Error for multiple gpu training

[udasinnayan]

In the config file trainer part, if I set auto_select_gpus: true and devices: 2, I got an error like AttributeError: Can't pickle local object 'FeatureExtractor.get_features.<locals>.hook', it comes from this line trainer.fit(model=model, datamodule=datamodule) in anomalib/tools/train.py script but for single GPU it works without an error. Any comments about this error / is it possible to use multiple GPUs for training and testing?

[ashwinvaidya17]

Hi @udasinnayan that's a known issue. This might probably be solved when me move towards torchfx but the only option is to use single GPU for now.

[samet-akcay]

@ashwinvaidya17, I'm wondering if this is the case even when we use timm's feature extractor

[ashwinvaidya17]

I tested this on the current main branch and it still does not work with timm's feature extractor. I am closing as it will require a significant change.