[RLlib] RLTrainer is all you need. #31490
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Signed-off-by: Kourosh Hakhamaneshi <kourosh@anyscale.com>
kouroshHakha
requested review from
sven1977,
gjoliver,
avnishn,
ArturNiederfahrenhorst,
smorad,
maxpumperla and
krfricke
as code owners
2. multi-gpus tests pass now Signed-off-by: Kourosh Hakhamaneshi <kourosh@anyscale.com>
Signed-off-by: Kourosh Hakhamaneshi <kourosh@anyscale.com>
Signed-off-by: Kourosh Hakhamaneshi <kourosh@anyscale.com>
Signed-off-by: Kourosh Hakhamaneshi <kourosh@anyscale.com>
Signed-off-by: Kourosh Hakhamaneshi <kourosh@anyscale.com>
avnishn
reviewed
Jan 6, 2023
rllib/core/rl_trainer/rl_trainer.py
Outdated
| # rerun make_optimizers to update the params and optimizer | ||
| self.make_optimizers() | ||
|
|
||
| def make_module(self) -> RLModule: |
avnishn
approved these changes
Jan 6, 2023
Signed-off-by: Kourosh Hakhamaneshi <kourosh@anyscale.com>
kouroshHakha
commented
Jan 6, 2023
| @@ -5,19 +5,13 @@ | |||
| import unittest | |||
|
|
|||
| import ray | |||
There was a problem hiding this comment.
You can ignore everything that is under optim. Since these tests are removed from CI anyway.
Signed-off-by: Kourosh Hakhamaneshi <kourosh@anyscale.com>
Signed-off-by: Kourosh Hakhamaneshi <kourosh@anyscale.com>
|
Test failures do not seem relevant |
AmeerHajAli
pushed a commit
that referenced
this pull request
Jan 12, 2023
moved rl_optimizer logic into rl_trainer Signed-off-by: Kourosh Hakhamaneshi <kourosh@anyscale.com>
tamohannes
pushed a commit
to ju2ez/ray
that referenced
this pull request
Jan 25, 2023
moved rl_optimizer logic into rl_trainer Signed-off-by: Kourosh Hakhamaneshi <kourosh@anyscale.com> Signed-off-by: tmynn <hovhannes.tamoyan@gmail.com>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Signed-off-by: Kourosh Hakhamaneshi kourosh@anyscale.com
Why are these changes needed?
The RLOptimizer seems to have become this shallow module that just adds to the complexity of the system.
It basically is responsible for two things: 1) defining framework optimizers, 2) containing the loss logic.
defining the optimizer inside this module creates a lot of un-necessary complexities when it comes to multi-agent RLOptimizers. By moving this logic to RLTrainer we can get rid of these complexities.
Since compute_loss is also a state-less function we can easily move that to RLTrainer as well. Now all users have to do is to extend RLTrainer directly to customize the training phase of their algorithm. For example BCOptimizer will now become part of BCRLTrainer's implementation where all I have to do is to optionally override the
_configure_optimizer()and writecompute_loss. compute_loss will be written as a multi-agent loss. This is where the first-class-ness of MARL comes into play. It will make very complicated MARL communication patterns possible and also extremely easy.Related issue number
Checks
git commit -s) in this PR.scripts/format.shto lint the changes in this PR.