clip

Function Usage

Clipping API of the tensor class. It is used to limit elements in a tensor between the specified minimum value and maximum value (elements smaller than the minimum value will be replaced with the minimum value, and elements greater than the maximum value will be replaced with the maximum value).

It is supported by the Atlas inference product and Atlas 200I/500 A2 inference product.

  • The input and output tensors must be on the device.
  • The shape of the input tensor cannot exceed four dimensions.

Prototype

clip(inputTensor: Tensor, minVal: float, maxVal: float)

Parameters

Parameter

Input/Output

Description

inputTensor

Input

Tensor class, supporting dtype.float16, dtype.float32, and dtype.uint8 data types.

minVal

Input

Float type, minimum value after clipping in a tensor. Elements smaller than the minimum value will be replaced with the minimum value. The value of minVal must be less than or equal to that of maxVal.

maxVal

Input

Float type, maximum value after clipping in a tensor. Elements greater than the maximum value will be replaced with the maximum value. The value of maxVal must be greater than or equal to that of minVal.

Response Parameters

Clipped tensor data