t_emd error in ViTset code about few-shot-diffusion-models HOT 5 OPEN

Hi, how to run the preprocess_datasets.py with dataset cifar? I find that the type of dataset in it is h5py. How did you prepare the dataset?

from few-shot-diffusion-models.

Hi, how to run the preprocess_datasets.py with dataset cifar? I find that the type of dataset in it is h5py. How did you prepare the dataset?

run this code to download cifar100,

from torchmeta.datasets import Omniglot, DoubleMNIST, TripleMNIST, CUB, FC100
dataset = FC100("./", meta_split='train', download=True)

This code will have an errror, but the h5py file arleady download, you ignore that and continue run preprocess_datasets.py

from few-shot-diffusion-models.

run this code to download cifar100,

from torchmeta.datasets import Omniglot, DoubleMNIST, TripleMNIST, CUB, FC100
dataset = FC100("./", meta_split='train', download=True)

This code will have an errror, but the h5py file arleady download, you ignore that and continue run preprocess_datasets.py

Thank you.
I met the same error about the shape of t_emb as you mentioned. It seems that there are some other mistakes in this code instead of running directly.

from few-shot-diffusion-models.

Maybe only unet or vit is correct. I fix vit-set code error and sample code error, and i get 46 FID on Vit-set LAC T, although paper is 40.71.
I won't take more tests，and there is no need to waste time to modify code errors.
If you need help about this, you can email me [email protected]

from few-shot-diffusion-models.

Maybe only unet or vit is correct. I fix vit-set code error and sample code error, and i get 46 FID on Vit-set LAC T, although paper is 40.71. I won't take more tests，and there is no need to waste time to modify code errors. If you need help about this, you can email me [email protected]

Hello, can you tell me how to fix the vit-set code error and sample code error? I met the same problems now.
Thank you.

from few-shot-diffusion-models.