Log

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 logarithm of the elements in the input tensor x element-wise, and saves the result to the output tensor y. It supports input tensors of multiple types and supports the scaling (shift/scale) operation on the input tensor x.

• Computation formula:

  $$y=lo{g}_{base}(shift+scale\ast x)$$

Parameter Description

Parameter Name	Input/Output/Attribute	Description	Data Type	Data Format
x	Input	Input tensor, which is x in the formula.	INT8, INT16, INT32, INT64, UINT8, FLOAT16, FLOAT, DOUBLE BOOL, BFLOAT16, COMPLEX64, COMPLEX128	ND
base	Optional attribute	Base of the logarithm. The default value is -1.0 (indicating the natural logarithm base e).	FLOAT	-
scale	Optional attribute	Scaling factor of the input tensor x. The default value is 1.0.	FLOAT	-
shift	Optional attribute	Offset of the input tensor x. The default value is 0.0.	FLOAT	-
y	Output	Output tensor after the logarithm computation, which is y in the formula.	If the input type is an integer type (INT type), the output type is FLOAT; in other cases, the output type is the same as that of the input tensor x	ND

Constraint Description

• The parameter base must be greater than 0. The default value -1 indicates the base e.
• If the input value is in the specific range (0,0.01] or [0.95,1.05), the output precision may be unstable
• Integer inputs are automatically converted to the FLOAT type for output

Invocation Description

Invocation Method	Sample Code	Description
aclnn invocation	test_aclnn_log test_aclnn_inplace_log	Invokes the Log operator through the aclnnLog or aclnnInplaceLog API.
aclnn invocation	test_aclnn_log2 test_aclnn_inplace_log2	Invokes the Log operator through the aclnnLog2 or aclnnInplaceLog2 API.
aclnn invocation	test_aclnn_log10 test_aclnn_inplace_log10	Invokes the Log operator through the aclnnLog10 or aclnnInplaceLog10 API.
Graph-mode invocation	test_geir_log	Invokes the Log operator by building a graph through the operator IR.