TF的計算圖分爲三種:靜態計算圖(1.x)、動態計算圖(2.x)和 AutoGraph(2.x)
AutoGraph 可以把動態圖轉化爲靜態圖保存,通過@tf.function實現,有以下三個注意事項:
- 使用 tf 內部函數,避免直接使用python函數,因爲無法嵌入進計算圖
- 避免定義 tf.Variable, 以爲它是動態的,每次迭代都會更新
- 不可以修改列表字典等數據結構
代碼:
import tensorflow as tf
fashion = tf.keras.datasets.fashion_mnist
(x_train, y_train),(x_test, y_test) = fashion.load_data()
x_train, x_test = x_train / 255.0, x_test / 255.0
#子類化模型
class MyModel(tf.keras.Model):
def __init__(self):
super(MyModel,self).__init__()
self.D1 = tf.keras.layers.Flatten()
self.D2 = tf.keras.layers.Dense(128,activation='relu')
self.D3 = tf.keras.layers.Dense(10,activation='softmax')
@tf.function(input_signature=[tf.TensorSpec([None,28,28],tf.float32)])
def call(self,inputs):
x = self.D1(inputs)
x = self.D2(x)
x = self.D3(x)
return x
model = MyModel()
model.compile(optimizer='adam',
loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False),
metrics=['sparse_categorical_accuracy'])
model.fit(x_train, y_train, batch_size=32, epochs=5, validation_data=(x_test, y_test), validation_freq=1)
model.summary()
tf.saved_model.save(model,'my_saved_model')
保存結果:
保存後可供後續跨平臺使用