Add auto-detection for AMD GPUs #14176
Merged
Conversation
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
charliermarsh
added
the
enhancement
charliermarsh
force-pushed
the
charlie/amd
branch
from
b968f25 to
0f3f9c2
Compare
charliermarsh
force-pushed
the
charlie/amd
branch
from
0f3f9c2 to
ed8862b
Compare
charliermarsh
temporarily deployed
to
uv-test-registries
GitHub Actions
Inactive
zanieb
reviewed
Jun 21, 2025
zanieb
approved these changes
Jun 21, 2025
Co-authored-by: Zanie Blue <[email protected]>
charliermarsh
temporarily deployed
to
uv-test-registries
GitHub Actions
Inactive
zanieb
added a commit
that referenced
this pull request
Jun 21, 2025
Allows `--torch-backend=auto` to detect AMD GPUs. The approach is fairly well-documented inline, but I opted for `rocm_agent_enumerator` over (e.g.) `rocminfo` since it seems to be the recommended approach for scripting: https://rocm.docs.amd.com/projects/rocminfo/en/latest/how-to/use-rocm-agent-enumerator.html. Closes #14086. ``` root@rocm-jupyter-gpu-mi300x1-192gb-devcloud-atl1:~# ./uv-linux-libc-11fb582c5c046bae09766ceddd276dcc5bb41218/uv pip install torch --torch-backend=auto Resolved 11 packages in 251ms Prepared 2 packages in 6ms Installed 11 packages in 257ms + filelock==3.18.0 + fsspec==2025.5.1 + jinja2==3.1.6 + markupsafe==3.0.2 + mpmath==1.3.0 + networkx==3.5 + pytorch-triton-rocm==3.3.1 + setuptools==80.9.0 + sympy==1.14.0 + torch==2.7.1+rocm6.3 + typing-extensions==4.14.0 ``` --------- Co-authored-by: Zanie Blue <[email protected]>
tmeijn
pushed a commit
to tmeijn/dotfiles
that referenced
this pull request
Jun 24, 2025
This MR contains the following updates: | Package | Update | Change | |---|---|---| | [astral-sh/uv](https://github.com/astral-sh/uv) | patch | `0.7.13` -> `0.7.14` | MR created with the help of [el-capitano/tools/renovate-bot](https://gitlab.com/el-capitano/tools/renovate-bot). **Proposed changes to behavior should be submitted there as MRs.** --- ### Release Notes <details> <summary>astral-sh/uv (astral-sh/uv)</summary> ### [`v0.7.14`](https://github.com/astral-sh/uv/blob/HEAD/CHANGELOG.md#0714) [Compare Source](astral-sh/uv@0.7.13...0.7.14) ##### Enhancements - Add XPU to `--torch-backend` ([#​14172](astral-sh/uv#14172)) - Add ROCm backends to `--torch-backend` ([#​14120](astral-sh/uv#14120)) - Remove preview label from `--torch-backend` ([#​14119](astral-sh/uv#14119)) - Add `[tool.uv.dependency-groups].mygroup.requires-python` ([#​13735](astral-sh/uv#13735)) - Add auto-detection for AMD GPUs ([#​14176](astral-sh/uv#14176)) - Show retries for HTTP status code errors ([#​13897](astral-sh/uv#13897)) - Support transparent Python patch version upgrades ([#​13954](astral-sh/uv#13954)) - Warn on empty index directory ([#​13940](astral-sh/uv#13940)) - Publish to DockerHub ([#​14088](astral-sh/uv#14088)) ##### Performance - Make cold resolves about 10% faster ([#​14035](astral-sh/uv#14035)) ##### Bug fixes - Don't use walrus operator in interpreter query script ([#​14108](astral-sh/uv#14108)) - Fix handling of changes to `requires-python` ([#​14076](astral-sh/uv#14076)) - Fix implied `platform_machine` marker for `win_amd64` platform tag ([#​14041](astral-sh/uv#14041)) - Only update existing symlink directories on preview uninstall ([#​14179](astral-sh/uv#14179)) - Serialize Python requests for tools as canonicalized strings ([#​14109](astral-sh/uv#14109)) - Support netrc and same-origin credential propagation on index redirects ([#​14126](astral-sh/uv#14126)) - Support reading `dependency-groups` from pyproject.tomls with no `[project]` ([#​13742](astral-sh/uv#13742)) - Handle an existing shebang in `uv init --script` ([#​14141](astral-sh/uv#14141)) - Prevent concurrent updates of the environment in `uv run` ([#​14153](astral-sh/uv#14153)) - Filter managed Python distributions by platform before querying when included in request ([#​13936](astral-sh/uv#13936)) ##### Documentation - Replace cuda124 with cuda128 ([#​14168](astral-sh/uv#14168)) - Document the way member sources shadow workspace sources ([#​14136](astral-sh/uv#14136)) - Sync documented PyTorch integration index for CUDA and ROCm versions from PyTorch website ([#​14100](astral-sh/uv#14100)) </details> --- ### Configuration 📅 **Schedule**: Branch creation - At any time (no schedule defined), Automerge - At any time (no schedule defined). 