深度学习2.0-5.tensorflow的基础操作之前向传播(张量)实战

前向传播(张量)实战

# 将无关信息屏蔽掉
import os
# 取值有四个：0，1，2，3，分别对应INFO，WARNING，ERROR，FATAL
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'
import tensorflow as tf
from tensorflow import keras
from tensorflow.keras import datasets

# 加载数据集 x:[60k,28,28]  y:[60k]
(x, y), _ = datasets.mnist.load_data()
# x: [0~255] => [0~1.]
x = tf.convert_to_tensor(x, dtype=tf.float32) / 255
y = tf.convert_to_tensor(y, dtype=tf.int32)

print(x.shape, y.shape, x.dtype, y.dtype)
# 查看数据范围
print(tf.reduce_min(x), tf.reduce_max(x))
print(tf.reduce_min(y), tf.reduce_max(y))

# from_tensor_slices表示切分传入Tensor的第一个维度，生成相应的dataset
train_db = tf.data.Dataset.from_tensor_slices((x,y)).batch(128)
train_iter = iter(train_db)
sample = next(train_iter)
print('batch:', sample[0].shape, sample[1].shape)      # batch: (128, 28, 28) (128,)


# 构建权值
# X: [b,784] => [b,256] => [b,128] => [b,10]
# W: [dim_in,dim_out]   b: [dim_out]
# truncated_normal表示正态分布,设置为均值为0，方差为0.1的范围，可解决梯度爆炸与梯度离散的情况
w1 = tf.Variable(tf.random.truncated_normal([784, 256], stddev=0.1))
b1 = tf.Variable(tf.zeros([256]))
w2 = tf.Variable(tf.random.truncated_normal([256, 128], stddev=0.1))
b2 = tf.Variable(tf.zeros([128]))
w3 = tf.Variable(tf.random.truncated_normal([128, 10], stddev=0.1))
b3 = tf.Variable(tf.zeros([10]))

lr = 1e-3
for epoch in range(10):         # 迭代整个数据集
    for step, (x, y) in enumerate(train_db):        # for every batch
        # x:[128,28,28]  y:[128]
        # [b,28,28] => [b,28*28]
        x = tf.reshape(x, [-1, 28*28])

        # 求梯度
        # GradientTape默认只会跟踪类型为tf.Variable
        with tf.GradientTape() as tape:
            # x: [b, 28*28]
            # h1 = x@w1 + b1
            # [b, 784]@[784, 256] + [256] => [b, 256] + [256] => [b, 256] + [b, 256]
            h1 = x @ w1 + tf.broadcast_to(b1, [x.shape[0], 256])
            h1 = tf.nn.relu(h1)
            # [b, 256] => [b, 128]
            h2 = h1 @ w2 + b2
            h2 = tf.nn.relu(h2)
            # [b, 128] => [b, 10]
            out = h2 @ w3 + b3

            # compute loss
            # out: [b, 10]
            # y: [b] => [b, 10]
            y_onehot = tf.one_hot(y, depth=10)

            # mse = mean(sum(y-out)^2)
            # [b, 10]
            loss = tf.square(y_onehot - out)
            # mean: scalar
            loss = tf.reduce_mean(loss)

        # 计算梯度,返回对应的梯度列表
        grads = tape.gradient(loss, [w1, b1, w2, b2, w3, b3])
        # print(grads)
        # 更新可训练参数
        # w1.assign_sub表示原地更新，保持tf.Variable
        w1.assign_sub(lr * grads[0])
        b1.assign_sub(lr * grads[1])
        w2.assign_sub(lr * grads[2])
        b2.assign_sub(lr * grads[3])
        w3.assign_sub(lr * grads[4])
        b3.assign_sub(lr * grads[5])

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

运算结果：

(60000, 28, 28) (60000,) <dtype: 'float32'> <dtype: 'int32'>
tf.Tensor(0.0, shape=(), dtype=float32) tf.Tensor(1.0, shape=(), dtype=float32)
tf.Tensor(0, shape=(), dtype=int32) tf.Tensor(9, shape=(), dtype=int32)
batch: (128, 28, 28) (128,)
0 0 loss 0.3418075442314148
0 100 loss 0.20874373614788055
0 200 loss 0.19508568942546844
0 300 loss 0.17686240375041962
0 400 loss 0.17524364590644836
1 0 loss 0.1492575705051422
1 100 loss 0.15576913952827454
1 200 loss 0.15462367236614227
1 300 loss 0.14648546278476715
1 400 loss 0.14724934101104736
2 0 loss 0.12436133623123169
2 100 loss 0.13543009757995605
2 200 loss 0.13396985828876495
2 300 loss 0.1286272406578064
2 400 loss 0.13028298318386078
3 0 loss 0.10929928719997406
3 100 loss 0.12242629379034042
3 200 loss 0.12099750339984894
3 300 loss 0.11676524579524994
3 400 loss 0.11878874152898788
4 0 loss 0.09928615391254425
4 100 loss 0.11343791335821152
4 200 loss 0.11198092997074127
4 300 loss 0.10829611867666245
4 400 loss 0.11033062636852264
5 0 loss 0.09216360747814178
5 100 loss 0.1067122370004654
5 200 loss 0.10542497783899307
5 300 loss 0.10182454437017441
5 400 loss 0.1037207618355751
6 0 loss 0.08672191202640533
6 100 loss 0.10142513364553452
6 200 loss 0.10031310468912125
6 300 loss 0.09669476002454758
6 400 loss 0.09842309355735779
7 0 loss 0.0824439525604248
7 100 loss 0.09712380170822144
7 200 loss 0.09613628685474396
7 300 loss 0.09249486774206161
7 400 loss 0.09409850835800171
8 0 loss 0.07895340770483017
8 100 loss 0.09356103837490082
8 200 loss 0.09263540804386139
8 300 loss 0.08899327367544174
8 400 loss 0.09041593968868256
9 0 loss 0.07604382187128067
9 100 loss 0.09052439779043198
9 200 loss 0.08964134752750397
9 300 loss 0.08601444214582443
9 400 loss 0.08727019280195236