Toybrick

标题: 做混合量化hybrid_quantization [打印本页]

作者: fjndfazp    时间: 2020-5-30 17:37
标题: 做混合量化hybrid_quantization
在做common_function_demos\hybrid_quantization里面的混合量化step1之后,会生成一个cfg配置文件,根据下面的解释,我要怎么设置每一层的量化方式?
%YAML 1.2
---
# add layer name and corresponding quantized_dtype to customized_quantize_layers, e.g conv2_3: float32
customized_quantize_layers: {}

比如:截图中的,如果attach_concat_1/out0_0不量化,是不是直接写下面这样?如果有多个层都需要不做量化呢?有没有举例说明一下
customized_quantize_layers: {attach_concat_1/out0_0: float32}



作者: jefferyzhang    时间: 2020-5-30 17:46
参看rknn-toolkit里的sample,有完整混合量化的sample的
作者: brunolin    时间: 2021-2-19 13:40
customized_quantize_layers: {
    attach_concat_1/out0_0: float32,
    layer_000/xxx_000: float32,
    layer_111/xxx_111: float32
}




欢迎光临 Toybrick (https://t.rock-chips.com/) Powered by Discuz! X3.3