Add logic to stream weights in EmbeddingKVDB #2930
Open
Conversation
This file contains hidden or 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
…ytorch#2928) Summary: X-link: facebookresearch/FBGEMM#1138 X-link: pytorch/FBGEMM#4053 As titled, add this option all the way to gate the upcoming changes of raw embedding streaming in SSDTBE. Differential Revision: D73691088
Summary: Gated by enable_raw_embedding_streaming Add the logic to send the passed in tensors to `TrainingParameterServerService` thrift service in EmbeddingKVDB The passed in - `table_names` to get the table FQN when streaming - `table_offsets` to get the global row id across TBEs. - `table_sizes` to get size of each table in TBE to infer which table a specific row belongs to. - `ps_server_port` is the port that runs the local `TrainingParameterServerService` to stream tensors to. It creates a new thread `weights_stream_thread_` in EmbeddingKBDB to stream the weights out of trainers asynchronously. Differential Revision: D73792631
facebook-github-bot
added
the
CLA Signed
|
This pull request was exported from Phabricator. Differential Revision: D73792631 |
chouxi
added a commit
to chouxi/FBGEMM
that referenced
this pull request
Apr 30, 2025
Summary: X-link: pytorch/torchrec#2930 Gated by enable_raw_embedding_streaming Add the logic to send the passed in tensors to `TrainingParameterServerService` thrift service in EmbeddingKVDB The passed in - `table_names` to get the table FQN when streaming - `table_offsets` to get the global row id across TBEs. - `table_sizes` to get size of each table in TBE to infer which table a specific row belongs to. - `ps_server_port` is the port that runs the local `TrainingParameterServerService` to stream tensors to. It creates a new thread `weights_stream_thread_` in EmbeddingKBDB to stream the weights out of trainers asynchronously. Differential Revision: D73792631
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.
Summary:
Gated by enable_raw_embedding_streaming
Add the logic to send the passed in tensors to
TrainingParameterServerServicethrift service in EmbeddingKVDBThe passed in
table_namesto get the table FQN when streamingtable_offsetsto get the global row id across TBEs.table_sizesto get size of each table in TBE to infer which table a specific row belongs to.ps_server_portis the port that runs the localTrainingParameterServerServiceto stream tensors to.It creates a new thread
weights_stream_thread_in EmbeddingKBDB to stream the weights out of trainers asynchronously.Differential Revision: D73792631