MXNet
的深度学习引擎有两个主要的高级接口:Gluon API
和Module API
。下面提供了每个教程。
两者之间的区别是命令式编程和符号式编程风格。Gluon
可以很容易地原型、构建和训练深度学习模型,而不牺牲训练速度,通过启用(1)直觉命令的Python
代码开发和(2)通过使用杂交特征自动生成符号执行图来更快地执行。
TL;DR:如果您不熟悉深度学习或MXNet
,您应该从Gluon
教程开始。
Gluon
和Module
教程使用Python
,但您也可以在下面的其他语言API
教程部分找到各种其他MXNet教程,例如R
,Scala
和C++
。
示例脚本和应用程序以及贡献信息如下。
Python API教程
1、Gluon
(1)、Introductuon
(1)、Basics
- Manipulate data the MXNet way with ndarray
- Automatic differentiation with autograd
- Linear regression with gluon
- Serialization - saving, loading and checkpointing
(2)、Neural Networks
- Multilayer perceptrons in gluon
- Multi-class object detection using CNNs in gluon
- Advanced RNNs with gluon
(3)、Advanced
- Plumbing: A look under the hood of gluon
- Designing a custom layer with gluon
- Fast, portable neural networks with Gluon HybridBlocks
- Training on multiple GPUs with gluon
(2)、Applications
- Creating custom operators with numpy
- Handwritten digit recognition (MNIST)
- Hybrid network example
- Neural network building blocks with gluon
- Simple autograd example
2、Module
(1)、Introductuon
(1)、Basics
- Imperative tensor operations on CPU/GPU
- NDArray Indexing
- Symbol API
- Module API
- Iterators - Loading data
(2)、Neural Networks
- Linear regression
- MNIST - handwriting recognition
- Large scale image classification
(3)、Advanced
- NDArray in Compressed Sparse Row Storage Format
- Sparse Gradient Updates
- Train a Linear Regression Model with Sparse Symbols
(2)、Applications
- Connectionist Temporal Classification
- Distributed key-value store
- Fine-tuning a pre-trained ImageNet model with a new dataset
- Generative Adversarial Networks
- Matrix factorization in recommender systems
- Text classification (NLP) on Movie Reviews
2、Module
用ndarray操作MXNet的数据方式 使用autograd自动分化 用胶子线性回归 序列化 - 保存,加载和检查点 其他语言API教程
其他语言教程
C
- MNIST with the MXNet C++ API
Scala
- Setup your MXNet with Scala on InelliJ
- MNIST with the Scala API
- Use Scala to build a Long Short-Term Memory network that generates Barack Obama’s speech patterns
R
- NDArray: Vectorized Tensor Computations on CPUs and GPUs with R
- Symbol API with R
- Custom Iterator
- Callback Function
- Five minute neural network
- MNIST with R
- Classify images via R with a pre-trained model
- Char RNN Example with R
- Custom loss functions in R
原创文章,转载请注明 :Mxnet入门教程以及Gluon/Module教程 - MXNet中文网
原文出处: https://mxnets.com/news/12.html
问题交流群 :168117787