Error on import pykeops: Unable to find file.

[johnsoltis]

Hello,

I have been struggling to install and get pykeops to run. I encounter the error below whenever I try to run 'import pykeops'.

Traceback (most recent call last):
File "/home/jsoltis/cuda_test.py", line 7, in
import pykeops
File "/home/jsoltis/miniforge3/envs/gpu_env/lib/python3.10/site-packages/pykeops/init.py", line 2, in
import keopscore
File "/home/jsoltis/miniforge3/envs/gpu_env/lib/python3.10/site-packages/keopscore/init.py", line 14, in
from keopscore.config import *
File "/home/jsoltis/miniforge3/envs/gpu_env/lib/python3.10/site-packages/keopscore/config/init.py", line 10, in
cuda_config = CUDAConfig()
File "/home/jsoltis/miniforge3/envs/gpu_env/lib/python3.10/site-packages/keopscore/config/cuda.py", line 56, in init
self.set_use_cuda()
File "/home/jsoltis/miniforge3/envs/gpu_env/lib/python3.10/site-packages/keopscore/config/cuda.py", line 146, in set_use_cuda
self.get_cuda_include_path()
File "/home/jsoltis/miniforge3/envs/gpu_env/lib/python3.10/site-packages/keopscore/config/cuda.py", line 329, in get_cuda_include_path
cuda_h_path = self.get_include_file_abspath("cuda.h")
File "/home/jsoltis/miniforge3/envs/gpu_env/lib/python3.10/site-packages/keopscore/config/cuda.py", line 344, in get_include_file_abspath
tmp_file = tempfile.NamedTemporaryFile(dir=self.get_build_folder()).name
File "/home/jsoltis/miniforge3/envs/gpu_env/lib/python3.10/tempfile.py", line 575, in NamedTemporaryFile
file = _io.open(dir, mode, buffering=buffering,
File "/home/jsoltis/miniforge3/envs/gpu_env/lib/python3.10/tempfile.py", line 572, in opener
fd, name = _mkstemp_inner(dir, prefix, suffix, flags, output_type)
File "/home/jsoltis/miniforge3/envs/gpu_env/lib/python3.10/tempfile.py", line 256, in _mkstemp_inner
fd = _os.open(file, flags, 0o600)
FileNotFoundError: [Errno 2] No such file or directory: '/home/jsoltis/.cache/keops2.3/Linux_exp-9-58_4.18.0-513.24.1.el8_9.x86_64_p3.10.20_CUDA_VISIBLE_DEVICES_0/tmpych3azll'

Attached is a copy of the environment file: gpu_env.yml

Torch is able to detect the GPU correctly. I tried a variant of the environment where I also installed cmake in addition to the other files (i.e., conda install conda-forge::cmake), but that gave the exact same error.

Any help would be greatly appreciated.