TensorFlow2-实战-手写数字识别（二）：模型版【初始化参数】--＞【循环（①根据参数W、B通过模型前向计算，计算出输入X对应的输出Y；②计算Loss；③计算梯度；④利用梯度下降来更新参数）】

喜欢ヅ旅行 2023-09-29 10:20 3阅读 0赞

import os
    import pandas as pd
    import tensorflow as tf
    from tensorflow import keras
    from tensorflow.keras import layers, optimizers, datasets
    
    os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'
    
    # 读取数据集
    mnist_dataset = datasets.mnist.load_data()
    dataset_train, datasset_val = mnist_dataset[0], mnist_dataset[1]
    (x, y) = dataset_train
    (x_val, y_val) = datasset_val
    print("type(x) = {0}, type(y) = {1}".format(type(x), type(y)))
    
    # 将ndarray数组转为Tensor格式
    x = tf.convert_to_tensor(x, dtype=tf.float32) / 255.
    print("x.shape = {0}, x[0] = \n{1}".format(x.shape, pd.DataFrame(x[0].numpy())))  # x.shape =  (60000, 28, 28)
    y = tf.convert_to_tensor(y, dtype=tf.int32)
    print("y.shape = {0}, y = {1}".format(y.shape, y))  # y.shape = (60000,), y = [5 0 4 ... 5 6 8]
    
    # 将y转为one-hot编码
    y = tf.one_hot(y, depth=10)
    print("将y转为one-hot编码之后：y.shape = {0}, y = \n{1}".format(y.shape, y))  # 将y转为one-hot编码之后：y.shape = (60000, 10)
    
    # 从(x, y)中抽取训练数据集
    train_dataset = tf.data.Dataset.from_tensor_slices((x, y))
    train_dataset = train_dataset.batch(200)
    
    # 通过keras接口构建模型
    model = keras.Sequential([
        layers.Dense(512, activation='relu'),
        layers.Dense(256, activation='relu'),
        layers.Dense(10)])
    
    # 构建优化器
    optimizer = optimizers.SGD(learning_rate=0.001)
    
    
    # 训练一个epoch
    def train_epoch(epoch):
        # Step4.每个循环一个batch
        for batch_idx, (x, y) in enumerate(train_dataset):
            # Tensorflow使用梯度带（tf.GradientTape）来记录正向运算过程，然后反向传播自动得到梯度值。
            with tf.GradientTape() as tape:
                # 将特征值进行维度变换
                x = tf.reshape(x, (-1, 28 * 28))  # [b, 28, 28] => [b, 784]
                # 步骤一. 根据参数，通过模型前向计算，得到输入特征值对应的输出
                out = model(x)  # [b, 784] => [b, 10]
                # 步骤二. 计算Loss
                loss = tf.reduce_sum(tf.square(out - y)) / x.shape[0]
            # 步骤三. 计算模型中类型为Variable的参数的梯度
            grads = tape.gradient(loss, model.trainable_variables)
            # 步骤四. 利用梯度下降来更新参数【optimize and update w1, w2, w3, b1, b2, b3】【w' = w - lr * grad】
            optimizer.apply_gradients(zip(grads, model.trainable_variables))
    
            if batch_idx % 100 == 0:
                print("epoch = {0}, batch_idx = {1}, loss = {2}".format(epoch, batch_idx, loss.numpy()))
    
    
    def train():
        for epoch in range(30):
            train_epoch(epoch)
    
    
    if __name__ == '__main__':
        train()

