Skip to content

Enhancements to allow FP8 constants - #121

Merged
vineet-g merged 3 commits into
apple:mainfrom
vineet-g:vineet/fp8_constants
Sep 25, 2026
Merged

vineet-g merged 3 commits into
apple:mainfrom
vineet-g:vineet/fp8_constants

Conversation

@vineet-g

@vineet-g vineet-g commented Sep 25, 2026 •

Copy link
Copy Markdown
Contributor

Support FP8 constants in full/empty/full_like lowering.

  • Numpy has no FP8 data types leading to failure
  • These data types are now handled via DenseElementsAttr.get_splat + coreai.ConstantOp

@gokulkrishna98

Copy link
Copy Markdown
Contributor

Hi @vineet-g, could you please provide informative description ?

@gokulkrishna98

gokulkrishna98 commented Sep 25, 2026 •

Copy link
Copy Markdown
Contributor

Hi Vineet, if possible, please add numerical validation test for the following change.

Comment thread coreai_torch/_utils.py
Comment thread coreai_torch/_utils.py Outdated
Comment thread coreai_torch/_utils.py Outdated
@vineet-g
vineet-g merged commit 3ec2506 into apple:main Sep 25, 2026
2 checks passed
@vineet-g
vineet-g deleted the vineet/fp8_constants branch September 25, 2026 20:33
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants