scs-python inference 时报错，

[UsersNGT]

你好，请问在使用rtx3090 进行inference时， scs报错，找不到_scs_gpu.py是什么原因，是scs不支持cuda11.0么？

[jiaweihe1996]

Did you install the gpu version of scs solver? As

# Install scs-gpu
pip uninstall scs
cd scs-python
python setup.py install --scs --gpu

[UsersNGT]

Did you install the gpu version of scs solver? As

# Install scs-gpu
pip uninstall scs
cd scs-python
python setup.py install --scs --gpu

是的，安装了，在v100服务器，cuda10上能运行，在3090，cuda11 就报错了，没有找到原因。。所以想请教一下

[jiaweihe1996]

I test only on CUDA10.0, and don't know what's wrong with CUDA11. Could you provide detailed error message when you install scs-gpu? Maybe you could install a newer version of scs-python from https://github.com/bodono/scs-python. Another option is to use CPU version of SCS solver, with only a little bit slower. See https://github.com/jiaweihe1996/GMTracker/issues/4.

[UsersNGT]

this is error message in scs-gpu, now only use cpu version.....

[jiaweihe1996]

From this error message, it seems you do not install scs-gpu correctly. Does any error appear when you run python setup.py install --scs --gpu?

[UsersNGT]

root@09acf0732b67:/app/scs# python setup.py install --scs --gpu Namespace(blas64=False, extraverbose=False, float32=False, gpu=True, gpu_atrans=True, int32=False, scs=True) running install running bdist_egg running egg_info writing scs.egg-info/PKG-INFO writing dependency_links to scs.egg-info/dependency_links.txt writing requirements to scs.egg-info/requires.txt writing top-level names to scs.egg-info/top_level.txt blas_mkl_info: libraries = ['mkl_rt', 'pthread'] library_dirs = ['/opt/conda/lib'] define_macros = [('SCIPY_MKL_H', None), ('HAVE_CBLAS', None)] include_dirs = ['/opt/conda/include'] blas_opt_info: libraries = ['mkl_rt', 'pthread'] library_dirs = ['/opt/conda/lib'] define_macros = [('SCIPY_MKL_H', None), ('HAVE_CBLAS', None)] include_dirs = ['/opt/conda/include'] lapack_mkl_info: libraries = ['mkl_rt', 'pthread'] library_dirs = ['/opt/conda/lib'] define_macros = [('SCIPY_MKL_H', None), ('HAVE_CBLAS', None)] include_dirs = ['/opt/conda/include'] lapack_opt_info: libraries = ['mkl_rt', 'pthread'] library_dirs = ['/opt/conda/lib'] define_macros = [('SCIPY_MKL_H', None), ('HAVE_CBLAS', None)] include_dirs = ['/opt/conda/include'] Supported SIMD extensions in this NumPy install: baseline = SSE,SSE2,SSE3 found = SSSE3,SSE41,POPCNT,SSE42,AVX,F16C,FMA3,AVX2,AVX512F,AVX512CD,AVX512_SKX not found = AVX512_KNL,AVX512_KNM,AVX512_CNL {'libraries': ['mkl_rt', 'pthread'], 'library_dirs': ['/opt/conda/lib'], 'define_macros': [('SCIPY_MKL_H', None), ('HAVE_CBLAS', None)], 'include_dirs': ['/opt/conda/include']} {'libraries': ['mkl_rt', 'pthread'], 'library_dirs': ['/opt/conda/lib'], 'define_macros': [('SCIPY_MKL_H', None), ('HAVE_CBLAS', None)], 'include_dirs': ['/opt/conda/include']} Adding scs 2.1.2 to easy-install.pth file

Installed /opt/conda/lib/python3.7/site-packages/scs-2.1.2-py3.7-linux-x86_64.egg Processing dependencies for scs==2.1.2 Searching for scipy==1.6.3 Best match: scipy 1.6.3 Adding scipy 1.6.3 to easy-install.pth file

Using /opt/conda/lib/python3.7/site-packages Searching for numpy==1.21.2 Best match: numpy 1.21.2 Adding numpy 1.21.2 to easy-install.pth file Installing f2py script to /opt/conda/bin Installing f2py3 script to /opt/conda/bin Installing f2py3.7 script to /opt/conda/bin

Using /opt/conda/lib/python3.7/site-packages Finished processing dependencies for scs==2.1.2

this is installing message, appear seems no any error.

[jiaweihe1996]

Sorry. I cannot solve it. :( Maybe you could submit an issue in scs-python repo and ask them for help.

[UsersNGT]

ok ,but thank you i have a question also, when i try to your Repetition training experiment, use det-npy and gt-npy. yesterday found you open source training code，in code you only used gt-npy , this why?

[jiaweihe1996]

Because if using detection bounding boxes, the track id is hard to assign. You need to assign gt ids to the detection boxes by iou or other metrics. Now using gt boxes to train is a common practice in MOT. Some papers use augmentation, like random translation and scaling the boxes, on gt boxes to mimic detection distribution. However, in our paper, we don't use any augmentations.