Skip to content

feat: enable FlashSAC multi-GPU training#648

Merged
TATP-233 merged 5 commits into
mainfrom
refactor/offpolicy-distributed-learner-contract
Jun 29, 2026
Merged

feat: enable FlashSAC multi-GPU training#648
TATP-233 merged 5 commits into
mainfrom
refactor/offpolicy-distributed-learner-contract

Conversation

@TATP-233

Copy link
Copy Markdown
Collaborator

Summary

  • add a learner capability contract for generic off-policy multi-GPU routing
  • enable FlashSAC on the shared MultiGPUOffPolicyRunner with sync_sgd/local_sgd hooks
  • keep FlashSAC reward normalization rank0-ordered with broadcast, and sync obs/global learner state
  • update EN/ZH docs and tests for FlashSAC multi-GPU dispatch

Validation

  • make test-all

@TATP-233 TATP-233 requested a review from caozx1110 as a code owner June 29, 2026 14:06
@TATP-233 TATP-233 merged commit 840e4d2 into main Jun 29, 2026
6 checks passed
@TATP-233 TATP-233 deleted the refactor/offpolicy-distributed-learner-contract branch June 29, 2026 14:51
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant