Merge pull request #536 from ooooo-create/accuracy_copysign

【Hackathon 9th No.9】Skip int8 for copysign

Author: cangtianhuang

tester/api_config/torch_error_skip.txt

@@ -3098,3 +3098,4 @@ paddle.std(x=Tensor([3, 3, 477218589],"float16"), axis=0, unbiased=False, )
 paddle.std(x=Tensor([3, 477218589, 3],"float16"), axis=0, )
 paddle.std(x=Tensor([3, 477218589, 3],"float16"), axis=0, unbiased=False, )
 paddle.index_put(Tensor([110, 42, 56, 56],"float64"), tuple(Tensor([16, 16],"int64"),Tensor([16, 16],"int64"),Tensor([1, 16],"int64"),), Tensor([56],"float64"), True, )
+paddle.copysign(Tensor([12, 20, 2],"int8"), Tensor([12, 20, 2],"int8"), )