Open lsj1111 opened 2 months ago
Hi, are you using the correct cuda and Pytorch version that matches your GPU arch?
Hi, are you using the correct cuda and Pytorch version that matches your GPU arch?
year,I have solved this problem. In the official documentation of detectron2, it seems that only cuda11.3 is supported, so I used cuda11.3 and caused the above problem, but then I found that cuda11.6 can also use detectron2, so the problem was solved. .
I configured it according to the author's method:
conda create --name maskdino python=3.8 -y conda activate maskdino conda install pytorch==1.9.0 torchvision==0.10.0 cudatoolkit=11.1 -c pytorch -c nvidia pip install -U opencv-python
under your working directory
git clone git@github.com:facebookresearch/detectron2.git cd detectron2 pip install -e . pip install git+https://github.com/cocodataset/panopticapi.git pip install git+https://github.com/mcordts/cityscapesScripts.git
cd .. git clone git@github.com:facebookresearch/MaskDINO.git cd MaskDINO pip install -r requirements.txt cd maskdino/modeling/pixel_decoder/ops sh make.sh
But I encountered some problems:
RuntimeError:nvrte:error:invalid value for -gpu-architecture (-arch) nvrtc compilation failed:
define NANint as float(0x7fffffff)
define PoS INFINITY_int_as float(0x7f800000)
define NEG INFINITYint as float(0xff800000)
template
_device_Tmaximum(T a,Tb){returnisnan(a)?a:(a>b?a: b);
template
device -Tminimum(Ta,Tb){return isnan(a)?a:(a<b?a: b);