skills.rest
matlab
matlabOFFICIAL
Explore all skills by this author →
WebsiteGitHub

matlab-deploy-ai-model

Generate C/C++ or CUDA code from PyTorch and LiteRT AI models using MATLAB Coder.

995|122|Updated Apr 3, 2026
Search Tags:#Deployment#CUDA#matlab#pyTorch#Code generation#Simulink#litert
One-click install
npx skills add https://github.com/matlab/matlab-agentic-toolkit --skill matlab-deploy-ai-model
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: matlab-deploy-ai-model
Source: https://github.com/matlab/matlab-agentic-toolkit/tree/main/skills-catalog/code-generation/matlab-deploy-ai-model
Command: npx skills add https://github.com/matlab/matlab-agentic-toolkit --skill matlab-deploy-ai-model

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

Deploying a trained PyTorch or LiteRT model as standalone C/C++ or CUDA code requires navigating multiple MATLAB APIs, data layouts, and verification steps, and choosing the wrong import path (e.g., dlnetwork converters) breaks the code generation workflow.

Core Features & Use Cases

  • Model Loading and Code Generation: Load .pt2 files with loadPyTorchExportedProgram or .tflite files with loadLiteRTModel, write a codegen entry-point, and generate MEX, static/dynamic libraries, or executables with MATLAB Coder or GPU Coder.
  • Numeric Verification: Compare MATLAB inference output against the original PyTorch or LiteRT model using pyrun, then verify MEX output with matlab.unittest tolerances before generating production code.
  • Simulink Integration: Use the dedicated PyTorch ExportedProgram or LiteRT blocks from dlosslib, or a MATLAB Function block, with slbuild code generation support.
  • Use Case: You have a ResNet model exported as model.pt2 and need a CUDA-accelerated library for deployment. This Skill walks you through loading the model, writing the entry-point with coder.Constant, verifying MEX numerics, and generating the final CUDA library.

Quick Start

Ask your agent to generate a verified MEX and then a C++ static library from your PyTorch .pt2 or LiteRT .tflite model file using MATLAB Coder.

Frequently Asked Questions about matlab-deploy-ai-model

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I generate C++ code from a PyTorch model in MATLAB?▼

Export the model to .pt2 with torch.export, load it with loadPyTorchExportedProgram, write an entry-point function calling invoke, and run codegen with coder.config. Generate and verify a MEX first, then produce a static library, DLL, or executable.

What is the difference between loadPyTorchExportedProgram and importNetworkFromPyTorch?▼

loadPyTorchExportedProgram loads a .pt2 file for direct C/C++/CUDA code generation via invoke, preserving PyTorch's NCHW layout. importNetworkFromPyTorch returns an editable dlnetwork for Deep Learning Toolbox workflows like quantization, pruning, and transfer learning.

Can I generate CUDA code from a LiteRT .tflite model?▼

Yes. Load the .tflite file with loadLiteRTModel, write an entry-point calling invoke, and use coder.gpuConfig with codegen to produce CUDA MEX, libraries, or executables. GPU Coder is required for CUDA targets.

Does loadLiteRTModel support stateful LSTM models?▼

No, stateful operations are not supported in R2026a. Rewrite the model in Python as stateless with explicit hidden and cell state inputs/outputs, then emulate stateful behavior in MATLAB by feeding states back in a loop.

Why does my PyTorch model inference fail with a shape mismatch in MATLAB?▼

PyTorchExportedProgram preserves PyTorch's NCHW layout, while MATLAB images from imread are HWC. Check inputSpecifications for the expected shape and permute image data to NCHW, for example [1, 3, 224, 224], before calling invoke.

How do I use a PyTorch or LiteRT model in Simulink?▼

Add the dedicated PyTorch ExportedProgram or LiteRT block from the dlosslib library and set its ModelFilePath parameter; the block auto-detects input and output shapes. Alternatively, use a MATLAB Function block with loadPyTorchExportedProgram or loadLiteRTModel and invoke.

Related Skills

Explore More Software Engineering Skills→
yunseo-kimyunseo-kim
1

torch-export

Export PyTorch models to ExportedProgram format with dynamic shapes and control flow.

Community
Advanced
matlabmatlab
1.0k

simulink-generate-embedded-code

Configure Simulink models for Embedded Coder and AUTOSAR production code generation.

Official
Advanced
mindspore-aimindspore-ai
258

cpu-optimization-x64

Optimizes C++ x86_64 CPU code with SIMD vectorization and cache-friendly layouts.

Official
Advanced
amdamd
154

quark-torch-ptq

Quantize PyTorch transformer models into deployment artifacts using AMD Quark PTQ workflows.

Official
Advanced