-
Bug
-
Resolution: Done
-
Critical
-
None
-
None
-
None
Importing torchaudio package failed on ROCm with the following error. It is looking for libtorch_cuda.so instead of libtorch_hip.so that is present on the system. This issue was noticed as part of testing vllm-0.10.0 on ROCm.
(.venv) (app-root) /work$ python Python 3.12.9 (main, Jun 20 2025, 00:00:00) [GCC 11.5.0 20240719 (Red Hat 11.5.0-5)] on linux Type "help", "copyright", "credits" or "license" for more information. >>> import torchaudio Traceback (most recent call last): File "<stdin>", line 1, in <module> File "/opt/app-root/.venv/lib64/python3.12/site-packages/torchaudio/__init__.py", line 2, in <module> from . import _extension # noqa # usort: skip ^^^^^^^^^^^^^^^^^^^^^^^^ File "/opt/app-root/.venv/lib64/python3.12/site-packages/torchaudio/_extension/__init__.py", line 38, in <module> _load_lib("libtorchaudio") File "/opt/app-root/.venv/lib64/python3.12/site-packages/torchaudio/_extension/utils.py", line 60, in _load_lib torch.ops.load_library(path) File "/opt/app-root/.venv/lib64/python3.12/site-packages/torch/_ops.py", line 1392, in load_library ctypes.CDLL(path) File "/usr/lib64/python3.12/ctypes/__init__.py", line 379, in __init__ self._handle = _dlopen(self._name, mode) ^^^^^^^^^^^^^^^^^^^^^^^^^ OSError: libtorch_cuda.so: cannot open shared object file: No such file or directory
The latest base image from quay (quay.io/aipcc/base-images/rocm) was used for testing and wheels were installed from the builder cache https://gitlab.com/api/v4/projects/72176068/packages/pypi/simple/
On the same system, in a different virtual environment, I installed upstream torchaudio (2.7.1) wheel and the import didn't give any error.
- mentioned on
(1 mentioned on)
