Matlab Deploy AI Model

MathWorksOfficialCreated by MathWorks

Skill
Tool integration
Software development
BSD-2-Clause
Low risk

From the author

Generate C/C++ or CUDA code from an AI model (PyTorch, LiteRT) using MATLAB Coder or GPU Coder. Use when the user wants to integrate an AI model into an application with code generation as the end goal — generating MEX, CUDA MEX, static library, dynamic library, or executable — or using the model in Simulink for simulation and code generation. Covers PyTorch ExportedProgram (.pt2) via loadPyTorchExportedProgram and LiteRT (.tflite) via loadLiteRTModel (R2026a+). Keywords: PyTorch, torch, .pt2, ExportedProgram, loadPyTorchExportedProgram, invoke, codegen, MEX, CUDA, GPU, C, C++, deploy, AI model, deep learning model, LiteRT, TFLite, TensorFlow Lite, Simulink, slbuild, PyTorch ExportedProgram block, MATLAB Function block, dlosslib, loadLiteRTModel.

— the description by MathWorks

From the skill’s files

Taken from the skill’s own files (its description, section headings and stated usage). Not an AutoGineer summary, and not written by the author as one.

Generate C/C++ or CUDA code from an AI model (PyTorch, LiteRT) using MATLAB Coder or GPU Coder. Use when the user wants to integrate an AI model into an application with code generation as the end goal — generating MEX, CUDA MEX, static library, dynamic library, or executable — or using the model in Simulink for simulation and code generation. Covers PyTorch ExportedProgram (.pt2) via loadPyTorchExportedProgram and LiteRT (.tflite) via loadLiteRTModel (R2026a+). Keywords: PyTorch, torch, .pt2, ExportedProgram, loadPyTorchExportedProgram, invoke, codegen, MEX, CUDA, GPU, C, C++, deploy, AI model, deep learning model, LiteRT, TFLite, TensorFlow Lite, Simulink, slbuild, PyTorch ExportedProgram block, MATLAB Function block, dlosslib, loadLiteRTModel. Generate deployable C/C++ or CUDA code from an AI model using MATLAB Coder or GPU Coder. The workflow follows a common pattern regardless of…

What it does

When to Use
When NOT to Use
Supported Frameworks
Generic Workflow
Key Functions
Conventions

When to use it

Use when the user wants to integrate an AI model into an application with code generation as the end goal — generating MEX, CUDA MEX, static library, dynamic library, or executable — or using the…
User wants to generate C/C++/CUDA code from an AI model (PyTorch, LiteRT)
User has a model file (.pt2, .tflite) and wants to load it into MATLAB
User wants MEX acceleration for an AI model
User wants to generate CUDA code or GPU-accelerated MEX from an AI model
User wants to deploy an AI model to hardware

Advantages

The skill’s files do not state this.

Usage examples

The skill’s files do not state this.

Install

Download the skill as a ZIP:
https://autogineer.com/agent-skills/matlab/matlab-deploy-ai-model/download

Unzip it into the cross-client skills folder of your project:

mkdir -p .agents/skills && unzip matlab-deploy-ai-model.zip -d .agents/skills

Documentation by MathWorks

Generate C/C++/CUDA Code from an AI Model

Generate deployable C/C++ or CUDA code from an AI model using MATLAB Coder or GPU Coder. The workflow follows a common pattern regardless of model framework: load, inspect, write entry-point, generate MEX, verify, then generate production code.

When to Use

User wants to generate C/C++/CUDA code from an AI model (PyTorch, LiteRT)
User has a model file (.pt2, .tflite) and wants to load it into MATLAB
User wants MEX acceleration for an AI model
User wants to generate CUDA code or GPU-accelerated MEX from an AI model
User wants to deploy an AI model to hardware
User wants to use a PyTorch or LiteRT model in Simulink (simulation or code generation)
User wants to verify AI model numerics between the source framework and MATLAB

When NOT to Use

General MATLAB Coder usage (codegen syntax, config tuning, writing codegen-ready code)
Editable dlnetwork for Deep Learning Toolbox workflows (quantization, compression, transfer learning) — use importNetworkFromPyTorch (PyTorch), importNetworkFromTensorFlow (SavedModel), or importNetworkFromKeras (.keras/.h5) which return a dlnetwork. For deployment of an editable dlnetwork with model compression (INT8 quantization via dlquantizer, pruning, projection) or exportNetworkToSimulink workflows — use matlab-deploy-embedded-ai (Pattern 1).
Training or fine-tuning — this skill is for inference code generation only

Supported Frameworks

Framework	Model format	Load function	Status
PyTorch	`.pt2`	`loadPyTorchExportedProgram`	Supported (R2026a+)
LiteRT / TFLite	`.tflite`	`loadLiteRTModel`	Supported (R2026a+)

For PyTorch-specific details (API routing, entry-point pattern, export workflow, data layout, common mistakes): see references/pytorch-workflow.md.

