examples icon indicating copy to clipboard operation
examples copied to clipboard

how to dump a int4 model

Open baodingge opened this issue 2 years ago • 0 comments

hi我使用代码中提示的dump.py脚本,不加载模型文件,生成的mgb模型速度远低于resnet50.int4.mge的模型 另外参考int8的模型,https://github.com/MegEngine/Models/tree/master/official/quantization#quantization-aware-training-qat dump成量化模型需要quantize。麻烦请教一下是否是我操作有误best yours

baodingge avatar Nov 07 '23 07:11 baodingge