Not compile with GPU support in detectron2 - pytorch

Question

enter image description here
Detectron2 ran faster- RCNN when the error, from the error, should be the network RPN part caused the error.
The GPU should be running because the backbone part did not report an error.
How can I solve this problem？

Answer 1

The reason for this error is the server cuda version with pytorch
Cuda version mismatches, such as between 10.1 and 10.0. So you should check whether the pytorch cuda version is the same as the machine cuda version

Answer 2

"nvcc not found" or "Not compiled with GPU support" or "Detectron2 CUDA Compiler: not available".
CUDA is not found when building detectron2.
You should make sure
import torch
from torch.utils.cpp_extension import CUDA_HOME
print(torch.cuda.is_available(), CUDA_HOME)
should return
print (True, a directory with cuda) at the time you build detectron2.
Most models can run inference (but not training) without GPU support. To use CPUs, set MODEL.DEVICE='cpu' in the config.
and for in-depth solution you may visit this link:
https://detectron2.readthedocs.io/en/latest/tutorials/install.html#common-installation-issues