Unstructured-IO / Unstructured-IO/unstructured-api

Why do I get "Cannot Initialize NVML"? How can I initialize the GPU kernel?

Open
#418 1 comment 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
951
Forks
192
Avg merge
7d 8h
Merged PRs (30d)
2

Description

Describe the bug
A clear and concise description of what the bug is.

When I run unstrctured docker container locally, I get this error:

2024-05-20 12:44:22 /home/notebook-user/.local/lib/python3.10/site-packages/torch/cuda/init.py:619: UserWarning: Can't initialize NVML
2024-05-20 12:44:22 warnings.warn("Can't initialize NVML")

I assume that its because its running the CPU version of unstructured and not the GPU version. How can i fix this? running in a GPU would help with hi-res parsing

To Reproduce
Please provide as much info as possible:

  • Filetype:
  • Any additional API parameters:

Environment:

  • Using the hosted API or self hosting?
  • How are you calling the API? (Langchain, SDKs, cUrl, etc.)

Additional context
Add any other context about the problem here.

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 reproducing the warning in the unstructured Docker container and inspect how the container exposes CUDA and NVML to PyTorch. The issue is complete when the GPU execution path is either made to work or the required runtime configuration is documented, with a reproducible environment and file type included.

Written by the indexing model from the issue text.

Assessment

Tech stack
docker, python, pytorch
Domain
infrastructure, machine-learning
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
20/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.