Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Un-scale output distances #1499

Merged
merged 6 commits into from
May 10, 2023

Conversation

achirkin
Copy link
Contributor

@achirkin achirkin commented May 9, 2023

Reverts the scaling of input data and queries that normally happens for int8/uint8 types before distance computation.
Reuses the ivf-pq post-processing step for that.

Along the way, replaces the custom fragment_scale helper with the utils::mapping that is shared among several ANN methods.

Solves #1457

@achirkin achirkin requested review from a team as code owners May 9, 2023 15:48
@achirkin achirkin added 3 - Ready for Review non-breaking Non-breaking change bug Something isn't working labels May 9, 2023
Copy link
Contributor

@tfeher tfeher left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks Artem for the fix, just a minor thing, otherwise it looks good to me.

Tagging @enp1s0 for visibility.

cpp/include/raft/neighbors/detail/cagra/cagra_search.cuh Outdated Show resolved Hide resolved
@cjnolet
Copy link
Member

cjnolet commented May 10, 2023

/merge

@rapids-bot rapids-bot bot merged commit 2014112 into rapidsai:branch-23.06 May 10, 2023
@ahendriksen
Copy link
Contributor

ahendriksen commented May 10, 2023

It looks like this PR increased the binary size of raft_conda_cpp_cuda11_x86_64.tar.gz by 100MB (from 650MB to 750MB).
See:

Does anybody understand why? It is a significant enough change that might be worth investigating.

EDIT: I just noticed that the binary sizes of the rapids-nightly channel are substantially smaller (~250MB). This might be due to bz2 compression.

@achirkin
Copy link
Contributor Author

achirkin commented May 10, 2023

The archives contain tests, right? We add two more sets of instantiations to cagra in tests here, which is a lot (also see #1459). Hopefully, #1428 will improve this?

@ahendriksen
Copy link
Contributor

Ah yes, that makes sense. I also don't see an increase in the wheels. Thanks for pointing out!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
3 - Ready for Review bug Something isn't working CMake cpp non-breaking Non-breaking change
Projects
None yet
Development

Successfully merging this pull request may close these issues.

4 participants