解决 AMD GPU 上的“没有注册 Op-Kernel 来支持 Op”问题

Question

model = Sequential([
    LSTM(units=50,return_sequences=True,input_shape=(lookback,1)),
    Dense(units=1)
])
model.compile(loss=keras.losses.MSE,optimizer=keras.optimizers.Adam(),metrics=['accuracy'])

model.fit(trainx,trainy,validation_data=(testx,testy),epochs=20,batch_size=1)

model.compile 抛出错误

InvalidArgumentError: Graph execution error:

No OpKernel was registered to support Op 'CudnnRNN' used by {{node CudnnRNN}} with these attrs: [seed=0, dropout=0, T=DT_FLOAT, input_mode="linear_input", direction="unidirectional", rnn_mode="lstm", seed2=0, is_training=true]
Registered devices: [CPU, GPU]
...
Registered kernels:
  <no registered kernels>

     [[CudnnRNN]]
     [[sequential_2/lstm_6/PartitionedCall]] [Op:__inference_train_function_15994]

我有AMD GPU和tensorflow-directml

print(tf. __version__) 

====>2.10.0

print(tf.config.list_physical_devices('GPU'))

====>[PhysicalDevice(name='/physical_device:GPU:0', device_type='GPU')]

注意上面的代码运行完美 如果我在CPU上运行它

with tf.device('/CPU:0'):
    model.fit(trainx,trainy,validation_data=(testx,testy),epochs=2,batch_size=1,verbose=1)

有没有办法在AMD GPU上运行它

Answer 1

同样的错误： tensorflow.python.framework.errors_impl.InvalidArgumentError：图形执行错误：

没有注册 OpKernel 来支持 {{node CudnnRNN}} 使用的 Op 'CudnnRNN'，其属性如下：[seed=0, dropout=0, T=DT_FLOAT, input_mode="linear_input", Direction="unidirection", rnn_mode= “lstm”，seed2=0，is_training=true] 注册设备：[CPU、GPU] 注册内核：

 [[CudnnRNN]]

有人可以帮忙吗？