Makes ops.split in torch consistent with other backends

[james77777778]

Fixes #544

Unit test has been updated to verify the behavior consistency across backends.

...
        # test zero dimension
        x = np.ones(shape=(0,))
        self.assertEqual(len(knp.split(x, 2)), 2)
        self.assertEqual(len(knp.Split(2)(x)), 2)
...

[codecov]

Codecov Report

Patch coverage: 100.00% and project coverage change: -3.80% ⚠️

Comparison is base (b4019bc) 83.63% compared to head (4b56db2) 79.83%.
Report is 2 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #914      +/-   ##
==========================================
- Coverage   83.63%   79.83%   -3.80%     
==========================================
  Files         318      318              
  Lines       28391    28591     +200     
  Branches     5409     5447      +38     
==========================================
- Hits        23745    22827     -918     
- Misses       3147     4297    +1150     
+ Partials     1499     1467      -32     

Flag	Coverage Δ
keras_core	79.76% <100.00%> (-3.77%)	⬇️
keras_core-jax	?
keras_core-numpy	60.46% <0.00%> (-0.27%)	⬇️
keras_core-tensorflow	66.90% <0.00%> (-0.18%)	⬇️
keras_core-torch	69.30% <100.00%> (-0.14%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files Changed	Coverage Δ
keras_core/backend/torch/numpy.py	95.17% <100.00%> (+0.04%)	⬆️

... and 34 files with indirect coverage changes

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[fchollet]

Thanks for the fix!

[fchollet]

Typo: integer

Use the following:

Received indices_or_sections={indices_or_sections}
(interpreted as a number of sections) and axis={axis},
but input dimension x.shape[{axis}]={x.shape[axis]}
is not divisible by {indices_or_sections}.
Full input shape: x.shape={x.shape}

[james77777778]

Thanks! Fixed.

[fchollet]

LGTM

The JAX CI failure seems persistent, despite not occurring with local testing and having no clear connection to the commit where it started failing. We'll investigate it. There's something weird going on.

[james77777778]

The JAX CI failure seems persistent, despite not occurring with local testing and having no clear connection to the commit where it started failing. We'll investigate it. There's something weird going on.

I could not reproduce these JAX CI failure in my local env, too.

It appears that this run has started to fail:
https://github.com/keras-team/keras-core/actions/runs/6230566801/job/16910778781