Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Unable to update Opset Version due to Batch Normalization #859

Closed
zjgarvey opened this issue Oct 9, 2024 · 1 comment
Closed

Unable to update Opset Version due to Batch Normalization #859

zjgarvey opened this issue Oct 9, 2024 · 1 comment
Assignees

Comments

@zjgarvey
Copy link
Collaborator

zjgarvey commented Oct 9, 2024

Removing the opset update, these models pass ORT inference, but also require basic ORT optimizations to clean up BatchNormalization Nodes with missing outputs.

Including an update soon to remove the opset version updating and apply basic ORT optimizations for these models. They should now report a PASS.

@zjgarvey zjgarvey self-assigned this Oct 9, 2024
zjgarvey added a commit to nod-ai/SHARK-TestSuite that referenced this issue Oct 9, 2024
Resolves issues <nod-ai/SHARK-ModelDev#860>
<nod-ai/SHARK-ModelDev#859> and
<nod-ai/SHARK-ModelDev#863>

- 5 tests should newly pass with this change (the vit models listed
under "no_opset_update")
- 1 test "resnetv2_50x1_bit.goog_in21k_ft_in1k_vaiq" should go from an
import fail to a Numerics fail.
@zjgarvey
Copy link
Collaborator Author

partially resolved with nod-ai/SHARK-TestSuite#365

There are a few models left commented out that require this change, but for some reason, fixing them seems to cause a test runner crash (OOM) for me locally on import phase. Since those models OOM on import anyway, we need to fix issue #861 before uncommenting the two remaining models which need the fix.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

1 participant