Error when running model.py, something to do with RandomRotation?

[kbbernander]

Error when running model.py, something to do with RandomRotation?

[kbbernander]

Thanks for your answer and support. I tried updating PIL from 5.4.1 to 7.0.0, and executing the code inside a virtual environment but still the same problem occurs.

Here are the package version (Python 3.7 is used)

(e2cnn_2) karl@IT-LL-karbe297:~$ pip list
Package Version

---

attrs 19.3.0
backcall 0.1.0
bleach 3.1.0
decorator 4.4.1
defusedxml 0.6.0
e2cnn 0.1
entrypoints 0.3
importlib-metadata 1.4.0
ipykernel 5.1.3
ipython 7.11.1
ipython-genutils 0.2.0
jedi 0.15.2
Jinja2 2.10.3
json5 0.8.5
jsonschema 3.2.0
jupyter-client 5.3.4
jupyter-core 4.6.1
jupyterlab 1.2.5
jupyterlab-server 1.0.6
MarkupSafe 1.1.1
mistune 0.8.4
more-itertools 8.1.0
nbconvert 5.6.1
nbformat 5.0.3
notebook 6.0.2
numpy 1.18.1
pandocfilters 1.4.2
parso 0.5.2
pexpect 4.7.0
pickleshare 0.7.5
Pillow 7.0.0
pip 20.0.1
pkg-resources 0.0.0
prometheus-client 0.7.1
prompt-toolkit 3.0.2
ptyprocess 0.6.0
Pygments 2.5.2
pyrsistent 0.15.7
python-dateutil 2.8.1
pyzmq 18.1.1
scipy 1.4.1
Send2Trash 1.5.0
setuptools 45.1.0
six 1.14.0
terminado 0.8.3
testpath 0.4.4
torch 1.4.0
torchvision 0.5.0
tornado 6.0.3
traitlets 4.3.3
wcwidth 0.1.8
webencodings 0.5.1
wheel 0.33.6
zipp 2.0.0

[Gabri95]

Hi @Tarnekar! I am happy you liked our work!

Unfortunately, I have never had this problem.
It seems its source is in Pytorch's code where PIL functionalities are used.
Is it possible that you have an old version of PIL which is not fully compatible with your current version of torchvision?

Could you give some more details about your Python environment?

[kbbernander]

That was indeed the problem; using torch 1.3, torchvision 0.4 and pillow 6.2.1 and all the code runs. Although there is a warning that torchvision 0.4.0 has a requirement of torch==1.2.0.

Thanks!

[Gabri95]

I was indeed able to reproduce your problem.
Using torch 1.3 and torchvision 0.4 (and pillow 6.2.1), the following code runs.
However, I get your error when I update to the newest (recently released) versions of torch (1.4) and torchvision (0.5).

import torch
from torch.utils.data import DataLoader
from torchvision.transforms import RandomRotation
from torchvision.transforms import ToTensor
from torchvision.transforms import Compose
from torchvision.datasets import MNIST
from PIL import Image


transform = Compose([
    RandomRotation(180., resample=Image.BILINEAR, expand=False),
    ToTensor(),
])

mnist = MNIST("./", train=True, transform=transform, download=True)
train_loader = DataLoader(mnist, batch_size=1)


for x, t in enumerate(train_loader):
    print("hello world")
    break

For this reason, I think this is a problem with the new version of Pytorch.
I would suggest you to temporarily downgrade it to 1.3 if you want to use transforms.RandomRotation.
You can probably find more support on Pytorch's forum.

Hope this helps!