[PR #1443] [MERGED] docs: install CPU PyTorch first on machines without an NVIDIA GPU #1405

New issue

Closed

opened 2026-10-18 14:19:27 +00:00 by bsakaguchi · 0 comments

bsakaguchi commented

2026-10-18 14:19:27 +00:00

Owner

📋 Pull Request Information

Original PR: https://github.com/open-webui/docs/pull/1443
Author: @silentoplayz
Created: 9/30/2026
Status: ✅ Merged
Merged: 10/4/2026
Merged by: @silentoplayz

Base: main ← Head: docs/pip-cpu-torch

📝 Commits (1)

db8bdff docs: install CPU PyTorch first on machines without an NVIDIA GPU

📊 Changes

1 file changed (+10 additions, -0 deletions)

View changed files

📝 docs/getting-started/quick-start/tab-python/_PythonCompat.md (+10 -0)

📄 Description

Summary

On Linux x86_64, pip install open-webui pulls in the default PyTorch build from PyPI, which depends on the NVIDIA CUDA libraries whatever GPU the machine has. On AMD or CPU-only machines those libraries are downloaded and never used (reported in https://github.com/open-webui/open-webui/issues/29490).

This adds a short tip under the Python version note on the Quick Start page, so it shows once above every Python install tab (pip, venv, uv, Conda): install the CPU build of PyTorch first in the same environment, or the ROCm build for GPU acceleration on AMD.

https://github.com/open-webui/open-webui/issues/29490

Checklist

I have reviewed the relevant documentation and matched the existing style.
This PR meets Open WebUI's contribution standards: it is accurate, relevant to users, narrowly scoped, maintainable, and not promotional content, advertising, lead generation, SEO placement, or a request to list a product, service, provider, integration, gateway, tool, or company primarily for visibility.
I understand that PRs that do not meet these standards may be closed without review and will not be merged. Repeated, low-quality, off-topic, promotional, or intentionally misleading submissions may result in the contributor being blocked from future participation in Open WebUI repositories.

Notes for reviewers

Checked in clean Python 3.11 and 3.12 containers with uv pip install --dry-run open-webui (0.11.4): the default resolve adds torch==2.14.0 and 18 nvidia-* and cuda-* packages. With torch 2.14.1+cpu installed first from the CPU index, the same dry run adds none and keeps the CPU build.

_{🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.}

## 📋 Pull Request Information **Original PR:** https://github.com/open-webui/docs/pull/1443 **Author:** [@silentoplayz](https://github.com/silentoplayz) **Created:** 9/30/2026 **Status:** ✅ Merged **Merged:** 10/4/2026 **Merged by:** [@silentoplayz](https://github.com/silentoplayz) **Base:** `main` ← **Head:** `docs/pip-cpu-torch` --- ### 📝 Commits (1) - [`db8bdff`](https://github.com/open-webui/docs/commit/db8bdffd68e27672d3388e889ce67431ed9d3aea) docs: install CPU PyTorch first on machines without an NVIDIA GPU ### 📊 Changes **1 file changed** (+10 additions, -0 deletions) <details> <summary>View changed files</summary> 📝 `docs/getting-started/quick-start/tab-python/_PythonCompat.md` (+10 -0) </details> ### 📄 Description ## Summary On Linux x86_64, `pip install open-webui` pulls in the default PyTorch build from PyPI, which depends on the NVIDIA CUDA libraries whatever GPU the machine has. On AMD or CPU-only machines those libraries are downloaded and never used (reported in https://github.com/open-webui/open-webui/issues/29490). This adds a short tip under the Python version note on the Quick Start page, so it shows once above every Python install tab (pip, venv, uv, Conda): install the CPU build of PyTorch first in the same environment, or the ROCm build for GPU acceleration on AMD. ## Related issue or discussion https://github.com/open-webui/open-webui/issues/29490 ## Checklist - [x] I have reviewed the relevant documentation and matched the existing style. - [x] This PR meets Open WebUI's contribution standards: it is accurate, relevant to users, narrowly scoped, maintainable, and not promotional content, advertising, lead generation, SEO placement, or a request to list a product, service, provider, integration, gateway, tool, or company primarily for visibility. - [x] I understand that PRs that do not meet these standards may be closed without review and will not be merged. Repeated, low-quality, off-topic, promotional, or intentionally misleading submissions may result in the contributor being blocked from future participation in Open WebUI repositories. ## Notes for reviewers Checked in clean Python 3.11 and 3.12 containers with `uv pip install --dry-run open-webui` (0.11.4): the default resolve adds `torch==2.14.0` and 18 `nvidia-*` and `cuda-*` packages. With `torch 2.14.1+cpu` installed first from the CPU index, the same dry run adds none and keeps the CPU build. --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>

bsakaguchi

2026-10-18 14:19:27 +00:00