https://www.zhihu.com/question/362455124?sort=created
微软的模型压缩工具：distiller（重点）
https://github.com/NervanaSystems/distiller (支持PyTorch，官方是在PyTorch1.3上测试的，在GitHub上搜PyTorch Pruning最多星)
微软的AutoML工具：nni中也有模型压缩的模块，https://github.com/microsoft/nni/blob/master/examples/model_compress/QAT_torch_quantizer.py
https://nni.readthedocs.io/zh/latest/Compressor/Quantizer.html
https://github.com/microsoft/nni/blob/master/examples/model_compress/DoReFaQuantizer_torch_mnist.py
Pytorch自带的量化工具（PyTorch>1.3）
https://zhuanlan.zhihu.com/p/81026071
https://github.com/pytorch/glow/blob/master/docs/Quantization.md
https://github.com/pytorch/QNNPACK
nni中的量化从例子看起来还蛮好用的（Pytorch的官方量化文档看晕了，不适合刚入手量化的小白）
https://github.com/microsoft/nni/blob/master/examples/model_compress/DoReFaQuantizer_torch_mnist.py
nni中有4种量化方式：
1. Naive Quantizer，2. QAT Quantizer，3. DoReFa Quantizer，4.BNN Quantizer
其中，1貌似是最low的，推理的时候32位变成8位，不想用，4，二进制神经网络是啥？按位运算挺好的，不知道部署时候是否有坑
2，Quantization and Training of Neural Networks for Efficient Integer-Arithmetic-Only Inference（谷歌2018）nni中说不支持批量归一化折叠，不知道有没有影响
3，DoReFa-Net: Training Low Bitwidth Convolutional Neural Networks with Low Bitwidth Gradients（Face++ 2018）例子看起来简单https://arxiv.org/abs/1606.06160，https://blog.csdn.net/langzi453/article/details/88172080
nni中的量化只是模拟不是加速？https://github.com/microsoft/nni/issues/2332
https://github.com/microsoft/nni/blob/master/examples/model_compress/auto_pruners_torch.py
paper with code 的量化github排名：
https://paperswithcode.com/search?q_meta=&q=Quantizer
第一名：Quantization and Training of Neural Networks for Efficient Integer-Arithmetic-Only Inference（谷歌2018）
https://paperswithcode.com/paper/quantization-and-training-of-neural-networks Tensorflow劝退
第二名：Training with Quantization Noise for Extreme Model Compression（2020）
https://paperswithcode.com/paper/training-with-quantization-noise-for-extreme PyTorch实现
模型压缩的benchmark: https://paperswithcode.com/task/model-compression
模型压缩的benchmark:https://paperswithcode.com/task/quantization
DoReFa还有这个人的：https://github.com/666DZY666/model-compression\
QAT还有这个人的：https://github.com/Xilinx/brevitas
