Investigating torch==1.10

[ksaur]

Runs ok locally; let's check the pipeline.

[codecov-commenter]

Codecov Report

Merging #554 (3798d02) into main (7705f71) will decrease coverage by 5.80%.
The diff coverage is 100.00%.

❗ Current head 3798d02 differs from pull request most recent head d53b33e. Consider uploading reports for the commit d53b33e to get more accurate results

@@            Coverage Diff             @@
##             main     #554      +/-   ##
==========================================
- Coverage   90.59%   84.79%   -5.81%     
==========================================
  Files          78       78              
  Lines        4540     4544       +4     
  Branches      940      941       +1     
==========================================
- Hits         4113     3853     -260     
- Misses        240      499     +259     
- Partials      187      192       +5     

Flag	Coverage Δ
unittests	84.79% <100.00%> (-5.65%)	⬇️

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

Impacted Files	Coverage Δ
hummingbird/ml/_topology.py	69.58% <100.00%> (-15.59%)	⬇️
...rator_converters/_label_encoder_implementations.py	100.00% <100.00%> (ø)
.../ml/operator_converters/_linear_implementations.py	75.55% <100.00%> (-22.23%)	⬇️
...tor_converters/_one_hot_encoder_implementations.py	92.45% <100.00%> (-3.78%)	⬇️
.../ml/operator_converters/_scaler_implementations.py	82.60% <100.00%> (ø)
...rd/ml/operator_converters/_tree_implementations.py	98.27% <100.00%> (ø)
...ummingbird/ml/containers/sklearn/tvm_containers.py	17.24% <0.00%> (-75.87%)	⬇️
...mmingbird/ml/operator_converters/sparkml/linear.py	42.85% <0.00%> (-57.15%)	⬇️
...bird/ml/operator_converters/sparkml/discretizer.py	52.94% <0.00%> (-35.30%)	⬇️
... and 15 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 7705f71...d53b33e. Read the comment docs.

[ksaur]

Ok generating the new model with static dims makes the tests pass locally, but the test still hangs locally (meaning it prints the "warnings" summary", and also hangs in the pipeline.

Ex:

test_extra_conf.py ........sssss............ssssssssssss                                                                                                                              [100%]

========== warnings summary ===========================================
../../.local/lib/python3.7/site-packages/onnx/mapping.py:27
  /home/kasaur/.local/lib/python3.7/site-packages/onnx/mapping.py:27: DeprecationWarning: `np.object` is a deprecated alias for the builtin `object`. To silence this warning, use `object` by itself. Doing this will not modify any behavior and is safe. 
  Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations
    int(TensorProto.STRING): np.dtype(np.object)

../../.local/lib/python3.7/site-packages/sklearn/experimental/enable_hist_gradient_boosting.py:17
  /home/kasaur/.local/lib/python3.7/site-packages/sklearn/experimental/enable_hist_gradient_boosting.py:17: UserWarning: Since version 1.0, it is not needed to import enable_hist_gradient_boosting anymore. HistGradientBoostingClassifier and HistGradientBoostingRegressor are now stable and can be normally imported from sklearn.ensemble.
    "Since version 1.0, "

-- Docs: https://docs.pytest.org/en/stable/warnings.html
======= 20 passed, 17 skipped, 2 warnings in 7.08s =========================================

and then......hangs....forever. That was a new one.

I am not sure what to try next. It only behaves this way for torch==1.10 on onnx tests on linux/mac. Passes everything on Windows. Passes everything if torch <1.10.

[interesaaat]

Since we are at it, can you also please remove torch.IntTensor. It is not necessary here anymore.