Describe the bug
arm64 builds of the cuVS benchmark images are failing like this:
The following packages are incompatible
└─ cuvs-bench-cpu =25.4 * is not installable because there are no viable options
├─ cuvs-bench-cpu [25.04.00a87|25.04.00a88|25.04.00a89|25.04.00a91|25.04.00a92] would require
│ └─ sklearn >=1.5 *, which does not exist (perhaps a missing channel);
└─ cuvs-bench-cpu 25.04.00a93 would require
└─ mkl =2023 *, which does not exist (perhaps a missing channel).
(example build link)
Steps/Code to reproduce bug
Look at any recent 25.04 cuVS benchmark builds here.
Expected behavior
N/A
Additional context
There are not any linux-aarch64 builds of mkl on conda-forge (for any verison): https://anaconda.org/conda-forge/mkl/files?version=2023.2.0
Looks like this was introduced by rapidsai/cuvs#260
I guess that this package is never installed in an arm64 environment in cuVS CI, and that suggestions like rapidsai/cuvs#260 (comment) were not applied everywhere
Describe the bug
arm64builds of the cuVS benchmark images are failing like this:(example build link)
Steps/Code to reproduce bug
Look at any recent 25.04 cuVS benchmark builds here.
Expected behavior
N/A
Additional context
There are not any
linux-aarch64builds ofmklonconda-forge(for any verison): https://anaconda.org/conda-forge/mkl/files?version=2023.2.0Looks like this was introduced by rapidsai/cuvs#260
I guess that this package is never installed in an arm64 environment in cuVS CI, and that suggestions like rapidsai/cuvs#260 (comment) were not applied everywhere