Import Error: DLL loaded failed when import torchvision

[Ieunn]

🐛 Describe the bug

The error happened when I'm doing this:

import torch
print(torch.version)
2.0.0+cu118
import torchvision
Traceback (most recent call last):
File "", line 1, in
File "C:\Coding\Anaconda\envs\stable-diffusion-webui\lib\site-packages\torchvision_init_.py", line 6, in
from torchvision import datasets, io, models, ops, transforms, utils
File "C:\Coding\Anaconda\envs\stable-diffusion-webui\lib\site-packages\torchvision\datasets_init_.py", line 1, in
from ._optical_flow import FlyingChairs, FlyingThings3D, HD1K, KittiFlow, Sintel
File "C:\Coding\Anaconda\envs\stable-diffusion-webui\lib\site-packages\torchvision\datasets_optical_flow.py", line 10, in
from PIL import Image
File "C:\Coding\Anaconda\envs\stable-diffusion-webui\lib\site-packages\PIL\Image.py", line 103, in
from . import _imaging as core
ImportError: DLL load failed while importing _imaging

The answers on Google said it's caused by unmatch of versions for pytorch and torchvision, but I've remove pytorch and torchvision several times and install them either by command line from official website or by pip locally install. But I still got this error whatever I do.

Versions

PyTorch version: 2.0.0+cu118
Is debug build: False
CUDA used to build PyTorch: 11.8
ROCM used to build PyTorch: N/A

OS: Microsoft Windows 11
GCC version: Could not collect
Clang version: Could not collect
CMake version: Could not collect
Libc version: N/A

Python version: 3.10.10 | packaged by conda-forge | (main, Mar 24 2023, 20:00:38) [MSC v.1934 64 bit (AMD64)] (64-bit runtime)
Python platform: Windows-10-10.0.22621-SP0
Is CUDA available: True
CUDA runtime version: 12.1.66
CUDA_MODULE_LOADING set to: LAZY
GPU models and configuration: GPU 0: NVIDIA GeForce RTX 4080 Laptop GPU
Nvidia driver version: 531.29
cuDNN version: Could not collect
HIP runtime version: N/A
MIOpen runtime version: N/A
Is XNNPACK available: True

Versions of relevant libraries:
[pip3] mypy-extensions==1.0.0
[pip3] numpy==1.23.3
[pip3] open-clip-torch==2.16.0
[pip3] pytorch-lightning==1.9.4
[pip3] torch==2.0.0+cu118
[pip3] torchaudio==2.0.0
[pip3] torchdiffeq==0.2.3
[pip3] torchmetrics==0.11.4
[pip3] torchsde==0.2.5
[pip3] torchvision==0.15.1+cu118
...
[conda] numpy 1.23.3 pypi_0 pypi
[conda] open-clip-torch 2.16.0 pypi_0 pypi
[conda] pytorch-cuda 11.8 h24eeafa_3 pytorch
[conda] pytorch-lightning 1.9.4 pypi_0 pypi
[conda] pytorch-mutex 1.0 cuda pytorch
[conda] torch 2.0.0+cu118 pypi_0 pypi
[conda] torchaudio 2.0.0 pypi_0 pypi
[conda] torchdiffeq 0.2.3 pypi_0 pypi
[conda] torchmetrics 0.11.4 pypi_0 pypi
[conda] torchsde 0.2.5 pypi_0 pypi
[conda] torchvision 0.15.1+cu118 pypi_0 pypi

[pmeier]

File "C:\Coding\Anaconda\envs\stable-diffusion-webui\lib\site-packages\PIL\Image.py", line 103, in
from . import _imaging as core
ImportError: DLL load failed while importing _imaging

The error is coming from PIL (install name Pillow) not from torchvision. The reason you are seeing this when importing torchvision is because we are importing PIL.

[Ieunn]

File "C:\Coding\Anaconda\envs\stable-diffusion-webui\lib\site-packages\PIL\Image.py", line 103, in
from . import _imaging as core
ImportError: DLL load failed while importing _imaging

The error is coming from PIL (install name Pillow) not from torchvision. The reason you are seeing this when importing torchvision is because we are importing PIL.

Yes I notced that and tried to reinstall pillow once but it didn't work.

[pmeier]

What do you get when running python -c 'import PIL'?

[Ieunn]

What do you get when running python -c 'import PIL'?

seems nothing happened...I get into next command line without error prompt.

[pmeier]

What about python -c 'from PIL import Image'?

[Ieunn]

What about python -c 'from PIL import Image'?

Okay now I got a similar error prompt as what I got when import torchvision.

Traceback (most recent call last):
File "", line 1, in
File "C:\Coding\Anaconda\envs\deep_learning\lib\site-packages\PIL\Image.py", line 100, in
from . import _imaging as core
ImportError: DLL load failed while importing _imaging

So does it mean the problem is caused by pillow and I should try to find and reinstall some matched version of pillow?

[pmeier]

So does it mean the problem is caused by pillow

Yes.

I should try to find and reinstall some matched version of pillow?

Not sure what you mean here. This is not a version problem. PIL.Image is the core namespace of the package and thus wasn't added recently or the like. Basically everything PIL>=8 should be fine with torchvision. What version do you have?

I suspect PIL is not correctly installed and you need to fix that. How did you install in the first place?

In any case, this is out of scope for torchvision. Thus, I'm closing this.