KEMBAR78
Add BFloat16 support to BatchedUnaryEmbeddingBag for FusedDebiasArch in VDD by billyzhong-meta · Pull Request #1832 · pytorch/FBGEMM · GitHub
Skip to content

Conversation

@billyzhong-meta
Copy link
Contributor

Summary: We are currently testing HAMMER fusion optimizations in VDD for four modules (DedicatedTaskMLPs, MMoE, MoEDense, DebiasArch). FusedDebiasArch relies on BatchedUnaryEmbeddingBag which does not currently support BFloat16, which is necessary for VDD. Thus, we add support for it in this diff

Reviewed By: jianyuh

Differential Revision: D46777003

…in VDD

Summary: We are currently testing HAMMER fusion optimizations in VDD for four modules (```DedicatedTaskMLPs```, ```MMoE```, ```MoEDense```, ```DebiasArch```). ```FusedDebiasArch``` relies on ```BatchedUnaryEmbeddingBag``` which does not currently support ```BFloat16```, which is necessary for VDD. Thus, we add support for it in this diff

Reviewed By: jianyuh

Differential Revision: D46777003

fbshipit-source-id: fe9eda4cc2d94dc0d820cc2443dec1a6c447c280
@netlify
Copy link

netlify bot commented Jun 15, 2023

Deploy Preview for pytorch-fbgemm-docs canceled.

Name Link
🔨 Latest commit a078872
🔍 Latest deploy log https://app.netlify.com/sites/pytorch-fbgemm-docs/deploys/648b8d391d3be50008e04eef

@facebook-github-bot
Copy link
Contributor

This pull request was exported from Phabricator. Differential Revision: D46777003

@facebook-github-bot
Copy link
Contributor

This pull request has been merged in 0e613f6.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants