Different Model Outputs on different Systems

[FruitPunch17]

Hey,
I trained a reverse distillation model on windows and tested it. After reviewing the performance I wanted to port it to a linux system.
However, for some reason unknown to me, the model has a different output on the linux system. I installed the same packages through pip freeze on windows and install on linux, but this did not change the output. So I just wanted to ask whether someone knows if this might be a problem with anomalib.

Best regards