Efficient_Ad downloading and running on Imagenette

[ghost]

I am running anomalib's efficient_ad after modifying the config file with the custom dataset just like the way I did this for other models in anomalib. I used the following command to run it:

python tools/train.py --model efficient_ad --config /path_to_config/stfpm_TCL_config.yaml

The config file is modified as follows for the custom dataset:

When I run, it first starts downloading Imagenette2 (more than 1 GB). What is wrong?

[abc-125]

it first starts downloading Imagenette2

EfficientAD uses the ImageNet dataset (or Imagenette2) during the training to penalize the student for imitating the teacher for out-of-distribution images, so downloading this dataset is an expected behavior.

Then, runs inference on that instead of my custom dataset.

Can you please provide more details on this? Which version of the library are you using, logs, etc.

[ghost]

Thanks for clarifying the first point. I am sorry I was mistaken in the second point of inference on Imagenetter2. It was just taking a lot of time during the training but eventually was trained successfully on my custom dataset. I have edited the question to remove this second point.