an "import cudf" leads to an error

[MarcoMagl]

I have a python file that I can run successfully up to completion.
I would like to use scalene to profile it and improve it.
Unfortunately, I get the following error:

    import cudf
  File "/apps/USE/easybuild/release/2023.1/software/cuDF/23.10.0-foss-2023a-CUDA-12.2.0-python-3.10.8/lib/python3.10/site-packages/cudf/__init__.py", line 9, in <module>
    _setup_numba()
  File "/apps/USE/easybuild/release/2023.1/software/cuDF/23.10.0-foss-2023a-CUDA-12.2.0-python-3.10.8/lib/python3.10/site-packages/cudf/utils/_numba.py", line 106, in _setup_numba
    versions = safe_get_versions()
  File "/apps/USE/easybuild/release/2023.1/software/cuDF/23.10.0-foss-2023a-CUDA-12.2.0-python-3.10.8/lib/python3.10/site-packages/cudf/utils/_ptxcompiler.py", line 106, in safe_get_versions
    driver_version, runtime_version = get_versions()
  File "/apps/USE/easybuild/release/2023.1/software/cuDF/23.10.0-foss-2023a-CUDA-12.2.0-python-3.10.8/lib/python3.10/site-packages/cudf/utils/_ptxcompiler.py", line 64, in get_versions
    versions = [int(s) for s in cp.stdout.strip().split()]
  File "/apps/USE/easybuild/release/2023.1/software/cuDF/23.10.0-foss-2023a-CUDA-12.2.0-python-3.10.8/lib/python3.10/site-packages/cudf/utils/_ptxcompiler.py", line 64, in <listcomp>
    versions = [int(s) for s in cp.stdout.strip().split()]
ValueError: invalid literal for int() with base 10: b'Scalene

I did not see any similar issue reported and as a new user of scalene, I am not sure on which side the error would be. Does someone have an idea?

Thanks!

[emeryberger]

This issue (which appears to be some interference of printing to stdout instead of stderr) went away for me after running it once as root (a.k.a., superuser). That said, I'll change these print statements to go to stderr.

In any event: I first created a virtual environment and installed Scalene and cudf (getting its dependencies first).

$ sudo apt install nvidia-cuda-toolkit
$ python3 -m venv test-cudf
$ source ../test-cudf/bin/activate
(test-cudf) $ python3 -m pip install scalene cudf

Then I ran the code as root.

$ sudo bash
# source ../test-cudf/bin/activate
(test-cudf) # python3 -m scalene --cpu --gpu testcudf.py 

It continues to run fine afterwards when not running as root.

(test-cudf) $ python3 -m scalene --cpu --gpu testcudf.py