ponder-lab / Hybridize-Functions-Refactoring / #1619
72%

DEFAULT BRANCH: main
Ran
Jobs 1
Files 31
Run time 1min
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst
coverage: 72.219%. Remained the same
#1619

push

travis-pro

web-flow 
Extend precision audit to fixtures 123 + 134-143 (`tf.range` with `@tf.function`) (#540)

* Extend precision audit to fixtures 105-108 (`ragged.constant`)

Cover `tf.ragged.constant(pylist)` with the symmetric audit helper. Each fixture passes a flat one-dim list (`[1, 2]` for `a`, `[2, 2]` for `b`); the ragged-constant path in `RaggedConstant.java` emits no ragged-marker `null` entries for these inputs (the `for (Long i = 0L; i < R; i++) shape.add(null);` loop runs zero times when `ragged_rank = 0`), so Layer 1 is a plain `INT32, (NumericDim(2))` and Layer 2 passes it through unchanged.

The numerical assertion is clean; whether `tf.ragged.constant([1, 2])` returns a `RaggedTensor` or a regular `Tensor` at runtime determines whether the same Layer-2 spec-emission concern as #524 / #533 applies here. Treating as IDEAL pending verification.

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

* Extend precision audit to fixtures 109-111 (`sparse.SparseTensor`)

Sister batch to #534 (`sparse.eye`). Each fixture constructs a `tf.sparse.SparseTensor(indices=[[0,0],[1,2]], values=[1,2], dense_shape=[3,4])`—two non-zero values in a (3, 4) sparse matrix. Ariadne emits `INT32, (NumericDim(3), NumericDim(4))`, matching the static `dense_shape`. The inference algorithm passes that through unchanged.

The numerical assertion is clean. The runtime-semantics finding from #533 applies identically here: a `tf.function(input_signature=[TensorSpec(shape=(3, 4), dtype=int32)])` rejects a `SparseTensor` argument at runtime, since TF distinguishes `SparseTensor` from `Tensor` at the API level. The precise emission is `tf.SparseTensorSpec`. Inline TODO comments at fixtures 109-111 reference #533; the numerical assertions pin current behavior unchanged.

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

* Extend precision audit to fixtures 112-117 (`ragged.range`)

Apply the two-layer audit helper to fixtures 112-117, all exercising `tf.ragged.range(starts, lim... (continued)

1188 of 1645 relevant lines covered (72.22%)

0.72 hits per line

Jobs
ID Job ID Ran Files Coverage
1 #1619.1 31
72.22
Source Files on build #1619