Comments (2)
您好,因为我们继承的是DKD( https://github.com/megvii-research/mdistiller/blob/a08d46f10d6102bd6e3f258ca5ac880b020ea259/mdistiller/engine/cfg.py )的实验设置,而他们设置KD的CE_WEIGHT也是来自于Hinton的设置,即CE_WEIGHT=0.1、KD_WEIGHT=0.9;而其他算法都是其作者设计的CE_WEIGHT=1.0。在实验中我们其实也注意到了,CE_WEIGHT=1.0会比0.1略好一些,但是我们的实验是为了验证我们的logit标准化可以优化KD,所以保留了KD算法的CE_WEIGHT=0.1的设置
Hi, it is because we inherited the experimental setup of DKD (https://github.com/megvii-research/mdistiller/blob/a08d46f10d6102bd6e3f258ca5ac880b020ea259/mdistiller/engine/cfg.py) and the CE_WEIGHT of KD they set is also derived from Hinton's setting, i.e., CE_WEIGHT=0.1, KD_WEIGHT=0.9, while the other algorithms are designed by their authors, i.e., CE_WEIGHT=1.0.
In the experiment, we actually noticed that CE_WEIGHT=1.0 would be slightly better than 0.1. But our experiment was to verify that our logit standardization can improve those logit-based KD. So we kept the KD algorithm setting of CE_WEIGHT=0.1. Hope this help you.
from logit-standardization-kd.
感谢解惑!这对我非常有帮助!
from logit-standardization-kd.
Related Issues (12)
- Distributed training for ImageNet HOT 2
- 关于蒸馏Vit的问题 HOT 2
- Teacher model configurations
- cannot reproduce the results? HOT 12
- 迁移实验 & 学生模型 HOT 5
- 实现层面 HOT 4
- 交叉熵损失的作用是什么 HOT 2
- 怎么迁移到多标签分类任务上? HOT 2
- config HOT 1
- 想问一下logit具体的计算方式 HOT 4
- Missing configuration to reproduce "KD + Ours" result of ResNet-18 student with ResNet-34 teacher HOT 8
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 logit-standardization-kd.