NVIDIA / NVIDIA/GenerativeAIExamples

Internal server error for role orders in LLM inference

Open
#113 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Jupyter Notebook
Stars
4.2k
Forks
1.1k
Avg merge
10h 15m
Merged PRs (30d)
1

Description

Hi,

I am getting 500s with description

chat messages must alternate roles between 'user' and 'assistant'.  Message may have a leading 'system' role message

and

"Internal Server Error\",\"status\":500,\"detail\":\"Last message role should be 'user'

I think this order validation is unnecessary. Most of the models (e.g. mixtrals, llama3, gemma etc) are perfectly fine with any order of roles. This claim can be validated on groq's playground: https://console.groq.com/playground

This issue currently breaks some existing patterns, like continuations (without the user explicitly saying "continue") or in some cases running agents with observations etc.

Contributor guide

Open the contributing guide

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 the validation that produces the quoted role-order and last-message errors, then reproduce a request with continuation or agent-observation messages. Done means supported non-alternating role sequences no longer return HTTP 500 while the intended validation behavior remains covered.

Written by the indexing model from the issue text.

Assessment

Domain
ai, api
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.