RuntimeError: grad can be implicitly created only for scalar outputs

[Shn9909]

I encountered this strange error. Here is the output, thank you.
Before, it was showing that the error cannot run on CPU and GPU at the same time, I added . cuda() after loss, it starts showing this error.

Traceback (most recent call last):
File "D:/xiangmu/ENAS-pytorch-master/main.py", line 56, in
main(args)
File "D:/xiangmu/ENAS-pytorch-master/main.py", line 35, in main
trnr.train()
File "D:\xiangmu\ENAS-pytorch-master\trainer.py", line 223, in train
self.train_shared(dag=dag)
File "D:\xiangmu\ENAS-pytorch-master\trainer.py", line 317, in train_shared
loss.backward()
File "C:\Users\sunhaonan.conda\envs\enas\lib\site-packages\torch_tensor.py", line 307, in backward
torch.autograd.backward(self, gradient, retain_graph, create_graph, inputs=inputs)
File "C:\Users\sunhaonan.conda\envs\enas\lib\site-packages\torch\autograd_init_.py", line 150, in backward
grad_tensors_ = make_grads(tensors, grad_tensors)
File "C:\Users\sunhaonan.conda\envs\enas\lib\site-packages\torch\autograd_init_.py", line 51, in _make_grads
raise RuntimeError("grad can be implicitly created only for scalar outputs")
RuntimeError: grad can be implicitly created only for scalar outputs