Issues · wala/ML · GitHub

Labels Milestones

Loop-iterated layer calls over a list field lose their output type

#599

· khatchad opened

on Jun 24, 2026

`numpy.array` tensor type is not propagated to a callee parameter but `tf.constant`-wrapped is

#598

· khatchad opened

on Jun 23, 2026

Generator argument caller-walk is k-CFA-depth-bounded, missing values constructed more than k frames up

tensor inference

#597

· khatchad opened

on Jun 23, 2026

User-defined Keras `Layer`/`Model` subclass forward result is not tensor-typed, dropping chained parameters

#595

· khatchad opened

on Jun 22, 2026

Add a concise factory for constructing `TensorType` from numeric dimensions

good first issue

#594

· khatchad opened

on Jun 22, 2026

`ReadDataSets` shape parameter position and name disagree

#593

· khatchad opened

on Jun 18, 2026

·

Tensor shape and dtype inference

`Reduction.getDTypes`/`getDefaultDTypes` pass a parameter index where a value number is expected

#592

· khatchad opened

on Jun 18, 2026

·

Tensor shape and dtype inference

`Eye.getShapes` prepends batch dimensions by mutating shared shape lists (aliasing + double-prepend)

#591

· khatchad opened

on Jun 18, 2026

·

Tensor shape and dtype inference

Degrade shape/dtype inference to ⊤ on malformed or non-numeric constant arguments

#590

· khatchad opened

on Jun 18, 2026

·

Tensor shape and dtype inference

Recognize `tf.SparseTensorSpec` on the input side so sparse-spec parameters are typed

tensor inference

#589

· khatchad opened

on Jun 18, 2026

`NamedTuple` constructor synthesis treats non-annotated class attributes as tuple fields

tensor inference

#586

· khatchad opened

on Jun 17, 2026

·

Tensor shape and dtype inference

Fold arithmetic shape-argument dims via the analysis and unify the two extraction paths

#581

· khatchad opened

on Jun 17, 2026

·

Tensor shape and dtype inference