allenai

/

Molmo-7B-D-0924

like 496

Follow

Ai2 1.91k

I am trying to perform few-shot VQA with this model, however it's much more cumbersome than with Qwen2-VL because the tokenizer doesn't contain a chat template. Would it be possible to add one, so we can easily parse multi-turn conversations?
See:

• https://huggingface.co/Qwen/Qwen2-VL-7B-Instruct/blob/main/chat_template.json
• https://huggingface.co/docs/transformers/main/en/chat_templating