[Bug]: Unable to initialize model

[OneBugMaker]

Describe the bug

problem

When I tried to initialize the model in anomalib, an error occurred. This happened during the initialization phase without any further operations. Both "Padim" and "Patchcore" encountered the same issue. The detailed error message is as follows:

---

RuntimeError Traceback (most recent call last)
~\AppData\Local\Temp\ipykernel_21240\545706316.py in <cell line: 1>()
----> 1 model = Padim()

D:\ML\0-PHD2\27_IDD\MyIDD\new_19_14_42\anomalib\src\anomalib\models\image\padim\lightning_model.py in init(self, backbone, layers, pre_trained, n_features)
48 super().init()
49
---> 50 self.model: PadimModel = PadimModel(
51 backbone=backbone,
52 pre_trained=pre_trained,

D:\ML\0-PHD2\27_IDD\MyIDD\new_19_14_42\anomalib\src\anomalib\models\image\padim\torch_model.py in init(self, layers, backbone, pre_trained, n_features)
99 self.idx: torch.Tensor
100 self.loss = None
--> 101 self.anomaly_map_generator = AnomalyMapGenerator()
102
103 self.gaussian = MultiVariateGaussian()

D:\ML\0-PHD2\27_IDD\MyIDD\new_19_14_42\anomalib\src\anomalib\models\image\padim\anomaly_map.py in init(self, sigma)
24 super().init()
25 kernel_size = 2 * int(4.0 * sigma + 0.5) + 1
---> 26 self.blur = GaussianBlur2d(kernel_size=(kernel_size, kernel_size), sigma=(sigma, sigma), channels=1)
27
28 @staticmethod

D:\ML\0-PHD2\27_IDD\MyIDD\new_19_14_42\anomalib\src\anomalib\models\components\filters\blur.py in init(self, sigma, channels, kernel_size, normalize, border_type, padding)
66 self.kernel = normalize_kernel2d(self.kernel)
67 self.kernel.unsqueeze_(0).unsqueeze_(0)
---> 68 print(self.kernel.shape, self.channels)
69 self.kernel = self.kernel.expand(self.channels, -1, -1, -1)
70 self.kernel = self.kernel.expand(self.channels, -1, -1, -1)

RuntimeError: expand(torch.FloatTensor{[1, 1, 1, 33, 33]}, size=[1, -1, -1, -1]): the number of sizes provided (4) must be greater or equal to the number of dimensions in the tensor (5)

environment

python==3.10.0
torch==2.1.0+cu121
anomalib==1.1.0dev

Dataset

MVTec

Model

PatchCore

Steps to reproduce the behavior

When I tried to initialize the model in anomalib, an error occurred.

This happened during the initialization phase without any further operations.

Both "Padim" and "Patchcore" encountered the same issue.

OS information

OS information:

• OS: [Windows 11]
• Python version: [ 3.10.0]
• Anomalib version: [1.1.0dev]
• PyTorch version: [2.1.0]
• CUDA/cuDNN version: [12.1]
• GPU models and configuration: [GeForce RTX 2060]

Expected behavior

Resolve this bug and successfully use Padim and Patchcore models

Screenshots

Pip/GitHub

GitHub

What version/branch did you use?

No response

Configuration YAML

i do not know

Logs

i do not know

Code of Conduct

• I agree to follow this project's Code of Conduct

[samet-akcay]

Can you show the full code you ran? It is a bit hard to find out the problem from model = Padim()

[OneBugMaker]

Thank you for your comment.
I imported the required modules in Jupyter Notebook and attempted to initialize the model, but encountered an error, as shown in the screenshot.

[samet-akcay]

I have tried it on both MacOS and Linux, which worked fine. Can you check if your installation is correct?

On MacOS

On Linux

[OneBugMaker]

Thank you for your efforts！
The good news is that this issue has been resolved. Here are the specific steps taken to address it:

Open the file: anomalib/src/anomalib/models/components/filters/blur.py
Locate line 67: self.kernel.unsqueeze_(0).unsqueeze_(0)
Replace it with: self.kernel.unsqueeze_(0)

[luizfelippesr]

#1944

[ghost]

Thanks OneBugMaker. Faced same issue. Resolved.
In our case the file is located in:
C:\Users\ADMIN\anaconda3\envs\anomalib_env\Lib\site-packages\anomalib\models\components\filters\blur.py