Update logic to only track policy version in the main controller #181
Conversation
This file contains hidden or 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
meta-cla
bot
added
the
CLA Signed
joecummings
approved these changes
Sep 18, 2025
apps/grpo/main.py
Outdated
JenniferWang
force-pushed
the
change-policy-versioning-logic
branch
from
b8934b3 to
29a2143
Compare
JenniferWang
changed the title
joecummings
reviewed
Sep 19, 2025
joecummings
reviewed
Sep 19, 2025
src/forge/actors/replay_buffer.py
Outdated
There was a problem hiding this comment.
nit: Move this up to be the first method.
|
I'd like to test on #183 as well... |
joecummings
reviewed
Sep 19, 2025
src/forge/actors/replay_buffer.py
Outdated
There was a problem hiding this comment.
Similar to my other comment, could we leave this out? Or at least switch it to be debug only?
Ritesh1905
reviewed
Sep 19, 2025
Ritesh1905
approved these changes
Sep 19, 2025
JenniferWang
force-pushed
the
change-policy-versioning-logic
branch
from
eac4e78 to
a0ccc7e
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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Outdated suggestions cannot be applied.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.
Before
Policy version is tracked in two places: main controller and Policy Actor (which increments the version number upon
update_weightscall). This feels fragile and unnecessary.After
This PR is not to debate the need for a versioning number but rather assuming the concept is valid, only track it in the controller.
Tests
Run on both grpo and toy_rl. But mainly look at progress on the toy_rl
https://www.internalfb.com/phabricator/paste/view/P1954914879
No regression
Before: https://meta.wandb.io/jiyue/sumdigits-training/runs/omqmpkq1?nw=nwuserjiyue
After: https://meta.wandb.io/jiyue/sumdigits-training/runs/xv5dduh5?nw=nwuserjiyue
