Thanks for releasing the codes! I was reading your paper, but still have some questions about LanguageBind used in Video-LLaVA:
1) Were the weights of the image/video encoder initialized from LanguageBind trainable or frozen, during Video-LLaVA training?
2) Which version of LanguageBind from the model zoo did you initialize the weights from, e.g. LanguageBind_Video_V1.5_FT or LanguageBind_Video_FT?
Hi,
Thanks for releasing the codes! I was reading your paper, but still have some questions about LanguageBind used in Video-LLaVA:
1) Were the weights of the image/video encoder initialized from LanguageBind trainable or frozen, during Video-LLaVA training? 2) Which version of LanguageBind from the model zoo did you initialize the weights from, e.g. LanguageBind_Video_V1.5_FT or LanguageBind_Video_FT?