For LiteRT-specific details (API routing, entry-point pattern, variable-size inputs, Simulink integration, conventions): see references/litert-workflow.md. For converting TensorFlow/Keras/.h5 to .tflite: see references/tensorflow-to-litert-conversion.md.

Generic Workflow

The code generation workflow follows the same steps for any framework:

1. Load and Inspect

Load the model and check its input/output specifications to determine expected shapes and types.

2. Write Entry-Point Function

Create a codegen-compatible entry-point function that:

Loads the model from a file path
Runs inference on an input
Returns the output

The model file path must be wrapped with coder.Constant so it's known at compile time.

3. Verify Numerics

Compare MATLAB inference output against the source framework to confirm correct loading. Use the same input data in both environments and compare with tolerance.

4. Generate MEX (First!)

Always generate MEX before lib/exe to verify on the host machine:

CPU MEX:

cfg = coder.config("mex");
codegen -config cfg -args {coder.Constant("model_file"), input} entryPoint

CUDA MEX (GPU acceleration):

cfg = coder.gpuConfig("mex");
codegen -config cfg -args {coder.Constant("model_file"), input} entryPoint

For CPU MEX SIMD acceleration (SIMDAcceleration = 'Full' for AVX2 on Intel/AMD), see the matlab-generate-code skill. For the DNN- inference-specific MEX AVX2 ceiling, see references/dnn-codegen-options.md.

5. Verify MEX Output

Compare MEX output against MATLAB reference using matlab.unittest with tolerance:

refOut = entryPoint("model_file", input);
mexOut = entryPoint_mex("model_file", input);
testCase = matlab.unittest.TestCase.forInteractiveUse;
testCase.verifyThat(mexOut, matlab.unittest.constraints.IsEqualTo(refOut, ...
    'Within', matlab.unittest.constraints.AbsoluteTolerance(single(1e-5))));

6. Generate Library/Executable

Once MEX is verified, generate production code:

cfgLib = coder.config("lib");
cfgLib.TargetLang = "C++";  % set to "C++" for C++ output; default is "C"
codegen -config cfgLib -args {coder.Constant("model_file"), input} entryPoint

For DLL: coder.config("dll"). For executable: coder.config("exe").

CUDA variants: Replace coder.config with coder.gpuConfig.

Performance tuning:

Generic knobs (SIMD instruction sets, reduction-loop vectorization, multithreaded loops): see the matlab-generate-code skill.
MATLAB Coder ↔ Simulink Coder property naming duality and slbuild set_param patterns: see the matlab-deploy-embedded-code skill.
DNN-inference-specific knobs (DLTargetLibrary / DeepLearningConfig to disable third-party DL libraries, LargeConstantGeneration to serialize weights to data files): see references/dnn-codegen-options.md.

7. Use in Simulink

For Simulink integration, use the dedicated PyTorch ExportedProgram block from dlosslib — set ModelFilePath to the .pt2 file and it auto-detects input/output shapes. No entry-point function or coder.Constant needed.

Pre/post-processing can be done with Simulink blocks around the dedicated block. If you need everything in a single block, use a MATLAB Function block with loadPyTorchExportedProgram + invoke (same pattern as the entry-point, but the model path is a string literal — no coder.Constant).

Both paths support slbuild code generation (requires fixed-step solver + ERT or GRT target). See references/simulink-workflow.md for full details.

