Shuyang Gu

hi, i meet the same problem when i train from scratch, did you find out the reason?

> I am not sure about linear lr rule works well in gan trainings. I think you can just use 0.002 or use 4 sample per gpus. Thanks for your...

> You can get 3.81 FID for ffhq at 256px. (from https://arxiv.org/abs/2006.06676) Yes, but I found in the tensorflow implementation, there are implementation bugs on compute the FID. They compute...

> > > You can get 3.81 FID for ffhq at 256px. (from https://arxiv.org/abs/2006.06676) > > > > > > Yes, but I found in the tensorflow implementation, there are...

In VQ-Diffusion, the largest dataset is LAION-person, which contains about 42M images, we trained it for 2 epochs. In improved VQ-Diffusion, the largest dataset is ITHQ dataset, which contains about...

Sorry for the late reply. I think you may notice two important things. First, in the sampling process, the truncation rate is extremely important, we searched the best truncation rate...

(1) I think it will only have a slight difference. (2) we only trained it for 100 epochs, I'm not sure if more epochs will improve the performance. (3) we...

Sorry i haven't meet this problem, did you check if the network load the model correctly?

You may refer to https://github.com/cientgu/GIQA ,you can achieve DS by exchanging the training and testing when calculating the QS of GMM-GIQA.

I am glad you are interested in our work. 1. we use python 3.6, PyTorch 1.4 and sklearn 0.22.2 for our code. 2. we have provided the command examples to...