haotian-liu / LLaVA

[NeurIPS'23 Oral] Visual Instruction Tuning (LLaVA) built towards GPT-4V level capabilities and beyond.
https://llava.hliu.cc
Apache License 2.0
20.42k stars 2.26k forks source link

Eval fixes #1743

Closed lpullela closed 1 month ago

lpullela commented 1 month ago

updating eval script to include dynamic paths and base model