Skip to content

scikit-learn-intelex-gpu is documented for conda-forge but missing from the channel #3211

Description

@cakedev0

Summary

The GPU installation docs currently recommend installing scikit-learn-intelex-gpu from conda-forge, but the package does not appear to exist in the live conda-forge index. This makes conda/pixi environment solves fail when following the documented conda installation path.

Docs page:
https://uxlfoundation.github.io/scikit-learn-intelex/latest/oneapi-gpu.html

The page shows:

conda install -c conda-forge scikit-learn-intelex-gpu

Reproducer

With a pixi environment using only conda-forge:

[workspace]
channels = ["conda-forge"]
platforms = ["linux-64"]

[feature.intel.dependencies]
python = "3.12.*"
scikit-learn = "*"
scikit-learn-intelex = "*"
scikit-learn-intelex-gpu = "*"

Running:

pixi run -e intel python -m sklbench --config configs/sklearnex.json

fails during solve with:

Cannot solve the request because of: No candidates were found for scikit-learn-intelex-gpu *.

A direct package search also fails:

pixi search scikit-learn-intelex-gpu
Using channels: conda-forge
Error: No packages found matching 'scikit-learn-intelex-gpu'

while the CPU package is available:

pixi search scikit-learn-intelex

returns scikit-learn-intelex-2025.11.0 for linux-64.

I also checked conda-forge channeldata.json on 2026-05-28: packages["scikit-learn-intelex-gpu"] is null, while packages["scikit-learn-intelex"] exists.

Expected behavior

Either:

  1. scikit-learn-intelex-gpu should be published to conda-forge for the documented platforms, or
  2. the docs should be updated to describe the currently supported installation path for GPU support.

For reference, PyPI does currently have scikit-learn-intelex-gpu==2026.0.0, with wheels for Linux and Windows.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions