Uh oh!
There was an error while loading. Please reload this page.
Serialize NamedData in PTE file - #8835
Merged
Merged
Conversation
1. Serialize NamedData in PTE file
2. Add NamedDataStore to EdgeProgramManager
---
Serializing NamedData is slightly different to constant/delegate data as each segment comes with its own alignment.
**An example:**
Given NamedData = {"key": data}. Data is 250 bytes.
- BackendA requires data with alignment=3
- BackendB requires data with alignment=4
Then, data0 should be serialized with alignment of lcm(3, 4) = 12
At serialization, ExecuTorch has a 'segment_alignment' that defaults to 128. Data is now serialized to lcm(12, 128) = 384.
Inside the DataSegment, we want to store the original size of the data (250). The offset of the subsequent DataSegment would be 384 bytes after the start of this one.
**Design**
Introduce a new dataclass 'AlignedData' that stores the buffer and any alignment that's required. This is used when assembling Program.segments to ensure we get lcm(buffer_alignment, segment_alignment).
Note: The default segment_alignment can be overridden inside 'ExecutorchBackendConfig'.
Differential Revision: [D69764150](https://our.internmc.facebook.com/intern/diff/D69764150/)
[ghstack-poisoned]lucylq
requested review from
JacobSzwejbka, larryliu0820 and tarun292
as code ownersFebruary 28, 2025 18:22
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/8835
Note: Links to docs will display an error until the docs builds have been completed. ❌ 1 Cancelled JobAs of commit 521a5c8 with merge base 4df0ade ( CANCELLED JOB - The following job was cancelled. Please retry:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
facebook-github-bot
commented
Feb 28, 2025
Contributor
This pull request was exported from Phabricator. Differential Revision: D69764150 |
mcr229
approved these changes
Feb 28, 2025
lucylqtemporarily deployed
to
cherry-pick-bot
February 28, 2025 21:41 — with 
GitHub Actions
Inactive
lucylq added a commit
that referenced
this pull request
Feb 28, 2025
1. Serialize NamedData in PTE file
2. Add NamedDataStore to EdgeProgramManager
---
Serializing NamedData is slightly different to constant/delegate data as each segment comes with its own alignment.
**An example:**
Given NamedData = {"key": data}. Data is 250 bytes.
- BackendA requires data with alignment=3
- BackendB requires data with alignment=4
Then, data0 should be serialized with alignment of lcm(3, 4) = 12
At serialization, ExecuTorch has a 'segment_alignment' that defaults to 128. Data is now serialized to lcm(12, 128) = 384.
Inside the DataSegment, we want to store the original size of the data (250). The offset of the subsequent DataSegment would be 384 bytes after the start of this one.
**Design**
Introduce a new dataclass 'AlignedData' that stores the buffer and any alignment that's required. This is used when assembling Program.segments to ensure we get lcm(buffer_alignment, segment_alignment).
Note: The default segment_alignment can be overridden inside 'ExecutorchBackendConfig'.
Differential Revision: [D69764150](https://our.internmc.facebook.com/intern/diff/D69764150/)
ghstack-source-id: 269030449
Pull Request resolved: #8835
Co-authored-by: lucylq <lfq@meta.com>
This was referenced Feb 28, 2025
kedarnath03 pushed a commit
to kedarnath03/executorch
that referenced
this pull request
Jun 25, 2025
1. Serialize NamedData in PTE file
2. Add NamedDataStore to EdgeProgramManager
---
Serializing NamedData is slightly different to constant/delegate data as each segment comes with its own alignment.
**An example:**
Given NamedData = {"key": data}. Data is 250 bytes.
- BackendA requires data with alignment=3
- BackendB requires data with alignment=4
Then, data0 should be serialized with alignment of lcm(3, 4) = 12
At serialization, ExecuTorch has a 'segment_alignment' that defaults to 128. Data is now serialized to lcm(12, 128) = 384.
Inside the DataSegment, we want to store the original size of the data (250). The offset of the subsequent DataSegment would be 384 bytes after the start of this one.
**Design**
Introduce a new dataclass 'AlignedData' that stores the buffer and any alignment that's required. This is used when assembling Program.segments to ensure we get lcm(buffer_alignment, segment_alignment).
Note: The default segment_alignment can be overridden inside 'ExecutorchBackendConfig'.
Pull Request resolved: pytorch/executorch#8835
Internal:
<< DO NOT EDIT BELOW THIS LINE >>
**GitHub Author**: pytorchbot <soumith+bot@pytorch.org>
**GitHub Repo**: [pytorch/executorch](https://github.com/pytorch/executorch)
**GitHub Pull Request**: [#8835](pytorch/executorch#8835)
Initially generated by: https://www.internalfb.com/intern/sandcastle/job/18014400207802303/
This was imported as part of a Diff Train.
Please review this as soon as possible. Since it is a direct copy of a commit on
GitHub, there shouldn't be much to do.
diff-train-source-id: 314ab18
Co-authored-by: lucylq <lfq@meta.com>
Differential Revision: [D69764150](https://our.internmc.facebook.com/intern/diff/D69764150/)
ghstack-source-id: 269082799kedarnath03 pushed a commit
to kedarnath03/executorch
that referenced
this pull request
Jun 25, 2025
1. Serialize NamedData in PTE file
2. Add NamedDataStore to EdgeProgramManager
---
Serializing NamedData is slightly different to constant/delegate data as each segment comes with its own alignment.
**An example:**
Given NamedData = {"key": data}. Data is 250 bytes.
- BackendA requires data with alignment=3
- BackendB requires data with alignment=4
Then, data0 should be serialized with alignment of lcm(3, 4) = 12
At serialization, ExecuTorch has a 'segment_alignment' that defaults to 128. Data is now serialized to lcm(12, 128) = 384.
Inside the DataSegment, we want to store the original size of the data (250). The offset of the subsequent DataSegment would be 384 bytes after the start of this one.
**Design**
Introduce a new dataclass 'AlignedData' that stores the buffer and any alignment that's required. This is used when assembling Program.segments to ensure we get lcm(buffer_alignment, segment_alignment).
Note: The default segment_alignment can be overridden inside 'ExecutorchBackendConfig'.
Differential Revision: [D69764150](https://our.internmc.facebook.com/intern/diff/D69764150/)
ghstack-source-id: 269030449
Pull Request resolved: pytorch/executorch#8835
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
Sign up for freeto join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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.
Stack from ghstack (oldest at bottom):
Serializing NamedData is slightly different to constant/delegate data as each segment comes with its own alignment.
An example:
Given NamedData = {"key": data}. Data is 250 bytes.
Then, data0 should be serialized with alignment of lcm(3, 4) = 12
At serialization, ExecuTorch has a 'segment_alignment' that defaults to 128. Data is now serialized to lcm(12, 128) = 384.
Inside the DataSegment, we want to store the original size of the data (250). The offset of the subsequent DataSegment would be 384 bytes after the start of this one.
Design
Introduce a new dataclass 'AlignedData' that stores the buffer and any alignment that's required. This is used when assembling Program.segments to ensure we get lcm(buffer_alignment, segment_alignment).
Note: The default segment_alignment can be overridden inside 'ExecutorchBackendConfig'.
Differential Revision: D69764150