dqn-pytorch

dqn-pytorch copied to clipboard

According to this [blog](https://danieltakeshi.github.io/2016/11/25/frame-skipping-and-preprocessing-for-deep-q-networks-on-atari-2600-games/) which has nicely explained the preprocessing done in actual DQN by Deep Mind, We should skip 3 consecutive frames and consider only 4rth frame. But you...

kartikpaigwar

I get the following error when running the DQN in train mode: ` File "dqn.py", line 507, in optimize target_values[non_final_mask] = reward_batch[non_final_mask] + target_pred.max(1)[0] * gamma RuntimeError: The expanded size...

thomaspinder

In code snippet below from `select_action` method, the input hidden state is taken from `self.train_hidden_state`. But the output hidden state is saved in `self.dqn_hidden_state`. It seems the output hidden state...

parthjaggi

Missing batch_first attribute for LSTM model.

5

By default the `BATCH_SIZE = 32`. Input to the LSTM from the CNN is of the shape `(32, 64, 16)`. The semantics of LSTM input are `(seq_len, batch_size, input_size)`. But...

parthjaggi