ArgMin

English | 简体中文

Product Support Status

Product	Supported
Ascend 950PR&950DT products	√
Atlas A3 products	√
Atlas A2 products	√
Atlas 200I/500 A2 inference products	×
Atlas inference products	√
Atlas training products	√

Function Description

Returns the indices of the minimum values of a tensor along the specified axis and saves them to out. If keepdim is false, the corresponding axis is not retained; if keepdim is true, the dimension value of the specified axis is retained as 1.

Function Prototype

Each operator is divided into a two-phase API. You must first call the "aclnnArgMinGetWorkspaceSize" API to obtain the workspace size required for the computation and the executor that contains the operator computation process, and then call the "aclnnArgMin" API to perform the computation.

aclnnStatus aclnnArgMinGetWorkspaceSize(
  const aclTensor* self,
  int64_t          dim,
  bool             keepdim,
  aclTensor*       out,
  uint64_t*        workspaceSize,
  aclOpExecutor**  executor)

aclnnStatus aclnnArgMin(
  void          *workspace,
  uint64_t       workspaceSize,
  aclOpExecutor *executor,
  aclrtStream    stream)

aclnnArgMinGetWorkspaceSize

• Parameter Description:

  Parameter Name	Input/Output	Description	Usage Instructions	Data Type	Data Format	Dimension (shape)	Non-contiguous Tensor
  self	Input	Input tensor.	-	FLOAT, FLOAT16, BFLOAT16, DOUBLE, INT8, INT16, INT32, INT64, UINT8	ND	-	√
  dim	Input	The dimension on which the minimum value is computed.	The value range is [-self.dim(), self.dim()).	-	-	-	-
  keepdim	Input	Specifies whether the dimensions of the output tensor shape are the same as those of the input tensor.	-	BOOL	-	-	-
  out	Output	Output tensor.	-	INT32, INT64	ND	-	√
  workspaceSize	Output	Returns the size of the workspace that needs to be allocated on the Device side.	-	-	-	-	-
  executor	Output	Returns the op executor, which contains the operator computation process.	-	-	-	-	-

  • Atlas training products: The data type BFLOAT16 is not supported.
  • Atlas inference products: The data type BFLOAT16 is not supported. When the input data type of self is INT8, INT16, INT32, INT64, or UINT8, the dimensions of the input tensor shape support a maximum of 7 dimensions.

• Return Values:

  aclnnStatus: returns a status code. For details, see aclnn Return Codes.

  The first-phase API completes input parameter validation and reports an error in the following scenarios:

  Return Code	Error Code	Description
  ACLNN_ERR_PARAM_NULLPTR	161001	The passed-in self or out is a null pointer.
  ACLNN_ERR_PARAM_INVALID	161002	The data types and data formats of self and out are not within the supported range.
  		The deduced data type cannot be converted to the type of the specified output out.
  		dim exceeds the dimensions of the tensor array.
  		The shape of the passed-in out does not satisfy InferShape.

aclnnArgMin

• Parameter Description:

  Parameter Name	Input/Output	Description
  workspace	Input	The memory address of the workspace allocated on the Device side.
  workspaceSize	Input	The size of the workspace allocated on the Device side, obtained by the first-phase API aclnnArgMinGetWorkspaceSize.
  executor	Input	The op executor, which contains the operator computation process.
  stream	Input	Specifies the Stream that executes tasks.

• Return Values:

  aclnnStatus: returns a status code. For details, see aclnn Return Codes.

Constraint Description

• Deterministic computation:

  • aclnnArgMin uses a deterministic implementation by default.

• For the input self, the scenario where the reduce axis is 0 is not supported

Invocation Description

Invocation Method	Invocation Sample	Description
aclnn invocation	test_aclnn_arg_min	Invokes the ArgMin operator through the aclnnArgMin API.