More model structure details expecially parameter numbers of the adapters

[Stephan-XUE]

Hello, first, thanks for your work on the multimodal model using adapter-based fine-tuning, which inspires me greatly.
I would like to know more details about the model structure, especially the parameter numbers of the adapters, for which you only give a ratio to the default setting with sequential FF adapters of downsample factor 4 in the ablation study.