Few doubts on training !

[kpbhat25]

Hi ,
Can you please help me with understanding following issue :
I am trying to understand the code and run it.
Dataset : Colon/normal
1.
Had to import SwinUNetModel in script_util.py and artifusion.py , was it missed when pushing the latest code ?

File "/ArtiFusion/model_training/scripts/improved_diffusion/swinunet_flops.py", line 538, in forward
timesteps = x[1]
IndexError: index 1 is out of bounds for dimension 0 with size 1

I increased the batch size to 4 , this error goes of but , the patch embedding has error bcoz it take x.shape as B,C,H,W.
Can you check it?

What is the thing i missed ? (only data and logger path was to be filled in)

[zhenqi-he]

Hi, thanks for your questions.

For the issues you mentioned,

1, SwinUNetModel should be ArtiFusionModel. I changed the name of SwinUNetModel to ArtiFusionModel and forgot to update it in this repo.

2, We will recheck this size error, as we use batch_size = 1 during the training.

Great Thanks for pointing these errors out. We will fix these errors soon.