llava 7b

7B open vision-language model — LLaVA 1.5, a Vicuna/Llama-2 LLM paired with a CLIP visual encoder for image understanding and visual Q&A. One of the models that opened this space, and still light enough to matter: the single recipe here runs it on an 8 GB RTX 3060 Ti, with 72 tokens/s recorded at 4-bit. Llama 2 Community License.