1. No. BentoML doesn’t do training.
2. Yes.
3. BentoML speeds up INFERENCE on the architecture and engineering point of view. Using GPU will further speed up the process but that is up to your resource allocation. Not everyone has access to GPU for inference