Question on CogVLM image input resolution at inference time.

[aldoz-mila]

Hi all, I am currently playing with the CogVLM17b model (using this checkpoint: https://huggingface.co/THUDM/cogvlm-chat-hf). What is the resizing resolution of the vision encoder? The ending resolution during the training is 490 × 490 according to the paper, I assume this is also the image resolution used during inference? Thanks!