lm-sys / lm-sys/FastChat

trust_remote_code=True

Open
#2,034 6 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

Some models will not run and I the error to set trust_remote_code=true.

Some of the adapters in model_adapters include this flag but many don't.

The problem seems to be more prevalent in running quantized and modified models - TheBloke/Vicuna-7B-v1-3-SuperHOT-8K-fp16 for example.

One of the issues I cannot tell which of the adapters is being selected. I don't know Python well enough to read through code in detail.

Any clues would be much appreciated.

Edit: This is from the CLI when running either the cli or the worker.

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 by tracing model_adapters and the CLI and worker entry points to determine which adapter is selected for the reported quantized or modified models. Compare the adapters that include trust_remote_code=True with those that do not, then verify the affected models can run through both CLI paths. The issue does not name specific files or tests.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
cli, 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.