[export] Initial serialization v2 #102707
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/102707
Note: Links to docs will display an error until the docs builds have been completed. ❌ 1 New Failure, 2 Unrelated FailuresAs of commit 0d44425: BROKEN TRUNK - The following jobs failed but were present on the merge base 12cd1db:👉 Rebase onto the `viable/strict` branch to avoid these failures
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
|
@angelayi has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator. |
There was a problem hiding this comment.
Can add some high-level descriptions about this PR in the pr description? Just to keep a record in case we need to find out why we do it in this way in the future and also encourage people to add more meaningful reviews.Example description could include 1. what's the motivation of this change? 2. how users are supposed to user this API 3. Does the test cover all corner cases? Anything that's in your mind but not tested yet? 4. what are the implementation choices we made and why we choose this implementation instead of the alternatives? 5. Is there any lilmitations of current design? Feel free to skip some if you find meaningless and it would be better if new points could be added if you feel that may be helpful for people to understand better.
I understand it's discussed offline and adding more descriptions could slow down our development a little bit but would definitely be helpful to keep everyone (e.g. those who didn't participate in the discussion like me) on the same page and help us maintain the code in the long run :).
|
@angelayi has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator. |
|
This pull request was exported from Phabricator. Differential Revision: D46362466 |
Summary: v2 of #102125 because of git issues corresponding deserialization diff: #102716 Implementing serialization of the exported program to a python dataclass, and then from that dataclass to json. This is split into a couple of sections: - `serialize(ep: ep.ExportedProgram, opset_version: Dict[str, int]) -> Tuple[bytes, bytes]` -- takes an exported program object, a dictionary mapping opset namespaces to versions, and returns the serialized exported program in bytes, and separately the state dict serialized in bytes - `GraphModuleSerializer` class that serializes torch.fx.GraphModule to the schema.GraphModule dataclass - `ExportedProgramSerializer` class that serializes torch._export.exported_program.ExportedProgram to the schema.ExportedProgram dataclass Serialization TODOs: - [x] pytree spec: #102577 - [ ] higher order ops - [ ] node metadata (specifically nn_module_stack/source_fn) - [ ] constraints - [ ] graph module metadata The tests are not super comprehensive, but that's because I think it'll be better tested + easier to test once deserialization is implemented. Pull Request resolved: #102707 Reviewed By: zhxchen17 Differential Revision: D46362466 Pulled By: angelayi fbshipit-source-id: 1d3fc157a7a5c2e615dbcc7f0e87d76f2f4c43ed
|
This pull request was exported from Phabricator. Differential Revision: D46362466 |
Summary: v2 of #102125 because of git issues corresponding deserialization diff: #102716 Implementing serialization of the exported program to a python dataclass, and then from that dataclass to json. This is split into a couple of sections: - `serialize(ep: ep.ExportedProgram, opset_version: Dict[str, int]) -> Tuple[bytes, bytes]` -- takes an exported program object, a dictionary mapping opset namespaces to versions, and returns the serialized exported program in bytes, and separately the state dict serialized in bytes - `GraphModuleSerializer` class that serializes torch.fx.GraphModule to the schema.GraphModule dataclass - `ExportedProgramSerializer` class that serializes torch._export.exported_program.ExportedProgram to the schema.ExportedProgram dataclass Serialization TODOs: - [x] pytree spec: #102577 - [ ] higher order ops - [ ] node metadata (specifically nn_module_stack/source_fn) - [ ] constraints - [ ] graph module metadata The tests are not super comprehensive, but that's because I think it'll be better tested + easier to test once deserialization is implemented. Pull Request resolved: #102707 Reviewed By: zhxchen17 Differential Revision: D46362466 Pulled By: angelayi fbshipit-source-id: 033cb9a22d905d944e182dba3b191df4c52413c8
