If you have some data and want to train or run a small custom model but don't have powerful enough hardware for training, fine-tuning services can be a good solution. Once training (requiring more resources than inference) is done, the custom model can then run locally. For larger models, there is a
