Detectron2 coco evaluator

[Nurungyi]

📚 Documentation Issue

This issue category is for problems about existing documentation, not for asking how-to questions.

• Provide a link to an existing documentation/comment/tutorial:
  In tutorial, to execute cocoevaluator, trainer.model, val_loader, evaluator are parameters of inference_on_dataset.
  (below is a tutorial code)

  from detectron2.evaluation import COCOEvaluator, inference_on_dataset
  from detectron2.data import build_detection_test_loader
  evaluator = COCOEvaluator("balloon_val")
  val_loader = build_detection_test_loader(cfg, "balloon_val")
  print(inference_on_dataset(trainer.model, val_loader, evaluator))

Although I think that both (trainer.model, predictor.model) are same in an evaluation, instead of trainer.model, predictor.model is better. While trainer.model can be done after trainer should be finished, predictor.model can be executed with pre trained weight without execution of trainer.

Is it correct??

Thank you!!

• How should the above documentation/comment/tutorial improve: