[5590225] Fixed regression introduced by PR #364 (FP64-to-FP32 conversion)

[gcunhase]

What does this PR do?

Type of change: ?

Overview: Fixed regression introduced by #364.

1. Detect ops not supported in lower precision and block them from being converted.
2. Postponed infer_shapes to PrecisionConverter.convert() to fix type propagation mismatches.
3. Created utils to get the model's opset version.
4. Moved lib_test_models.py one level up (from _test_utils.onnx.quantization to _test_utils.onnx), as the model zoo is also needed in the autocast testing workflow.

Usage

$ python -m modelopt.onnx.autocast --onnx_path=$MODEL_NAME.onnx
$ python -m modelopt.onnx.quantization --onnx_path=$MODEL_NAME.onnx

Testing

• Added unittest in quantization and autocast.
• Verified with models in bug 5590225 and 5532019.

Before your PR is "Ready for review"

• Make sure you read and follow Contributor guidelines and your commits are signed.
• Is this change backward compatible?: Yes
• Did you write any new necessary tests?: Yes
• Did you add or update any necessary documentation?: No
• Did you update Changelog?: No

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 73.49%. Comparing base (47ddd14) to head (23f0994).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #462      +/-   ##
==========================================
+ Coverage   73.45%   73.49%   +0.04%     
==========================================
  Files         180      180              
  Lines       18141    18171      +30     
==========================================
+ Hits        13325    13355      +30     
  Misses       4816     4816              

☔ 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.