FilesLast commitLast update
1 month ago
11 days ago
2 months ago
11 days ago
3 months ago
27 days ago
11 days ago
1 month ago
5 days ago
README

ComplexAbs

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

  • Operator function: computes the absolute value (magnitude) of each element in the input complex tensor.

  • Computation formula:

    For the complex input xi=ai+bi⋅jx_i = a_i + b_i \cdot j, the magnitude is output:

yi=∣xi∣=ai2+bi2y_{i}=|x_{i}|=\sqrt{a_{i}^{2}+b_{i}^{2}}

Parameter Description

Parameter Name Input/Output/Attribute Description Data Type Data Format
x Input Input parameter for the complex absolute value computation, which is the input tensor x in the formula. COMPLEX32, COMPLEX64, COMPLEX128 ND
y Output Output parameter for the complex absolute value computation, which is the output tensor y in the formula. The output data type is determined by the input data type: FLOAT is output when the input is COMPLEX64, FLOAT16 is output when the input is COMPLEX32, and DOUBLE is output when the input is COMPLEX128. FLOAT, FLOAT16, DOUBLE ND
Tout Attribute Output data type. This attribute is retained for compatibility with the TensorFlow ComplexAbs operator and is currently not used. The output data type is deduced from the input data type. Type, and the default value is DT_FLOAT -

Constraint Description

  • The input supports only the COMPLEX32, COMPLEX64, and COMPLEX128 types, and other data types are not supported.
  • The output data type is determined by the input data type and cannot be explicitly specified: COMPLEX64 corresponds to the FLOAT output, COMPLEX32 corresponds to the FLOAT16 output, and COMPLEX128 corresponds to the DOUBLE output.
  • The shapes of the input and output are the same. Dynamic shape (DynamicShape) and dynamic dimension (DynamicRank) are supported.
  • Empty tensors (with 0 in the shape) are not supported.
  • The data format supports ND. Non-contiguous tensors are supported, and the number of dimensions is not greater than 8.

Invocation Description

Invocation Method Invocation Sample Description
Graph-mode invocation operator IR Invokes the ComplexAbs operator by building a graph through the operator IR.