pypto.rsqrt

Applicable Products

Ascend 950PR/Ascend 950DT: Supported
Atlas A3 training products/Atlas A3 inference products: Supported
Atlas A2 training products/Atlas A2 inference products: Supported

Description

Computes the reciprocal square root of each element in the input tensor, performing element-wise operations. Returns NaN when the input is negative and Inf when the input is zero.

Prototype

pypto.rsqrt(input, precision_type=pypto.PrecisionType.INTRINSIC) -> Tensor

Parameters

Parameter	Type	Description
input	Tensor	Source operand. Supported type: Tensor. Supported data types of Tensor: DT_FP32, DT_FP16, and DT_BF16. Empty tensors are not supported. The shape supports only 1 to 4 dimensions. The shape size must not exceed 2147483647 (that is, INT32_MAX).
precision_type	PrecisionType, optional	Precision mode of the reciprocal square root operation. The default value is `PrecisionType.INTRINSIC`. INTRINSIC: Directly uses chip instructions for computation, which is faster. HIGH_PRECISION: Uses a higher-precision computation method to reduce precision loss.

Return Value

Returns a tensor. Its shape and data type are the same as those of the input tensor, and its elements are the reciprocal square roots of the corresponding elements of the input tensor.

Constraints

Tensor inputs do not support the TileOpFormat.TILEOP_NZ format.

Examples

TileShape Setting Example

Note: Before calling this operation API, set the TileShape through set_vec_tile_shapes.

The dimensions of the TileShape must be the same as those of the output.

Example 1: If the input shape is [m, n], the output is [m, n], and the TileShape is set to [m1, n1], then m1 and n1 are used to tile the m and n axes, respectively.

pypto.set_vec_tile_shapes(4, 16)

API Call Example

x = pypto.tensor([2, 2], pypto.DT_FP32)
y = pypto.rsqrt(x)

The results are as follows:

Input data x: [[1.0  4.0], [16.0  9.0]]
Output data y: [[1.0  0.5], [0.25  0.33333]]

High-Precision Mode Example

x = pypto.tensor([2, 2], pypto.DT_FP16)
y = pypto.rsqrt(x, pypto.PrecisionType.HIGH_PRECISION)

Instruction Mode Example

x = pypto.tensor([2, 2], pypto.DT_FP16)
y = pypto.rsqrt(x, pypto.PrecisionType.INTRINSIC)