Add Claude Code coding-agent loop-owning GRPO examples - #6601
Draft
sergiopaniego wants to merge 1 commit into
Draft
Add Claude Code coding-agent loop-owning GRPO examples#6601sergiopaniego wants to merge 1 commit into
sergiopaniego wants to merge 1 commit into
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
What does this PR do?
Adds a loop-owning GRPO training example for the real Claude Code agent, mirroring the opencode example. Two scripts under
examples/scripts/openenv/:claude.py: local subprocess sandbox (one node, no tunnel)claude_code_hf_sandbox.py: one remote HF sandbox per rolloutClaude Code speaks the Anthropic Messages API, so an in-sandbox Anthropic-to-OpenAI shim sits in front of the interception proxy. Capture still happens at the vLLM seam, so training correctness is unchanged.
AsyncGRPOTrainer+HarnessRolloutWorker(harness_adapter=None) train on the trace, with a held-out DeepCoder verifier as the reward.Depends on the
claude_code_envenvironment (huggingface/OpenEnv#1030). Draft until that merges, then the uv headers point at the released package.Before submitting
AI writing disclosure