Support sending using lengths to TBE instead of just offsets #2557
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…arded & Quantized EBC & PEA (pytorch#2553) Summary: Pull Request resolved: pytorch#2553 Reduce the number of TBEs created per Sharded & Quantized EBC & PEA. We realized we are creating a separate TBE for different datatype for every EBC or PEA module. However, FBGEMM's TBE module actually supports having embedding_tables with different data types all within the same TBE. This optimization can help reduce the # of inputs to the merge net after model split - which we realized was the portion with the most regression for TorchRec Inference Differential Revision: D64066446
Differential Revision: D64906767
facebook-github-bot
added
the
CLA Signed
|
This pull request was exported from Phabricator. Differential Revision: D64906767 |
PaulZhang12
added a commit
to PaulZhang12/FBGEMM
that referenced
this pull request
Nov 26, 2024
…#2557) Summary: Pull Request resolved: pytorch/torchrec#2557 Differential Revision: D64906767
facebook-github-bot
pushed a commit
to pytorch/FBGEMM
that referenced
this pull request
Nov 27, 2024
Summary: X-link: facebookresearch/FBGEMM#508 X-link: pytorch/torchrec#2557 Here we modify FBGEMM inference TBE to have the forward logic live in `_forward_impl`. This change makes it easy for subclasses of TBE to extend without having to call `super()`, which is not TorchScriptable. An example of a subclass using this is D66515313. Having subclasses TorchScriptable is vital for inference as TBE is generally not FX traced through. Reviewed By: sryap Differential Revision: D64906767 fbshipit-source-id: 41bab272c2611fc97dece7f89fdbbd820671843c
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Differential Revision: D64906767