目录
import tensorflow as tf
from tensorflow import keras
from tensorflow.keras import datasets
import os# do not print irrelevant information
# os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'# x: [60k,28,28], [10,28,28]
# y: [60k], [10k]
(x, y), (x_test, y_test) = datasets.mnist.load_data()# transform Tensor
# x: [0~255] ==》 [0~1.]
x = tf.convert_to_tensor(x, dtype=tf.float32) / 255.
y = tf.convert_to_tensor(y, dtype=tf.int32)x_test = tf.convert_to_tensor(x_test, dtype=tf.float32) / 255.
y_test = tf.convert_to_tensor(y_test, dtype=tf.int32)f'x.shape: {x.shape}, y.shape: {y.shape}, x.dtype: {x.dtype}, y.dtype: {y.dtype}'"x.shape: (60000, 28, 28), y.shape: (60000,), x.dtype: <dtype: 'float32'>, y.dtype: <dtype: 'int32'>"f'min_x: {tf.reduce_min(x)}, max_x: {tf.reduce_max(x)}''min_x: 0.0, max_x: 1.0'f'min_y: {tf.reduce_min(y)}, max_y: {tf.reduce_max(y)}''min_y: 0, max_y: 9'# batch of 128
train_db = tf.data.Dataset.from_tensor_slices((x, y)).batch(128)
test_db = tf.data.Dataset.from_tensor_slices((x_test, y_test)).batch(128)
train_iter = iter(train_db)
sample = next(train_iter)
f'batch: {sample[0].shape,sample[1].shape}''batch: (TensorShape([128, 28, 28]), TensorShape([128]))'# [b,784] ==> [b,256] ==> [b,128] ==> [b,10]
# [dim_in,dim_out],[dim_out]
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]))# learning rate
lr = 1e-3for epoch in range(10): # iterate db for 10
# tranin every train_db
for step, (x, y) in enumerate(train_db):
# x: [128,28,28]
# y: [128]
# [b,28,28] ==> [b,28*28]
x = tf.reshape(x, [-1, 28 * 28])
with tf.GradientTape(
) as tape: # only data types of tf.variable are logged
# 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 = x@w2 + b2 # b2 can broadcast automatic
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)
# compute gradients
grads = tape.gradient(loss, [w1, b1, w2, b2, w3, b3])
# w1 = w1 - lr * w1_grad
# w1 = w1 - lr * grads[0] # not in situ update
# in situ update
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(f'epoch:{epoch}, step: {step}, loss:{float(loss)}')
# [w1,b1,w2,b2,w3,b3]
total_correct, total_num = 0, 0
for step, (x, y) in enumerate(test_db):
# [b,28,28] ==> [b,28*28]
x = tf.reshape(x, [-1, 28 * 28])
# [b,784] ==> [b,256] ==> [b,128] ==> [b,10]
h1 = tf.nn.relu(x @ w1 + b1)
h2 = tf.nn.relu(h1 @ w2 + b2)
out = h2 @ w3 + b3
# out: [b,10] ~ R
# prob: [b,10] ~ (0,1)
prob = tf.nn.softmax(out, axis=1)
# [b,10] ==> [b]
pred = tf.argmax(prob, axis=1)
pred = tf.cast(pred, dtype=tf.int32)
# y: [b]
# [b], int32
correct = tf.cast(tf.equal(pred, y), dtype=tf.int32)
correct = tf.reduce_sum(correct)
total_correct += int(correct)
total_num += x.shape[0]
acc = total_correct / total_num
print(f'test acc: {acc}')epoch:0, step: 0, loss:0.4985736012458801
epoch:0, step: 100, loss:0.22939381003379822
epoch:0, step: 200, loss:0.2018660604953766
epoch:0, step: 300, loss:0.18181894719600677
epoch:0, step: 400, loss:0.1831897795200348
test acc: 0.1153
epoch:1, step: 0, loss:0.1674182116985321
epoch:1, step: 100, loss:0.17186065018177032
epoch:1, step: 200, loss:0.16210347414016724
epoch:1, step: 300, loss:0.1499405801296234
epoch:1, step: 400, loss:0.15070970356464386
test acc: 0.1769
epoch:2, step: 0, loss:0.14020009338855743
epoch:2, step: 100, loss:0.14754906296730042
epoch:2, step: 200, loss:0.13924123346805573
epoch:2, step: 300, loss:0.1308508813381195
epoch:2, step: 400, loss:0.1306917369365692
test acc: 0.235
epoch:3, step: 0, loss:0.12297296524047852
epoch:3, step: 100, loss:0.13165466487407684
epoch:3, step: 200, loss:0.12420644611120224
epoch:3, step: 300, loss:0.1179303377866745
epoch:3, step: 400, loss:0.11716334521770477
test acc: 0.2927
epoch:4, step: 0, loss:0.11098697036504745
epoch:4, step: 100, loss:0.12046296894550323
epoch:4, step: 200, loss:0.11333265155553818
epoch:4, step: 300, loss:0.10868857055902481
epoch:4, step: 400, loss:0.10756760835647583
test acc: 0.3386
epoch:5, step: 0, loss:0.1022152453660965
epoch:5, step: 100, loss:0.1120707243680954
epoch:5, step: 200, loss:0.10497119277715683
epoch:5, step: 300, loss:0.10168357938528061
epoch:5, step: 400, loss:0.10033649206161499
test acc: 0.379
epoch:6, step: 0, loss:0.09566861391067505
epoch:6, step: 100, loss:0.10548736900091171
epoch:6, step: 200, loss:0.09834134578704834
epoch:6, step: 300, loss:0.0961376205086708
epoch:6, step: 400, loss:0.09474694728851318
test acc: 0.4168
epoch:7, step: 0, loss:0.09054075181484222
epoch:7, step: 100, loss:0.1001550704240799
epoch:7, step: 200, loss:0.09303966909646988
epoch:7, step: 300, loss:0.09163998067378998
epoch:7, step: 400, loss:0.09031815826892853
test acc: 0.453
epoch:8, step: 0, loss:0.08635123074054718
epoch:8, step: 100, loss:0.0957597866654396
epoch:8, step: 200, loss:0.08867798745632172
epoch:8, step: 300, loss:0.08790989965200424
epoch:8, step: 400, loss:0.08668653666973114
test acc: 0.4831
epoch:9, step: 0, loss:0.08282895386219025
epoch:9, step: 100, loss:0.09203790128231049
epoch:9, step: 200, loss:0.0850382000207901
epoch:9, step: 300, loss:0.08473993837833405
epoch:9, step: 400, loss:0.0835738554596901
test acc: 0.506515 测试（张量）- 实战