Skip to content

Conversation

@KyleHerndon
Copy link
Contributor

Extending the work in #2464

dims = dims[:-2] + (dims[-2] * dims[-1] * mul,)
layout = BlockScaledI4Layout(shape=dims, d=new_d, qs=new_qs, m=new_m)
return PlanarQuantizedTensor(shape=dims, layout=layout)
elif isinstance(unpacked, TensorScaledLayout):
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It might look like this is deleted, but there's another copy in quantized_impls. I do not know why.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yet another reason to refactor all the variants for a single op into one place.

@KyleHerndon KyleHerndon requested a review from sogartar October 28, 2025 00:21
@github-actions
Copy link
Contributor

Coverage report

Click to see where and how coverage changed

FileStatementsMissingCoverageCoverage
(new stmts)
Lines missing
  sharktank/sharktank/ops
  default_impls.py 390-391, 422-423, 472-480, 1011-1012, 1198-1199
  quantized_impls.py 351-352
Project Total  

This report was generated by python-coverage-comment-action

Copy link
Contributor

@sogartar sogartar left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The description should say that this change removes quantized_tensor_layout_of_type in favor of another dispatch mechanism.

It is not sufficient to say:

Extending the work in #2464.

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

Successfully merging this pull request may close these issues.

3 participants