lm-sys / lm-sys/FastChat

the shape of params lm_head.wegiht is not compatible between base weight and delta weight

Open
#510 2 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
39.5k
Forks
4.8k
PR merge metrics
No merged PRs in 30d

Description

when running apply_delta.py of weight version v1.1, I find that the the shape of params lm_head.wegiht is [32000, 5120] in base weight, and is [32001, 5120] in delta weight. However, the shape of model.embed_tokends.weight are the same for these two weights.

Have anyone met this problem before?

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

Start with apply_delta.py and reproduce the v1.1 run using the reported base and delta weights. Inspect how lm_head.wegiht and model.embed_tokends.weight shapes are loaded and compared; done means identifying the cause of the [32000, 5120] versus [32001, 5120] mismatch and documenting or fixing the behavior.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
machine-learning
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.