Skip to content

[TAO-0000][Test] Validate AutoML dataclass bounds - #25

Closed
ramanathan831 wants to merge 2 commits into
NVIDIA-TAO:mainfrom
ramanathan831:dev/ram/automl-dataclass-unit-test
Closed

[TAO-0000][Test] Validate AutoML dataclass bounds#25
ramanathan831 wants to merge 2 commits into
NVIDIA-TAO:mainfrom
ramanathan831:dev/ram/automl-dataclass-unit-test

Conversation

@ramanathan831

Copy link
Copy Markdown
Contributor

Summary

  • add repository-wide AutoML dataclass metadata validation
  • fill missing transformer-layer and SegFormer weight-decay bounds found by the tests

Root cause

AutoML-enabled numeric fields could omit finite search bounds without CI detection.

Validation

  • python3 -m pytest -q nvidia_tao_core/tests/config/test_automl_config_validation.py
  • 35 passed, 11 skipped

@ramanathan831

Copy link
Copy Markdown
Contributor Author

Closing as a duplicate migration. The retained PR is #27, which contains the preferred/current-main version of this change. The source branch is being kept intact.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant