We're having issues with our DGL dependency since DGL uses conda labels to split CUDA version so we need a different channel specification per CUDA version, which dependencies.yaml currently does not allow. By adding support, we would be able to better handle DGL testing for CUDA 12.
Related: dmlc/dgl#7344 rapidsai/cugraph#4346
We're having issues with our DGL dependency since DGL uses conda labels to split CUDA version so we need a different channel specification per CUDA version, which
dependencies.yamlcurrently does not allow. By adding support, we would be able to better handle DGL testing for CUDA 12.Related: dmlc/dgl#7344 rapidsai/cugraph#4346