ETRecord: capture delegate _delegate_info_meta into ETRecord#21379
Open
Gasoonjia wants to merge 1 commit into
Open
ETRecord: capture delegate _delegate_info_meta into ETRecord#21379Gasoonjia wants to merge 1 commit into
_delegate_info_meta into ETRecord#21379Gasoonjia wants to merge 1 commit into
Conversation
Summary:
Backends can attach arbitrary per-delegate info during lowering via
`PreprocessResult._delegate_info_meta`, which `to_backend` copies onto
`lowered_module.meta["_delegate_info_meta"]`. Until now that info dead-ended on
the lowered module and never reached the ETRecord. This diff carries it the last
mile so it can be inspected after an ETRecord is parsed back.
What changed:
- `ETRecord` gains a `delegate_info: Dict[str, Any]` field, stored as a single
JSON blob (reserved name `delegate_info`), keyed by `f"{method}/{lowered_name}"`.
- Extraction lives entirely inside ETRecord (no delegate-specific logic in
`_program.py`). `_maybe_extract_delegate_info` is called from
`add_edge_dialect_program` and `add_executorch_program`, gated so it runs only
when (a) the ETRecord has no delegate_info yet and (b) the incoming graph
actually carries it. All recorded graphs are the same model, so it is captured
exactly once. It walks lowered submodules (recursing into control-flow
submodules) and skips non-JSON-serializable values (with a warning) so a
backend cannot break ETRecord saving.
- Covers both flows: `to_edge_transform_and_lower(...)` (captured at
`add_edge_dialect_program`) and `to_edge(...).to_backend(...)` (captured at
`add_executorch_program`, since the delegated graph is only available there).
- Example backend now logs a minimal, non-static fingerprint
(`sha256(processed_bytes)`) via `_delegate_info_meta` as the reference pattern
for other backends (Arm/XNNPACK) to follow.
Note: `exir/program/_program.py` is intentionally untouched.
Differential Revision: D113189648
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/21379
Note: Links to docs will display an error until the docs builds have been completed. ❌ 4 New Failures, 41 Unrelated FailuresAs of commit 210a94d with merge base e2c9a48 ( NEW FAILURES - The following jobs have failed:
FLAKY - The following jobs failed but were likely due to flakiness present on trunk:
BROKEN TRUNK - The following jobs failed but were present on the merge base:👉 Rebase onto the `viable/strict` branch to avoid these failures
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
meta-cla
Bot
added
the
CLA Signed
Contributor
|
@Gasoonjia has exported this pull request. If you are a Meta employee, you can view the originating Diff in D113189648. |
This PR needs a
|
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
1 participant
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.
Summary:
Backends can attach arbitrary per-delegate info during lowering via
PreprocessResult._delegate_info_meta, whichto_backendcopies ontolowered_module.meta["_delegate_info_meta"]. Until now that info dead-ended onthe lowered module and never reached the ETRecord. This diff carries it the last
mile so it can be inspected after an ETRecord is parsed back.
What changed:
ETRecordgains adelegate_info: Dict[str, Any]field, stored as a singleJSON blob (reserved name
delegate_info), keyed byf"{method}/{lowered_name}"._program.py)._maybe_extract_delegate_infois called fromadd_edge_dialect_programandadd_executorch_program, gated so it runs onlywhen (a) the ETRecord has no delegate_info yet and (b) the incoming graph
actually carries it. All recorded graphs are the same model, so it is captured
exactly once. It walks lowered submodules (recursing into control-flow
submodules) and skips non-JSON-serializable values (with a warning) so a
backend cannot break ETRecord saving.
to_edge_transform_and_lower(...)(captured atadd_edge_dialect_program) andto_edge(...).to_backend(...)(captured atadd_executorch_program, since the delegated graph is only available there).(
sha256(processed_bytes)) via_delegate_info_metaas the reference patternfor other backends (Arm/XNNPACK) to follow.
Note:
exir/program/_program.pyis intentionally untouched.Differential Revision: D113189648