how to dump a int4 model

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

Start with the dump.py script mentioned in the issue and compare its output with resnet50.int4.mge. Read the linked MegEngine quantization-aware training documentation to understand the expected quantization flow. Done means identifying whether the model dump procedure is incorrect and documenting the correct operation or reproducing a confirmed issue.

Written by the indexing model from the issue text.