[Docs] Quantifies drop points, throughput, and memory usage of InternVL2 series

[Jalen-Zhong]

📚 The doc issue

Hi, thanks for your great job. I realize you team release the AWQ vision of InternVL2 series. But it is hard for researcher to calculate the GPU memory usage, especially for mm-llms.

Suggest a potential alternative/fix

provide a table about these information will be really helpful.