8. Deploy to Hardware (Optional — requires Embedded Coder)

For embedded deployment, use the same entry-point function with an Embedded Coder configuration. See the matlab-deploy-embedded-code skill for ERT config, hardware settings, PIL/SIL verification, and target-specific options. Ask the user to install the skill if it is not installed

Key Functions

Function	Purpose	Package	Since
`coder.Constant`	Make argument a compile-time constant	MATLAB Coder	R2011a
`coder.gpuConfig`	Create GPU (CUDA) code generation config	GPU Coder	R2017b
`codegen`	Generate code	MATLAB Coder	R2011a
`loadPyTorchExportedProgram`	Load .pt2 into MATLAB	MATLAB Coder Support Package for PyTorch and LiteRT Models	R2026a
`loadLiteRTModel`	Load .tflite into MATLAB	MATLAB Coder Support Package for PyTorch and LiteRT Models	R2026a

Conventions

Always check the model's input specifications for correct input shape and type
Always generate MEX first, verify, then proceed to lib/exe
Always use coder.Constant for the model file path argument
Input data is typically single-precision (check model input specs to confirm)
Do NOT use importNetworkFromPyTorch, importNetworkFromTensorFlow, or importNetworkFromKeras in this skill — they return a dlnetwork on a different path. If the user needs a dlnetwork for quantization, projection, pruning, or exportNetworkToSimulink before code generation, route to matlab-deploy-embedded-ai instead

References

references/pytorch-workflow.md — Full PyTorch-specific workflow: API routing, entry-point pattern, export guidance, common mistakes, and conventions. Consult for any PyTorch/.pt2 model code generation task. Links to deeper PyTorch references (API signatures, data layout, numeric verification, supported models).
references/export-pytorch-models.md — Exporting an eager-mode PyTorch model to .pt2 with torch.export (upstream of loading). Consult when the user has a PyTorch model but no .pt2 file yet, or hits torch.export SerializeError / kwarg-mismatch errors. Links to pytorch-export-patterns.md (per-source templates) and pytorch-export-gotchas.md (torch 2.11 serialization fixes).
references/simulink-workflow.md — Simulink integration: dedicated PyTorch ExportedProgram block (Path A) vs MATLAB Function block (Path B), block mask parameters, code generation config, and key differences from command-line codegen.
references/litert-workflow.md — Full LiteRT-specific workflow: API routing (loadLiteRTModel → inputSpecifications → invoke), entry-point pattern, variable-size input handling, Simulink integration, and conventions. Consult for any LiteRT/.tflite model code generation task.
references/tensorflow-to-litert-conversion.md — Converting TensorFlow SavedModel/Keras/.h5 to .tflite via the Python tf.lite.TFLiteConverter API. Consult when the user has a TensorFlow model but no .tflite file yet.
references/litert-numeric-verification.md — Verifying MEX numerics against MATLAB reference for LiteRT models (tolerance guidance, common mismatches).
references/codegen-workflow.md — Shared code generation steps for both PyTorch and LiteRT: MEX generation, MEX verification, library/executable targets, coder.Constant usage, and coder.DeepLearningConfig notes.
references/dnn-codegen-options.md — DNN-INFERENCE-SPECIFIC codegen options: DLTargetLibrary / DeepLearningConfig('none') for the plain-C DL path, LargeConstantGeneration for serializing large DNN weights to data files, and MEX SIMD ceiling in a DNN-inference context. Read this when the generic file's knobs need DNN-specific framing (e.g., "the MEX SIMD cap matters because inference is the target").

Files

ATTRIBUTION.mdAdded by AutoGineer1.0 KiB
LICENSELicence file added by AutoGineer1.4 KiB
SKILL.md10.8 KiB
manifest.yaml617 B
references/codegen-workflow.md4.0 KiB
references/dnn-codegen-options.md6.9 KiB
references/export-pytorch-models.md4.6 KiB
references/litert-numeric-verification.md3.8 KiB
references/litert-workflow.md14.7 KiB
references/pytorch-api-reference.md2.6 KiB
references/pytorch-data-layout.md4.1 KiB
references/pytorch-export-gotchas.md4.7 KiB
references/pytorch-export-patterns.md5.2 KiB
references/pytorch-numeric-verification.md2.6 KiB
references/pytorch-supported-models.md3.4 KiB
references/pytorch-workflow.md5.7 KiB
references/simulink-workflow.md12.7 KiB
references/tensorflow-to-litert-conversion.md5.8 KiB

references/litert-workflow.md

Generate C/C++/CUDA Code from a LiteRT Model

Complete workflow for generating C/C++ or CUDA code from a LiteRT model (.tflite) using the MATLAB Coder Support Package for PyTorch and LiteRT Models (R2026a+).

