Issues building docker image from NGC container nvcr.io/nvidia/pytorch:22*-py3

Accelerated Computing NGC GPU Cloud Feature Requests
1

After building docker image from nvcr.io/nvidia/pytorch:22.02-py3, AND also installing in the Docker File apt-get install -y libboost-all-dev, the pytorch within container doesn’t work any more. The error reporting is strange and fatal ! Any Idea how to solve this usue since libboost-all-dev is needed for other software in the Docker file !

python
Python 3.8.12 | packaged by conda-forge | (default, Oct 12 2021, 21:59:51)
[GCC 9.4.0] on linux
Type “help”, “copyright”, “credits” or “license” for more information.
import torch
Traceback (most recent call last):
File “”, line 1, in
File “/opt/conda/lib/python3.8/site-packages/torch/init.py”, line 198, in
_load_global_deps()
File “/opt/conda/lib/python3.8/site-packages/torch/init.py”, line 151, in _load_global_deps
ctypes.CDLL(lib_path, mode=ctypes.RTLD_GLOBAL)
File “/opt/conda/lib/python3.8/ctypes/init.py”, line 373, in init
self._handle = _dlopen(self._name, mode)
OSError: /opt/hpcx/ompi/lib/libmpi.so.40: undefined symbol: opal_hwloc201_hwloc_get_type_depth

2

Hi, I also faced the same problem. Did you solve the problem?

3

I am also facing the same problem. how did u solve the problem?

4

I somehow managed to live without using libboost-all-dev … then it works.

5

Faced the same problem.
Adding

ENV PATH="${PATH}:/opt/hpcx/ompi/bin"
ENV LD_LIBRARY_PATH="${LD_LIBRARY_PATH}:/opt/hpcx/ompi/lib"

before installing libboost-all-dev in my Dockerfile fixed the issue for me.

7 Likes