NVIDIA-NeMo / NVIDIA-NeMo/Curator

Curator pip install: Transformers version conflict: nemo_toolkit[asr] requires transformers<4.55 vllm requires transformers>=4.55

Open
#1,340 2 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

bug
Dominant language
Python
Stars
1.8k
Forks
328
Avg merge
4d 5h
Merged PRs (30d)
30

Description

Describe the bug

ALL Extras - Install Only (pip) - Py3.12 / Py3.10
Root causes:
Transformers version conflict:
nemo_toolkit[asr] requires transformers<4.55
vllm requires transformers>=4.55
These are mutually exclusive - pip can't resolve this
flash-attn build issues:
Requires packaging module at build time
Requires torch pre-installed at build time

Why uv works but pip doesn't
uv uses a lockfile (uv.lock) that:
Pre-resolves all dependencies
Uses override-dependencies in pyproject.toml to force transformers>=4.55.2
Has no-build-isolation-package = ["flash-attn"] so torch is available during build
pip doesn't have these mechanisms - it tries to resolve on-the-fly and can't handle build-time dependencies or forced overrides.

Expected behavior

Expected Pip to install without issue

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 with pyproject.toml and uv.lock, then inspect the dependency declarations for nemo_toolkit[asr], vllm, transformers, and flash-attn. Reproduce the pip installation on Python 3.10 and 3.12, including the all extras, and compare it with the documented uv setup. Done means pip resolves the dependencies and completes the flash-attn build without the reported conflict or missing build requirements.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
build-system
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.