lm-sys / lm-sys/FastChat

Inference on peft trained falcon model gives nonsense

Open
#2,099 1 comment 2 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

I have a peft adapter model for a finetuned Falcon7b model, running it gives gibberish. The problem appears to be in get_generate_stream_function https://github.com/lm-sys/FastChat/blob/ae8abd20cbe1821100595d56d9ac4b137b31b53c/fastchat/model/model_adapter.py#L262 The model is detected as peft instead of falcon. Hardcoding is_falcon = True gives the expected output.

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 in fastchat/model/model_adapter.py at get_generate_stream_function around line 262, then reproduce inference with a PEFT adapter on a fine-tuned Falcon7b model. Compare model detection with the is_falcon=True behavior and verify that inference produces the expected output rather than gibberish.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
ai, machine-learning
Issue type
Bug
Difficulty
3/5
Estimated time
1-2 days
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
45/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.