[ExecuTorch][Weight Sharing] Track Named Data Store in EdgeProgramManager #9151
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
…ager
We enable Backends to return Named Data by adding NamedDataStoreOutput to the preprocess result. This is a completely BC change, as no backends with an implemented preprocess will see any change if nothing is explicitly implemented.
For backend developers to leverage the new NamedDataStore, they can initialize a new NamedDataStore() within preprocess, add_named_data to the data store, and return the NamedDataStore.get_named_data_store_output() in the preprocess result like such:
```
def preprocess(ExportedProgram, List[CompileSpecs]) -> PreprocessResult:
named_data_store = NamedDataStore()
for node in exported_program.graph.nodes:
named_data_store.add_named_data("name", bytes)
return PreprocessResult(
processed_bytes=bytes,
debug_handle_map={},
data_store_output= named_data_store.get_named_data_store_output()
)
```
Under the hood, the data store output is embedded in the loweredbackendmodule, (serializing loweredbackendmodule by itself with the a named_data_store_output is still a todo). But via the EdgeProgramManager path, we add the named_data_store_outputs to the edge_program_manger's named data store to keep track of all the named data returned by backends.
Differential Revision: [D70451660](https://our.internmc.facebook.com/intern/diff/D70451660/)
[ghstack-poisoned]
mcr229
requested review from
JacobSzwejbka,
tarun292 and
larryliu0820
as code owners
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/9151
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit a815591 with merge base 630d0cc ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
facebook-github-bot
added
the
CLA Signed
This was referenced Mar 11, 2025
|
This pull request was exported from Phabricator. Differential Revision: D70451660 |
cccclai
approved these changes
Mar 12, 2025
…eProgramManager"
We enable Backends to return Named Data by adding NamedDataStoreOutput to the preprocess result. This is a completely BC change, as no backends with an implemented preprocess will see any change if nothing is explicitly implemented.
For backend developers to leverage the new NamedDataStore, they can initialize a new NamedDataStore() within preprocess, add_named_data to the data store, and return the NamedDataStore.get_named_data_store_output() in the preprocess result like such:
```
def preprocess(ExportedProgram, List[CompileSpecs]) -> PreprocessResult:
named_data_store = NamedDataStore()
for node in exported_program.graph.nodes:
named_data_store.add_named_data("name", bytes)
return PreprocessResult(
processed_bytes=bytes,
debug_handle_map={},
data_store_output= named_data_store.get_named_data_store_output()
)
```
Under the hood, the data store output is embedded in the loweredbackendmodule, (serializing loweredbackendmodule by itself with the a named_data_store_output is still a todo). But via the EdgeProgramManager path, we add the named_data_store_outputs to the edge_program_manger's named data store to keep track of all the named data returned by backends.
Differential Revision: [D70451660](https://our.internmc.facebook.com/intern/diff/D70451660/)
[ghstack-poisoned]
|
This pull request was exported from Phabricator. Differential Revision: D70451660 |
…eProgramManager"
We enable Backends to return Named Data by adding NamedDataStoreOutput to the preprocess result. This is a completely BC change, as no backends with an implemented preprocess will see any change if nothing is explicitly implemented.
For backend developers to leverage the new NamedDataStore, they can initialize a new NamedDataStore() within preprocess, add_named_data to the data store, and return the NamedDataStore.get_named_data_store_output() in the preprocess result like such:
```
def preprocess(ExportedProgram, List[CompileSpecs]) -> PreprocessResult:
named_data_store = NamedDataStore()
for node in exported_program.graph.nodes:
named_data_store.add_named_data("name", bytes)
return PreprocessResult(
processed_bytes=bytes,
debug_handle_map={},
data_store_output= named_data_store.get_named_data_store_output()
)
```
Under the hood, the data store output is embedded in the loweredbackendmodule, (serializing loweredbackendmodule by itself with the a named_data_store_output is still a todo). But via the EdgeProgramManager path, we add the named_data_store_outputs to the edge_program_manger's named data store to keep track of all the named data returned by backends.
Differential Revision: [D70451660](https://our.internmc.facebook.com/intern/diff/D70451660/)
[ghstack-poisoned]
|
This pull request was exported from Phabricator. Differential Revision: D70451660 |
…eProgramManager"
We enable Backends to return Named Data by adding NamedDataStoreOutput to the preprocess result. This is a completely BC change, as no backends with an implemented preprocess will see any change if nothing is explicitly implemented.
