Skip to content

eval_per_class

Return a sorted per-class view over evaluation output.

Inputs

  • checkpoint
  • dataset_path
  • evaluation knobs from eval_run
  • sort_by
  • ascending

Outputs

  • ordered per-class metric rows plus the aggregate eval context

Examples

{"checkpoint":"run_demo","dataset_path":"~/data/mini_yolo"}
{"checkpoint":"~/models/retail.pt","dataset_path":"~/data/retail","sort_by":"precision"}
{"checkpoint":"~/models/factory.pt","dataset_path":"~/data/factory","ascending":false}

Common Errors

  • same failure modes as eval_run

eval_run, eval_error_analysis