You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Commit 925c2da
Browse filesBrowse the repository at this point in the historyBrowse files
Fix Java and Python CI failures from the Dataset Specs merge
- Java: upstream #2774 added FromSerializedBuilder.withOutputDataset(
DeserializeDataset), but this branch renamed DeserializeDataset to
DenseOwningDataset. Update the remaining references in CagraIndex,
the java22 CagraIndexImpl and CagraIndexDeprecatedBuilderIT.
- Python: common/dataset.pxd cimported cuvsKMeansType from
cuvs.cluster.kmeans.kmeans, which imports that module when
common.dataset initializes. That loops through cuvs.cluster ->
neighbors -> cagra back into the half-initialized common.dataset
(ImportError: cannot import name make_device_padded_dataset). Declare
the enum locally from cuvs/cluster/kmeans.h instead.
Verification
- Full cuvs Python suite on GPU against a local libcuvs build:
16663 passed, 168 skipped, 2 xfailed.
- Java not compiled locally (no JDK 22 or jextract).
0 commit comments