torchrec icon indicating copy to clipboard operation
torchrec copied to clipboard

Published 20 hours ago • verified publisher icon pytorch

provide correct named parameters and buffers for fused and dense kernels

Open joshuadeng opened this issue 3 years ago • 1 comments

Summary: For batched dense embedding we incorrectly provide a named_buffers implementation due to inheritance. For batched dense embeddings the weight values should be provided only through named_parameters rather than named_buffers for fused embeddings.

Differential Revision: D38599839

joshuadeng avatar Aug 11 '22 18:08 joshuadeng

This pull request was exported from Phabricator. Differential Revision: D38599839

facebook-github-bot avatar Aug 11 '22 18:08 facebook-github-bot

This pull request was exported from Phabricator. Differential Revision: D38599839

facebook-github-bot avatar Aug 15 '22 19:08 facebook-github-bot

This pull request was exported from Phabricator. Differential Revision: D38599839

facebook-github-bot avatar Aug 15 '22 20:08 facebook-github-bot

This pull request was exported from Phabricator. Differential Revision: D38599839

facebook-github-bot avatar Aug 17 '22 22:08 facebook-github-bot

This pull request was exported from Phabricator. Differential Revision: D38599839

facebook-github-bot avatar Aug 18 '22 00:08 facebook-github-bot