Bogdan Mazoure

@astooke Any chance to get a short explanation about the discrepancy between batch_T+warmup_T and the actual sample dimension? For `batch_T=2`, `warmup_T=40`, I get `[49,batch_B,3,84,84]` dimensions of sample.

Update: From what I get in the R2D1 code, the SequenceBuffer samples in chunks of `self._batch_size`, so if you set `self._batch_size =4` in that chunk of code, then this will...

This makes sense and it is what I ended up doing. However, is the code expected to run much slower? E.g. on 8 CPUs it took me 30 minutes. So...

Which networks do you need, and on which envs?

Hey, Thanks for going through to analyze this! Yeah, some environments definitively have inconsistencies. You're right, I did notice a lot of impact in the different inits, so you could...