Add verify_backward to enable testing bacward ops #1383
Conversation
This file contains 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
ndrakulicTT
requested review from
nvukobratTT,
pilkicTT and
dgolubovicTT
as code owners
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## main #1383 +/- ##
=======================================
Coverage ? 43.40%
=======================================
Files ? 48
Lines ? 7860
Branches ? 0
=======================================
Hits ? 3412
Misses ? 4448
Partials ? 0 ☔ View full report in Codecov by Sentry. |
|
||||||||||||||
|
||||||||||||||
|
||||||||||||||
|
||||||||||||||
vkovinicTT
reviewed
Mar 7, 2025
Comment on lines
+332
to
+333
| if not isinstance(framework_model, torch.nn.Module): | ||
| raise TypeError(f"Framework model must be of type {torch.nn.Module}, but got {type(framework_model)}") |
There was a problem hiding this comment.
So for now we just support backward verification for torch?
| compiled_output: torch.Tensor, | ||
| framework_model: torch.nn.Module, | ||
| compiled_model: CompiledModel, | ||
| original_model: torch.nn.Module = None, |
There was a problem hiding this comment.
Do we really need original_model?
Comment on lines
+387
to
+388
| fw = _squeeze_tensor(fw) | ||
| co = _squeeze_tensor(co) |
There was a problem hiding this comment.
Can we just use regular squeeze instead?
| VerifyConfig, | ||
| VerifyTensorMetadata, | ||
| should_waive_gradient, | ||
| AutomaticValueChecker, |
There was a problem hiding this comment.
Please just remove import AutomaticValueChecker if we don't use it here?
Comment on lines
+172
to
+176
| # NOTE: We probably need two framework models with the same state_dict to compare the outputs | ||
| # But for now it works without that for some reason? | ||
| # model_for_compile = Matmul() | ||
| # model_for_compile.eval() if not training else model_for_compile.train() | ||
| # model_for_compile.load_state_dict(framework_model.state_dict()) |
There was a problem hiding this comment.
Discussed offline, we probably don't need 2 framework models.
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.
Ticket
Fixes #1356
Problem description
There was no way to in one place verify gradients from backward pass
What's changed
Added verify_backward:
Checklist