search

katerakelly / oyster

Implementation of Efficient Off-policy Meta-learning via Probabilistic Context Variables (PEARL)
MIT License
472 stars 125 forks source link

The code seems not working well with mujoco-200 #29

Open Cubism-star opened 2 years ago

Cubism-star commented 2 years ago

I've install mujoco-200 following the instruction, and I run "python launch_experiment.py ./configs/cheetah-dir.json". It should work as it is mentioned in Readme("For the task distributions in which the reward function varies (Cheetah, Ant, Humanoid), install MuJoCo200."), but still report lacking dependencies of Mujoco131. In fact, I think the code contains a mandatory check on the version of mujoco. Maybe the README is inconsistent with the actual situation.

ZiyiLiubird commented 2 years ago

You should download mujoco131 files and uncompress them in .mujoco folder. In other words, .mujoco folder should contain mujoco200 and mujoco131 at the same time.