626 - INFO: [validation] Epoch:8/10, val_loss:[0.00418], val_top1:[0.91538], val_top5:[0.99846]
2021-08-18 20:56:10,626 - INFO: 最优top1测试精度:0.92615 (epoch=5)
2021-08-18 20:56:13,795 - INFO: Epoch:9/10, batch:900, train_loss:[0.01744], acc_top1:[1.00000], acc_top5:[1.00000](18.37s)
2021-08-18 20:56:22,685 - INFO: Epoch:9/10, batch:910, train_loss:[0.01865], acc_top1:[1.00000], acc_top5:[1.00000](8.89s)
2021-08-18 20:56:32,232 - INFO: Epoch:9/10, batch:920, train_loss:[0.01607], acc_top1:[1.00000], acc_top5:[1.00000](9.55s)
2021-08-18 20:56:40,905 - INFO: Epoch:9/10, batch:930, train_loss:[0.03191], acc_top1:[0.98438], acc_top5:[1.00000](8.67s)
2021-08-18 20:56:49,723 - INFO: Epoch:9/10, batch:940, train_loss:[0.02905], acc_top1:[1.00000], acc_top5:[1.00000](8.82s)
2021-08-18 20:56:58,554 - INFO: Epoch:9/10, batch:950, train_loss:[0.01862], acc_top1:[1.00000], acc_top5:[1.00000](8.83s)
2021-08-18 20:57:08,666 - INFO: Epoch:9/10, batch:960, train_loss:[0.04437], acc_top1:[0.98438], acc_top5:[1.00000](10.11s)
2021-08-18 20:57:19,214 - INFO: Epoch:9/10, batch:970, train_loss:[0.00877], acc_top1:[1.00000], acc_top5:[1.00000](10.55s)
2021-08-18 20:57:29,104 - INFO: Epoch:9/10, batch:980, train_loss:[0.02090], acc_top1:[1.00000], acc_top5:[1.00000](9.89s)
2021-08-18 20:57:40,243 - INFO: Epoch:9/10, batch:990, train_loss:[0.07159], acc_top1:[0.96875], acc_top5:[1.00000](11.14s)
2021-08-18 20:57:49,314 - INFO: Epoch:9/10, batch:1000, train_loss:[0.02380], acc_top1:[1.00000], acc_top5:[1.00000](9.07s)
2021-08-18 20:58:05,972 - INFO: [validation] Epoch:9/10, val_loss:[0.00425], val_top1:[0.92769], val_top5:[0.99692]
2021-08-18 20:58:06,635 - INFO: 已保存当前测试模型(epoch=9)为最优模型:Zodiac_Mobilenetv2_final
2021-08-18 20:58:06,636 - INFO: 最优top1测试精度:0.92769 (epoch=9)
2021-08-18 20:58:08,393 - INFO: Epoch:10/10, batch:1010, train_loss:[0.01303], acc_top1:[1.00000], acc_top5:[1.00000](19.08s)
2021-08-18 20:58:17,196 - INFO: Epoch:10/10, batch:1020, train_loss:[0.01460], acc_top1:[1.00000], acc_top5:[1.00000](8.80s)
2021-08-18 20:58:27,992 - INFO: Epoch:10/10, batch:1030, train_loss:[0.02274], acc_top1:[1.00000], acc_top5:[1.00000](10.80s)
2021-08-18 20:58:38,393 - INFO: Epoch:10/10, batch:1040, train_loss:[0.01915], acc_top1:[1.00000], acc_top5:[1.00000](10.40s)
2021-08-18 20:58:48,539 - INFO: Epoch:10/10, batch:1050, train_loss:[0.00959], acc_top1:[1.00000], acc_top5:[1.00000](10.15s)
2021-08-18 20:58:57,457 - INFO: Epoch:10/10, batch:1060, train_loss:[0.01160], acc_top1:[1.00000], acc_top5:[1.00000](8.92s)
2021-08-18 20:59:06,237 - INFO: Epoch:10/10, batch:1070, train_loss:[0.01434], acc_top1:[1.00000], acc_top5:[1.00000](8.78s)
2021-08-18 20:59:14,633 - INFO: Epoch:10/10, batch:1080, train_loss:[0.02018], acc_top1:[1.00000], acc_top5:[1.00000](8.40s)
2021-08-18 20:59:24,413 - INFO: Epoch:10/10, batch:1090, train_loss:[0.01969], acc_top1:[1.00000], acc_top5:[1.00000](9.78s)
2021-08-18 20:59:33,290 - INFO: Epoch:10/10, batch:1100, train_loss:[0.01351], acc_top1:[1.00000], acc_top5:[1.00000](8.88s)
2021-08-18 20:59:42,958 - INFO: Epoch:10/10, batch:1110, train_loss:[0.02126], acc_top1:[1.00000], acc_top5:[1.00000](9.67s)
2021-08-18 20:59:53,134 - INFO: Epoch:10/10, batch:1120, train_loss:[0.02153], acc_top1:[1.00000], acc_top5:[1.00000](10.18s)
2021-08-18 21:00:02,480 - INFO: [validation] Epoch:10/10, val_loss:[0.00408], val_top1:[0.92769], val_top5:[1.00000]
2021-08-18 21:00:02,481 - INFO: 最优top1测试精度:0.92769 (epoch=9)
2021-08-18 21:00:02,481 - INFO: 训练完成，最终性能accuracy=0.92769(epoch=9), 总耗时1294.77s, 已将其保存为：Zodiac_Mobilenetv2_final
2021-08-18 21:00:02,482 - INFO: 训练完毕，结果路径D:\Workspace\ExpResults\HW06TransferLearning\Zodiac_Mobilenetv2.
2021-08-18 21:00:02,482 - INFO: Done.