打印结果：

type(x) = <class 'numpy.ndarray'>, type(y) = <class 'numpy.ndarray'>
    x.shape = (60000, 28, 28), x[0] = 
         0    1    2    3         4  ...         23   24   25   26   27
    0   0.0  0.0  0.0  0.0  0.000000 ...   0.000000  0.0  0.0  0.0  0.0
    1   0.0  0.0  0.0  0.0  0.000000 ...   0.000000  0.0  0.0  0.0  0.0
    2   0.0  0.0  0.0  0.0  0.000000 ...   0.000000  0.0  0.0  0.0  0.0
    3   0.0  0.0  0.0  0.0  0.000000 ...   0.000000  0.0  0.0  0.0  0.0
    4   0.0  0.0  0.0  0.0  0.000000 ...   0.000000  0.0  0.0  0.0  0.0
    5   0.0  0.0  0.0  0.0  0.000000 ...   0.498039  0.0  0.0  0.0  0.0
    6   0.0  0.0  0.0  0.0  0.000000 ...   0.250980  0.0  0.0  0.0  0.0
    7   0.0  0.0  0.0  0.0  0.000000 ...   0.000000  0.0  0.0  0.0  0.0
    8   0.0  0.0  0.0  0.0  0.000000 ...   0.000000  0.0  0.0  0.0  0.0
    9   0.0  0.0  0.0  0.0  0.000000 ...   0.000000  0.0  0.0  0.0  0.0
    10  0.0  0.0  0.0  0.0  0.000000 ...   0.000000  0.0  0.0  0.0  0.0
    11  0.0  0.0  0.0  0.0  0.000000 ...   0.000000  0.0  0.0  0.0  0.0
    12  0.0  0.0  0.0  0.0  0.000000 ...   0.000000  0.0  0.0  0.0  0.0
    13  0.0  0.0  0.0  0.0  0.000000 ...   0.000000  0.0  0.0  0.0  0.0
    14  0.0  0.0  0.0  0.0  0.000000 ...   0.000000  0.0  0.0  0.0  0.0
    15  0.0  0.0  0.0  0.0  0.000000 ...   0.000000  0.0  0.0  0.0  0.0
    16  0.0  0.0  0.0  0.0  0.000000 ...   0.000000  0.0  0.0  0.0  0.0
    17  0.0  0.0  0.0  0.0  0.000000 ...   0.000000  0.0  0.0  0.0  0.0
    18  0.0  0.0  0.0  0.0  0.000000 ...   0.000000  0.0  0.0  0.0  0.0
    19  0.0  0.0  0.0  0.0  0.000000 ...   0.000000  0.0  0.0  0.0  0.0
    20  0.0  0.0  0.0  0.0  0.000000 ...   0.000000  0.0  0.0  0.0  0.0
    21  0.0  0.0  0.0  0.0  0.000000 ...   0.000000  0.0  0.0  0.0  0.0
    22  0.0  0.0  0.0  0.0  0.000000 ...   0.000000  0.0  0.0  0.0  0.0
    23  0.0  0.0  0.0  0.0  0.215686 ...   0.000000  0.0  0.0  0.0  0.0
    24  0.0  0.0  0.0  0.0  0.533333 ...   0.000000  0.0  0.0  0.0  0.0
    25  0.0  0.0  0.0  0.0  0.000000 ...   0.000000  0.0  0.0  0.0  0.0
    26  0.0  0.0  0.0  0.0  0.000000 ...   0.000000  0.0  0.0  0.0  0.0
    27  0.0  0.0  0.0  0.0  0.000000 ...   0.000000  0.0  0.0  0.0  0.0
    
