Post #3534991

Running LLMs locally, thinking about drivers and quantization, and hardware constraints and all, feels a lot like selecting sound card in the 90s. I hope we get a future that this is abstracted away like 99.999% of the time.