[bitnami/pytorch] RuntimError : Couldn't load custom C++ ops. This can happen if your PyTorch and torchvision versions are incompatible,

[zhouenxian]

Name and Version

bitnami/pytorch:2.1.1

What architecture are you using?

arm64

What steps will reproduce the bug?

I pull the image ''bitnami/pytorch:2.1.1" from DockerHub. Then I run a container to deploy my service based on pytorch. When I
use the service to predict , an error occur in the container : Couldn't load custom C++ ops. This can happen if your PyTorch and torchvision versions are incompatible

What is the expected behavior?

What do you see instead?

Beside , when I try to import torchvison , a Warning occur /opt/bitnami/python/lib/python3.11/site-packages/torchvision/io/image.py:13: UserWarning: Failed to load image Python extension: '/opt/bitnami/python/lib/python3.11/site-packages/torchvision/image.so: undefined symbol: _ZN5torch3jit17parseSchemaOrNameERKSs'If you don't plan on using image functionality fromtorchvision.io, you can ignore this warning. Otherwise, there might be something wrong with your environment. Did you havelibjpegorlibpnginstalled before buildingtorchvisionfrom source?

Additional information

No response

[zhouenxian]

Besides , when I depoly the same service based on the images "bitnami/pytorch:2.0.1" on ubuntu18.4 with amd64 architecture， anything is well ,this error doesn't happen . But when I use "bitnami/pytorch:2.1.1" deploy the same service on ubuntu18.4 with amd64 architecture. Unfortunately, this bug happen 。I found that this bug exists in the latest images, whether it is amd64 or ARM64

[github-actions[bot]]

This Issue has been automatically marked as "stale" because it has not had recent activity (for 15 days). It will be closed if no further activity occurs. Thanks for the feedback.

[fmulero]

Hi @zhouenxian I've just reproduced your issue and it seems related to the compilation phase. I'll open an internal task to address it but we are a small team and I can't give you any ETA. We'll keep you posted on.