手写数字问题实战（层） - 好文

FashionMNIST
import tensorflow as tf from tensorflow import keras from tensorflow.keras
import datasets, layers, optimizers, Sequential, metrics def preprocess(x, y):
"""数据处理函数""" x = tf.cast(x, dtype=tf.float32) / 255. y = tf.cast(y,
dtype=tf.int32) return x, y # 加载数据 (x, y), (x_test, y_test) =
datasets.fashion_mnist.load_data() print(x.shape, y.shape) # 处理train数据
batch_size = 128 db = tf.data.Dataset.from_tensor_slices((x, y)) db =
db.map(preprocess).shuffle(10000).batch(batch_size) # 处理test数据 db_test =
tf.data.Dataset.from_tensor_slices((x_test, y_test)) db_test =
db_test.map(preprocess).batch(batch_size) # 生成train数据的迭代器 db_iter = iter(db)
sample = next(db_iter) print(f'batch: {sample[0].shape,sample[1].shape}') #
设计网络结构 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], 330=32*10+10 ]) model.build(input_shape=[None, 28 * 28])
model.summary() # 调试 # w = w - lr*grad optimizer = optimizers.Adam(lr=1e-3) #
优化器，加快训练速度 def main(): """主运行函数""" for epoch in range(10): 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)
y_onehot = tf.one_hot(y, depth=10) # [b] loss_mse =
tf.reduce_mean(tf.losses.MSE(y_onehot, logits)) loss_ce = tf.reduce_mean(
tf.losses.categorical_crossentropy(y_onehot, logits, from_logits=True)) grads =
tape.gradient(loss_ce, model.trainable_variables)
optimizer.apply_gradients(zip(grads, model.trainable_variables)) if step % 100
== 0: print(epoch, step, f'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]) # [b,10] logits = model(x) # logits
--> prob [b,10] prob = tf.nn.softmax(logits, axis=1) # [b,10] --> [b], int32
pred = tf.argmax(prob, axis=1) pred = tf.cast(pred, dtype=tf.int32) # pred:[b]
# y:[b] # correct: [b], True: equal; False: not equal correct = tf.equal(pred,
y) correct = tf.reduce_sum(tf.cast(correct, dtype=tf.int32)) total_correct +=
int(correct) total_num += x.shape[0] acc = total_correct / total_num
print(epoch, f'test acc: {acc}') if __name__ == '__main__': main() (60000, 28,
28) (60000,) batch: (TensorShape([128, 28, 28]), TensorShape([128])) Model:
"sequential" _________________________________________________________________
Layer (type) Output Shape Param #
================================================================= dense (Dense)
multiple 200960
_________________________________________________________________ dense_1
(Dense) multiple 32896
_________________________________________________________________ dense_2
(Dense) multiple 8256
_________________________________________________________________ dense_3
(Dense) multiple 2080
_________________________________________________________________ dense_4
(Dense) multiple 330
================================================================= Total params:
244,522 Trainable params: 244,522 Non-trainable params: 0
_________________________________________________________________ 0 0 loss:
(2.317634105682373, 0.240666925907135) 0 100 loss: (0.5373745560646057,
22.973751068115234) 0 200 loss: (0.43246397376060486, 23.848735809326172) 0 300
loss: (0.46746426820755005, 25.622543334960938) 0 400 loss:
(0.35051512718200684, 25.582551956176758) 0 test acc: 0.8411 1 0 loss:
(0.3928898572921753, 25.729997634887695) 1 100 loss: (0.3873934745788574,
28.322147369384766) 1 200 loss: (0.3231426477432251, 23.980030059814453) 1 300
loss: (0.38196271657943726, 26.83534812927246) 1 400 loss:
(0.32163679599761963, 23.470108032226562) 1 test acc: 0.8522 2 0 loss:
(0.37968552112579346, 24.739742279052734) 2 100 loss: (0.3641986846923828,
32.74003601074219) 2 200 loss: (0.28391164541244507, 27.59665870666504) 2 300
loss: (0.35625365376472473, 32.84230422973633) 2 400 loss: (0.2914122939109802,
28.2105712890625) 2 test acc: 0.8643 3 0 loss: (0.32174795866012573,
28.54513168334961) 3 100 loss: (0.33406931161880493, 35.23341369628906) 3 200
loss: (0.2978339195251465, 31.21166229248047) 3 300 loss: (0.34828150272369385,
37.68536376953125) 3 400 loss: (0.2958236336708069, 34.39887619018555) 3 test
acc: 0.8657 4 0 loss: (0.2884419858455658, 28.723865509033203) 4 100 loss:
(0.30416643619537354, 42.262481689453125) 4 200 loss: (0.2996847927570343,
35.41472244262695) 4 300 loss: (0.312608003616333, 43.87290954589844) 4 400
loss: (0.2881354093551636, 39.22924041748047) 4 test acc: 0.8671 5 0 loss:
(0.27839434146881104, 35.540130615234375) 5 100 loss: (0.2806701958179474,
48.4008903503418) 5 200 loss: (0.2766285836696625, 42.87168884277344) 5 300
loss: (0.2809426784515381, 52.62693786621094) 5 400 loss: (0.26189112663269043,
47.885108947753906) 5 test acc: 0.8735 6 0 loss: (0.2579110264778137,
43.07761764526367) 6 100 loss: (0.2582871615886688, 59.50879669189453) 6 200
loss: (0.28057757019996643, 48.108917236328125) 6 300 loss:
(0.27566733956336975, 59.61842346191406) 6 400 loss: (0.25613951683044434,
58.044837951660156) 6 test acc: 0.8777 7 0 loss: (0.2313823103904724,
51.310028076171875) 7 100 loss: (0.2618938088417053, 68.77056884765625) 7 200
loss: (0.2880491614341736, 57.71855163574219) 7 300 loss: (0.2529357075691223,
68.95218658447266) 7 400 loss: (0.22080641984939575, 67.93892669677734) 7 test
acc: 0.8817 8 0 loss: (0.2174786627292633, 67.77418518066406) 8 100 loss:
(0.24183037877082825, 73.86817932128906) 8 200 loss: (0.2777296304702759,
70.0999755859375) 8 300 loss: (0.24576255679130554, 75.03533935546875) 8 400
loss: (0.22443264722824097, 77.09339904785156) 8 test acc: 0.8838 9 0 loss:
(0.1982460469007492, 67.42424011230469) 9 100 loss: (0.18205790221691132,
84.63094329833984) 9 200 loss: (0.2785370349884033, 78.5411376953125) 9 300
loss: (0.2484734058380127, 94.95791625976562) 9 400 loss: (0.20044317841529846,
85.65432739257812) 9 test acc: 0.8833

热门工具换一换