kleiti

PostTrainingQuantConfig(quant_level='auto', device='npu', backend="onnxrt_dml_ep") produces fp32 ops.

1

The below PostTrainingQuantConfig produces fp32 ops for NPU using 2.4.1. Models with int8 and fp16 ops would be preferred for NPU. conf=PostTrainingQuantConfig(quant_level='auto', device='npu', backend="onnxrt_dml_ep", quant_format="QOperator", approach="static", excluded_precisions=['bf16'])