OpenBMB / OpenBMB/VoxCPM

is there more stable way to use for 5090 in window?

Open
#258 3 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
37.8k
Forks
4.3k
Avg merge
7m
Merged PRs (30d)
1

Description

hi, I got it worked for 5090 in window but it still appears problems after a few generations. I reseach it and found out it is a known limitation of torch.compile + CUDA Graphs when the inference runs inside Gradio's background worker thread (anyio.to_thread.run_sync).
The first 1–2 generations sometimes succeed (fresh compile), but subsequent ones hit the thread-local storage assertion. is there anyway I can get this worked with optimize=True (torch.compile + CUDA Graphs) at maximum speed on RTX 5090 in window?

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 reproducing the failure on an RTX 5090 on Windows with optimize=True, using the Gradio background worker path described in the report. Investigate the interaction between torch.compile, CUDA Graphs, and anyio.to_thread.run_sync across repeated generations; done means multiple generations remain stable at the requested optimization level.

Written by the indexing model from the issue text.

Assessment

Tech stack
python, pytorch
Domain
machine-learning
Issue type
Bug
Difficulty
5/5
Estimated time
Over a week
Activity status
Quiet
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.