deformable_detr代码的运行、训练过程

deformable_detr代码的运行

其他参见官方github
1.coco数据集可以采用软链接的形式
2.文件权限的问题chmod
3.关于MultiScaleDeformableAttention模块不存在,参考issue10 @JarintotionDin 的回答

I met this problem too, and I solved it by changing 'CUDA_HOME=CUDA_HOME:/usr/local/cuda-11.0' to 'export CUDA_HOME=/usr/local/cuda-11.0'.
if torch.cuda.is_available() and CUDA_HOME is not None:
extension = CUDAExtension
sources += source_cuda
define_macros += [("WITH_CUDA", None)]
extra_compile_args["nvcc"] = [
"-DCUDA_HAS_FP16=1",
"-D__CUDA_NO_HALF_OPERATORS__",
"-D__CUDA_NO_HALF_CONVERSIONS__",
"-D__CUDA_NO_HALF2_OPERATORS__",
]
else:
raise NotImplementedError('Cuda is not availabel')

When I use 'CUDA_HOME=CUDA_HOME:/usr/local/cuda-11.0' , the CUDA_HOME is None!

你可能感兴趣的:(deformable,detr,深度学习)