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

Refuce f16 #261

Merged
merged 19 commits into from
Aug 20, 2024
Merged

Refuce f16 #261

merged 19 commits into from
Aug 20, 2024

Conversation

CongMa13
Copy link
Collaborator

@CongMa13 CongMa13 commented Aug 7, 2024 •
edited
Loading

  • Support f16/f16, bf16/bf16, f16/f32, bf16/f32 reduction
  • Support independent C and D
  • Add another CK instance with InSrcVectorDim==1 which supports far right dim
  • Support reducing a tensor to a scalar
  • Support permutation of output of reduction
  • Fix bugs

@CongMa13
Support independent tensor C and D
a1b9782 
Tensor C and D can point to the same tensor or two distinct tensors
@CongMa13 CongMa13 marked this pull request as ready for review August 13, 2024 14:40
@CongMa13 CongMa13 requested a review from a team as a code owner August 19, 2024 18:06
cgmillette
cgmillette approved these changes Aug 19, 2024
View reviewed changes
Copy link
Collaborator

@cgmillette cgmillette left a comment

Choose a reason for hiding this comment

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

LGTM pending CI

@CongMa13 CongMa13 merged commit e778abc into ROCm:develop Aug 20, 2024
10 of 14 checks passed
@CongMa13 CongMa13 deleted the refuce_f16 branch August 20, 2024 20:49
CongMa13 added a commit to CongMa13/hipTensor that referenced this pull request Aug 20, 2024
@CongMa13
Reduce f16 (ROCm#261)
6ff07da 
- Support f16/f16, bf16/bf16, f16/f32, bf16/f32 reduction
- Use f32 as compute type for f16/f16, bf16/bf16
- Support independent C and D
- Add another CK instance with InSrcVectorDim==1 which supports far right dim
- Support reducing a tensor to a scalar
- Support permutation of output of reduction
- Rename label [Reduced Dims] to [Output Dims] in reduction test config file 
- Commnet out all test cases that beta != 0. Will add the test cases back when the CK fix is merged into amd_master
- Fix bugs
CongMa13 added a commit that referenced this pull request Sep 24, 2024
@CongMa13
Reduce f16 (#261) (#266)
d965ac3 
- Support f16/f16, bf16/bf16, f16/f32, bf16/f32 reduction
- Use f32 as compute type for f16/f16, bf16/bf16
- Support independent C and D
- Add another CK instance with InSrcVectorDim==1 which supports far right dim
- Support reducing a tensor to a scalar
- Support permutation of output of reduction
- Rename label [Reduced Dims] to [Output Dims] in reduction test config file 
- Commnet out all test cases that beta != 0. Will add the test cases back when the CK fix is merged into amd_master
- Fix bugs
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@cgmillette cgmillette cgmillette approved these changes

@bragadeesh bragadeesh Awaiting requested review from bragadeesh bragadeesh is a code owner

@mkarunan mkarunan Awaiting requested review from mkarunan

@dlangbe dlangbe Awaiting requested review from dlangbe dlangbe is a code owner

@afanfa afanfa Awaiting requested review from afanfa afanfa is a code owner

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
2 participants
@CongMa13 @cgmillette