What's Changed
- multi-constrained sbsrl safe cartpole experiment by @lucavignola in https://github.com/yardenas/safe-learning/pull/235
- Add rccar rae by @yardenas in https://github.com/yardenas/safe-learning/pull/236
- uncertainty critic and real data by @lucavignola in https://github.com/yardenas/safe-learning/pull/237
- Add tasks for nonepisodic humanoid and implementation via terminations by @yardenas in https://github.com/yardenas/safe-learning/pull/238
- Match performance sac by @yardenas in https://github.com/yardenas/safe-learning/pull/239
- Sbsrl priors g2g by @lucavignola in https://github.com/yardenas/safe-learning/pull/240
- Normalize disagreement by @lucavignola in https://github.com/yardenas/safe-learning/pull/241
- sbsrl offline by flipping the uncertainty constraint by @lucavignola in https://github.com/yardenas/safe-learning/pull/242
- Remove some files by @yardenas in https://github.com/yardenas/safe-learning/pull/243
- Load the behavior action when training the actor critic by @yardenas in https://github.com/yardenas/safe-learning/pull/244
- Terminate humanoid by @yardenas in https://github.com/yardenas/safe-learning/pull/246
- sbsrl new sgd update for SAC by @lucavignola in https://github.com/yardenas/safe-learning/pull/245
- compute sgd step in parallel over ensemble dimension by @lucavignola in https://github.com/yardenas/safe-learning/pull/247
- sbsrl_offline - sooper compatibility by @lucavignola in https://github.com/yardenas/safe-learning/pull/248
- Migrate to UV by @yardenas in https://github.com/yardenas/safe-learning/pull/250
UV migration, results for online RL done
Full Changelog: yardenas/safe-learning@release/0.2.1...release/0.2.2
Contributors
yardenas and lucavignola