Question about the incompatibility of cuda12.1 version with cuda12.0 version on the server

Hi all,
I would like to know if the dgl under the cuda12.1 version is not compatible with the cuda12.0 server? Because my server’s cuda version is 12.0 and when I ran the code, there came an error:

Traceback (most recent call last):
File “/home/awp_Admin_ZixCQwyA/gxy/matbench-discovery-1.1.1/models/alignn/train_alignn2.py”, line 11, in
from alignn.config import TrainingConfig
File “/home/awp_Admin_ZixCQwyA/miniconda3/envs/mat/lib/python3.11/site-packages/alignn/config.py”, line 9, in
from alignn.models.modified_cgcnn import CGCNNConfig
File “/home/awp_Admin_ZixCQwyA/miniconda3/envs/mat/lib/python3.11/site-packages/alignn/models/modified_cgcnn.py”, line 5, in

import dgl
File “/home/awp_Admin_ZixCQwyA/miniconda3/envs/mat/lib/python3.11/site-packages/dgl/init.py”, line 14, in
from .backend import backend_name, load_backend # usort: skip
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File “/home/awp_Admin_ZixCQwyA/miniconda3/envs/mat/lib/python3.11/site-packages/dgl/backend/init.py”, line 122, in
load_backend(get_preferred_backend())
File “/home/awp_Admin_ZixCQwyA/miniconda3/envs/mat/lib/python3.11/site-packages/dgl/backend/init.py”, line 51, in load_b
ackend
from …_ffi.base import load_tensor_adapter # imports DGL C library
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File “/home/awp_Admin_ZixCQwyA/miniconda3/envs/mat/lib/python3.11/site-packages/dgl/_ffi/base.py”, line 50, in
_LIB, _LIB_NAME, _DIR_NAME = _load_lib()
^^^^^^^^^^^
File “/home/awp_Admin_ZixCQwyA/miniconda3/envs/mat/lib/python3.11/site-packages/dgl/_ffi/base.py”, line 39, in _load_lib
lib = ctypes.CDLL(lib_path[0])
^^^^^^^^^^^^^^^^^^^^^^^^
File “/home/awp_Admin_ZixCQwyA/miniconda3/envs/mat/lib/python3.11/ctypes/init.py”, line 376, in init
self._handle = _dlopen(self._name, mode)
^^^^^^^^^^^^^^^^^^^^^^^^^
OSError: libcudart.so.12: cannot open shared object file: No such file or directory

Looking forward to any answers! Thank you so much!!!

CUDA 12.0 is not part of DGL’s official support list so the behavior might be unknown, and CUDA 12.0 doesn’t seem to be supported by PyTorch either. You may need to build DGL (and probably PyTorch) from scratch.

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.