Skip to content

[coati] add vf_coef argument for PPOTrainer#3318

Merged
ht-zhou merged 1 commit intohpcaitech:mainfrom
zhang-yi-chi:feature/vf_coef_for_ppo_trainer
Apr 11, 2023
Merged

[coati] add vf_coef argument for PPOTrainer#3318
ht-zhou merged 1 commit intohpcaitech:mainfrom
zhang-yi-chi:feature/vf_coef_for_ppo_trainer

Conversation

@zhang-yi-chi
Copy link
Copy Markdown
Contributor

@zhang-yi-chi zhang-yi-chi commented Mar 29, 2023

📌 Checklist before creating the PR

  • I have created an issue for this PR for traceability
  • The title follows the standard format: [doc/gemini/tensor/...]: A concise description
  • I have added relevant tags if possible for us to better distinguish different PRs

🚨 Issue number

#3317

📝 What does this PR do?

  1. Adding an argument vf_coef for value loss which is common in standard PPO implementation.
  2. Setting default multiplier of value loss to 0.5 as shown in the original paper.

💥 Checklist before requesting a review

  • I have linked my PR to an issue (instruction)
  • My issue clearly describes the problem/feature/proposal, with diagrams/charts/table/code if possible
  • I have performed a self-review of my code
  • I have added thorough tests.
  • I have added docstrings for all the functions/methods I implemented

⭐️ Do you enjoy contributing to Colossal-AI?

  • 🌝 Yes, I do.
  • 🌚 No, I don't.

Tell us more if you don't enjoy contributing to Colossal-AI.

@zhang-yi-chi zhang-yi-chi changed the title [coati]: add vf_coef argument for PPOTrainer [coati] add vf_coef argument for PPOTrainer Mar 29, 2023
Comment thread applications/Chat/coati/models/loss.py
@ht-zhou ht-zhou self-assigned this Apr 6, 2023
@ht-zhou ht-zhou merged commit e6a132a into hpcaitech:main Apr 11, 2023
jamesthesnake added a commit to jamesthesnake/ColossalAI that referenced this pull request Apr 11, 2023
[chat]: add vf_coef argument for PPOTrainer (hpcaitech#3318)
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