nvrtc: error: invalid value for --gpu-architecture (-arch)

[error]

nvrtc: error: invalid value for --gpu-architecture (-arch)

 

[solution]

1.

detectron2/detectron2/engine/train_loop.py 

line 491 : 주석처리

     #with autocast(dtype=self.precision):  

 

2. 

detectron2/detectron2/engine/train_loop.py 

line 435~ : False를 True로 변경

class AMPTrainer(SimpleTrainer):

    def __init__(

        ...

        ...

        zero_grad_before_foward = True

    ):