Floating-point errors handling

In numpy you can raise floating points error using seterr([all, divide, over, under, invalid]). Is there something similar in Pytorch?