🚦 **Automerge**: Disabled by config. Please merge this manually once you are satisfied. ♻ **Rebasing**: Whenever MR becomes conflicted, or you tick the rebase/retry checkbox. 🔕 **Ignore**: Close this MR and you won't be reminded about this update again. --- - [ ] <!-- rebase-check -->If you want to rebase/retry this MR, check this box --- This MR has been generated by [Renovate Bot](https://github.com/renovatebot/renovate). <!--renovate-debug:eyJjcmVhdGVkSW5WZXIiOiI0MC42Mi4xIiwidXBkYXRlZEluVmVyIjoiNDAuNjIuMSIsInRhcmdldEJyYW5jaCI6Im1haW4iLCJsYWJlbHMiOlsiUmVub3ZhdGUgQm90Il19-->
sai-rekhawar
pushed a commit
to sai-rekhawar/cloe-nessy-py
that referenced
this pull request
Jul 1, 2025
This PR contains the following updates: | Package | Type | Update | Change | |---|---|---|---| | [ghcr.io/astral-sh/uv](https://github.com/astral-sh/uv) | final | patch | `0.7.13` -> `0.7.15` | --- ### Release Notes <details> <summary>astral-sh/uv (ghcr.io/astral-sh/uv)</summary> ### [`v0.7.15`](https://github.com/astral-sh/uv/blob/HEAD/CHANGELOG.md#0715) [Compare Source](astral-sh/uv@0.7.14...0.7.15) ##### Enhancements - Consistently use `Ordering::Relaxed` for standalone atomic use cases ([#​14190](astral-sh/uv#14190)) - Warn on ambiguous relative paths for `--index` ([#​14152](astral-sh/uv#14152)) - Skip GitHub fast path when rate-limited ([#​13033](astral-sh/uv#13033)) - Preserve newlines in `schema.json` descriptions ([#​13693](astral-sh/uv#13693)) ##### Bug fixes - Add check for using minor version link when creating a venv on Windows ([#​14252](astral-sh/uv#14252)) - Strip query parameters when parsing source URL ([#​14224](astral-sh/uv#14224)) ##### Documentation - Add a link to PyPI FAQ to clarify what per-project token is ([#​14242](astral-sh/uv#14242)) ##### Preview features - Allow symlinks in the build backend ([#​14212](astral-sh/uv#14212)) ### [`v0.7.14`](https://github.com/astral-sh/uv/blob/HEAD/CHANGELOG.md#0714) [Compare Source](astral-sh/uv@0.7.13...0.7.14) ##### Enhancements - Add XPU to `--torch-backend` ([#​14172](astral-sh/uv#14172)) - Add ROCm backends to `--torch-backend` ([#​14120](astral-sh/uv#14120)) - Remove preview label from `--torch-backend` ([#​14119](astral-sh/uv#14119)) - Add `[tool.uv.dependency-groups].mygroup.requires-python` ([#​13735](astral-sh/uv#13735)) - Add auto-detection for AMD GPUs ([#​14176](astral-sh/uv#14176)) - Show retries for HTTP status code errors ([#​13897](astral-sh/uv#13897)) - Support transparent Python patch version upgrades ([#​13954](astral-sh/uv#13954)) - Warn on empty index directory ([#​13940](astral-sh/uv#13940)) - Publish to DockerHub ([#​14088](astral-sh/uv#14088)) ##### Performance - Make cold resolves about 10% faster ([#​14035](astral-sh/uv#14035)) ##### Bug fixes - Don't use walrus operator in interpreter query script ([#​14108](astral-sh/uv#14108)) - Fix handling of changes to `requires-python` ([#​14076](astral-sh/uv#14076)) - Fix implied `platform_machine` marker for `...
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
Summary
Allows
--torch-backend=autoto detect AMD GPUs. The approach is fairly well-documented inline, but I opted forrocm_agent_enumeratorover (e.g.)rocminfosince it seems to be the recommended approach for scripting: https://rocm.docs.amd.com/projects/rocminfo/en/latest/how-to/use-rocm-agent-enumerator.html.Closes #14086.
Test Plan