Critical: API Routing

Three different pipelines exist for .tflite models. Choose based on the goal:

	Deep Learning Toolbox path (editable dlnetwork)	MATLAB Coder Support Package for PyTorch and LiteRT Models (code generation)	LiteRT Interpreter path (bit-true on-target)
Load	`importNetworkFromTensorFlow`	`loadLiteRTModel`	`loadTFLiteModel`
Object	`dlnetwork`	`LiteRTModel`	`TFLiteModel`
Inference	`predict(net, dlarray(...))`	`invoke(model, input)`	`predict(model, input)`
Codegen	`coder.loadDeepLearningNetwork` + `DeepLearningConfig`	Entry-point calls `loadLiteRTModel` + `invoke`	Generates code that calls LiteRT C++ API at runtime
Data layout (2-D images)	HWCN (MATLAB convention)	NHWC (LiteRT convention preserved)	HWCN
Since	R2022b	R2026a	R2024a
Use for	Transfer learning, compression (DLT tools)	Standalone C/C++/CUDA code generation	Bit-true inference via LiteRT interpreter on-target
Limitations	Converts model to dlnetwork; no support for quantized tensorflow models	—	No standalone C/CUDA — requires LiteRT runtime on target; must be built on target HW; not supported on macOS

Rule: If the goal is to keep the model's source of truth in TensorFlow/LiteRT and generate standalone C/C++/CUDA code from it, use the MATLAB Coder Support Package for PyTorch and LiteRT Models path (loadLiteRTModel + invoke). Do not use importNetworkFromTensorFlow, importNetworkFromKeras, coder.loadDeepLearningNetwork, or dlarray on this path — those convert the model to a dlnetwork first. If the user needs quantization (dlquantizer), projection, pruning, or exportNetworkToSimulink and is OK with translating the model to a dlnetwork using the Deep Learning Toolbox, then route to the matlab-deploy-embedded-ai skill (Pattern 1), which covers dlnetwork-based workflows.

loadTFLiteModel path: Use when the goal is bit-true inference that matches the LiteRT interpreter exactly on the target hardware. The generated code calls the LiteRT C++ API at runtime rather than producing standalone C/C++/CUDA — so the LiteRT runtime library must be present on the target and the code must be built on (or cross-compiled for) the target hardware. This path is not supported on macOS. Choose loadLiteRTModel instead when you need portable standalone code generation.

Obtaining a .tflite Model

LiteRT models (.tflite files) can originate from multiple frameworks:

From TensorFlow/Keras: Convert using tf.lite.TFLiteConverter in Python. See references/tensorflow-to-litert-conversion.md for complete workflows covering SavedModel, .keras, and legacy .h5 formats.

From PyTorch: Convert using litert_torch — a Google AI Edge workflow for converting PyTorch models directly to LiteRT format without going through TensorFlow. This is also a fallback route if loadPyTorchExportedProgram fails due to unsupported operations — convert the model to .tflite via litert_torch and use loadLiteRTModel instead.

Note (R2026a): After converting with litert_torch, the resulting .tflite file may contain keep_stablehlo_constant metadata that prevents loadLiteRTModel from loading it. Strip this metadata before use in MATLAB:

from ai_edge_litert import schema_py_generated as schema
import flatbuffers

with open(tflite_path, "rb") as f:
    buf = f.read()

model = schema.ModelT.InitFromPackedBuf(buf, 0)
model.metadata = [
    m for m in model.metadata
    if (m.name.decode() if isinstance(m.name, bytes) else m.name)
    != "keep_stablehlo_constant"
]

builder = flatbuffers.Builder(len(buf) + 1024)
packed = model.Pack(builder)
builder.Finish(packed, b"TFL3")

with open(tflite_path, "wb") as f:
    f.write(bytes(builder.Output()))

This requires the ai_edge_litert and flatbuffers pip packages.

From JAX: JAX models can also be converted to .tflite format. This workflow will be covered in a future skill update.

Workaround: Stateful Models (R2026a)

loadLiteRTModel does not support models with stateful operations (e.g., stateful LSTM/GRU). Loading such a model produces:

Models containing 'stateful' operations or tfl.assign_variable, tfl.call_once, tfl.read_variable, tfl.var_handle operations are not supported.

