RuntimeError: cuDNN error: CUDNN_STATUS_INTERNAL_ERROR in RNNLogic+

[yzhangee]

Dear Authors,

I tried to run this code on FB15k-237 on a single GPU (after revising config file) but face the following error. The related code for rnn seems correct. I searched this error online and did not fine a solution. So I wonder why this problem happens and how to solve it. My GPU is 3090 with CUDA 11.2.

Traceback (most recent call last):
  File "/home/Logic/RNNLogic/RNNLogic+/src/run_rnnlogic.py", line 143, in <module>
    main(parse_args())
  File "/home/Logic/RNNLogic/RNNLogic+/src/run_rnnlogic.py", line 58, in main
    solver_g.train(dataset, **cfg.generator.pre_train)
  File "/home/Logic/RNNLogic/RNNLogic+/src/trainer.py", line 327, in train
    loss = model.loss(inputs, target, mask, weight, hidden)
  File "/home/Logic/RNNLogic/RNNLogic+/src/generators.py", line 32, in loss
    logits, hidden = self.forward(inputs, inputs[:, 0], hidden)
  File "/home/Logic/RNNLogic/RNNLogic+/src/generators.py", line 27, in forward
    outputs, hidden = self.rnn(embedding, hidden)
  File "/home/anaconda3/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
    return forward_call(*input, **kwargs)
  File "/home/anaconda3/lib/python3.9/site-packages/torch/nn/modules/rnn.py", line 691, in forward
    result = _VF.lstm(input, hx, self._flat_weights, self.bias, self.num_layers,
RuntimeError: cuDNN error: CUDNN_STATUS_INTERNAL_ERROR