PKU-YuanGroup / Video-LLaVA

Video-LLaVA: Learning United Visual Representation by Alignment Before Projection
https://arxiv.org/pdf/2311.10122.pdf
Apache License 2.0
2.74k stars 194 forks source link

Huggingface data files - corrupted zip files? #118

Open henryyuanheng-wang opened 5 months ago

henryyuanheng-wang commented 5 months ago

Thanks for the awesome work and open-sourcing the code. I'm trying to perform LORA fine-tuning on custom video instruction dataset, and would like to know the required format for data. I tried to download the videochatgpt_tune zip files from Huggingface, but currently having issues with unzipping the files. Are there any chances the uploaded files are corrupted? Thanks for looking into it and looking forward to your response.

zli999 commented 3 weeks ago

The valley data from huggingface seems cannot unzip.

Aakriti05 commented 1 week ago

While doing training, I am getting error for some of the videos. I downloaded the data from hugging face. Why is this happening and will this affect my final training results ?

Video-LLaVA/DATA_ROOT/valley/073451_073500/1032751259.mp4, No such file or directory
Error with Error reading /opt/dlami/nvme/Video-LLaVA/DATA_ROOT/valley/073451_073500/1032751259.mp4