Love monarch, amazing primitives and feels so much lighter than Ray.
replyAnother awesome add on their part
So are we at a point where we can’t train llms for fun at home on less expensive AMD cards ?
replyWho was training LLMs on residential AMD GPUs?
replyI am. Most models that can be run locally have extraordinarily out of date training data for certain frameworks that predate breaking changes and important features, so their docs and API schemas need to be put into LoRAs.
replycan't or can?
replybenchmqarks?
reply