Xingyuan Zhang

[RL Unplugged] How to recover the sequential data

3

Hi, I am interested in testing my algorithm on `dm_control` dataset as benchmark and I am trapped in data loading part. Since `dm_control` is a set of sequential control tasks...

Why MOPO is given access to terminal function in rollout generation?

2

Hi, thanks for sharing your great work. However, I am confused about the rollout generation process. As I see in the code, the agent can access to a pre-defined terminal...

incompatible registration argument `tags` for latest version of gym

1

Hi, there. I find a compatibility issue. According to https://github.com/openai/gym/commit/a99e8d153e94f532e5728b01b02b42e076334bf7, the latest version of gym has remove the argument `tags` in the registration function, which makes it incompatible with the...

Hey @danijar. I just noticed that the code is using `TruncNormal` as the actor distribution instead of `TanhNormal` as in v1. I wonder did you make some ablations on these...

**Describe the bug** As the title says, when applying `d4rl.sequence_dataset` on a `v2` environment of d4rl-mujoco, you get an error of: ``` IndexError Traceback (most recent call last) Cell In...