e4s2022

After checking section D.1 in the Appendix of the LDM paper, I found the KL-reg term is applied to avoid arbitrarily high-variance latent space. Because the variance of the latent...

Could you please post the command line output information?

Indeed, the `./pretrained_ckpts/auxiliray/model.pth` was ignored by the `.gitignore` file due to an incorrectly configured rule. I've added it to the main branch, please check the latest update. Thank you for...

Didn't test with full CPU env, which is not recommended

300k iterations. 8 A100 GPU, each with batch size = 2. We use reconstruction as the proxy task during training. So, I think it is better to visualize the reconstruction...

what's the GCC version? Similar as #3

Sorry that I have no Windows machine. To be honest, I tested the installation steps with a pure clean Ubuntu environment, but didn’t find this issues… could you please downgrade...

I never tried WSL before. I think you should try to downgrade the GCC of WSL system first.

Seems like a ninja related error. My env is CUDA 11.3 + GCC 7.5.0

What's the file size? `ls pretrained_ckpts/facevid2vid/00000189-checkpoint.pth.tar -l` tells me `2112619148`. Please try to re-download the file.