[ExecuTorch][Weight Sharing][XNNPACK] load named data map data for xnnpack#9152
Merged
facebook-github-bot merged 5 commits intogh/mcr229/8/basefrom Mar 14, 2025
Merged
[ExecuTorch][Weight Sharing][XNNPACK] load named data map data for xnnpack#9152facebook-github-bot merged 5 commits intogh/mcr229/8/basefrom
facebook-github-bot merged 5 commits intogh/mcr229/8/basefrom
Conversation
…npack If data is serialized into the NamedDataMap, then we overload getConstantDataPtr to retrieve the data from the named data map. This should be done in a Backwards Compatible way. Meaning if no data is serialized into the named data map, then we are still loading the data from the flatbuffer payload. Since the runtime change here is being made before the AoT changes, All CI on this diff by itself should test that the changes made here are backwards compatitble. Note: We do not resolve Runtime Memory usage at this point. WeightCache will be implemented in the next diff. Meaning If we load via the same key across different methods, we still pack twice and allocate two instances for the packed weights. Differential Revision: [D70315209](https://our.internmc.facebook.com/intern/diff/D70315209/) [ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/9152
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 935f105 with merge base 630d0cc ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
This was referenced Mar 11, 2025
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D70315209 |
…data for xnnpack" If data is serialized into the NamedDataMap, then we overload getConstantDataPtr to retrieve the data from the named data map. This should be done in a Backwards Compatible way. Meaning if no data is serialized into the named data map, then we are still loading the data from the flatbuffer payload. Since the runtime change here is being made before the AoT changes, All CI on this diff by itself should test that the changes made here are backwards compatitble. Note: We do not resolve Runtime Memory usage at this point. WeightCache will be implemented in the next diff. Meaning If we load via the same key across different methods, we still pack twice and allocate two instances for the packed weights. Differential Revision: [D70315209](https://our.internmc.facebook.com/intern/diff/D70315209/) [ghstack-poisoned]
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D70315209 |
…data for xnnpack" If data is serialized into the NamedDataMap, then we overload getConstantDataPtr to retrieve the data from the named data map. This should be done in a Backwards Compatible way. Meaning if no data is serialized into the named data map, then we are still loading the data from the flatbuffer payload. Since the runtime change here is being made before the AoT changes, All CI on this diff by itself should test that the changes made here are backwards compatitble. Note: We do not resolve Runtime Memory usage at this point. WeightCache will be implemented in the next diff. Meaning If we load via the same key across different methods, we still pack twice and allocate two instances for the packed weights. Differential Revision: [D70315209](https://our.internmc.facebook.com/intern/diff/D70315209/) [ghstack-poisoned]
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D70315209 |
kirklandsign
approved these changes
Mar 13, 2025
…data for xnnpack" If data is serialized into the NamedDataMap, then we overload getConstantDataPtr to retrieve the data from the named data map. This should be done in a Backwards Compatible way. Meaning if no data is serialized into the named data map, then we are still loading the data from the flatbuffer payload. Since the runtime change here is being made before the AoT changes, All CI on this diff by itself should test that the changes made here are backwards compatitble. Note: We do not resolve Runtime Memory usage at this point. WeightCache will be implemented in the next diff. Meaning If we load via the same key across different methods, we still pack twice and allocate two instances for the packed weights. Differential Revision: [D70315209](https://our.internmc.facebook.com/intern/diff/D70315209/) [ghstack-poisoned]
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D70315209 |
…data for xnnpack" If data is serialized into the NamedDataMap, then we overload getConstantDataPtr to retrieve the data from the named data map. This should be done in a Backwards Compatible way. Meaning if no data is serialized into the named data map, then we are still loading the data from the flatbuffer payload. Since the runtime change here is being made before the AoT changes, All CI on this diff by itself should test that the changes made here are backwards compatitble. Note: We do not resolve Runtime Memory usage at this point. WeightCache will be implemented in the next diff. Meaning If we load via the same key across different methods, we still pack twice and allocate two instances for the packed weights. Differential Revision: [D70315209](https://our.internmc.facebook.com/intern/diff/D70315209/) [ghstack-poisoned]
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D70315209 |
SS-JIA
pushed a commit
that referenced
this pull request
Mar 15, 2025
…npack (#9294) This PR was created by the merge bot to help merge the original PR into the main branch. ghstack PR number: #9152 by @mcr229 ^ Please use this as the source of truth for the PR details, comments, and reviews ghstack PR base: https://github.com/pytorch/executorch/tree/gh/mcr229/8/base ghstack PR head: https://github.com/pytorch/executorch/tree/gh/mcr229/8/head Merge bot PR base: https://github.com/pytorch/executorch/tree/gh/mcr229/7/orig Merge bot PR head: https://github.com/pytorch/executorch/tree/gh/mcr229/8/orig @diff-train-skip-merge --------- Co-authored-by: Max Ren <maxren@meta.com>
SS-JIA
pushed a commit
that referenced
this pull request
Mar 15, 2025
…npack Pull Request resolved: #9152 If data is serialized into the NamedDataMap, then we overload getConstantDataPtr to retrieve the data from the named data map. This should be done in a Backwards Compatible way. Meaning if no data is serialized into the named data map, then we are still loading the data from the flatbuffer payload. Since the runtime change here is being made before the AoT changes, All CI on this diff by itself should test that the changes made here are backwards compatitble. Note: We do not resolve Runtime Memory usage at this point. WeightCache will be implemented in the next diff. Meaning If we load via the same key across different methods, we still pack twice and allocate two instances for the packed weights. ghstack-source-id: 271732048 @exported-using-ghexport Differential Revision: [D70315209](https://our.internmc.facebook.com/intern/diff/D70315209/)
DannyYuyang-quic
pushed a commit
to CodeLinaro/executorch
that referenced
this pull request
Apr 2, 2025
…npack (pytorch#9294) This PR was created by the merge bot to help merge the original PR into the main branch. ghstack PR number: pytorch#9152 by @mcr229 ^ Please use this as the source of truth for the PR details, comments, and reviews ghstack PR base: https://github.com/pytorch/executorch/tree/gh/mcr229/8/base ghstack PR head: https://github.com/pytorch/executorch/tree/gh/mcr229/8/head Merge bot PR base: https://github.com/pytorch/executorch/tree/gh/mcr229/7/orig Merge bot PR head: https://github.com/pytorch/executorch/tree/gh/mcr229/8/orig @diff-train-skip-merge --------- Co-authored-by: Max Ren <maxren@meta.com>
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 free
to 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):
If data is serialized into the NamedDataMap, then we overload getConstantDataPtr to retrieve the data from the named data map. This should be done in a Backwards Compatible way. Meaning if no data is serialized into the named data map, then we are still loading the data from the flatbuffer payload.
Since the runtime change here is being made before the AoT changes, All CI on this diff by itself should test that the changes made here are backwards compatitble.
Note: We do not resolve Runtime Memory usage at this point. WeightCache will be implemented in the next diff. Meaning If we load via the same key across different methods, we still pack twice and allocate two instances for the packed weights.
Differential Revision: D70315209