Use CCCL's mdspan implementation#1605
Conversation
divyegala
added
breaking
| void extract_centers(raft::resources const& res, | ||
| const index<int64_t>& index, | ||
| raft::device_matrix_view<float, uint32_t, raft::row_major> cluster_centers); | ||
| raft::device_matrix_view<float, int64_t, raft::row_major> cluster_centers); |
There was a problem hiding this comment.
We might need to change the signatures of ivf-pq getters to int64_t instead of uint32_t, right? For example things like list_sizes.
There was a problem hiding this comment.
Can you share a link reference?
There was a problem hiding this comment.
yes so we have n_lists() for example
cuvs/cpp/include/cuvs/neighbors/ivf_pq.hpp
Line 406 in ba67db1
Do we want to change this signature to int64_t?
There was a problem hiding this comment.
In other words, what I am trying to say is that we might be considering an index-wide migration to int64_t from uint32_t for all extent types in IVF-PQ.
There was a problem hiding this comment.
@tfeher had made of aware of plans for this, but maybe that is outside the scope of this PR.
There was a problem hiding this comment.
n_lists() returns lists_.size() and lists_ is an std::vector
cuvs/cpp/include/cuvs/neighbors/ivf_pq.hpp
Line 501 in ba67db1
It should be returning size_t. Anyway, the extent type updates in this PR are not related to index<IdxT> where IdxT is the index type, not the extent type.
This reverts commit 0f43b18.
|
/merge |
github-project-automation
Bot
moved this from Todo
to Done
in Unstructured Data Processing
Based on rapidsai/raft#2836 and rapidsai/cuvs#1605 Authors: - Divye Gala (https://github.com/divyegala) Approvers: - Bradley Dice (https://github.com/bdice) URL: #7555
4 participants
Based on rapidsai/raft#2836