IndexError in 001_getting_started.ipynb

[otte-c]

Hi,

The code in 001_getting_started.ipynb

# start training
engine = Engine()
engine.fit(model=model, datamodule=datamodule)

raises IndexError: index 4906065 is out of bounds for dimension 0 with size 4906065

I cloned the repo and installed the dependencies on a clean python 3.11.11 (macOS) via

pip install -e ".[full]"

Also tried

pip install "anomalib[full]==2.0.0b2"

on a clean python 3.11.11 but same error.

Does anyone have an idea how to fix it? Thanks!

[ryouhei0622]

Hi @otte-c,

I've encountered this issue as well.

For what it's worth, I was able to successfully run inference on my Mac (Apple Silicon) by explicitly setting the accelerator to "cpu".

Here's the configuration that worked for me:

engine = Engine(max_epochs=2, accelerator="cpu")

Hope this is helpful!

[samet-akcay]

@otte-c, I agree with @ryouhei0622. The problem is mps acceleration does not play nicely with some layers used in these models. Running these layers with mps returns None values. We might need to find these layers and compute them on cpu.

Meanwhile, you could set the accelerator to cpu