facebookresearch / LaViLa

Code release for "Learning Video Representations from Large Language Models"
MIT License
492 stars 46 forks source link

Pretrained weight of HowTo100M #13

Open HYUNJS opened 1 year ago

HYUNJS commented 1 year ago

May I ask if you plan to release the pre-trained weights of Narrator and dual-encoder of HowTo100M?

Thank you for sharing your great work!

zhaoyue-zephyrus commented 1 year ago

Hi @HYUNJS ,

Thanks for your interests in our work.

For the pre-trained weight of HT-100M Narrator, please check out the 3rd-person video narrating demo, or more precisely the following lines in demo_narrator_3rd_person.py.

For the dual-encoder, it might be hard to retrieve the pre-trained weights since I don't have access to the internal storage any more. I might reproduce it later when I have time but I cannot guarantee you a specific timeline. However, do feel free to let me know if you have any question about the details.

Best, Yue

HYUNJS commented 1 year ago

Thank you for your answer!

Seth-Park commented 1 year ago

Are model weights for the dual-encoder available yet?