model = torch.load(path, map_location=self.device)["model"] KeyError: 'model

[EricHuiK]

What is the motivation for this task?

python tools/inference/torch_inference.py
--weights /disk4t/workspace/code/anomalib-main/results/fastflow/mvtec/bottle/run/weights/lightning/model-v7.ckpt
--input /disk4t/workspace/data/mvtec-ad/bottle/test/contamination/019.png
--output /disk4t/workspace/data/val_bottle
--device cuda:0
--task segmentation
--visualization_mode full

Describe the solution you'd like

model = torch.load(path, map_location=self.device)["model"] KeyError: 'model

Additional context

No response

[samet-akcay]

Hi @EricHuiK, can you please set the export_mode to torch? By default, it is for lightning.
https://github.com/openvinotoolkit/anomalib/blob/cdedf05451cc57fb373866595e8ef23a820cf123/src/anomalib/models/fastflow/config.yaml#L70-L71

anomalib will then create the torch model rather than the lightning model.