Hi I am having a hard time trying to figure out how to run the benchmarks (i.e. the triplet eval) with a model that is a standard pytorch model but with custom weights (from a different training). For example this may be a resnet50 or ViT etc. It seems even to run the current repo you need a "model_dict.json"? and I'm not sure where this comes from either. Could you provide a simple example either showing how to modify the current model loading to use a custom model or provide any other suggestions?