Wrap tf variables in keras variables for TFSMLayer

[mattdangerw]

Fixes #20955

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 82.43%. Comparing base (b36a9ac) to head (7ac025c).
Report is 4 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master   #20995      +/-   ##
==========================================
- Coverage   82.45%   82.43%   -0.02%     
==========================================
  Files         562      562              
  Lines       53316    53404      +88     
  Branches     8259     8275      +16     
==========================================
+ Hits        43960    44023      +63     
- Misses       7340     7361      +21     
- Partials     2016     2020       +4     

Flag	Coverage Δ
keras	82.25% <100.00%> (-0.02%)	⬇️
keras-jax	63.93% <33.33%> (-0.11%)	⬇️
keras-numpy	58.87% <33.33%> (+0.02%)	⬆️
keras-openvino	32.64% <0.00%> (-0.05%)	⬇️
keras-tensorflow	64.37% <100.00%> (-0.11%)	⬇️
keras-torch	63.98% <33.33%> (-0.11%)	⬇️

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

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

🚀 New features to boost your workflow:

• ❄ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[hertschuh]

Thanks for the fix!

[hertschuh]

This shortcut needs to make sure the dtype matches or is None, otherwise it's incorrect.

I think you'll have to move this up:

if dtype is not None:
        dtype = standardize_dtype(dtype)

and then do:

if isinstance(x, tf.Variable) and (dtype is None or x.dtype == dtype):
  return x

[hertschuh]

I believe that's why the unit tests are failing.

[mattdangerw]

Thanks! Actually looks like we can just stick to the main path in convert_to_tensor, as tf.is_tensor(variable) == True. I think we were missing a conversion when comparing dtypes though, we were comparing tf dtype, to standardized string dtype in one spot.

Trying a fix.