Patch shape incompatible with model #16

I am running into an issue when finetuning with config adapted from the defaults, specifically an issue relating to the patch shape:

ValueError: Input spatial dims torch.Size([1, 1, 192, 198, 128]) must be divisible by [16, 16, 16]

Indeed, 198 is not divisible by 16. Is this a mistake on my end? I find the same patch size settings in the main train.json file.
Changing the patch size to [192, 192, 128] resolves the issue. It would be nice if you could let me know if this is correct, or if this is different from how the model was trained.

Thanks,
Daniël