Skip to content

Add onnxruntime-ep-cuda output for the onnxruntime feedstock - #117

Draft
hmaarrfk wants to merge 1 commit into
conda-forge:mainfrom
hmaarrfk:onnxruntime-ep-cuda
Draft

Add onnxruntime-ep-cuda output for the onnxruntime feedstock#117
hmaarrfk wants to merge 1 commit into
conda-forge:mainfrom
hmaarrfk:onnxruntime-ep-cuda

Conversation

@hmaarrfk

@hmaarrfk hmaarrfk commented Aug 9, 2026

Copy link
Copy Markdown
Contributor

onnxruntime 1.28.0 can build its CUDA execution provider as a standalone, Python-independent plugin library (onnxruntime_BUILD_CUDA_EP_AS_PLUGIN=ON, loaded via the EP plugin API into any onnxruntime core >= 1.24.4).

conda-forge/onnxruntime-feedstock#198 restructures the feedstock accordingly: the onnxruntime/onnxruntime-cpp packages become CPU-only and a new onnxruntime-ep-cuda output (one build per platform x CUDA version, no Python dimension) provides GPU support. Its CI currently fails only output validation for the new name.

Discussion/design: conda-forge/onnxruntime-feedstock#196, #197, #198 (cc @cbourjau).

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant