Comment by fartcannon

How do you train new models if your GPUs are being used for inference? I guess the training happens significantly less frequently?

Forgive my ignorance.