For backend developers to leverage the new NamedDataStore, they can initialize a new NamedDataStore() within preprocess, add_named_data to the data store, and return the NamedDataStore.get_named_data_store_output() in the preprocess result like such:
```
def preprocess(ExportedProgram, List[CompileSpecs]) -> PreprocessResult:
named_data_store = NamedDataStore()
for node in exported_program.graph.nodes:
named_data_store.add_named_data("name", bytes)
return PreprocessResult(
processed_bytes=bytes,
debug_handle_map={},
data_store_output= named_data_store.get_named_data_store_output()
)
```
Under the hood, the data store output is embedded in the loweredbackendmodule, (serializing loweredbackendmodule by itself with the a named_data_store_output is still a todo). But via the EdgeProgramManager path, we add the named_data_store_outputs to the edge_program_manger's named data store to keep track of all the named data returned by backends.
Differential Revision: [D70451660](https://our.internmc.facebook.com/intern/diff/D70451660/)
[ghstack-poisoned]
|
This pull request was exported from Phabricator. Differential Revision: D70451660 |
|
What is XNNPack serializing here? Is it just the unpacked tensors and their names? In that case could we just put them in the namedTensorStore instead of the named data store so that its plug and play with training .ptds? |
facebook-github-bot
temporarily deployed
to
cherry-pick-bot
GitHub Actions
Inactive
SS-JIA
pushed a commit
that referenced
this pull request
Mar 14, 2025
…ager (#9293) This PR was created by the merge bot to help merge the original PR into the main branch. ghstack PR number: #9151 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/mcr229/7/base ghstack PR head: https://github.com/pytorch/executorch/tree/gh/mcr229/7/head Merge bot PR base: https://github.com/pytorch/executorch/tree/main Merge bot PR head: https://github.com/pytorch/executorch/tree/gh/mcr229/7/orig @diff-train-skip-merge Co-authored-by: Max Ren <[email protected]>
SS-JIA
pushed a commit
that referenced
this pull request
Mar 15, 2025
…ager Pull Request resolved: #9151 We enable Backends to return Named Data by adding NamedDataStoreOutput to the preprocess result. This is a completely BC change, as no backends with an implemented preprocess will see any change if nothing is explicitly implemented. For backend developers to leverage the new NamedDataStore, they can initialize a new NamedDataStore() within preprocess, add_named_data to the data store, and return the NamedDataStore.get_named_data_store_output() in the preprocess result like such: ``` def preprocess(ExportedProgram, List[CompileSpecs]) -> PreprocessResult: named_data_store = NamedDataStore() for node in exported_program.graph.nodes: named_data_store.add_named_data("name", bytes) return PreprocessResult( processed_bytes=bytes, debug_handle_map={}, data_store_output= named_data_store.get_named_data_store_output() ) ``` Under the hood, the data store output is embedded in the loweredbackendmodule, (serializing loweredbackendmodule by itself with the a named_data_store_output is still a todo). But via the EdgeProgramManager path, we add the named_data_store_outputs to the edge_program_manger's named data store to keep track of all the named data returned by backends. ghstack-source-id: 271732049 @exported-using-ghexport Differential Revision: [D70451660](https://our.internmc.facebook.com/intern/diff/D70451660/)
DannyYuyang-quic
pushed a commit
to CodeLinaro/executorch
that referenced
this pull request
Apr 2, 2025
…ager (pytorch#9293) This PR was created by the merge bot to help merge the original PR into the main branch. ghstack PR number: pytorch#9151 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/mcr229/7/base ghstack PR head: https://github.com/pytorch/executorch/tree/gh/mcr229/7/head Merge bot PR base: https://github.com/pytorch/executorch/tree/main Merge bot PR head: https://github.com/pytorch/executorch/tree/gh/mcr229/7/orig @diff-train-skip-merge Co-authored-by: Max Ren <[email protected]>
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.
Stack from ghstack (oldest at bottom):
We enable Backends to return Named Data by adding NamedDataStoreOutput to the preprocess result. This is a completely BC change, as no backends with an implemented preprocess will see any change if nothing is explicitly implemented.
For backend developers to leverage the new NamedDataStore, they can initialize a new NamedDataStore() within preprocess, add_named_data to the data store, and return the NamedDataStore.get_named_data_store_output() in the preprocess result like such:
Under the hood, the data store output is embedded in the loweredbackendmodule, (serializing loweredbackendmodule by itself with the a named_data_store_output is still a todo). But via the EdgeProgramManager path, we add the named_data_store_outputs to the edge_program_manger's named data store to keep track of all the named data returned by backends.
Differential Revision: D70451660