How to solve RuntimeError: For batched 3-D input, hx and cx should also be 3-D but got (2-D, 2-D) tensors

[augooh]

No description provided.

[augooh]

I tried modify hidden = (h_.contiguous(), h_.contiguous()) to

        # cell_state = torch.zeros(self.num_layers, batch_size, self.hidden_size).to(self.device)
        # hidden_state = torch.randn(self.num_layers, batch_size, self.hidden_size).to(self.device)
        # cell_state = torch.randn(self.num_layers, batch_size, self.hidden_size).to(self.device)
        # hidden_state = h_.clone().detach().view(self.num_layers, batch_size, self.hidden_size).to(self.device)
        # cell_state = h_.clone().detach().view(self.num_layers, batch_size, self.hidden_size).to(self.device)

        # hidden = (hidden_state, cell_state)

There is no error, but the training results are poor.

[GabSabb]

I got the same issue and replaced your code to

hidden_state = h_.contiguous().detach().clone().view(self.num_layers, batch_size, self.hidden_size).to(self.device)
cell_state = h_.contiguous().detach().clone().view(self.num_layers, batch_size, self.hidden_size).to(self.device) 
hidden = (hidden_state, cell_state)

The training is poor as well: