Merge pull request #505 from Difers/fix_argsort

[Big tensor]add some argsort case

Author: wanghuancoder

report/big_tensor_gpu/error_config.txt

@@ -1713,7 +1713,10 @@ paddle.argmin(x=Tensor([357913942, 3, 2, 1],"float64"), axis=-1, dtype="int64",
 paddle.argmin(x=Tensor([357913942, 3, 4],"float16"), axis=1, keepdim=False, )
 paddle.argmin(x=Tensor([715827883, 3, 2, 1],"float16"), axis=-1, dtype="int64", )
 paddle.argmin(x=Tensor([760567127, 3],"int64"), dtype="int32", )
-paddle.argsort(Tensor([2, 1140850690],"float32"), descending=True, )
+paddle.argsort(Tensor([2, 1140850690],"float32"), descending=True, stable=True)
+paddle.argsort(Tensor([1, 1073741823],"float16"), descending=True, stable=True)
+paddle.argsort(Tensor([1, 1073741825],"float16"), descending=True, stable=True)
+paddle.argsort(Tensor([2, 2147483649],"float16"), descending=True, stable=True)
 paddle.assign(list[Tensor([100, 42949673],"float16"),], )
 paddle.assign(list[Tensor([429496730, 10],"float16"),], )
 paddle.atan2(Tensor([111, 116197, 333],"float16"), Tensor([116197, 333],"float16"), )

tester/api_config/8_big_tensor/big_tensor_1_8.txt

@@ -17545,7 +17545,10 @@ paddle.argmin(x=Tensor([760567127, 3],"int64"), axis=-1, )
 paddle.argmin(x=Tensor([760567127, 3],"int64"), axis=-2, )
 paddle.argmin(x=Tensor([760567127, 3],"int64"), axis=Tensor([1],"int64"), )
 paddle.argmin(x=Tensor([760567127, 3],"int64"), dtype="int32", )
-paddle.argsort(Tensor([2, 1140850690],"float32"), descending=True, )
+paddle.argsort(Tensor([2, 1140850690],"float32"), descending=True, stable=True)
+paddle.argsort(Tensor([1, 1073741823],"float16"), descending=True, stable=True)
+paddle.argsort(Tensor([1, 1073741825],"float16"), descending=True, stable=True)
+paddle.argsort(Tensor([2, 2147483649],"float16"), descending=True, stable=True)
 paddle.as_complex(Tensor([10, 10, 11408507, 2],"float32"), )
 paddle.as_complex(Tensor([10, 11408507, 10, 2],"float32"), )
 paddle.as_complex(Tensor([11408507, 10, 10, 2],"float32"), )

tester/api_config/torch_error_skip.txt

@@ -3056,4 +3056,5 @@ paddle.nn.functional.normalize(x=Tensor([2147483649, 2],"float16"), p=1.2, )
 paddle.nn.functional.normalize(x=Tensor([4, 25565282, 6, 7],"float16"), p=4, axis=3, )
 paddle.nn.functional.normalize(x=Tensor([4, 5, 30678338, 7],"float16"), p=4, axis=3, )
 paddle.nn.functional.normalize(x=Tensor([4, 5, 6, 35791395],"float16"), p=4, )
-paddle.dist(x=Tensor([10],"float16"), y=Tensor([429496730, 10],"float16"), p=4, )
+paddle.dist(x=Tensor([10],"float16"), y=Tensor([429496730, 10],"float16"), p=4, )
+paddle.argsort(Tensor([1, 1073741823],"float16"), descending=True, stable=True)

tester/base_config.yaml

@@ -32,6 +32,7 @@ paddle_error_dismiss:
     - "(InvalidArgument) The total number of elements for 'label' should be less than"
     - "(InvalidArgument) Illegal memory allocation, total allocated space must be greater than 0"
   paddle.linalg.matrix_rank: "(PreconditionNotMet) The element size of x should be <= INT_MAX(2147483647)"
+  paddle.argsort: "(PreconditionNotMet) The dimension being sorted should be less than 2^31"
 
 # some accuracy error can be considered tolerable
 special_accuracy_atol_rtol: