[export] Update schema version by angelayi · Pull Request #115712 · pytorch/pytorch

angelayi · 2023-12-12T23:33:21Z

Since pytorch 2.1 release we've made some BC breaking changes to the serialized schema. We should update it in time for the 2.2 release. Some of the changes include:

[export] Move serialized custom class objs to toplevel #114371 - custom class objects / pybinded objects are no longer saved directly to the ExportedProgram structure. Instead, the name is serialized inside of the program, and the actual bytes are stored. in a separate location from the exported program, allowing it to be saved to a different location.
[export] Update serialization schema to input/output specs. (#845) #111204 - GraphSignature structure changed and call_spec is removed from the GraphModule schema
[export] Remove call_spec argument from ExportedProgram ctor. #111407 - loss_outout -> loss_output
[export] Enable verifier [2/n] #113075 - example_inputs removed from the ExportedProgram structure (this originally did not store anything), dialect added to the ExportedProgram structure.
[export][retry] Move lifted tensors out of state_dict #113689 - tensor constants are now lifted as inputs to the graph, and their locations are stored in the GraphSignature
[export] Update schema #114172 - removed equality_constraints and added a SymExprHint for all symbolic expressions.

pytorch-bot · 2023-12-12T23:33:24Z

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/115712

📄 Preview Python docs built from this PR
📄 Preview C++ docs built from this PR
❓ Need help or want to give feedback on the CI? Visit the bot commands wiki or our office hours

Note: Links to docs will display an error until the docs builds have been completed.

✅ You can merge normally! (1 Unrelated Failure)

As of commit 79f299e with merge base e918461 ():

FLAKY - The following job failed but was likely due to flakiness present on trunk:

pull / linux-focal-cuda11.8-py3.10-gcc9 / test (distributed, 3, 3, linux.8xlarge.nvidia.gpu) (gh)
distributed/fsdp/test_fsdp_core.py::TestParityWithDDP::test_delayed_optim_step_offload_true_no_shard

This comment was automatically generated by Dr. CI and updates every 15 minutes.

tugsbayasgalan · 2023-12-12T23:35:41Z


 # NOTE: Please update this value if any modifications are made to the schema
-SCHEMA_VERSION = 1
+SCHEMA_VERSION = 2


not blocking: should we add comment here describing what is changed? In general, whenever there is the version update, there should be some sort of doc or tracker?

ooh. I'm not sure what should be done during a version bump. I can add it to this PR and make sure it's in the release notes?

What if we paste it in the code as comment as well?

zhxchen17 · 2023-12-13T16:32:06Z

@angelayi considering export is still before alpha release, should we keep it 0? I feel it's rather confusing to see this to be bumped right now.

angelayi · 2023-12-13T17:47:38Z

@zhxchen17 I think export is at alpha release, so it's up to us how we want to do versioning. From a user's pov, the schema changed between 2.1 and 2.2 release, which is why I think we should bump the schema to show this change.
It seems your opinion is that we should bump the version when export gets to beta/actual release.
And I think @gmagogsfm 's opinion is that we should start maintaining BC now, and add back in logic that allows loading the schema from 2.1.

zhxchen17 · 2023-12-13T17:53:23Z

@zhxchen17 I think export is at alpha release, so it's up to us how we want to do versioning. From a user's pov, the schema changed between 2.1 and 2.2 release, which is why I think we should bump the schema to show this change. It seems your opinion is that we should bump the version when export gets to beta/actual release. And I think @gmagogsfm 's opinion is that we should start maintaining BC now, and add back in logic that allows loading the schema from 2.1.

well if there's no major user of the schema in real life, I'd rather make the breaking change than taking the cost to maintain it.

Also I'm not sure we're at alpha release really:

angelayi · 2023-12-13T17:54:00Z

Is alpha != prototype?

gmagogsfm · 2023-12-13T18:12:25Z

Alpha == Prototype, but Alpha is still open sourced, so we need more guarantee than private prototype. It is debatable whether we want to take on the responsibility of maintaining BC-safe loader since it involves additional eng work. But I think it is a bare minimum for us to update schema version per PyTorch release since we have breaking changes. @zhxchen17 thoughts?

angelayi · 2023-12-15T19:11:13Z

@pytorchbot merge

pytorchmergebot · 2023-12-15T19:13:20Z

Merge started

Your change will be merged once all checks pass (ETA 0-4 Hours).

Learn more about merging in the wiki.

Questions? Feedback? Please reach out to the PyTorch DevX Team

Advanced Debugging

Check the merge workflow status
here

Since pytorch 2.1 release we've made some BC breaking changes to the serialized schema. We should update it in time for the 2.2 release. Some of the changes include: * #114371 - custom class objects / pybinded objects are no longer saved directly to the `ExportedProgram` structure. Instead, the name is serialized inside of the program, and the actual bytes are stored. in a separate location from the exported program, allowing it to be saved to a different location. * #111204 - `GraphSignature` structure changed and `call_spec` is removed from the `GraphModule` schema * #111407 - `loss_outout` -> `loss_output` * #113075 - `example_inputs` removed from the `ExportedProgram` structure (this originally did not store anything), `dialect` added to the `ExportedProgram` structure. * #113689 - tensor constants are now lifted as inputs to the graph, and their locations are stored in the `GraphSignature` * #114172 - removed `equality_constraints` and added a `SymExprHint` for all symbolic expressions. Pull Request resolved: #115712 Approved by: https://github.com/gmagogsfm

Since pytorch 2.1 release we've made some BC breaking changes to the serialized schema. We should update it in time for the 2.2 release. Some of the changes include: * pytorch#114371 - custom class objects / pybinded objects are no longer saved directly to the `ExportedProgram` structure. Instead, the name is serialized inside of the program, and the actual bytes are stored. in a separate location from the exported program, allowing it to be saved to a different location. * pytorch#111204 - `GraphSignature` structure changed and `call_spec` is removed from the `GraphModule` schema * pytorch#111407 - `loss_outout` -> `loss_output` * pytorch#113075 - `example_inputs` removed from the `ExportedProgram` structure (this originally did not store anything), `dialect` added to the `ExportedProgram` structure. * pytorch#113689 - tensor constants are now lifted as inputs to the graph, and their locations are stored in the `GraphSignature` * pytorch#114172 - removed `equality_constraints` and added a `SymExprHint` for all symbolic expressions. Pull Request resolved: pytorch#115712 Approved by: https://github.com/gmagogsfm

Since pytorch 2.1 release we've made some BC breaking changes to the serialized schema. We should update it in time for the 2.2 release. Some of the changes include: * #114371 - custom class objects / pybinded objects are no longer saved directly to the `ExportedProgram` structure. Instead, the name is serialized inside of the program, and the actual bytes are stored. in a separate location from the exported program, allowing it to be saved to a different location. * #111204 - `GraphSignature` structure changed and `call_spec` is removed from the `GraphModule` schema * #111407 - `loss_outout` -> `loss_output` * #113075 - `example_inputs` removed from the `ExportedProgram` structure (this originally did not store anything), `dialect` added to the `ExportedProgram` structure. * #113689 - tensor constants are now lifted as inputs to the graph, and their locations are stored in the `GraphSignature` * #114172 - removed `equality_constraints` and added a `SymExprHint` for all symbolic expressions. Pull Request resolved: #115712 Approved by: https://github.com/gmagogsfm

