Fix the random handoff and change default param #14
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
… (mode, sandbox_slice, original/previous, expert_model).\n- Default external.mode=level_feedback; sandbox_slice=1 (supports 0/None/'all').\n- Handoff handled in CoMLRL trainer with strict modes; expose magrpo/grpo.handoff.\n- Update HumanEval/CHE splits (HE train 33:163, eval :32; CHE train 16:, eval :16).\n- Set output.save_final_model=false by default.\n- Set wandb.dir and output.base_dir to storage paths by trainer/mode:\n * ST GRPO: output_st_grpo, ST MAGRPO: output_st_magrpo\n * MT GRPO: output_mt_grpo, MT MAGRPO: output_mt_magrpo\n- Rename expert model key to external.expert_model (used only for expert_edits).\n- Simplify YAML comments to section headers only.\n- Read external.* in train_magrpo.py and train_grpo.py; defaults adjusted.\n- README: clarify external keys and sandbox_slice semantics.
…xceptions; minimal default tags
…train_magrpo.py and train_grpo.py
… avoid UnboundLocalError
ryankamiri
pushed a commit
that referenced
this pull request
Sep 24, 2025
* make random default * reset train num and levelfeedback as default * delete files no use * fix che train too less * Config and external overhaul:\n\n- Introduce unified external section (mode, sandbox_slice, original/previous, expert_model).\n- Default external.mode=level_feedback; sandbox_slice=1 (supports 0/None/'all').\n- Handoff handled in CoMLRL trainer with strict modes; expose magrpo/grpo.handoff.\n- Update HumanEval/CHE splits (HE train 33:163, eval :32; CHE train 16:, eval :16).\n- Set output.save_final_model=false by default.\n- Set wandb.dir and output.base_dir to storage paths by trainer/mode:\n * ST GRPO: output_st_grpo, ST MAGRPO: output_st_magrpo\n * MT GRPO: output_mt_grpo, MT MAGRPO: output_mt_magrpo\n- Rename expert model key to external.expert_model (used only for expert_edits).\n- Simplify YAML comments to section headers only.\n- Read external.* in train_magrpo.py and train_grpo.py; defaults adjusted.\n- README: clarify external keys and sandbox_slice semantics. * Remove unnecessary try/except; robust sandbox_slice parsing without exceptions; minimal default tags * Fix: define external_cfg before use; remove duplicate assignments in train_magrpo.py and train_grpo.py * Fix: handle dataset load failure in train_magrpo.py (return early) to avoid UnboundLocalError * Configs: reduce num_train_epochs by 20% (rounded) across all YAMLs
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.
No description provided.