search

magic-research / PLLaVA

Official repository for the paper PLLaVA
568 stars 37 forks source link

Training Dataset #71

Closed linxid closed 3 months ago

linxid commented 3 months ago

I found in paper and DATA Readme, training model only use video dataset. But I found train_corpus = "videochat2_instruction_full" in config_pllava_nframe.py.

So, what dataset used to training model. Thanks.

linxid commented 3 months ago

Has been resolved, train_corpus will be overwritten.