Update KJT stride calculation logic to be based off of inverse_indices for VBE KJTs. #3119
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
facebook-github-bot
added
the
CLA Signed
|
This pull request was exported from Phabricator. Differential Revision: D76997485 |
jd7-tr
added a commit
to jd7-tr/torchrec
that referenced
this pull request
Jun 20, 2025
…s for VBE KJTs. (pytorch#3119) Summary: For VBE KJTs, ppdate the `_maybe_compute_stride_kjt` logic to calculate stride based off of `inverse_indices` when its set. Currently, stride of VBE KJT with `stride_per_key_per_rank` is calculated as the max "stride per key". This is different from the batch size of the EBC output KeyedTensor which is based off of inverse_indices. This causes issues in IR module serialization. Reviewed By: TroyGarden Differential Revision: D76997485
TroyGarden
approved these changes
Jun 21, 2025
jd7-tr
added a commit
to jd7-tr/torchrec
that referenced
this pull request
Jun 21, 2025
…s for VBE KJTs. (pytorch#3119) Summary: For VBE KJTs, ppdate the `_maybe_compute_stride_kjt` logic to calculate stride based off of `inverse_indices` when its set. Currently, stride of VBE KJT with `stride_per_key_per_rank` is calculated as the max "stride per key". This is different from the batch size of the EBC output KeyedTensor which is based off of inverse_indices. This causes issues in IR module serialization. Reviewed By: TroyGarden Differential Revision: D76997485
jd7-tr
added a commit
to jd7-tr/torchrec
that referenced
this pull request
Jul 9, 2025
…s for VBE KJTs. (pytorch#3119) Summary: For VBE KJTs, ppdate the `_maybe_compute_stride_kjt` logic to calculate stride based off of `inverse_indices` when its set. Currently, stride of VBE KJT with `stride_per_key_per_rank` is calculated as the max "stride per key". This is different from the batch size of the EBC output KeyedTensor which is based off of inverse_indices. This causes issues in IR module serialization. Reviewed By: TroyGarden Differential Revision: D76997485
|
This pull request was exported from Phabricator. Differential Revision: D76997485 |
jd7-tr
added a commit
to jd7-tr/torchrec
that referenced
this pull request
Jul 9, 2025
…s for VBE KJTs. (pytorch#3119) Summary: For VBE KJTs, ppdate the `_maybe_compute_stride_kjt` logic to calculate stride based off of `inverse_indices` when its set. Currently, stride of VBE KJT with `stride_per_key_per_rank` is calculated as the max "stride per key". This is different from the batch size of the EBC output KeyedTensor which is based off of inverse_indices. This causes issues in IR module serialization. Reviewed By: TroyGarden Differential Revision: D76997485
|
This pull request was exported from Phabricator. Differential Revision: D76997485 |
jd7-tr
added a commit
to jd7-tr/torchrec
that referenced
this pull request
Jul 9, 2025
…s for VBE KJTs. (pytorch#3119) Summary: For VBE KJTs, ppdate the `_maybe_compute_stride_kjt` logic to calculate stride based off of `inverse_indices` when its set. Currently, stride of VBE KJT with `stride_per_key_per_rank` is calculated as the max "stride per key". This is different from the batch size of the EBC output KeyedTensor which is based off of inverse_indices. This causes issues in IR module serialization. Reviewed By: TroyGarden Differential Revision: D76997485
|
This pull request was exported from Phabricator. Differential Revision: D76997485 |
pytorch#2952) Summary: # Context * Currently torchrec IR serializer does not support exporting variable batch KJT, because the `stride_per_rank_per_rank` and `inverse_indices` fields are needed for deserializing VBE KJTs but they are included in the KJT's PyTree flatten/unflatten function. * The diff updates KJT's PyTree flatten/unflatten function to include `stride_per_rank_per_rank` and `inverse_indices`. # Ref Reviewed By: TroyGarden Differential Revision: D74295924
…s for VBE KJTs. (pytorch#3119) Summary: For VBE KJTs, ppdate the `_maybe_compute_stride_kjt` logic to calculate stride based off of `inverse_indices` when its set. Currently, stride of VBE KJT with `stride_per_key_per_rank` is calculated as the max "stride per key". This is different from the batch size of the EBC output KeyedTensor which is based off of inverse_indices. This causes issues in IR module serialization. Reviewed By: TroyGarden Differential Revision: D76997485
|
This pull request was exported from Phabricator. Differential Revision: D76997485 |
|
This pull request has been reverted by beaf74c. |
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:
For VBE KJTs, ppdate the
_maybe_compute_stride_kjtlogic to calculate stride based off ofinverse_indiceswhen its set.Currently, stride of VBE KJT with
stride_per_key_per_rankis calculated as the max "stride per key". This is different from the batch size of the EBC output KeyedTensor which is based off of inverse_indices. This causes issues in IR module serialization.Differential Revision: D76997485