fix: weakref error while saving torch model

[ab93]

• sustain trainer var in code to avoid weakref (temp fix)
• add uuid to model metadata
• fix logging
• Use SigmoidThreshold as a the static threshold estimator

[codecov-commenter]

Codecov Report

Merging #101 (216cd5d) into main (1866591) will decrease coverage by 0.55%. The diff coverage is 60.00%.

@@            Coverage Diff             @@
##             main     #101      +/-   ##
==========================================
- Coverage   86.93%   86.38%   -0.55%     
==========================================
  Files          19       19              
  Lines         949      955       +6     
  Branches       96       99       +3     
==========================================
  Hits          825      825              
- Misses        106      109       +3     
- Partials       18       21       +3     

Impacted Files	Coverage Δ
numaprom/udsink/train_rollout.py	84.07% <50.00%> (-4.72%)	:arrow_down:
numaprom/tools.py	86.33% <100.00%> (ø)
numaprom/udf/postprocess.py	96.51% <100.00%> (ø)

:mega: We’re building smart automated test selection to slash your CI/CD build times. Learn more