AddQuantDequantForInferencePass¶

class paddle.fluid.contrib.slim.quantization.quantization_pass. AddQuantDequantForInferencePass ( scope, place, quant_bits=8 ) [source]

When export quant model, it will traverse to find the output of each op, and then insert the quant/dequant op after it.

apply ( graph )

apply¶

Parameters

graph (IrGraph) – the target graph.