FilesLast commitLast update
2 months ago
11 days ago
11 days ago
2 months ago
2 months ago
22 days ago
1 month ago
18 days ago
README

NpuClearFloatStatus

English | 简体中文

Applicable Products

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

Function

  • Operator function: Clears the floating-point overflow status register of each AI core of the NPU, and the output is fixed to 8 float32 zero values.

  • Formula:

data=zeros(8,dtype=float32)data = zeros(8, dtype=float32)

Parameters

Parameter name Input/Output/Attribute Description Data type Data format
addr Input Address placeholder, with shape (8,); the data content does not participate in the computation. FLOAT ND
data Output Fixed output of 8 float32 zero values, with shape (8,). FLOAT ND

Constraints

  • The data type of addr must be float32.
  • The output data is fixed to 8 float32 zero values, regardless of the input data content.
  • addr serves only as a placeholder for the operator input interface, and its data content does not participate in the computation.

Invocation Description

Invocation method Invocation sample Description
Graph-mode invocation test_geir_npu_clear_float_status See operator invocation to complete operator compilation and verification.