NVIDIA / MinkowskiEngine

Minkowski Engine is an auto-diff neural network library for high-dimensional sparse tensors
https://nvidia.github.io/MinkowskiEngine
Other
2.47k stars 367 forks source link

install error in pytorch 1.5.1 cuda 10.3 minkowski0.4.3 #422

Open emilypython opened 2 years ago

emilypython commented 2 years ago

NVCC src/coords_manager.cu nvcc fatal : Unsupported gpu architecture 'compute_30' make: [Makefile:167:objs/cuda/coords_manager.o] 错误 1 make: 正在等待未完成的任务.... Traceback (most recent call last): File "setup.py", line 188, in run_command(*compile_args) File "setup.py", line 90, in run_command subprocess.check_call(args) File "/home/msy/anaconda3/envs/sparsem/lib/python3.7/subprocess.py", line 363, in check_call raise CalledProcessError(retcode, cmd) subprocess.CalledProcessError: Command '('make', '-j12', 'PYTHON=/home/msy/anaconda3/envs/sparsem/bin/python', 'BLAS=openblas')' returned non-zero exit status 2


To Reproduce Steps to reproduce the behavior. If the code is not attached and cannot be reproduced easily, the bug report will be closed without any comments.


Desktop (please complete the following information):

JimmyWong1998 commented 2 years ago

Hi, I also met this problem. Have you solved that?

serser commented 2 years ago

Commenting out SM_30 in the Makefile works.

serser commented 2 years ago

but met https://github.com/NVIDIA/MinkowskiEngine/issues/177 on cuda11. :(

mellody11 commented 3 weeks ago

but met #177 on cuda11. :(

Have you solved the problem? Thanks.