Inference Script

[ahmedheakl]

Hello team,

Nice work! Can you please provide a short script for evaluation? I tried similar codes as LLaVa-NeXt and Qwen2-VL, but your processor returns a list of token_ids when I do processor.apply_chat_template(prompt). It seems that your processor is just a Qwen2TokenizerFast. Can you please fix?

[xiangyue9607]

Hi, if you are using HF inference / generation. You might need to use the HF version of our checkpoint, which is available here: https://huggingface.co/neulab/Pangea-7B-hf

[yueqis]

Hi,
Thank you for your feedback! Could you provide specific code and logs you are attempting? Thanks!