georgecazenavette / glad Goto Github PK

Thanks for your amazing work. There are different class or functions which are imported in the buffer_mtt.py from utils, but they are not exist in the utils.py. would you please check and upload a new utils files. Thanks.

Hi @GeorgeCazenavette

I hope all is well. I am wondering if it would be possible for you to upload the Torch tensors containing the distilled dataset for the GLaD paper (CVPR 2023) distillation method, as you did for your previous CVPR 2022 paper available here?

Many thanks.

Hi, thank you for the open-source code. However, when run the command:
python distill_dc.py --dataset=CIFAR10 --layer=16 --ipc=1 --eval_it 1 --space wp
there is an error:
Traceback (most recent call last): File "glad/distill_dc.py", line 236, in <module> main(args) File "glad/distill_dc.py", line 185, in main gan_backward(latents=latents, f_latents=f_latents, image_syn=image_syn, G=G, args=args) File "glad/glad_utils.py", line 435, in gan_backward latents.grad = torch.cat(latents_grad_list) TypeError: expected Tensor as element 0 in argument 0, but got NoneType

there's no 'get_daparam' function in utils

Hi, I noticed there are two args: '--rand_gan_con' and '--rand_gan_un'. What exactly the meaning of these args? And how to use a random generative prior?
if args.rand_gan_con: network_pkl = "../stylegan_xl/random_conditional_{}.pkl".format(res) key = "G" elif args.rand_gan_un: network_pkl = "../stylegan_xl/random_unconditional_{}.pkl".format(res) key = "G"
Thanks a lot!