if __name__ == '__main__':
    # 设置输入样本的维度    
    input_spec = InputSpec(shape=[None] + args['input_size'], dtype='float32', name='image')
    label_spec = InputSpec(shape=[None, 1], dtype='int64', name='label')
    
    # 载入模型
    network = paddle.vision.models.mobilenet_v2(num_classes=args['class_dim'])
    model = paddle.Model(network, input_spec, label_spec)                   # 模型实例化
    model.load(deployment_checkpoint_model)                                # 载入调优模型的参数
    model.prepare(loss=paddle.nn.CrossEntropyLoss(),                        # 设置loss
                  metrics=paddle.metric.Accuracy(topk=(1,5)))               # 设置评价指标
    
    # 执行评估函数，并输出验证集样本的损失和精度
    print('开始评估...')
    avg_loss, avg_acc_top1, avg_acc_top5 = eval(model, val_reader(), verbose=1)
    print('\r [验证集] 损失: {:.5f}, top1精度:{:.5f}, top5精度为:{:.5f} \n'.format(avg_loss, avg_acc_top1, avg_acc_top5), end='')    
    avg_loss, avg_acc_top1, avg_acc_top5 = eval(model, test_reader(), verbose=1)
    print('\r [测试集] 损失: {:.5f}, top1精度:{:.5f}, top5精度为:{:.5f}'.format(avg_loss, avg_acc_top1, avg_acc_top5), end='')

开始评估...
 [验证集] 损失: 0.00425, top1精度:0.92769, top5精度为:0.99692 
 [测试集] 损失: 0.00516, top1精度:0.91212, top5精度为:0.99545]

模型名称	Baseline模型	ImageNet预训练	learning_rate	best_epoch	top-1 acc	top-5 acc	loss	test_top1_acc	单batch时间/总训练时间(s)	可训练参数/总参数
Zodiac_Alexnet	Alexnet	否	0.001	3/10	0.48769	0.87231	0.02458	0.48485	9.06/1177.53	11,172,042/11,191,242
Zodiac_Resnet18	ResNet18	是
Zodiac_Resnet18_withoutPretrained	ResNet18	否	0.01	50/50	0.60769	0.90769	0.02072	0.60909	0.26/374.98	11,172,042/11,191,242
Zodiac_Resnet50	ResNet50	是
Zodiac_Resnet50_withoutPretrained	ResNet50	否	0.01	45/50	0.54154	0.89538	0.04756	0.55152	11.24/7868.96	23,479,500/23,585,740
Zodiac_VGG16	VGG16	是
Zodiac_VGG16_withoutPretrained	VGG16	否	0.001	45/50	0.55231	0.90000	0.03685	0.86429	14.75/8727.87	134,309,708/134,309,708
Zodiac_Mobilenetv2	Mobilenetv2	是
Zodiac_Mobilenetv2_withoutPretrained	Mobilenetv2	否	0.001	44/50	0.52615	0.87077	0.03536	0.51667	10.44/5838.82s	2,205,132/2,273,356

