Image-Text-to-Text
Safetensors
qwen2_5_vl
historical
conversational