Since pytorch 2.1 release we've made some BC breaking changes to the serialized schema. We should update it in time for the 2.2 release. Some of the changes include: * pytorch#114371 - custom class objects / pybinded objects are no longer saved directly to the `ExportedProgram` structure. Instead, the name is serialized inside of the program, and the actual bytes are stored. in a separate location from the exported program, allowing it to be saved to a different location. * pytorch#111204 - `GraphSignature` structure changed and `call_spec` is removed from the `GraphModule` schema * pytorch#111407 - `loss_outout` -> `loss_output` * pytorch#113075 - `example_inputs` removed from the `ExportedProgram` structure (this originally did not store anything), `dialect` added to the `ExportedProgram` structure. * pytorch#113689 - tensor constants are now lifted as inputs to the graph, and their locations are stored in the `GraphSignature` * pytorch#114172 - removed `equality_constraints` and added a `SymExprHint` for all symbolic expressions. Pull Request resolved: pytorch#115712 Approved by: https://github.com/gmagogsfm

[export] Update schema version

79f299e

angelayi marked this pull request as ready for review December 12, 2023 23:33

angelayi requested review from avikchaudhuri, gmagogsfm, tugsbayasgalan and zhxchen17 as code owners December 12, 2023 23:33

angelayi added the topic: not user facing topic category label Dec 12, 2023

github-actions Bot added ciflow/inductor module: export labels Dec 12, 2023

tugsbayasgalan reviewed Dec 12, 2023

View reviewed changes

angelayi added release notes: export and removed topic: not user facing topic category labels Dec 12, 2023

gmagogsfm approved these changes Dec 15, 2023

View reviewed changes

pytorch-bot Bot added the ciflow/trunk Trigger trunk jobs on your pull request label Dec 15, 2023

pytorchmergebot added the merging label Dec 15, 2023

pytorchmergebot added Merged and removed merging labels Dec 15, 2023

pytorchmergebot closed this in 1f3bdf4 Dec 15, 2023

This was referenced Dec 15, 2023

[cherry-pick][export] Update schema version #115952

Merged

[v.2.2.0] Release Tracker #115300

Closed

github-actions Bot deleted the angelayi/schema_version1 branch February 19, 2024 02:02

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[export] Update schema version#115712

[export] Update schema version#115712
angelayi wants to merge 1 commit intomainfrom
angelayi/schema_version1

angelayi commented Dec 12, 2023 •

edited

Loading

Uh oh!

pytorch-bot Bot commented Dec 12, 2023 •

edited

Loading

Uh oh!

tugsbayasgalan Dec 12, 2023

Uh oh!

angelayi Dec 12, 2023

Uh oh!

tugsbayasgalan Dec 12, 2023

Uh oh!

zhxchen17 commented Dec 13, 2023

Uh oh!

angelayi commented Dec 13, 2023 •

edited

Loading

Uh oh!

zhxchen17 commented Dec 13, 2023

Uh oh!

angelayi commented Dec 13, 2023

Uh oh!

gmagogsfm commented Dec 13, 2023

Uh oh!

angelayi commented Dec 15, 2023

Uh oh!

pytorchmergebot commented Dec 15, 2023

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

Conversation

angelayi commented Dec 12, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

pytorch-bot Bot commented Dec 12, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/115712

✅ You can merge normally! (1 Unrelated Failure)

Uh oh!

tugsbayasgalan Dec 12, 2023

Choose a reason for hiding this comment

Uh oh!

angelayi Dec 12, 2023

Choose a reason for hiding this comment

Uh oh!

tugsbayasgalan Dec 12, 2023

Choose a reason for hiding this comment

Uh oh!

zhxchen17 commented Dec 13, 2023

Uh oh!

angelayi commented Dec 13, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

zhxchen17 commented Dec 13, 2023

Uh oh!

angelayi commented Dec 13, 2023

Uh oh!

gmagogsfm commented Dec 13, 2023

Uh oh!

angelayi commented Dec 15, 2023

Uh oh!

pytorchmergebot commented Dec 15, 2023

Merge started

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

angelayi commented Dec 12, 2023 •

edited

Loading

pytorch-bot Bot commented Dec 12, 2023 •

edited

Loading

angelayi commented Dec 13, 2023 •

edited

Loading