com.microsoft - QGemm#

QGemm - 1#

Version

  • name: QGemm (GitHub)

  • domain: com.microsoft

  • since_version: 1

  • function:

  • support_level: SupportType.COMMON

  • shape inference: True

This version of the operator has been available since version 1 of domain com.microsoft.

Summary

Attributes

  • alpha - FLOAT : Scalar multiplier for the product of input tensors A * B.

  • transA - INT : Whether A should be transposed

  • transB - INT : Whether B should be transposed

Inputs

Between 6 and 9 inputs.

  • A (heterogeneous) - TA:

  • a_scale (heterogeneous) - T:

  • a_zero_point (heterogeneous) - TA:

  • B (heterogeneous) - TB:

  • b_scale (heterogeneous) - T:

  • b_zero_point (heterogeneous) - TB:

  • C (optional, heterogeneous) - TC:

  • y_scale (optional, heterogeneous) - T:

  • y_zero_point (optional, heterogeneous) - TYZ:

Outputs

  • Y (heterogeneous) - TY:

Type Constraints

  • T in ( tensor(float) ): Constrain scale types to float tensors.

  • TA in ( tensor(int8), tensor(uint8) ): Constrain input A and its zero point types to 8 bit tensors.

  • TB in ( tensor(int8), tensor(uint8) ): Constrain input B and its zero point types to 8 bit tensors.

  • TC in ( tensor(int32) ): Constrain input C to 32 bit integer tensors.

  • TYZ in ( tensor(int8), tensor(uint8) ): Constrain output zero point types to 8 bit tensors.

  • TY in ( tensor(float), tensor(int8), tensor(uint8) ): Constrain output type to float32 or 8 bit tensors.

Examples