模型名称	Baseline模型	ImageNet预训练	learning_rate	best_epoch	top-1 acc	top-5 acc	loss	test_top1_acc	单batch时间/总训练时间(s)	可训练参数/总参数
Zodiac_Alexnet	Alexnet	否	0.001	3/10	0.48769	0.87231	0.02458	0.48485	9.06/1177.53	11,172,042/11,191,242
Zodiac_Resnet18	ResNet18	是	0.01	7/10	0.83385	0.97538	0.01093	0.81364	10.20/1389.07	11,172,042/11,191,242
Zodiac_Resnet18_withoutPretrained	ResNet18	否	0.01	50/50	0.60769	0.90769	0.02072	0.60909	12.28/7220.98	11,172,042/11,191,242
Zodiac_Resnet50	ResNet50	是	0.01	8/10	0.94615	0.99846	0.00343	0.94545	11.67/1455.60	23,479,500/23,585,740
Zodiac_Resnet50_withoutPretrained	ResNet50	否	0.01	45/50	0.54154	0.89538	0.04756	0.55152	11.24/7868.96	23,479,500/23,585,740
Zodiac_VGG16	VGG16	是	0.001	8/10	0.89385	0.98615	0.00581	0.86515	16.69/2243.58	134,309,708/134,309,708
Zodiac_VGG16_withoutPretrained	VGG16	否	0.001	45/50	0.55231	0.90000	0.03685	0.86429	14.75/8727.87	134,309,708/134,309,708
Zodiac_Mobilenetv2	Mobilenetv2	是	0.001	9/10	0.92769	0.99692	0.00425	0.91212	10.18/1294.77	2,205,132/2,273,356
Zodiac_Mobilenetv2_withoutPretrained	Mobilenetv2	否	0.001	44/50	0.52615	0.87077	0.03536	0.51667	10.44/5838.82s	2,205,132/2,273,356

【课后作业06】迁移学习和恢复训练（十二生肖） `习题答案`¶

【实验目的】¶

【实验要求】¶

1. 实验结果¶

2. 实验结果分析¶

【参考代码】¶

【实验一】数据集准备¶

1.0 处理数据集中样本命名的非法字符¶

1.1 生产图像列表及类别标签¶

【实验二】全局参数设置及数据基本处理¶

2.1 导入依赖及全局参数配置¶

2.2 数据集定义及数据预处理¶

2.3 设置数据提供器¶

2.4 定义过程可视化函数¶

2.5 定义日志输出函数¶

【实验三】模型训练与评估¶

3.1 配置网络¶

3.2 定义优化方法¶

3.3 定义验证函数¶

3.4 模型训练及在线测试¶

3.5 离线测试¶

【实验结果及评价】¶

1. 实验结果¶

2. 实验结果分析¶

【课后作业06】 迁移学习和恢复训练（十二生肖） 习题答案¶

【实验目的】¶

【实验要求】¶

1. 实验结果¶

2. 实验结果分析¶

【参考代码】¶

【实验一】 数据集准备¶

1.0 处理数据集中样本命名的非法字符¶

1.1 生产图像列表及类别标签¶

【实验二】 全局参数设置及数据基本处理¶

2.1 导入依赖及全局参数配置¶

2.2 数据集定义及数据预处理¶

2.3 设置数据提供器¶

2.4 定义过程可视化函数¶

2.5 定义日志输出函数¶

【实验三】 模型训练与评估¶

3.1 配置网络¶

3.2 定义优化方法¶

3.3 定义验证函数¶

3.4 模型训练及在线测试¶

3.5 离线测试¶

【实验结果及评价】¶

1. 实验结果¶

2. 实验结果分析¶

【课后作业06】迁移学习和恢复训练（十二生肖） `习题答案`¶

【实验一】数据集准备¶

【实验二】全局参数设置及数据基本处理¶

【实验三】模型训练与评估¶