tensorflow(12)---参数提取

上一篇文章简单讲了怎么断点续训，但是有时候吧，想要看看训练过程中那些参数都长啥样，从开始到结束各个参数如何变化，在网络训练过程中只有accuracy和loss可以看到，这时候需要使用参数提取的方法。

在上一篇基础上，加入一些新内容

import tensorflow as tf
import os
from tensorflow import keras
import numpy as np

# 设置打印格式，threshold=超过多少不打印,np.inf无限大
np.set_printoptions(threshold=np.inf)
mnist = keras.datasets.mnist
(x_train,y_train),(x_test,y_test) = mnist.load_data()
x_train,x_test = x_train/255,x_test/255

model = keras.models.Sequential([
    keras.layers.Flatten(),
    keras.layers.Dense(128,activation='relu'),
    keras.layers.Dense(10,activation='softmax')
])

model.compile(optimizer='adam',
              loss=keras.losses.SparseCategoricalCrossentropy(from_logits=False),
              metrics=['sparse_categorical_accuracy'])

checkpoint_save_path = './临时文件/mnist.ckpt'
# 生成ckpt的同时会生成索引文件
if os.path.exists(checkpoint_save_path+'.index'):
    print('-----------load the model----------------')
    model.load_weights(checkpoint_save_path)

cp_callback = keras.callbacks.ModelCheckpoint(filepath=checkpoint_save_path,save_weights_only=True,save_best_only=True)
history = model.fit(x_train,y_train,batch_size=32,epochs=5,validation_data=(x_test,y_test),validation_freq=1,callbacks=[cp_callback])
model.summary()

查看一下训练过程中各参数

print(model.trainable_variables)

然后把参数保存在文件中

with open('./临时文件/weights.txt','w')as f:
    for v in model.trainable_variables:
        f.write(str(v.name)+'\n')
        f.write(str(v.shape) + '\n')
        f.write(str(v.numpy()) + '\n')

 完整代码如下

import tensorflow as tf
import os
from tensorflow import keras
import numpy as np

# 设置打印格式，threshold=超过多少不打印,np.inf无限大
np.set_printoptions(threshold=np.inf)
mnist = keras.datasets.mnist
(x_train,y_train),(x_test,y_test) = mnist.load_data()
x_train,x_test = x_train/255,x_test/255

model = keras.models.Sequential([
    keras.layers.Flatten(),
    keras.layers.Dense(128,activation='relu'),
    keras.layers.Dense(10,activation='softmax')
])

model.compile(optimizer='adam',
              loss=keras.losses.SparseCategoricalCrossentropy(from_logits=False),
              metrics=['sparse_categorical_accuracy'])

checkpoint_save_path = './临时文件/mnist.ckpt'
# 生成ckpt的同时会生成索引文件
if os.path.exists(checkpoint_save_path+'.index'):
    print('-----------load the model----------------')
    model.load_weights(checkpoint_save_path)

cp_callback = keras.callbacks.ModelCheckpoint(filepath=checkpoint_save_path,save_weights_only=True,save_best_only=True)
history = model.fit(x_train,y_train,batch_size=32,epochs=5,validation_data=(x_test,y_test),validation_freq=1,callbacks=[cp_callback])
model.summary()
print(model.trainable_variables)
    with open('./临时文件/weights.txt','w')as f:
        for v in model.trainable_variables:
            f.write(str(v.name)+'\n')
            f.write(str(v.shape) + '\n')
            f.write(str(v.numpy()) + '\n')

打开刚才保存的文件看一下