Videochatgpt tuning data encounters some error

[Lexarymade]

dear author, while using the vcg tuning data downloaded from your Baidu Disk to tune my own video language model, I encounter the error: [h264 @ 0x37c3dec0] mmco: unref short failure when I'm encoding the video frames, I'm not so sure about the dataset, is there some damages showed up in the datasets?

[Lexarymade]

I checked my code on the NeXT-QA tuning dataset, which is work