Running N2V prediction on a GPU cluster fails

[Eddymorphling]

Hi folks,
I have a trained 3D-N2V that I would like to use to predict on a folder that has around 100, 3D tiff stacks. I am trying to run the predictions using GPU compute resources on my local HPC cluster using the CLI I found in the N2V-wiki. Here is what I try to run

python /home/N2V_predict.py --fileName=*.tif --dataPath='/Raw_data' --tile=4 --dims=ZYX --baseDir='/Analysis/N2V_model' --name=N2V-3D

Strangely this job fails and I see that my GPU runs out of memory. Does it try to load all 100 files at once and runs out of memory? I am not sure. When I run this job on an interactive-GPU node using the napari-n2v plugin, it works well as it only run predictions on one file at a time. Any clue how I can get it working using a non-interactive HPC cluster run? Thank you.