深度学习代码生成基础知识
您可以将 MATLAB® Coder™ 与 Deep Learning Toolbox™ 结合使用，从经过训练的 CNN 生成 C++ 代码。然后您可以将生成的代码部署到使用 Intel® 或 ARM® 处理器的嵌入式平台。您还可以从不依赖任何第三方库的经过训练的 CNN 生成泛型 C 或 C++ 代码。
函数
codegen | 从 MATLAB 代码生成 C/C++ 代码 |
coder.loadDeepLearningNetwork | 加载深度学习网络模型 |
coder.DeepLearningConfig | 创建深度学习代码生成配置对象 |
coder.DeepLearningCodeConfig | Parameters to configure deep learning code generation that does not depend on third-party libraries (自 R2021a 起) |
coder.ai.enableParameterUpdate | Enables run-time update of network parameters (自 R2025a 起) |
coder.ARMNEONConfig | Parameters to configure deep learning code generation with the ARM Compute Library |
coder.CMSISNNConfig | Parameters to configure deep learning code generation with the CMSIS-NN library for Cortex-M targets (自 R2022a 起) |
coder.MklDNNConfig | Parameters to configure deep learning code generation with the Intel Math Kernel Library for Deep Neural Networks |
analyzeNetworkForCodegen | 分析深度学习网络以进行代码生成 (自 R2022b 起) |
coder.regenerateDeepLearningParameters | Regenerate files containing network learnables and states parameters (自 R2021b 起) |
代码配置参数
| 深度学习目标库 | Target library for deep learning code generation |
| 可学习参数压缩 | Compression type |
| ARM Compute Library 版本 | Version of ARM Compute Library |
| ARM Compute 架构 | ARM architecture supported in the target hardware |
| 数据类型(ARM Compute) | Inference computation precision |
| 标定结果文件路径(ARM Compute) | Location of calibration MAT file |
| 数据类型(CMSIS-NN) | Inference computation precision |
| 标定结果文件路径(CMSIS-NN) | Location of calibration MAT file |
主题
- 使用 MATLAB Coder 进行深度学习的前提条件
为用于深度学习网络的代码生成安装相关产品并配置环境。
- 使用 MATLAB Coder 生成深度学习代码的工作流
根据预训练网络生成用于预测的代码。
- 加载预训练网络以用于代码生成
创建一个
dlnetwork对象或用于代码生成的目标检测器。 - Analyze Network for Code Generation
Check code generation compatibility of a deep learning network.
- 代码生成支持的网络和层
选择目标处理器支持的卷积神经网络。
- Code Generation for dlarray
Use deep learning arrays in MATLAB code intended for code generation.
- dlarray Limitations for Code Generation
Adhere to code generation limitations for deep learning arrays.
- 为深度学习网络生成泛型 C/C++ 代码
从不依赖第三方库的深度学习网络中生成用于预测的 C/C++ 代码。
- Update Network Parameters at Run Time
Update deep learning network parameters at run-time without regenerating code.
- Compress Networks Learnables in bfloat16 Format
Perform learnables compression and generate C/C++ code in Brain Floating Point format, bfloat16.
- Optimize C/C++ Code Performance for Deep Learning Applications without Deep Learning Libraries
Code generation configuration settings that optimize the performance of the generated C/C++ code for a deep learning network.
- 使用 MKL-DNN 的深度学习网络的代码生成
从深度学习网络生成用于预测的 C++ 代码，以 Intel CPU 为目标。
- Code Generation for Deep Learning Networks with ARM Compute Library
Generate C++ code for prediction from a deep learning network, targeting an ARM processor.
- Cross-Compile Deep Learning Code That Uses ARM Compute Library
Generate library or executable code on host computer for deployment on ARM hardware target.
- Generate int8 Code for Deep Learning Networks
Quantize and generate code for a pretrained convolutional neural network.
- Update Network Parameters After Code Generation
Perform post code generation updates of deep learning network parameters.
相关信息
- Deep Learning Toolbox 快速入门 (Deep Learning Toolbox)
- Deep Learning with GPU Coder (GPU Coder)
精选示例
在 Intel CPU 上使用变分自编码器生成数字图像
为经过训练以生成手绘形状数字的 VAE dlnetwork 生成代码。
Cross Compile Deep Learning Code for ARM Neon Targets
Generate library or executable code on host computer for deployment on ARM hardware target.
Post-Code-Generation Update of Deep Learning Network Parameters
Incrementally update the network learnables of a deep learning network application running on edge devices such as Raspberry Pi®. This example uses a cart-pole reinforcement learning application to illustrate:
