#1582
72%
main: 72%

Ran 22 May 2026 07:11PM UTC

Jobs 1

Files 31

Run time 1min

Badge

Embed ▾

Committed 22 May 2026 07:08PM UTC coverage: 72.219%. Remained the same

Build # #1582

Build Type

Pull #535

travis-pro

Committed by

khatchad

Commit Message

Extend precision audit to fixtures 100-104 (`keras.Input`)

The first batch where the Layer 1 / Layer 2 distinction is asymmetric for a reason *other* than ragged tensors. `tf.keras.Input(shape=(32,))` produces a rank-2 tensor: a dynamic batch dimension (the leading `None`) plus the static feature dimension (`32`). Ariadne (consumed at 0.44.0 + bump) emits the batch dim as raw `null` per the representation tracked at wala/ML#545; Hybridize's inference algorithm collapses that to `SymbolicDim("?")` in the output `TensorSpec`.

That collapse is correct here, unlike the ragged case. A `TensorSpec(shape=(?, 32), dtype=float32)` accurately represents the runtime contract—a `tf.keras.Input` parameter genuinely admits any batch size at call time, and `SymbolicDim("?")` is the right encoding for "any value." No #524-style flip target on the Layer 2 side. The Layer 1 raw `null` is hygiene only; flips when wala/ML#545's typed `DynamicDim` lands.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

Pull Request Pull Request #535: Extend precision audit to fixtures 100-104 (`keras.Input`)

Coverage Stats

1188 of 1645 relevant lines covered (72.22%)

0.72 hits per line

Jobs

ID	Job ID	Ran	Files	Coverage
1	#1582.1	22 May 2026 07:11PM UTC	31	72.22

ponder-lab / Hybridize-Functions-Refactoring / #1582
72%
main: 72%

README BADGES
x

Markdown

Textile

RDoc

HTML

Rst

Jobs

Source Files on build #1582

ponder-lab / Hybridize-Functions-Refactoring / #1582 72% main: 72%

README BADGES x

Markdown

Textile

RDoc

HTML

Rst

Jobs

Source Files on build #1582

ponder-lab / Hybridize-Functions-Refactoring / #1582
72%
main: 72%

README BADGES
x