[RLlib] Add separate learning rates for policy and `alpha` to SAC.#47078

Merged

sven1977 merged 43 commits intoray-project:masterfrom simonsays1980:add-actor-specific-learning-rate

Aug 21, 2024

+135-32

Commits on Aug 12, 2024

Added spearate learnign rates for policy, critic, and alpha to SAC. This imrpoves learning a bit.
simonsays1980
committed
Merge branch 'master' into add-actor-specific-learning-rate
sven1977
authored

Commits on Aug 13, 2024

Added an additional 'ciritc_lr', change 'policy_lr' to 'actor_lr', and set 'lr' to 'None' as requested in @sven1977 's review. Furthermore, changed all examples and tuning scripts.
simonsays1980
committed
Merge branch 'master' into add-actor-specific-learning-rate
simonsays1980
committed
Merge branch 'master' into add-actor-specific-learning-rate
simonsays1980
authored

Commits on Aug 14, 2024

Apply suggestions from code review
sven1977
authored

Commits on Aug 15, 2024

[RLlib; Offline RL] Implement twin-Q net option for CQL. (ray-project#47105 )
simonsays1980
committed
[core] remove unused GcsAio(Publisher|Subscriber) methods and subclasses. (ray-project#47057 )

rynewang
authored and
simonsays1980
committed
[Core] Fix a bug where we submit the actor creation task to the wrong submitter (ray-project#47109 )

jjyao
authored and
simonsays1980
committed
[doc][build] Update all changed files timestamp to latest (ray-project#47115 )

khluu
authored and
simonsays1980
committed
[serve] split test_proxy.py into unit and e2e tests (ray-project#47112 )

zcin
authored and
simonsays1980
committed
[Utility] add env_float utility into ray._private.ray_constants (ray-project#47117 )

hongpeng-guo
authored and
simonsays1980
committed
[Data] Fix progress bars not showing % progress (ray-project#47120 )

scottjlee
authored and
simonsays1980
committed
[data] change data17 to datal (ray-project#47082 )

aslonnie
authored and
simonsays1980
committed
[ci] change data build for all python versions to arrow 17 (ray-project#47121 )

can-anyscale
authored and
simonsays1980
committed
[doc][rllib] add missing public api references (ray-project#47111 )

can-anyscale
authored and
simonsays1980
committed
[Core] Clarify docstring for get_gpu_ids() that it is only called inside a task or actor

petern48
authored and
simonsays1980
committed
[doc][rllib] the rest of missing api references + lint checker (ray-project#47114 )

can-anyscale
authored and
simonsays1980
committed
Light up Ask AI button When Seach is Open (ray-project#47054 )

cristianjd
authored and
simonsays1980
committed
[serve] immediately send ping in router when receiving new replica set (ray-project#47053 )

zcin
authored and
simonsays1980
committed
[data] Add label to indicate if operator is backpressured (ray-project#47095 )

omatthew98
authored and
simonsays1980
committed
[Core] Add ray[adag] option to pip install (ray-project#47009 )

ruisearch42
authored and
simonsays1980
committed
[Doc] Run pre-commit on tune docs (ray-project#47108 )

peytondmurray
authored and
simonsays1980
committed
[release tests] update anyscale service utils (ray-project#46397 )

zcin
authored and
simonsays1980
committed
[core][experimental] Build an operation-based execution schedule for each actor to avoid deadlocks caused by NCCL operations (ray-project#46911 )

kevin85421
authored and
simonsays1980
committed
[serve] remove warnings about ongoing requests default change (ray-project#47085 )

zcin
authored and
simonsays1980
committed
[serve] __init__ functions have no return values (ray-project#47144 )

aslonnie
authored and
simonsays1980
committed
Merge branch 'master' into add-actor-specific-learning-rate
simonsays1980
committed
Merge branch 'master' into add-actor-specific-learning-rate
simonsays1980
committed
Turned off test 'self_play_with_policy_checkpoint' b/c it was failing constantly with TF in CI tests. This is old stack.
simonsays1980
committed

Commits on Aug 16, 2024

Commits on Aug 19, 2024

Commits on Aug 20, 2024