How to get model pred_scores？

[ikun0124]

I want to get model output, such as the prediction score of the model. When I run the python test.py --config xx, I only get the metric of model, but not the prediction of dataset.
I also try other method that load trained dataset and mode to predict output. But there are always bug in the sklearn.metric module. Besides I don't know how to call adaptive score calculation on non PL framework code.
I had to turn to the authors to solve the above problems.

[ntdgo]

Hi, try to get the output of trainer.test command in test.py script
https://github.com/openvinotoolkit/anomalib/blob/c82ddcd76bed47d196cffbee333a995e7f1df3ed/tools/test.py#L49

It should be return some information might you need:
https://github.com/Lightning-AI/lightning/blob/ca88f813a440bef61e611dc3c40343a2774cd21a/src/pytorch_lightning/trainer/trainer.py#L769-L773

[ikun0124]

Hi, I have seen the link you gave me. It says that test can only return metric result, but I need the prediction of the model.

[ikun0124]

@thanh-repo I have seen the link you gave me. It says that test can only return metric result, but I need the prediction of the model.

[ashwinvaidya17]

@loyal-ikun have you looked at the torch_inference.py script https://github.com/openvinotoolkit/anomalib/blob/main/tools/inference/torch_inference.py. Maybe that will give you an idea on how to approach this. You can also have a look at this method https://github.com/openvinotoolkit/anomalib/blob/c82ddcd76bed47d196cffbee333a995e7f1df3ed/anomalib/deploy/inferencers/base_inferencer.py#L54