Skip to content

[rocm] No GPU support after rebuild with ROCm 6.0

Description:

After updating ROCm to 6.0 and python-pytorch-opt-rocm to 2.1.2-2, pytorch doesn't seem to be built with GPU support (which is called CUDA).

$ python -c "import torch; torch.cuda.current_device()"
Traceback (most recent call last):
  File "<string>", line 1, in <module>
  File "/usr/lib/python3.11/site-packages/torch/cuda/__init__.py", line 769, in current_device
    _lazy_init()
  File "/usr/lib/python3.11/site-packages/torch/cuda/__init__.py", line 289, in _lazy_init
    raise AssertionError("Torch not compiled with CUDA enabled")
AssertionError: Torch not compiled with CUDA enabled

In addition, the following returned previously 1, but now it returns 0.

python -c "import torch; print(torch.cuda.device_count())"

Additional info:

  • package version(s): 2.1.2-2
  • config and/or log files:
  • link to upstream bug report, if any:

Steps to reproduce:

  1. Install python-pytorch-opt-rocm from testing.
  2. Run: python -c "import torch; torch.cuda.current_device()"
To upload designs, you'll need to enable LFS and have an admin enable hashed storage. More information