SkyworkAI / SkyworkAI/Skywork-R1V

is language model frozen for whole training?

Open
#28 4 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
3.2k
Forks
282
PR merge metrics
No merged PRs in 30d

Description

Hi, thank you for the great work on Skywork R1V2 — the results are impressive.

I was reading the paper and had a question regarding the training setup. Specifically, it's not entirely clear whether the language model (QwQ-32B) was kept frozen during the entire training process, including both the MPO and GRPO stages.

From Section 3.1 and Table 4, it seems like the adapter-only configuration yields the best performance, suggesting that the LLM might have been frozen. However, this isn't stated explicitly in the paper.

Could you kindly confirm:
Was the language model completely frozen throughout the entire training process?

Thanks again for sharing the model and for your contributions to the open-source community!

Contributor guide

No contributing guide indexed for this repository

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Read Section 3.1 and Table 4 of the referenced paper, then inspect the repository's training documentation or configuration entry points for the MPO and GRPO stages. Done means confirming whether QwQ-32B is frozen throughout both stages and recording that clarification in the appropriate project documentation or issue response.

Written by the indexing model from the issue text.

Assessment

Domain
documentation, machine-learning
Issue type
Documentation
Difficulty
3/5
Estimated time
1-2 days
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.