can I simply set vit_l14 to larger input image size, such as 336/448, without re-training for stage2/3, videochat2?

OpenGVLab / Ask-Anything

[CVPR2024 Highlight][VideoChatGPT] ChatGPT with video understanding! And many more supported LMs such as miniGPT4, StableLM, and MOSS.

https://vchat.opengvlab.com/

MIT License

2.85k stars 230 forks source link

can I simply set vit_l14 to larger input image size, such as 336/448, without re-training for stage2/3, videochat2? #189

Closed dragen1860 closed 3 weeks ago

dragen1860 commented 3 weeks ago

hi, i want to test whether larger input image size helps my custom finetunning model performance. Does anyone know can i simply feed larger input image size without re-training on stage2&3. thank you.

Andy1621 commented 3 weeks ago

Yes, you can~ Please follow the code:

dragen1860 commented 3 weeks ago

ok, I will try.