Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/yuyijie1995/dropblock_mxnet_bottom_implemention
用C++实现一个mxnet版本dropblock Op 最后可以用mx.sym.Dropblock()调用
https://github.com/yuyijie1995/dropblock_mxnet_bottom_implemention
Last synced: 2 months ago
JSON representation
用C++实现一个mxnet版本dropblock Op 最后可以用mx.sym.Dropblock()调用
- Host: GitHub
- URL: https://github.com/yuyijie1995/dropblock_mxnet_bottom_implemention
- Owner: yuyijie1995
- Created: 2019-02-22T08:16:32.000Z (almost 6 years ago)
- Default Branch: master
- Last Pushed: 2020-02-17T14:44:55.000Z (almost 5 years ago)
- Last Synced: 2024-08-01T22:40:00.957Z (5 months ago)
- Language: C++
- Size: 412 KB
- Stars: 5
- Watchers: 2
- Forks: 2
- Open Issues: 2
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
- Awesome-MXNet - DropBlock(c++ implementaion)
README
# dropblock_mxnet_bottom_implemention
用C++实现一个mxnet版本dropblock Op 最后可以用mx.sym.Dropblock()调用
dropblock_c++自定义OP实现
GPU版本要调用的话,OP名改成了mx.sym.GPUDropblock()
===================================================================================================================
## 参考博客
[MXNet中新增Operator](http://shuokay.com/2017/10/04/mxnet-add-op-in-backend/)
## 论文链接
[DropBlock: A regularization method for
convolutional networks](https://arxiv.org/pdf/1810.12890.pdf)
## 参考python端dropblock实现代码
[pytorch实现](https://github.com/miguelvr/dropblock)[mxnet实现](https://github.com/chenzx921020/DropBlock-mxnet)
-----------------------------------------------------------------------------------------------------------------
### 已实现功能
* 3,5,7的blocksize的dropblock
* gpu,cpu均能进行训练
* mnist训练集上测试通过
* 新增了维度检查功能 用户输入二维特征图是程序不会奔溃
* 加入了p_keep schedule 功能
* 在cifar100上完成了测试 大约有1~1.5个百分点的提升### dropblock_gpu和dropout_cpu mnist训练效果对比
![](https://github.com/yuyijie1995/dropblock_mxnet_bottom_implemention/blob/master/gpu_blocksize3.png)
![](https://github.com/yuyijie1995/dropblock_mxnet_bottom_implemention/blob/master/cpu_dropout.jpg)### TODO
* 尝试增加高宽不相等的特征图的dropblock功能
* 学习谷歌代码规范对代码进行优化