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.

Sign up for GitHub

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

Jump to bottom

Implement CuBlas/MKL int8, float mixed precision gemm_batch #506

Open
AidanBeltonS opened this issue May 31, 2024 · 0 comments
Open

Implement CuBlas/MKL int8, float mixed precision gemm_batch #506

AidanBeltonS opened this issue May 31, 2024 · 0 comments
Labels
feature A request to add a new feature

Comments

@AidanBeltonS
Copy link
Contributor

AidanBeltonS commented May 31, 2024 •
edited

Summary

The CuBlas backend currently does not support the gemm_batch combination for (int8, int8, float, float).
This should be implemented as there is an equivalent gemm batch operation that can be used within CuBlas.

The MKLCPU/GPU backends also have this combination set to unsupported for the same precision issues.

Problem statement

This was not added in #466 due to precision issues when using adaptaiveCPP. Tests passed locally with DPC++.
See #466 for details on how to implement this.
@AidanBeltonS AidanBeltonS mentioned this issue May 31, 2024
Merged
4 tasks
@AidanBeltonS AidanBeltonS changed the title Implement CuBlas int8, float mixed precision gemm_batch Implement CuBlas/MKL int8, float mixed precision gemm_batch Jun 5, 2024
@Rbiessy Rbiessy added the feature A request to add a new feature label Jul 11, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
feature A request to add a new feature
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
2 participants
@Rbiessy @AidanBeltonS