Closed DanielCoelho112 closed 2 years ago
The problem was the version of torchvision. This solves:
pip install torchvision==0.7.0
Hi @DanielCoelho112 ,
you can put in the requirements.txt file the required version, e.g.
https://iscompatible.readthedocs.io/en/latest/
That way other people do not have to worry about this.
The transforms applied to the images are saved using a pickle file. The purpose of saving the transforms is to associate the transforms with the model and then when using the model we automatically apply the right transformations to the inputs. The problem is that I'm having problems reading the transforms when the file is saved inside docker. I think this is due to different versions of the libraries. I'll look into it.