Pytorch ignite create_supervised_evaluator not returning the Custom Metrics

Question

I am trying to use create_supervised_evaluator from ignite.engine with my custom Metrics

evaluator = create_supervised_evaluator(model, metrics={'perform': Perform()}, device=device)

But evaluator.state.metrics.keys() is returning empty state dict.

I am using PyTorch == 1.7.1 and pytorch-ignite 0.4.2 . Can any one please let me know where am I going wrong .

Answer 1

Obviously you haven't run the evaluator, you must do that first

evaluator.run(data_loader, max_epochs=100)

Then doing this will return the metrics dict for you

evaluator.state.metrics