docs(installation): clarify NVIDIA driver vs CUDA Toolkit; add uv source install#318
Merged
Merged
Conversation
…rce install The GPU build of PyTorch bundles its own CUDA runtime, so users never need the CUDA Toolkit — only a recent NVIDIA driver. This is identical on Windows and Linux. The old docs told users to download and run the CUDA Toolkit installer and claimed CUDA 13 was unsupported (cu130 is now the stable default as of PyTorch 2.11/2.12). - Rewrite installation/cuda.md around "driver, not toolkit" with an nvidia-smi readiness check (it ships with the driver; its "CUDA Version" is the driver capability, not an installed toolkit) and driver-install steps for Windows and Linux. Note macOS uses MPS automatically. - installation.md: correct the manual-install GPU note (driver-only; Linux/ macOS get a GPU build automatically, Windows uses the CUDA index or --torch-backend auto). - installation.md: add a uv variant (uv venv && uv pip install .) to the manual-from-source section alongside the stock pip path, plus an editable- install note for people modifying the source. - mkdocs.yml: nav label "CUDA Setup" -> "GPU Acceleration". Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What & why
Installing the GPU build of PyTorch bundles its own CUDA runtime, so users never need the CUDA Toolkit — only a recent NVIDIA driver. This is identical on Windows and Linux. The existing docs were misleading: they told users to download and run NVIDIA's CUDA Toolkit installer, treated Windows and Linux as different, and claimed CUDA 13 was unsupported (cu130 is now PyTorch's stable default as of 2.11/2.12).
Changes
docs/installation/cuda.md— rewritten around "driver, not toolkit" (heading now NVIDIA GPU Acceleration):nvidia-smiis the readiness check, with the two clarifications that trip people up: it ships with the driver (not the toolkit), and its top-right "CUDA Version" is the driver's capability, not an installed toolkit (just confirm 12.x or newer).nouveau;ubuntu-drivers autoinstall/apt install nvidia-driver-XXX/ Driver Manager GUI).docs/installation.md— corrected the manual-install GPU note (driver-only; Linux/macOS get a GPU build automatically, Windows uses the CUDA index or--torch-backend auto); added a uv variant (uv venv && uv pip install .) to the manual-from-source section alongside the stockpippath, plus an editable-install note for people modifying the source.mkdocs.yml— nav label "CUDA Setup" → "GPU Acceleration".Docs-only; no code paths touched.
🤖 Generated with Claude Code