About GPUs and pre-training weights

[Mxwgreat]

Hello! You have done an excellent job and I am very interested in this work. I would like to know how many GPUs and the type of GPUs you used for training on CIFAR and MNIST datasets, and do you have any plans to release the pre-trained weights for GFDM models? I look forward to your reply in your busy schedule. Thanks!

[GabrielNobis]

Hello! To sample from pre-trained models please have a look at the section "Sample from Pretrained Models" in the README. The training on CIFAR10 with EMA was done on two A100 GPUs with 40G RAM in parallel. It took round about 50h to train for 1 mio iterations with batch size 128. The training on MNIST was done on one A100 GPU with 32GB RAM. It took around 17h with batch_size 1024.