I recall that the original LLAVA project also conducted fine-tuning on purely NLP datasets. If this is possible, it would allow me to avoid deploying an additional online machine just for a language modeling service. Yes, LLaVA on vLLM can handle text-only input. Is this the correct ...
period. Despite those who don’t likeThe Shadowand him in his dual role of Lamont Cranston and The Shadow, I’m quite fond of it. Let’s just skip past any mention ofThe Cat in the Hat… AhhhhRise of the Guardians