Comments (12)
关于与ncnn的区别联系参考 #6 , 量化会做 但是不是当前最重要的事情,单精度仍有不少潜力可以继续做下去。
from feathercnn.
了解,在做量化之前,先做出一个基于float32的稳定的版本是更重要的,如果后面做量化的话,是否可以交流一下呢?我目前也在看量化方面。
from feathercnn.
@JSnobody armv8.2以下指令集其实对int8计算不是很友好,理论上fp32和int8的乘加op吞吐率是一样的,所以一个稳定高效的fp32是更加重要的,当然后面avx512或者dot int8的指令是另外一回事了。 @turbo0628 加油鸭
from feathercnn.
@BUG1989 如果我没记错的话,caffe-int8-convert-tools 是你做的吧,这几天刚认识caffe-int8-convert-tools,准备仔细研究一下
from feathercnn.
@BUG1989 我看主流的框架和平台的int8量化,只针对训练后量化,主要分为tensorflow和tensorRT两个流派,有的是用tensorflow'的**和流程,有的采用tensorRT**结合KL散度校准,从结果上看,tensorRT的准确率会有提升。关于tensorRT和tensorflow量化你怎么看,推理框架也需要反量化嘛?
from feathercnn.
修改一下:tensorRT较比tensorflow准确率会有提升。
from feathercnn.
现在还有第三个流派是,trainning int8,鼻祖可以参考Ristretto caffe,分为两个步骤:
- 使用数据集统计出大概的量化scale;
- 将量化scale引入finetune流程,使权重反过来适配第一步中的量化scale,恢复精度。
框架有:TensorFlow、Intel Caffe、Xilinx dnn
from feathercnn.
@BUG1989 Ristretto caffe我前几天看到过,当时看的不仔细,通过你的介绍我学习了。
听你的描述,我还是有很多疑惑,不过,我先自己看看Ristretto caffe再讨论吧。
能否加个联系方式呢?
from feathercnn.
@BUG1989 这三个流派,你觉得哪个比较好,各自优缺点是什么?
from feathercnn.
我想请教下,如果不做量化,那大概从哪些方面做加速, 除了合并层这个方式以外,其他都是通过指令集来实现加速么?
from feathercnn.
@huangrichen11 有好多种方式
from feathercnn.
@huangrichen11 有好多种方式
有没有相关的资料,想了解下?对这方面没有接触过
from feathercnn.
Related Issues (20)
- Do you plan to support convert pytorch model to feathercnn model? HOT 2
- InitFromPath never returns HOT 1
- 请问是否支持ssd? HOT 1
- Running benchmark with SqueezeNet model, segmentation fault occured HOT 2
- Does this also support RPi3B with Raspbian OS (armv7l) HOT 3
- 请问转换caffe模型的文件是feather_convert_caffe还是caffe_model_convert HOT 5
- typo and why so large input buffer HOT 1
- supported layer mismatch between layer_factory.cpp and feather_convert_caffe.cc
- Model convert error - libprotobuf HOT 1
- Performance testing using experimental branch on Andriod
- build_linux.sh
- 未使用的局部变量
- Add support for tf or split+transpose in caffe.
- loadparam
- 这个项目还在维护吗?
- Layer type Deconvolution not registered
- Documentation or examples for ARM usage. HOT 3
- OpenCL or Vulkan port?
- Evaluation
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from feathercnn.