参考文章:
2020年低层视觉任务论文汇总
https://github.com/Kobaayyy/Awesome-ECCV2020-Low-Level-Vision
PyTorch 深度学习模型压缩开源库（含量化、剪枝、轻量化结构、BN融合）, https://github.com/666DZY666/model-compression, 最近又改名为micronet
https://blog.csdn.net/moxibingdao/article/details/106667518
666DZY666/model-compression，讲解注释比较详细，没有标明算法出处
https://github.com/666DZY666/micronet
基础教学
深度学习量化技术科普, 基础教学视频
https://www.bilibili.com/video/BV1cZ4y1u7T5/
Pytorch量化感知训练详解, QAT上手
https://my.oschina.net/u/4580321/blog/4750607
Pytorch实现卷积神经网络训练量化（QAT）, 讲解tflite、666DZY666/model-compression
https://zhuanlan.zhihu.com/p/164901397
https://blog.csdn.net/just_sort/article/details/107600007
模型量化原理及tflite示例，tflite论文的简单写法
https://www.bbsmax.com/A/nAJv1EDozr/
PyTorch如何量化模型（int8）并使用GPU（训练/Inference）？ - 知乎，分为训练后动态量化、训练后静态量化、量化意识训练
https://www.zhihu.com/question/431572414/answer/1601740370
PyTorch量化官方教程-v1.7.1
https://pytorch.org/docs/stable/quantization.html
PyTorch量化官方教程（中文）-v1.4.0，最高版本
https://pytorch.apachecn.org/docs/1.4/88.html
量化原理教学
神经网络量化----吐血总结，对称量化，非对称量化，随机量化，QAT
https://blog.csdn.net/weixin_41910772/article/details/109637956
Int8量化-介绍（一），含NVIDIA量化方法PPT链接，含python、ncnn量化解读
https://zhuanlan.zhihu.com/p/58182172
信息熵，交叉熵和相对熵解读，含KL散度解读，量化数值分布分析
https://www.cnblogs.com/liaohuiqiang/p/7673681.html
ncnn源码学习（六）：模型量化原理笔记，汇总性文章
https://blog.csdn.net/sinat_31425585/article/details/101607785
int8量化和tvm实现，mxnet+tensorRT实现量化和部署
https://sundrops.blog.csdn.net/article/details/90295078
pytorch量化实现
PyTorch模型量化工具学习
https://zhuanlan.zhihu.com/p/144025236
Batch Normalization Auto-fusion for PyTorch，BN-conv层量化融合
https://github.com/Ironteen/Batch-Normalization-fusion
QNNPACK (Quantized Neural Networks PACKage) is a mobile-optimized library，基于C++实现，类似ncnn用于嵌入式部署，有2年未更新了
https://github.com/pytorch/QNNPACK
MergeBN && Quantization PyTorch 官方解决方案，只做了简单介绍
https://zhuanlan.zhihu.com/p/143664360
PyTorch的量化，讲解了训练后量化、QAT，和简单使用，开发了一个pytorch规范库deepvac，由civilnet机构开发
https://zhuanlan.zhihu.com/p/299108528
https://github.com/DeepVAC/deepvac
详解Pytorch中的网络构造，civilnet机构的另一篇文章
https://zhuanlan.zhihu.com/p/53927068
Xilinx/brevitas量化库QAT，Xilinx出品，基于pytorch实现
https://github.com/Xilinx/brevitas
Xilinx/brevitas量化库QAT，教程地址
file:///home/robert/DeepLearning/ModelCompression/brevitas-master/docs/index.html
ncnn量化实现
https://github.com/Tencent/ncnn
NCNN Conv量化详解（一） - 知乎
https://zhuanlan.zhihu.com/p/71881443
NCNN量化详解（二） - 知乎
https://zhuanlan.zhihu.com/p/72375164
quantized int8 inference，pytorch->onnx->caffe->ncnn或pytorch->caffe->ncnn量化方法，训练后量化
https://github.com/Tencent/ncnn/wiki/quantized-int8-inference
caffe-int8-convert-tools工具，老方法
https://github.com/BUG1989/caffe-int8-convert-tools
caffe-int8-convert-tools工具量化精度说明，[WIP] new int8 implement, better accuracy
https://github.com/BUG1989/caffe-int8-convert-tools
ncnn2table、ncnn2int8，新的量化工具
https://github.com/Tencent/ncnn/tree/master/tools/quantize
tensorRT量化实现
tensorTR提供的量化工具QAT，基于pytorch开发，类似Xilinx/brevitas量化库QAT
https://github.com/NVIDIA/TensorRT/tree/master/tools/pytorch-quantization
TensorRT部署源码
https://github.com/NVIDIA/TensorRT
tflite量化实现
TensorFlow量化感知训练(QAT)工具学习
https://zhuanlan.zhihu.com/p/144870688
TensorFlow Lite models, samples, tutorials, tools and learning resources. 量化教程和资源
https://github.com/margaretmz/awesome-tensorflow-lite
TFLite代码实现
https://github.com/tensorflow/tensorflow/tree/master/tensorflow/lite
TFLite代码类型和结构体定义
https://github.com/tensorflow/tensorflow/blob/master/tensorflow/lite/toco/model.h
MobileNetV3Small defined and pre-trained in PyTorch to a TFLite，将Pytorch模型转成TFLite量化模型，19年7月提交
https://github.com/lain-m21/pytorch-to-tflite-example
Xilinx量化实现
深鉴科技DNNDK概览，已被赛灵思收购
https://blog.csdn.net/weixin_36474809/article/details/82585091
Xilinx技术ML实现网址参考
https://www.xilinx.com/applications/megatrends/machine-learning.html
微软AutoML工具
微软新工具 NNI 使用指南之体验篇
