Function smooth_quantize_ootb

tensorrt_llm/quantization/quantize.py:170–184

(
    model,
    quant_config: QuantConfig,
)

def smooth_quantize_ootb(
    model,
    quant_config: QuantConfig,
):
    quant_map = {
        ColumnLinear: Int8SmoothQuantLinear,
        RowLinear: Int8SmoothQuantRowLinear,
    }

    model = quantize_layers(
        model,
        quant_config,
        quant_map,
    )
    return model


def smooth_quantize_plugin(model, quant_mode):