[TorchRec][IR] Add IR serializer for KTRegroupAsDict Module #1900
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
facebook-github-bot
added
the
CLA Signed
|
This pull request was exported from Phabricator. Differential Revision: D56282744 |
TroyGarden
added a commit
to TroyGarden/torchrec
that referenced
this pull request
Apr 19, 2024
…port (pytorch#1900) Summary: reference: * D54009459 Differential Revision: D56282744
TroyGarden
force-pushed
the
export-D56282744
branch
from
deb11d0 to
ffbca7f
Compare
|
This pull request was exported from Phabricator. Differential Revision: D56282744 |
TroyGarden
added a commit
to TroyGarden/torchrec
that referenced
this pull request
Apr 19, 2024
…port (pytorch#1900) Summary: reference: * D54009459 Differential Revision: D56282744
TroyGarden
added a commit
to TroyGarden/torchrec
that referenced
this pull request
Apr 19, 2024
…port (pytorch#1900) Summary: reference: * D54009459 Differential Revision: D56282744
TroyGarden
force-pushed
the
export-D56282744
branch
from
ffbca7f to
3ff2015
Compare
|
This pull request was exported from Phabricator. Differential Revision: D56282744 |
TroyGarden
added a commit
to TroyGarden/torchrec
that referenced
this pull request
Apr 19, 2024
…port (pytorch#1900) Summary: reference: * D54009459 Differential Revision: D56282744
TroyGarden
force-pushed
the
export-D56282744
branch
from
3ff2015 to
ea9b83c
Compare
|
This pull request was exported from Phabricator. Differential Revision: D56282744 |
TroyGarden
added a commit
to TroyGarden/torchrec
that referenced
this pull request
Apr 19, 2024
…port (pytorch#1900) Summary: reference: * D54009459 Differential Revision: D56282744
TroyGarden
added a commit
to TroyGarden/torchrec
that referenced
this pull request
Apr 19, 2024
…port (pytorch#1900) Summary: reference: * D54009459 Differential Revision: D56282744
TroyGarden
force-pushed
the
export-D56282744
branch
from
ea9b83c to
628c83d
Compare
|
This pull request was exported from Phabricator. Differential Revision: D56282744 |
TroyGarden
force-pushed
the
export-D56282744
branch
from
628c83d to
b25c4f7
Compare
|
This pull request was exported from Phabricator. Differential Revision: D56282744 |
Summary: Pull Request resolved: pytorch#1900 # context * previously `KTRegroupAsDict` can't really supported by torch.export (IR) because this module has an intialization step as running the first batch. * during the export the `KTRegroupAsDict` module will be initialized by a fake_tensor which is wrong * if we initialize the module before torch.export, the device would be an issue. * another issue is that current torch.export [can't support conditional logic in training](https://pytorch.org/docs/stable/cond.html), where initialization step only runs once. > torch.cond is a prototype feature in PyTorch. It has limited support for input and output types and doesn’t support training currently. Please look forward to a more stable implementation in a future version of PyTorch. NOTE: this is more like a workaround solution, real solution needs support from pytorch compile for conditional logic # details * we treat the `KTRegroupAsDict` as another sparse_arch and do the model swap before and after torch.export. * more context: D59019375 Differential Revision: D56282744
TroyGarden
changed the title
|
This pull request was exported from Phabricator. Differential Revision: D56282744 |
TroyGarden
force-pushed
the
export-D56282744
branch
from
b25c4f7 to
15b0a4f
Compare
PaulZhang12
pushed a commit
that referenced
this pull request
Aug 19, 2024
Summary: Pull Request resolved: #1900 # context * previously `KTRegroupAsDict` can't really supported by torch.export (IR) because this module has an intialization step as running the first batch. * during the export the `KTRegroupAsDict` module will be initialized by a fake_tensor which is wrong * if we initialize the module before torch.export, the device would be an issue. * another issue is that current torch.export [can't support conditional logic in training](https://pytorch.org/docs/stable/cond.html), where initialization step only runs once. > torch.cond is a prototype feature in PyTorch. It has limited support for input and output types and doesn’t support training currently. Please look forward to a more stable implementation in a future version of PyTorch. NOTE: this is more like a workaround solution, real solution needs support from pytorch compile for conditional logic # details * we treat the `KTRegroupAsDict` as another sparse_arch and do the model swap before and after torch.export. * more context: D59019375 Reviewed By: PaulZhang12 Differential Revision: D56282744 fbshipit-source-id: b86f6eafa3d453735df6c9d00b33b16f70279dea
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.
context
KTRegroupAsDictcan't really supported by torch.export (IR) because this module has an intialization step as running the first batch.KTRegroupAsDictmodule will be initialized by a fake_tensor which is wrongNOTE: this is more like a workaround solution, real solution needs support from pytorch compile for conditional logic
details
KTRegroupAsDictas another sparse_arch and do the model swap before and after torch.export.