Uploaded image for project: 'AI Platform Core Components'
  1. AI Platform Core Components
  2. AIPCC-1897

Torch 2.7.0 fails with ImportError: libnccl.so.2

XMLWordPrintable

    • Icon: Bug Bug
    • Resolution: Done
    • Icon: Undefined Undefined
    • None
    • None
    • Wheel Package Index
    • None
    • False
    • Hide

      None

      Show
      None
    • False

       

      CUDA build of Torch 2.7.0 fails to import in the base image because the image is missing NVIDIA's NCCL library:

      >>> import torch           
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/opt/app-root/lib/python3.11/site-packages/torch/__init__.py", line 409, in <module>
    from torch._C import *  # noqa: F403
    ^^^^^^^^^^^^^^^^^^^^^^
ImportError: libnccl.so.2: cannot open shared object file: No such file or directory

       

       

              cheimes@redhat.com Christian Heimes
              cheimes@redhat.com Christian Heimes
              Ali Raza, Emilien Macchi
              Votes:
              0 Vote for this issue
              Watchers:
              3 Start watching this issue

                Created:
                Updated:
                Resolved: