Keras multi_gpu_model 返回错误“tensorflow_core._api.v2.config”没有属性“experimental_list_devices”

Question

我在两个 GPU 上训练我的 unet 模型时遇到问题，

该模型是一个我知道可行的简单 U-net 实现，因为它的 testet 不是 multi_gpu_model

train_generator = zip(image_generator, mask_generator)

with tf.device("/cpu:0"):
    # initialize the model
    model = unet((512,512,3))

# make the model parallel
model = multi_gpu_model(model, gpus=2)
model.compile(optimizer='adam', loss="mean_squared_error")

model.fit_generator(train_generator, steps_per_epoch=250, epochs=10)

:output
File "C:/Users/PycharmProjects/U-net/U-net.py", line 29, in <module>
    model = multi_gpu_model(model, gpus=2)
  File "C:\Users\AppData\Local\Programs\Python\Python37\lib\site-packages\keras\utils\multi_gpu_utils.py", line 150, in multi_gpu_model
    available_devices = _get_available_devices()
  File "C:\Users\AppData\Local\Programs\Python\Python37\lib\site-packages\keras\utils\multi_gpu_utils.py", line 16, in _get_available_devices
    return K.tensorflow_backend._get_available_gpus() + ['/cpu:0']
  File "C:\Users\AppData\Local\Programs\Python\Python37\lib\site-packages\keras\backend\tensorflow_backend.py", line 506, in _get_available_gpus
    _LOCAL_DEVICES = tf.config.experimental_list_devices()
AttributeError: module 'tensorflow_core._api.v2.config' has no attribute 'experimental_list_devices'

我也尝试过 tf.distributed.mirroredStrategy() 但也没有运气

任何帮助将不胜感激

Answer 1

experimental_list_devices 在 tf 2.1 中已弃用，请使用 tf.config.list_logical_devices 替换。

def _get_available_gpus():
"""Get a list of available gpu devices (formatted as strings).

# Returns
    A list of available GPU devices.
"""
global _LOCAL_DEVICES
if _LOCAL_DEVICES is None:
    if _is_tf_1():
        devices = get_session().list_devices()
        _LOCAL_DEVICES = [x.name for x in devices]
    else:
        devices = tf.config.list_logical_devices()
        _LOCAL_DEVICES = [x.name for x in devices]
return [x for x in _LOCAL_DEVICES if 'device:gpu' in x.lower()]

该链接有助于解决您的问题LINK