GitHub2 天
DeepSeek-VL2: Mixture-of-Experts Vision-Language Models for Advanced Multimodal Understanding
Note: You may need 80GB GPU memory to run this script with deepseek-vl2-small and even larger for deepseek-vl2.