Summary: v2 of #102125 because of git issues corresponding deserialization diff: #102716 Implementing serialization of the exported program to a python dataclass, and then from that dataclass to json. This is split into a couple of sections: - `serialize(ep: ep.ExportedProgram, opset_version: Dict[str, int]) -> Tuple[bytes, bytes]` -- takes an exported program object, a dictionary mapping opset namespaces to versions, and returns the serialized exported program in bytes, and separately the state dict serialized in bytes - `GraphModuleSerializer` class that serializes torch.fx.GraphModule to the schema.GraphModule dataclass - `ExportedProgramSerializer` class that serializes torch._export.exported_program.ExportedProgram to the schema.ExportedProgram dataclass Serialization TODOs: - [x] pytree spec: #102577 - [ ] higher order ops - [ ] node metadata (specifically nn_module_stack/source_fn) - [ ] constraints - [ ] graph module metadata The tests are not super comprehensive, but that's because I think it'll be better tested + easier to test once deserialization is implemented. Pull Request resolved: #102707 Reviewed By: zhxchen17 Differential Revision: D46362466 Pulled By: angelayi fbshipit-source-id: 22b0c38ddf3887e5966c0fe0b00c6984c30d98a9
|
This pull request was exported from Phabricator. Differential Revision: D46362466 |
angelayi
force-pushed
the
seralize2
branch
2 times, most recently
from
dcd8509
to
caafb61
Compare
Summary: v2 of #102125 because of git issues corresponding deserialization diff: #102716 Implementing serialization of the exported program to a python dataclass, and then from that dataclass to json. This is split into a couple of sections: - `serialize(ep: ep.ExportedProgram, opset_version: Dict[str, int]) -> Tuple[bytes, bytes]` -- takes an exported program object, a dictionary mapping opset namespaces to versions, and returns the serialized exported program in bytes, and separately the state dict serialized in bytes - `GraphModuleSerializer` class that serializes torch.fx.GraphModule to the schema.GraphModule dataclass - `ExportedProgramSerializer` class that serializes torch._export.exported_program.ExportedProgram to the schema.ExportedProgram dataclass Serialization TODOs: - [x] pytree spec: #102577 - [ ] higher order ops - [ ] node metadata (specifically nn_module_stack/source_fn) - [ ] constraints - [ ] graph module metadata The tests are not super comprehensive, but that's because I think it'll be better tested + easier to test once deserialization is implemented. Pull Request resolved: #102707 Reviewed By: zhxchen17 Differential Revision: D46362466 Pulled By: angelayi fbshipit-source-id: 32766639106abc0c4cea03bd298254140e7f3a1a
|
This pull request was exported from Phabricator. Differential Revision: D46362466 |
Summary: v2 of #102125 because of git issues corresponding deserialization diff: #102716 Implementing serialization of the exported program to a python dataclass, and then from that dataclass to json. This is split into a couple of sections: - `serialize(ep: ep.ExportedProgram, opset_version: Dict[str, int]) -> Tuple[bytes, bytes]` -- takes an exported program object, a dictionary mapping opset namespaces to versions, and returns the serialized exported program in bytes, and separately the state dict serialized in bytes - `GraphModuleSerializer` class that serializes torch.fx.GraphModule to the schema.GraphModule dataclass - `ExportedProgramSerializer` class that serializes torch._export.exported_program.ExportedProgram to the schema.ExportedProgram dataclass Serialization TODOs: - [x] pytree spec: #102577 - [ ] higher order ops - [ ] node metadata (specifically nn_module_stack/source_fn) - [ ] constraints - [ ] graph module metadata The tests are not super comprehensive, but that's because I think it'll be better tested + easier to test once deserialization is implemented. Pull Request resolved: #102707 Reviewed By: zhxchen17 Differential Revision: D46362466 Pulled By: angelayi fbshipit-source-id: 8e7d5cd4769bd6b4dcf64036dab43d54d7d4493a
|
This pull request was exported from Phabricator. Differential Revision: D46362466 |
