search

farewellthree / PPLLaVA

Official GPU implementation of the paper "PPLLaVA: Varied Video Sequence Understanding With Prompt Guidance"
Apache License 2.0
94 stars 4 forks source link

Training dataset #6

Closed zhuqiangLu closed 3 hours ago

zhuqiangLu commented 3 hours ago

Hi, may I ask what dataset is used for training

farewellthree commented 3 hours ago

Hi, all training settings can be found in the trainval.md as well as in the paper

zhuqiangLu commented 3 hours ago

Cheers, may I ask why these dataset specifically? I did a quick math and realized the dataset is quite large. I was thinking to reproduce the result with LORA, but considering the size of the dataset, it may take a while. My plan is to train PPLLaVA on a smaller dataset as a proxy. If you have tried training on a smaller dataset, please let me know which dataset it is and how PPLLaVA performs.

Best

zhuqiangLu commented 3 hours ago

Never mind my silly questions. I just found the training section in the paper. Closing this issue.