search

666DZY666 / micronet

micronet, a model compression and deploy lib. compression: 1、quantization: quantization-aware-training(QAT), High-Bit(>2b)(DoReFa/Quantization and Training of Neural Networks for Efficient Integer-Arithmetic-Only Inference)、Low-Bit(≤2b)/Ternary and Binary(TWN/BNN/XNOR-Net); post-training-quantization(PTQ), 8-bit(tensorrt); 2、 pruning: normal、regular and group convolutional channel pruning; 3、 group convolution structure; 4、batch-normalization fuse for quantization. deploy: tensorrt, fp32/fp16/int8(ptq-calibration)、op-adapt(upsample)、dynamic_shape
MIT License
2.2k stars 478 forks source link

将Conv2d全部替换为QuantizedConv2d后，是需要重新训练么？ #65

Open KiedaTamashi opened 3 years ago

KiedaTamashi commented 3 years ago

按照Readme中的方法将Conv2d替换后，是必须要重新训练或者finetune吗？ 如果我要加载已经训练好的模型，这会有影响吗 谢谢

xiaoguoer commented 3 years ago

需要finetune的，finetune进行的时候会不断更新量化的参数。

666DZY666 commented 3 years ago

请参照 https://github.com/666DZY666/micronet#quant_test_autopy 可加载预训练浮点模型。