This repository has been archived by the owner on Jun 14, 2022. It is now read-only.

Commits

Commits on May 10, 2022

Normalize Wrapped Reward for PreferenceComparisons (HumanCompatibleAI#425 )

committedMay 10, 2022

Commits on May 6, 2022

Remove redundant component of Xdummy (HumanCompatibleAI#434 )
AdamGleave
committedMay 6, 2022

Commits on Apr 25, 2022

Use permutations tests when comparing episode rewards (HumanCompatibleAI#430 )

ernestum
and
AdamGleave
committedApr 25, 2022

Commits on Apr 22, 2022

Upgrade to python 3.8 and ubuntu 20.04 (HumanCompatibleAI#431 )

ernestum
and
AdamGleave
committedApr 22, 2022

Commits on Apr 19, 2022

Remove mean_return helper and replace by sb3.common.evaluation.evaluate_policy. (HumanCompatibleAI#429 )
ernestum
committedApr 19, 2022

Commits on Apr 1, 2022

Remove dead reward input normalization code (HumanCompatibleAI#426 )
AdamGleave
committedApr 1, 2022

Commits on Mar 29, 2022

Refactoring of Behavior Cloning (HumanCompatibleAI#422 )

ernestum
and
AdamGleave
committedMar 29, 2022

Commits on Mar 23, 2022

Set deterministic_policy of exploration_wrapper to False (HumanCompatibleAI#423 )
yawen-d
committedMar 23, 2022

Commits on Mar 17, 2022

Fix typo in Docker (HumanCompatibleAI#419 )
AdamGleave
committedMar 17, 2022

Commits on Mar 16, 2022

Add wandb_additional_info in wb.wandb_config() (HumanCompatibleAI#417 )

committedMar 16, 2022

Commits on Mar 11, 2022

align main_trainer default configs with scripts.train_preference_comparisons configs (HumanCompatibleAI#418 )
yawen-d
committedMar 11, 2022

Commits on Mar 8, 2022

MCE IRL: fix sample-based occupancy measure computation (HumanCompatibleAI#414 )
AdamGleave
committedMar 8, 2022

Commits on Mar 2, 2022

Overhaul of the examples folder (HumanCompatibleAI#397 )

ernestum
and
AdamGleave
committedMar 2, 2022

Commits on Feb 28, 2022

Refactor stay_prob to switch_prob in ExplorationWrapper (HumanCompatibleAI#411 )

yawen-d
and
AdamGleave
committedFeb 28, 2022
Bits and pieces (HumanCompatibleAI#409 )
ernestum
committedFeb 28, 2022

Commits on Feb 24, 2022

Eliminate duplicate saving of DAgger demonstrations (HumanCompatibleAI#403 )

AdamGleave
and
qxcv
committedFeb 24, 2022

Commits on Feb 15, 2022

Train experts on demand (HumanCompatibleAI#405 )
ernestum
committedFeb 15, 2022

Commits on Feb 12, 2022

Rename key of original environment reward in info dict of RewardVecEnvWrapper (HumanCompatibleAI#404 )
yulonglin
committedFeb 12, 2022

Commits on Feb 3, 2022

Upgrade package versions and fix CI errors (HumanCompatibleAI#400 )
AdamGleave
committedFeb 3, 2022

Commits on Feb 1, 2022

Improvements-weight_decay_reward_logging (HumanCompatibleAI#382 )

committedFeb 1, 2022

Commits on Jan 29, 2022

Normalize observations with model architecture rather than VecNormalize (HumanCompatibleAI#393 )
AdamGleave
committedJan 29, 2022

Commits on Jan 21, 2022

Tweaks to docs (HumanCompatibleAI#391 )
AdamGleave
committedJan 21, 2022

Commits on Jan 11, 2022

Small improvements to preference comparisons (HumanCompatibleAI#371 )

committedJan 11, 2022

Commits on Dec 22, 2021

Add integration of Weights and Biases (HumanCompatibleAI#367 )

yawen-d
and
AdamGleave
committedDec 22, 2021

Commits on Dec 21, 2021

Saving checkpoints periodically for preference comparisons (HumanCompatibleAI#375 )
yawen-d
committedDec 21, 2021

Commits on Oct 19, 2021

Move back to mainline Darglint now HumanCompatibleAI#176 is in v1.8.1 (HumanCompatibleAI#370 )
AdamGleave
committedOct 19, 2021

Commits on Oct 8, 2021

Commits on Sep 19, 2021

Commits on Sep 18, 2021

Run Darglint on CI (accidentally disabled) (HumanCompatibleAI#355 )
AdamGleave
committedSep 18, 2021

Commits on Sep 17, 2021

Move adversarial algorithms into their own package (HumanCompatibleAI#353 )
AdamGleave
committedSep 17, 2021