Vision-CAIR / MiniGPT4-video

Official code for Goldfish model for long video understanding and MiniGPT4-video for short video understanding
https://vision-cair.github.io/Goldfish_website/
BSD 3-Clause "New" or "Revised" License
559 stars 60 forks source link

Are there any randomness in inference? #17

Open 2000ZRL opened 6 months ago

2000ZRL commented 6 months ago

I found that the inference results on my customized dataset are different if I run the evaluation code for multiple times. I fix the batch size to 4, and the "shuffle" of dataloader is set to False. I don't use "low_resource" during inference. However, if I fix the seed, the results will be consistent. Any clues on why this happened? Thanks.