BUG: PatchCore tends to go OOM easily

[SinChee]

OOM encounter in anomalib/models/components/dimensionality_reduction/random_projection.py line 132 when trying to run python main.py --mode train --data bras2021 --model patchcore

We are trying to replicate the experiments using 8 NVIDIA RTX 3090 GPU but are faced with problem from OOM.

Seems like there is a problem with memory allocation where it only uses 1 GPU instead of using the rest of the resources. How would one define how to train with more than 1 GPU?