Install from source results in shared object error

release/packaging
1

I am trying to install pytorch from source for development involving CUDA, however, I hit the following error on the shared object file. What am I missing?

    from torch.distributed.run import main
  File "/workspace/pytorch/torch/__init__.py", line 415, in <module>
    from torch._C import *  # noqa: F403
    ^^^^^^^^^^^^^^^^^^^^^^
ImportError: /workspace/pytorch/torch/lib/libtorch_python.so: undefined symbol: _ZTIN5torch8inductor28AOTIModelContainerRunnerCudaE

I have followed the steps from GitHub - pytorch/pytorch: Tensors and Dynamic neural networks in Python with strong GPU acceleration with pytorch from commit - fca2601c9d628e1bd2d75c7318cd22c4e8c832aa.

My environment is linux.

2

I’d suggest posting your question https://discuss.pytorch.org/

dev-discuss is closer to a blog ran by pytorch core developers

1 Like
3

Will do that, thank you for the pointer.

Have raised it here.

4

Hello @kmehant Please create an issue here Sign in to GitHub · GitHub

Also please provide more details such as error log and during what step when you build from source you see the error: GitHub - pytorch/pytorch: Tensors and Dynamic neural networks in Python with strong GPU acceleration

Also Please run the following and paste the output:

wget https://raw.githubusercontent.com/pytorch/pytorch/main/torch/utils/collect_env.py
# For security purposes, please check the contents of collect_env.py before running it.
python collect_env.py