SAC for Mujoco Environments (#6642)

This commit is contained in:
Michael Luo
2019-12-31 00:16:54 -08:00
committed by Eric Liang
parent cdc1ce4ebf
commit 1cb335487e
3 changed files with 48 additions and 7 deletions
+10 -2
View File
@@ -274,9 +274,17 @@ Soft Actor Critic (SAC)
SAC architecture (same as DQN)
RLlib's soft-actor critic implementation is ported from the `official SAC repo <https://github.com/rail-berkeley/softlearning>`__ to better integrate with RLlib APIs. Note that SAC has two fields to configure for custom models: ``policy_model`` and ``Q_model``, and currently has no support for non-continuous action distributions. It is also currently *experimental*.
RLlib's soft-actor critic implementation is ported from the `official SAC repo <https://github.com/rail-berkeley/softlearning>`__ to better integrate with RLlib APIs. Note that SAC has two fields to configure for custom models: ``policy_model`` and ``Q_model``, and currently has no support for non-continuous action distributions.
Tuned examples: `Pendulum-v0 <https://github.com/ray-project/ray/blob/master/rllib/tuned_examples/regression_tests/pendulum-sac.yaml>`__
Tuned examples: `Pendulum-v0 <https://github.com/ray-project/ray/blob/master/rllib/tuned_examples/regression_tests/pendulum-sac.yaml>`__, `HalfCheetah-v3 <https://github.com/ray-project/ray/blob/master/rllib/tuned_examples/halfcheetah-sac.yaml>`__
**MuJoCo results @500k steps:** `more details <https://github.com/ray-project/rl-experiments>`__
============= ========== ===================
MuJoCo env RLlib SAC Haarnoja et al SAC
============= ========== ===================
HalfCheetah 8752 ~9000
============= ========== ===================
**SAC-specific configs** (see also `common configs <rllib-training.html#common-parameters>`__):