It's the training of the LLM that requires huge amounts of data and processing power, once trained (depending on the model and parameters) they are fairly small in size and can be run on commercially available hardware.
MrSusan
joined 2 years ago
This person has no friend groups