swin-unetr model pamars

[Yffy123456]

I'm trying to calculate pamars its calculations and I'm having some problems

if name == 'main': x = torch.rand(1, 1, 96, 96, 96) model =SwinUNETR() model.cuda() y = model(x) print(y.shape)

/home/wf/anaconda3/envs/nnFormer0/bin/python /home/wf/anaconda3/envs/nnFormer0/lib/python3.8/site-packages/monai/networks/nets/swin_unetr.py Traceback (most recent call last): File "/home/wf/anaconda3/envs/nnFormer0/lib/python3.8/site-packages/monai/networks/nets/swin_unetr.py", line 990, in model =SwinUNETR() File "/home/wf/anaconda3/envs/nnFormer0/lib/python3.8/site-packages/monai/networks/nets/swin_unetr.py", line 113, in init self.swinViT = SwinTransformer( File "/home/wf/anaconda3/envs/nnFormer0/lib/python3.8/site-packages/monai/networks/nets/swin_unetr.py", line 934, in init layer = BasicLayer( File "/home/wf/anaconda3/envs/nnFormer0/lib/python3.8/site-packages/monai/networks/nets/swin_unetr.py", line 815, in init [ File "/home/wf/anaconda3/envs/nnFormer0/lib/python3.8/site-packages/monai/networks/nets/swin_unetr.py", line 816, in SwinTransformerBlock( File "/home/wf/anaconda3/envs/nnFormer0/lib/python3.8/site-packages/monai/networks/nets/swin_unetr.py", line 549, in init self.attn = WindowAttention( File "/home/wf/anaconda3/envs/nnFormer0/lib/python3.8/site-packages/monai/networks/nets/swin_unetr.py", line 437, in init coords = torch.stack(torch.meshgrid(coords_d, coords_h, coords_w, indexing="ij")) TypeError: stack(): argument 'tensors' (position 1) must be tuple of Tensors, not NoneType

Looking forward to getting your recovery

[tangy5]

HI @Yffy123456, the definition of SwinUNETR can be specified with several arguments. Below is an example:

import torch
from monai.networks.nets import SwinUNETR

x = torch.rand(1, 1, 96, 96, 96)
model = SwinUNETR(
    img_size=[96, 96, 96],
    in_channels=1,
    out_channels=3,
    feature_size=48,
)
y = model(x)
print(y.shape)

The output:

torch.Size([1, 3, 96, 96, 96])

Thanks.