TensorFlow实战之Fashion-Minst数据集

Fashion-Minst数据集是Minst数据集的升级版，具有更高的复杂性
如下代码实现Fashion-Minst数据集的训练及测试

import tensorflow as tf
from tensorflow import keras
from tensorflow.keras import datasets, layers, optimizers, Sequential, metrics
import os

# 避免出现一些不必要的警告
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'

# 下载数据集
(x, y), (x_test, y_test) = datasets.fashion_mnist.load_data()
print(x.shape, y.shape)
print(x_test.shape, y_test.shape)


# 定义一个预处理的函数
def preprocess(x, y):
    x = tf.cast(x, dtype=tf.float32) / 255.  # 转化到0到1的范围
    y = tf.cast(y, dtype=tf.int32)

    return x, y


batchsz = 128

# 构建数据集、进行预处理
db = tf.data.Dataset.from_tensor_slices((x, y))
db = db.map(preprocess).shuffle(10000).batch(batchsz)

db_test = tf.data.Dataset.from_tensor_slices((x_test, y_test))
db_test = db_test.map(preprocess).shuffle(10000).batch(batchsz)

# 得到一个迭代器
db_iter = iter(db)
sample = next(db_iter)
print('batch:', sample[0].shape, sample[1].shape)

model = Sequential([
    layers.Dense(256, activation=tf.nn.relu),  # [b,784] => [b, 256]
    layers.Dense(128, activation=tf.nn.relu),  # [b,256] => [b, 128]
    layers.Dense(64, activation=tf.nn.relu),  # [b,128] => [b, 64]
    layers.Dense(32, activation=tf.nn.relu),  # [b,64] => [b, 32]
    layers.Dense(10)  # [b,32] => [b, 10]， 参数量：330 = 32*10 + 10
])

model.build(input_shape=[None, 28 * 28])
model.summary()  # 调试的功能把网络结构打印出来
# w = w - lr*grad
optimizer = optimizers.Adam(lr=1e-3)  # 优化器


def main():
    for epoch in range(30):

        for step, (x, y) in enumerate(db):

            # x: [b,28,28] => [b,784]
            # y: [b]
            x = tf.reshape(x, [-1, 28 * 28])

            with tf.GradientTape() as  tape:
                # [b,784] => [b,10]
                logits = model(x)  # 完成前向传播
                y_onehot = tf.one_hot(y, depth=10)
                # [b]
                loss_mse = tf.reduce_mean(tf.losses.MSE(y_onehot, logits))  # MSE完成均方差的计算
                loss_ce = tf.losses.categorical_crossentropy(y_onehot, logits, from_logits=True)
                loss_ce = tf.reduce_mean(loss_ce)  # 为了求标量

            grads = tape.gradient(loss_ce, model.trainable_variables)
            optimizer.apply_gradients(zip(grads, model.trainable_variables))  # zip 对每一个地方的梯度在前，参数在后

            if step % 100 == 0:
                print(epoch, step, 'loss:', float(loss_ce), float(loss_mse))

        # test
        total_correct = 0
        total_num = 0
        for x, y in db_test:
            # x: [b,28,28] => [b,784]
            # y: [b]
            x = tf.reshape(x, [-1, 28 * 28])
            # y = tf.reshape(y, [128 * 128, ])
            # [b, 10]
            logits = model(x)
            # logits => prob ，[b, 10]
            prob = tf.nn.softmax(logits, axis=1)  # 完成概率的转换 在0和1之间 总和为1
            # [b, 10] => [b], int64
            pred = tf.argmax(prob, axis=1)  # 最大的
            pred = tf.cast(pred, dtype=tf.int32)
            # pred:[b]
            # y: [b]
            # correct: [b], True:equal, F: not equal

            # print(pred.dtype, y.dtype)  # 都是int32类型
            # print(pred.shape, y.shape)
            # y = tf.reshape(y, [-1, 128 * 128])
            correct = tf.equal(pred, y)
            correct = tf.reduce_sum(tf.cast(correct, dtype=tf.int32))

            total_correct += int(correct)  # tensor => numpy
            total_num += x.shape[0]

        acc = total_correct / total_num
        print(epoch, 'test acc:', acc)


# 避免全局变量的error
if __name__ == '__main__':
    main()

测试结果如下：