PPO1, A2C, VPG, DQN not training for Atari envs

SforAiDl / genrl

A PyTorch reinforcement learning library for generalizable and reproducible algorithm implementations with an aim to improve accessibility in RL

https://genrl.readthedocs.io

MIT License

403 stars 59 forks source link

Open sampreet-arthi opened 4 years ago

sampreet-arthi commented 4 years ago

DQN is also not training but that'll be addressed after DQN is restructured.

Sharad24 commented 4 years ago

It seems to me that this should be resolved in #175 itself?

cc: @ajaysub110 @Het-Shah

ajaysub110 commented 4 years ago

Yeah I agree

Het-Shah commented 4 years ago

Even current A2C is very fragile.

Sharad24 commented 4 years ago

What's the status here?

I tried running them but it seems like we dont have CUDA support, so I'm not sure if they're training or not.

sampreet-arthi commented 4 years ago

We should try trianing these after adding CUDA support. Most of these were tested a long time ago so we should test them again.