search

araffin / rl-baselines-zoo

A collection of 100+ pre-trained RL agents using Stable Baselines, training and hyperparameter optimization included.
https://stable-baselines.readthedocs.io/
MIT License
1.13k stars 208 forks source link

Trained model for SAC+HER on FetchReach-v1 #35

Closed gautams3 closed 5 years ago

gautams3 commented 5 years ago

Trained model of HER (with model_class SAC) for the FetchReach-v1 environments. Using the suggested hyperparams from hyperparams/her.yml itself.

Not sure how to get obs_rms.pkl and ret_rms.pkl for the environment, but it wasn't necessary for playback of trained model.

Running python enjoy.py --algo her --env FetchReach-v1 gives me these results

Success rate: 100.00% Mean reward: -1.81

araffin commented 5 years ago

Hello, Thanks for the pr. Two things: please add the folder with the config file and add an exception in the tests so CI passes (the docker image does not contain mujoco, that's why it fails)

gautams3 commented 5 years ago

Done. For the exception in test_enjoy, I tried to add a pytest.mark.skipif line, but it depended on the parameter from pytest parameterization of trained_model here, which was giving me an error.

So I had to remove the model from the trained_models dictionary before it got to test_enjoy