undefined symbol: _ZN6caffe28TypeMeta21_typeMetaDataInstanceIN3c107complexIfEEEEPKNS_6detail12TypeMetaDataEv“

[lcl-git-3d]

Hi, When I run: "python train.py --dataset lmo --batch_size 12 --num_inner_loops 10 --num_solver_steps 3". Program error: "ImportError: / home/ta/anaconda3 / envs/cir/lib/python3.8 / site - packages/lietorch_backends retaining - 38 - x86_64 - Linux - gnu. So: undefined symbol: _ZN6caffe28TypeMeta21_typeMetaDataInstanceIN3c107complexIfEEEEPKNS_6detail12TypeMetaDataEv“

My virtual environment: python=3.8 pytorch=1.8.1 torchvision=0.9.1 cuda=11.1. Both are the same as environment.ymal. How can I solve it?

[ji-min-song]

Hello, I have a same problem in RTX 4090. But it was not in RTX 3090 ti in previous experience. How about your HW setup?

p.s. It looks like CUDA error because when i setup anaconda environment with this github repo instruction. pytorch or torchvision was installed as cpu version.

[lcl-git-3d]

Hello, I have a same problem in RTX 4090. But it was not in RTX 3090 ti in previous experience. How about your HW setup?

RTX 3090 CUDA Vision:11.6.Can you solve this problem?

[ji-min-song]

i think you have to check your nvcc compiler.

[lcl-git-3d]

Thank you for your reply！I have checked my nvcc compiler,but I don't know what the problem is.Could you take a look at it for me, please? Thank you very much indeed！ When I checked the ncvv compiler, I found that the cuda version is 11.3, while the cuda version installed in my virtual environment according to Environment. yaml is 11.1. Is this the problem?

[rahul-martianaut]

Hello @lcl-git-3d . How did you solve this issue? I got the same error