Eagle: Vision-Centric High-Resolution Multimodal LLMs with Mixture of Encodersgithub.com/NVlabs1 pointtaikon2 years ago