In my experiment, I found that the model I wanted to replicate used the LLaVA-CC3M-Pretrain-595K dataset and required cc3m. csv. But I couldn't find this file in the file, so I want to ask for advice. Is it in the images.zip compressed file? Or do we need to generate it ourselves?