Alibaba Cloud releases Qwen-VL and Qwen-VL-Chat
Alibaba Cloud released Qwen-VL and Qwen-VL-Chat and published model weights on ModelScope and Hugging Face.
What happened
Event details
Alibaba Cloud released Qwen-VL and Qwen-VL-Chat on August 22, 2023. Qwen-VL used Qwen-7B as its language-model backbone, OpenCLIP ViT-bigG as its visual encoder, and a cross-attention layer to connect the two representations. It supported image captioning, visual question answering, text recognition, and object localization. Code and weights were available through ModelScope and Hugging Face.
Assessment
Why it matters
Qwen-VL was an early open-weight general visual language model for Chinese and English use cases and added image understanding and localization to the Qwen family.
Availability
Access notes
Code and model weights were publicly available through ModelScope and Hugging Face.