onnxruntime
6dbd41b5 - Quant tool: Use 2D scales and zp in default MatMulNBitsQuantizer (#26329)

Commit
64 days ago
Quant tool: Use 2D scales and zp in default MatMulNBitsQuantizer (#26329) ### Description #24828 updated the specs for the operator to use 2D scales and zero points but the default quantizer still produces 1D values. ### Motivation and Context Produce models that are consistent with the specs
Author
Parents
Loading