Xirui Li

Considering that the lr scheduler step 4 times with 4 GPUs, it seems logical that the global step should also be updated 4 times to ensure that the total training...

I solved the problem by adding the local host to "/etc/hosts" manually. Try this: [https://blog.csdn.net/lin_xiao_yi/article/details/132490694](https://blog.csdn.net/lin_xiao_yi/article/details/132490694)

Yes, it is normal. As we have two-stage token merging and unmerging operations around each self-attention module, they add computation overhead compared to processing each frame separately (w.o. VidToMe). VidToMe...

Hi, It seems a problem of transformers version. We use transformers==4.30.2. You can check whether your version matches.

Hi Jade, Thank you so much for the kind words and your interest in VidToMe! 🎉🚀 I'm excited about the idea of integrating it with the Hugging Face ecosystem. I...