    [28 rows x 28 columns]
    y.shape = (60000,), y = [5 0 4 ... 5 6 8]
    将y转为one-hot编码之后：y.shape = (60000, 10), y = 
    [[0. 0. 0. ... 0. 0. 0.]
     [1. 0. 0. ... 0. 0. 0.]
     [0. 0. 0. ... 0. 0. 0.]
     ...
     [0. 0. 0. ... 0. 0. 0.]
     [0. 0. 0. ... 0. 0. 0.]
     [0. 0. 0. ... 0. 1. 0.]]
    epoch = 0, batch_idx = 0, loss = 2.328511953353882
    epoch = 0, batch_idx = 100, loss = 1.009535312652588
    epoch = 0, batch_idx = 200, loss = 0.8182420134544373
    epoch = 1, batch_idx = 0, loss = 0.7303786277770996
    epoch = 1, batch_idx = 100, loss = 0.710280179977417
    epoch = 1, batch_idx = 200, loss = 0.6227332353591919
    epoch = 2, batch_idx = 0, loss = 0.5902734994888306
    epoch = 2, batch_idx = 100, loss = 0.6123036742210388
    epoch = 2, batch_idx = 200, loss = 0.5449265837669373
    epoch = 3, batch_idx = 0, loss = 0.5221412181854248
    epoch = 3, batch_idx = 100, loss = 0.5586401224136353
    epoch = 3, batch_idx = 200, loss = 0.49894028902053833
    epoch = 4, batch_idx = 0, loss = 0.47923576831817627
    epoch = 4, batch_idx = 100, loss = 0.5219769477844238
    epoch = 4, batch_idx = 200, loss = 0.46610695123672485
    epoch = 5, batch_idx = 0, loss = 0.44889944791793823
    epoch = 5, batch_idx = 100, loss = 0.49393826723098755
    epoch = 5, batch_idx = 200, loss = 0.44056564569473267
    epoch = 6, batch_idx = 0, loss = 0.4255021810531616
    epoch = 6, batch_idx = 100, loss = 0.4713611900806427
    epoch = 6, batch_idx = 200, loss = 0.41985756158828735
    epoch = 7, batch_idx = 0, loss = 0.40636977553367615
    epoch = 7, batch_idx = 100, loss = 0.45304715633392334
    epoch = 7, batch_idx = 200, loss = 0.40254470705986023
    epoch = 8, batch_idx = 0, loss = 0.39018934965133667
    epoch = 8, batch_idx = 100, loss = 0.4377802312374115
    epoch = 8, batch_idx = 200, loss = 0.38769394159317017
    epoch = 9, batch_idx = 0, loss = 0.3761257529258728
    epoch = 9, batch_idx = 100, loss = 0.42440712451934814
    epoch = 9, batch_idx = 200, loss = 0.37483566999435425
    epoch = 10, batch_idx = 0, loss = 0.3637397885322571
    epoch = 10, batch_idx = 100, loss = 0.41257765889167786
    epoch = 10, batch_idx = 200, loss = 0.3634665012359619
    epoch = 11, batch_idx = 0, loss = 0.35271918773651123
    epoch = 11, batch_idx = 100, loss = 0.4018203616142273
    epoch = 11, batch_idx = 200, loss = 0.3533402979373932
    epoch = 12, batch_idx = 0, loss = 0.3428434729576111
    epoch = 12, batch_idx = 100, loss = 0.39214104413986206
    epoch = 12, batch_idx = 200, loss = 0.3441687822341919
    epoch = 13, batch_idx = 0, loss = 0.333955317735672
    epoch = 13, batch_idx = 100, loss = 0.3832662105560303
    epoch = 13, batch_idx = 200, loss = 0.33575889468193054
    epoch = 14, batch_idx = 0, loss = 0.3258323669433594
    epoch = 14, batch_idx = 100, loss = 0.3750748336315155
    epoch = 14, batch_idx = 200, loss = 0.32803112268447876
    epoch = 15, batch_idx = 0, loss = 0.3183256685733795
    epoch = 15, batch_idx = 100, loss = 0.3674846291542053
    epoch = 15, batch_idx = 200, loss = 0.32090339064598083
    epoch = 16, batch_idx = 0, loss = 0.31145304441452026
    epoch = 16, batch_idx = 100, loss = 0.3604132831096649
    epoch = 16, batch_idx = 200, loss = 0.3143956661224365
    epoch = 17, batch_idx = 0, loss = 0.3050137758255005
    epoch = 17, batch_idx = 100, loss = 0.3537922203540802
    epoch = 17, batch_idx = 200, loss = 0.30833786725997925
    epoch = 18, batch_idx = 0, loss = 0.2989848256111145
    epoch = 18, batch_idx = 100, loss = 0.3475683629512787
    epoch = 18, batch_idx = 200, loss = 0.30268821120262146
    epoch = 19, batch_idx = 0, loss = 0.2933279275894165
    epoch = 19, batch_idx = 100, loss = 0.3417797088623047
    epoch = 19, batch_idx = 200, loss = 0.29743343591690063
    epoch = 20, batch_idx = 0, loss = 0.2880209684371948
    epoch = 20, batch_idx = 100, loss = 0.33625274896621704
    epoch = 20, batch_idx = 200, loss = 0.2924404740333557
    epoch = 21, batch_idx = 0, loss = 0.2830372750759125
    epoch = 21, batch_idx = 100, loss = 0.3309747576713562
    epoch = 21, batch_idx = 200, loss = 0.2876732051372528
    epoch = 22, batch_idx = 0, loss = 0.2783474326133728
    epoch = 22, batch_idx = 100, loss = 0.32599449157714844
    epoch = 22, batch_idx = 200, loss = 0.2831261456012726
    epoch = 23, batch_idx = 0, loss = 0.27390143275260925
    epoch = 23, batch_idx = 100, loss = 0.3213098645210266
    epoch = 23, batch_idx = 200, loss = 0.27885332703590393
    epoch = 24, batch_idx = 0, loss = 0.2696850299835205
    epoch = 24, batch_idx = 100, loss = 0.3168952465057373
    epoch = 24, batch_idx = 200, loss = 0.2747831344604492
    epoch = 25, batch_idx = 0, loss = 0.2656662166118622
    epoch = 25, batch_idx = 100, loss = 0.3127139210700989
    epoch = 25, batch_idx = 200, loss = 0.2708702087402344
    epoch = 26, batch_idx = 0, loss = 0.2618401050567627
    epoch = 26, batch_idx = 100, loss = 0.3087657690048218
    epoch = 26, batch_idx = 200, loss = 0.26713475584983826
    epoch = 27, batch_idx = 0, loss = 0.25819507241249084
    epoch = 27, batch_idx = 100, loss = 0.30497828125953674
    epoch = 27, batch_idx = 200, loss = 0.26355743408203125
    epoch = 28, batch_idx = 0, loss = 0.2547600567340851
    epoch = 28, batch_idx = 100, loss = 0.30135565996170044
    epoch = 28, batch_idx = 200, loss = 0.2601349949836731
    epoch = 29, batch_idx = 0, loss = 0.251463383436203
    epoch = 29, batch_idx = 100, loss = 0.2978888750076294
    epoch = 29, batch_idx = 200, loss = 0.25687384605407715
    
    Process finished with exit code 0