D4RL
D4RL copied to clipboard
Farama-Foundation
A collection of reference environments for offline reinforcement learning
For the creation of datasets, can maps of other sizes be created?
Sorry if I missed this information but is it possible to add new tasks to this repo/benchmark?
When I download a maze2d dataset with `env.get_dataset()`, the downloaded hdf5 file contains "timeouts" key: ``` >>> import d4rl, gym >>> env = gym.make('maze2d-large-v1') >>> dataset = env.get_dataset() Downloading dataset:...
### Question Currently, D4RL requires gym=0.25)? Is `terminals` equivalent to `done` (gym=0.25)?
When I download the halfcheetah dataset, it just shows me that > Downloading dataset: http://rail.eecs.berkeley.edu/datasets/offline_rl/gym_mujoco_v2/halfcheetah_medium-v2.hdf5 to /root/.d4rl/datasets/halfcheetah_medium-v2.hdf5 I don't know how many downloads. Could you add a time procedure when...
**Describe the bug** The docstring for `qlearning_dataset()` says: ``` terminate_on_end (bool): Set done=True on the last timestep in a trajectory. ``` However, if you look at the code, it does...
Currently the list of tasks wiki pages only show v0 environments for antmaze, However d4rl supports all v2 environments as well. I checked ! It would be nice to update...
Hi, I want to know how the maximum score is obtained for MuJoCo tasks? From the wiki (https://github.com/rail-berkeley/d4rl/wiki/Dataset-Reproducibility-Guide#gym-mujocogym-bullet), it seems that we use the stochastic SAC policy to obtain the...
### Question The paper says that Mujoco medium dataset was generated by training SAC and early stopping when "medium" level was reached. I am interested in knowing how this point...
Metadata
Owner
Metadata
A collection of reference environments for offline reinforcement learning