Summary: v2 of #102125 because of git issues corresponding deserialization diff: #102716 Implementing serialization of the exported program to a python dataclass, and then from that dataclass to json. This is split into a couple of sections: - `serialize(ep: ep.ExportedProgram, opset_version: Dict[str, int]) -> Tuple[bytes, bytes]` -- takes an exported program object, a dictionary mapping opset namespaces to versions, and returns the serialized exported program in bytes, and separately the state dict serialized in bytes - `GraphModuleSerializer` class that serializes torch.fx.GraphModule to the schema.GraphModule dataclass - `ExportedProgramSerializer` class that serializes torch._export.exported_program.ExportedProgram to the schema.ExportedProgram dataclass Serialization TODOs: - [x] pytree spec: #102577 - [ ] higher order ops - [ ] node metadata (specifically nn_module_stack/source_fn) - [ ] constraints - [ ] graph module metadata The tests are not super comprehensive, but that's because I think it'll be better tested + easier to test once deserialization is implemented. Pull Request resolved: #102707 Reviewed By: zhxchen17 Differential Revision: D46362466 Pulled By: angelayi fbshipit-source-id: 8627d9f783cea5af9c36b09f4216c7effc021593
|
This pull request was exported from Phabricator. Differential Revision: D46362466 |
|
@pytorchbot merge -f 'Landed internally' (Initiating merge automatically since Phabricator Diff has merged, using force because this PR might not pass merge_rules.json but landed internally) |
|
@pytorchbot merge -f 'Landed internally' (Initiating merge automatically since Phabricator Diff has merged, using force because this PR might not pass merge_rules.json but landed internally) |
8 similar comments
|
@pytorchbot merge -f 'Landed internally' (Initiating merge automatically since Phabricator Diff has merged, using force because this PR might not pass merge_rules.json but landed internally) |
|
@pytorchbot merge -f 'Landed internally' (Initiating merge automatically since Phabricator Diff has merged, using force because this PR might not pass merge_rules.json but landed internally) |
|
@pytorchbot merge -f 'Landed internally' (Initiating merge automatically since Phabricator Diff has merged, using force because this PR might not pass merge_rules.json but landed internally) |
|
@pytorchbot merge -f 'Landed internally' (Initiating merge automatically since Phabricator Diff has merged, using force because this PR might not pass merge_rules.json but landed internally) |
|
@pytorchbot merge -f 'Landed internally' (Initiating merge automatically since Phabricator Diff has merged, using force because this PR might not pass merge_rules.json but landed internally) |
|
@pytorchbot merge -f 'Landed internally' (Initiating merge automatically since Phabricator Diff has merged, using force because this PR might not pass merge_rules.json but landed internally) |
|
@pytorchbot merge -f 'Landed internally' (Initiating merge automatically since Phabricator Diff has merged, using force because this PR might not pass merge_rules.json but landed internally) |
|
@pytorchbot merge -f 'Landed internally' (Initiating merge automatically since Phabricator Diff has merged, using force because this PR might not pass merge_rules.json but landed internally) |
Merge startedYour change will be merged immediately since you used the force (-f) flag, bypassing any CI checks (ETA: 1-5 minutes). Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
v2 of #102126. mentally stacked on top of #102707 Pull Request resolved: #102716 Approved by: https://github.com/avikchaudhuri, https://github.com/zhxchen17
v2 of #102125 because of git issues
corresponding deserialization diff: #102716
Implementing serialization of the exported program to a python dataclass, and then from that dataclass to json. This is split into a couple of sections:
serialize(ep: ep.ExportedProgram, opset_version: Dict[str, int]) -> Tuple[bytes, bytes]-- takes an exported program object, a dictionary mapping opset namespaces to versions, and returns the serialized exported program in bytes, and separately the state dict serialized in bytesGraphModuleSerializerclass that serializes torch.fx.GraphModuleto the schema.GraphModule dataclass
ExportedProgramSerializerclass that serializes torch._export.exported_program.ExportedProgram to the schema.ExportedProgram dataclassSerialization TODOs:
The tests are not super comprehensive, but that's because I think it'll be better tested + easier to test once deserialization is implemented.