dreamer-pytorch
dreamer-pytorch copied to clipboard
yusukeurakami
pytorch-implementation of Dreamer (Model-based Image RL Algorithm)
Hi, thank you for the cool repository! I tried several tasks `walker walk`, `cheetah run`. They seem to work fine. But when I run `walker run`, the episode_reward cannot achieve...
dm_control now automatically installs MuJoCo.
Input should contain only the initial belief and state right? Here, input contains the entire sequence of beliefs and states instead. Not sure how this works with the algorithm
I tried running the agent on the Walker Walk environment and the KL Divergence loss seems to be growing exponentially and causing nans. But I have not made any changes...
Hi, Thanks you for your sharing~ I've implemented it for a while. I have some questions about the time because I spend lots of time on reaching 500K steps. However,...
Thank you for your sharing, but I have a small question. Why do you 1) use `F.softplus` for `variance(std_dev)` every time and 2) add a constant(min_std_dev). Is it to ensure...
Hi, I believe the reward loss should be based on `rewards[1:]` instead of `rewards[:-1] `: https://github.com/yusukeurakami/dreamer-pytorch/blob/7e9050e8c454309de40bd0d1a4ec0256ef600147/main.py#L209 If not, can you please explain your reasoning? Thanks,
Metadata
Owner
Metadata
pytorch-implementation of Dreamer (Model-based Image RL Algorithm)