I have tried different environment configurations and found that there will be problems in some cases.
The recommended configuration is python 3.7 + pytorch 1.6 + pytorch_scatter 1.4 + cuda 10.1.
The following configuration will have some problems, for example
"No module named 'marching_cubes" or " CUDA error: invalid device ordinal"
The problematic configurations include:
python 3.7 + pytorch 1.7.0 + pytorch_scatter 2.3 + cuda 11.0.
python 3.7 + pytorch 1.7.1 + pytorch_scatter 2.4 + cuda 11.0.
python 3.7 + pytorch 14.0 + pytorch_scatter 1.4 + cuda 10.0.
Hi @LiuXinqi12, thank you so much for this exploration! I really appreciate your dedication.
For further information, CUDA error: invalid device ordinal is discussed further in this issue here :)
I have tried different environment configurations and found that there will be problems in some cases. The recommended configuration is python 3.7 + pytorch 1.6 + pytorch_scatter 1.4 + cuda 10.1.
The following configuration will have some problems, for example "No module named 'marching_cubes" or " CUDA error: invalid device ordinal"
The problematic configurations include: python 3.7 + pytorch 1.7.0 + pytorch_scatter 2.3 + cuda 11.0. python 3.7 + pytorch 1.7.1 + pytorch_scatter 2.4 + cuda 11.0. python 3.7 + pytorch 14.0 + pytorch_scatter 1.4 + cuda 10.0.