question about processing caption_sharegtpvideo_300k

[yxsysu]

Hi authors, thanks for your great work.
When using the codes for stage3 training, it seems the codes cannot process the catpion_sharegptvideo_300k dataset.

The "img" item in the following config (instruction_data.py) seems make no difference.

   caption_sharegptvideo_300k=[
        f"{my_path}/video/caption/sharegptvideo/train_300k.json", 
        "{my_path}/shareGPTVideo/train_video_and_instruction/train_300k/videos",
        "video",
        "img", # read from image
    ],

Thanks for your reply.

[LiJiaqi96]

Same problem