[BUG] cannot import name 'IMG_EXTENSIONS' from 'tools.caption.utils'

[eastoc]

When we run

torchrun --nproc_per_node 1 --standalone -m tools.caption.caption_llava meta_info_fmin1.csv --dp-size 1 --tp-size 1 --model-path llava-v1.6-mistral-7b --prompt video

a error occurs:

cannot import name 'IMG_EXTENSIONS' from 'tools.caption.utils'

We found that the error is an package import error that happened in tools/caption/caption_llava.py

 from .utils import IMG_EXTENSIONS, PROMPTS, VID_EXTENSIONS, Timer, VideoTextDataset, collate_fn

“IMG_EXTENSIONS” and "VID_EXTENSIONS" should be imported from ../datasets/utils.py

The correct revision in caption_llava.py is shown

from .utils import PROMPTS, Timer, VideoTextDataset, collate_fn
from ..datasets.utils import IMG_EXTENSIONS, VID_EXTENSIONS

That had been verified on my machine.

[eastoc]

This bug solved at my fork

https://github.com/eastoc/Open-Sora/blob/main/tools/caption/caption_llava.py

[github-actions]

This issue is stale because it has been open for 7 days with no activity.

[zhengzangw]

fixed.