RuntimeError: cudnn RNN backward can only be called in training mode

1 ‘. Train()’ is not used. Some operations are prohibited when ‘. Eval()’

2. Add one line of the following code to the error reported by using LSTM

torch.backends.cudnn.enabled = False