https://www.jianshu.com/p/c76567718d03
量化论文与解读
Quantization and Training of Neural Networks for Efficient Integer-Arithmetic-Only Inference，TFLite量化方法，Google CVPR2018 int8量化算法
https://blog.csdn.net/just_sort/article/details/103704975
https://blog.csdn.net/mizhenpeng/article/details/81488244
TFLite量化方法，pytorch实现方案
https://github.com/skmhrk1209/QuanTorch
EasyQuant: Post-training Quantization via Scale Optimization，arXiv/2006.16669，含论文地址和源码地址
https://zhuanlan.zhihu.com/p/157214981
LSQ-Net: Learned Step Size Quantization，IBM量化方案，论文阅读，arXiv:1902.08153v3
https://github.com/zhutmost/lsq-net
https://blog.csdn.net/qq_37151108/article/details/108666779
基于可训练Step-size的低比特量化——LSQ: Learned Step-size Quantization
https://blog.csdn.net/nature553863/article/details/104275477
ResNet-s for CIFAR10/100 in pytorch，针对Cifar10的ResNet模型
https://github.com/akamaster/pytorch_resnet_cifar10
LQ-net implementation on pytorch，pytorch和tensorflow版本，微软，arXiv:1807.10029
https://github.com/Microsoft/LQ-Nets
https://github.com/pyjhzwh/LQ-net-pytorch
Awesome-Deep-Neural-Network-Compression，量化方案总结
https://github.com/csyhhu/Awesome-Deep-Neural-Network-Compression
mixed-precision-pytorch，Training with FP16 weights in PyTorch，使用FP16进行压缩
https://github.com/suvojit-0x55aa/mixed-precision-pytorch
Deep Compression: Compressing Deep Neural Networks with Pruning, Trained Quantization and Huffman Coding，arXiv:1510.00149v5，pytorch实现
https://github.com/mightydeveloper/Deep-Compression-PyTorch
PACT：PArameterized Clipping Activation for Quantized Neural Networks，arXiv:1805.06085v2，论文阅读
https://blog.csdn.net/qq_19784349/article/details/82979899
DoReFa-Net，量化方案实现，arXiv:1606.06160v3
https://github.com/zzzxxxttt/pytorch_DoReFaNet
https://github.com/Jzz24/pytorch_quantization
低比特量化之DoreFa-Net理论与实践
https://blog.csdn.net/just_sort/article/details/107476947
超分量化方案
PAMS: Quantized Super-Resolution via Parameterized Max Scale，arXiv/2011.04212
https://github.com/colorjam/PAMS
pytorch量化实现细节与技巧
Pytorch Tensor基本数学运算
https://blog.csdn.net/weicao1990/article/details/93738722
在python下实现C++的nth_element函数
https://numpy.org/doc/stable/reference/generated/numpy.ndarray.partition.html
torch.nn.functional.conv2d 函数详解，依次输入input/weight/bias等信息
https://blog.csdn.net/Li7819559/article/details/103813209
pytorch .detach() .detach_() 和 .data用于切断反向传播，用训练过程中的量化clamp和量化恢复
https://www.cnblogs.com/wanghui-garcia/p/10677071.html
Numpy/Pytorch之数据类型与强制类型转换
https://blog.csdn.net/qq_37385726/article/details/81774150
pytorch入坑一 | Tensor及其基本操作
https://zhuanlan.zhihu.com/p/36233589
Pytorch中tensor的打印精度
https://blog.csdn.net/wangpeng246300/article/details/111355945
pytorch：如何修改加载了预训练权重的模型的输入或输出--(修改torch.nn.DataParallel权重为cpu加载)
https://blog.csdn.net/qq_39852676/article/details/106928329
python使用matplotlib绘制柱状图教程，用于模型参数分析
https://www.jb51.net/article/104924.htm
https://blog.csdn.net/kaikai_sk/article/details/85332022
torch.distributions 详解，可参数化的概率分布和采样函数，这允许构造用于优化的随机计算图和随机梯度估计器。
https://blog.csdn.net/weixin_42018112/article/details/90899559
NVIDIA DALI训练数据导入
安装与使用说明
https://github.com/NVIDIA/DALI/releases
https://docs.nvidia.com/deeplearning/dali/user-guide/docs/installation.html
https://docs.nvidia.com/deeplearning/dali/user-guide/docs/index.html
FP32和FP16计算方法
关于CPU的浮点运算能力计算
https://www.jianshu.com/p/b9d7126b08cc?from
float 精度怎么算
https://jingyan.baidu.com/article/84b4f565ae145d60f6da323e.html
IEEE 754 单精度浮点数转换
http://www.styb.cn/cms/ieee_754.php
单精度浮点数 二进制的转换 C++实现
https://blog.csdn.net/zhanggusheng/article/details/52985261
FLOAT16 16BIT浮点数的解释
https://blog.csdn.net/ifenghua135792468/article/details/110450243
【QT】float double的范围与精度及Qt中的qfloat16
https://aidear.blog.csdn.net/article/details/77584948
Qt中float数组（int、double）与QByteArray二进制之间的无损转换，其实结构体等数据都可以转成二进制的
https://blog.csdn.net/weixin_39956356/article/details/97147446
快速浮点开方运算