tensorflow - 如何在选项 tensorlfow MirroredStrategy 期间不打印正在执行的操作

Question

感谢您对我的问题感兴趣。

我运行TensorFlow代码多GPU系统如下：

tf.debugging.set_log_device_placement(True)

策略 = tf.distribute.MirroredStrategy()

使用 strategy.scope(): sub_model = tf.keras.applications.VGG19(include_top=False, input_shape=(32, 32, 3), classes=10) sub_model.trainable = False

flat = tf.keras.layers.Flatten()(sub_model.layers[-1].output)
classify_ = tf.keras.layers.Dense(10, activation='softmax')(flat)
model = tf.keras.Model(inputs=sub_model.inputs, outputs=classify_)
model.compile(optimizer=tf.keras.optimizers.Adam(0.05), loss='categorical_crossentropy', metrics=['accuracy'])

lr_decay = tf.keras.callbacks.ReduceLROnPlateau()
check_pt = tf.keras.callbacks.ModelCheckpoint('./cache/best_param.hdf5', monitor='val_loss', save_best_only=True)

EPOCHS = 500
BATCH_SIZE = 300
CB = [lr_decay, check_pt]


history  = model.fit(x=x_train, 
                      y=y_train,
                      epochs=EPOCHS,
                      validation_data=(x_val,  y_val),
                      batch_size=BATCH_SIZE,
                      callbacks=CB,
                      verbose=False)

和大量的印刷品

Executing op __inference_train_function_3274 in device /job:localhost/replica:0/task:0/device:GPU:0
Executing op __inference_train_function_3274 in device /job:localhost/replica:0/task:0/device:GPU:0
Executing op __inference_train_function_3274 in device /job:localhost/replica:0/task:0/device:GPU:0
Executing op __inference_train_function_3274 in device /job:localhost/replica:0/task:0/device:GPU:0
Executing op __inference_train_function_3274 in device /job:localhost/replica:0/task:0/device:GPU:0
Executing op __inference_train_function_3274 in device /job:localhost/replica:0/task:0/device:GPU:0
Executing op __inference_train_function_3274 in device /job:localhost/replica:0/task:0/device:GPU:0
Executing op __inference_train_function_3274 in device /job:localhost/replica:0/task:0/device:GPU:0

我希望我的代码不打印这些打印件。我如何设置这个选项？

标签： tensorflowmulti-gpu