Parth Jaggi

Hi, any reason for delay in merging this PR. I wanted to use this feature so happy to help with this. If needed I can also help integrating the induction...

Cool. I am fixing it as well. Will try for a PR in a couple of days.

When checking for batch sizes of input to the forward method of LSTM using `print(x.shape)`, the following is obtained. ``` torch.Size([32, 4, 84, 84]) torch.Size([32, 4, 84, 84]) torch.Size([32, 4,...

Another issue that would need to be looked at would be `batch_size` when using `init_states` method to initialize `hidden_state` and `cell_state`. Hidden/Cell state semantics: `(n_layers, batch_size, hidden_size)` Since while training...