8000 fix grpo lora split module by hjh0119 · Pull Request #3635 · modelscope/ms-swift · GitHub
[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Skip to content

fix grpo lora split module #3635

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Mar 25, 2025
Merged

fix grpo lora split module #3635

merged 1 commit into from
Mar 25, 2025

Conversation

hjh0119
Copy link
Collaborator
@hjh0119 hjh0119 commented Mar 24, 2025

PR type

  • Bug Fix
  • New Feature
  • Document Updates
  • More Models or Datasets Support

PR information

peft model may start with a prefix like model.language_model, so using startswith might not work as expected

#3483

Experiment results

Paste your experiment result here(if needed).

@hjh0119 hjh0119 merged commit 3232b69 into modelscope:main Mar 25, 2025
1 of 2 checks passed
@hjh0119 hjh0119 deleted the grpo-split-fix branch March 25, 2025 01:58
tastelikefeet added a commit to tastelikefeet/swift that referenced this pull request Mar 25, 2025
…0325

* commit '20cfa3396705a50230d5cb64850e54ba5043ee2c': (23 commits)
  compat vllm0.8.1 (modelscope#3656)
  fix grpo vllm tp (modelscope#3658)
  fix label_names (modelscope#3657)
  set grpo multi turn max tokens (modelscope#3655)
  update docs (modelscope#3653)
  fix grpo epsilon(modelscope#3652)
  Fix template torch_dtype (modelscope#3651)
  [grpo] separate the epsilon (modelscope#3599)
  fix grpo pt ddp (modelscope#3648)
  fix prm (modelscope#3647)
  grpo reset prefix cache (modelscope#3640)
  fix grpo warning (modelscope#3630)
  support qwen2_5_vl_32b (modelscope#3642)
  fix reward model (modelscope#3641)
  fix grpo lora split module (modelscope#3635)
  fix grpo cosine reward (modelscope#3638)
  Support deepseek v3 0324 (modelscope#3637)
  update docs (modelscope#3633)
  fix (modelscope#3632)
  support train_sampler_random (modelscope#3631)
  ...

# Conflicts:
#	docs/source/Instruction/GRPO.md
#	docs/source/Instruction/命令行参数.md
#	docs/source_en/Instruction/Command-line-parameters.md
#	docs/source_en/Instruction/GRPO.md
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.

2 participants
0