Workaround: Rewrite the model in Python to be stateless by passing hidden and cell states as explicit inputs/outputs, then emulate stateful behavior in MATLAB by feeding states back in a loop.

Step 1 — Rewrite the model in Python:

import tensorflow as tf

# Original (fails): stateful=True stores state internally
# model = tf.keras.Sequential([
#     tf.keras.layers.LSTM(16, stateful=True, batch_input_shape=(1, 5, 3)),
#     tf.keras.layers.Dense(2)
# ])

# Workaround: stateful=False, pass states explicitly, unroll for tflite
input_seq = tf.keras.Input(batch_shape=(1, 5, 3), name='input_seq')
h_in = tf.keras.Input(batch_shape=(1, 16), name='h_in')
c_in = tf.keras.Input(batch_shape=(1, 16), name='c_in')

lstm_out, h_out, c_out = tf.keras.layers.LSTM(
    16, stateful=False, return_state=True, unroll=True
)(input_seq, initial_state=[h_in, c_in])
prediction = tf.keras.layers.Dense(2, name='prediction')(lstm_out)

model = tf.keras.Model(
    inputs=[input_seq, h_in, c_in],
    outputs=[prediction, h_out, c_out]
)

# Convert to tflite
converter = tf.lite.TFLiteConverter.from_keras_model(model)
tflite_model = converter.convert()
with open('stateless_lstm.tflite', 'wb') as f:
    f.write(tflite_model)

Key changes:

stateful=False — no internal state variables
return_state=True — outputs h and c explicitly
unroll=True — required for tflite conversion (avoids TensorListReserve ops)
States passed as explicit Input layers

Step 2 — Emulate stateful inference in MATLAB:

model = loadLiteRTModel('stateless_lstm.tflite');

% Initialize states to zero
h = zeros(1, 16, 'single');
c = zeros(1, 16, 'single');

% Process sequence of inputs, carrying state forward
for step = 1:numSteps
    input_seq = getNextInput();  % [1 5 3] single
    [pred, h, c] = invoke(model, input_seq, h, c);
end

Step 3 — Codegen entry-point:

function [pred, h_out, c_out] = statefulStep(modelPath, input_seq, h_in, c_in) %#codegen
    model = loadLiteRTModel(modelPath);
    [pred, h_out, c_out] = invoke(model, input_seq, h_in, c_in);
end

The caller maintains state between calls — the generated code itself is stateless. This pattern works for any RNN (LSTM, GRU) that was originally stateful.

Load and Inspect

model = loadLiteRTModel("model.tflite");
summary(model)
inSpecs = model.inputSpecifications;
outSpecs = model.outputSpecifications;

Use inputSpecifications to determine the expected input shape and type. LiteRT models preserve NHWC layout (batch, height, width, channels).

Variable-size batch dimension: Models converted from TensorFlow often have Inf as the first dimension (variable-size batch). Three options exist for codegen:

Option	When to use	Example
Fixed batch	Known batch size at deployment	`input = ones(1, 224, 224, 3, 'single')`
Unbounded variable	Batch varies with no upper limit (requires DMA)	`coder.typeof(single(0), [Inf 224 224 3], [1 0 0 0])`
Bounded variable	Batch varies up to a maximum (stack-allocates)	`coder.typeof(single(0), [8 224 224 3], [1 0 0 0])`

For the general concept of bounded/unbounded variable-size inputs with coder.typeof, see the matlab-generate-code skill. Below is the LiteRT-specific application:

% Fixed batch
input = ones(1, 224, 224, 3, 'single');
codegen -config cfg -args {coder.Constant("model.tflite"), input} mInvoke

% Unbounded variable batch
inputCT = coder.typeof(single(0), [Inf 224 224 3], [1 0 0 0]);
codegen -config cfg -args {coder.Constant("model.tflite"), inputCT} mInvoke

% Bounded variable batch (max 8)
inputCT = coder.typeof(single(0), [8 224 224 3], [1 0 0 0]);
codegen -config cfg -args {coder.Constant("model.tflite"), inputCT} mInvoke

Multi-signature models: A LiteRT model can contain multiple functions (signatures). Use FcnNames to discover them and FcnName to target one:

model = loadLiteRTModel("multi_sig.tflite");
model.FcnNames   % e.g. ["adder", "multiplier"]

% Inspect a specific function's I/O
summary(model, FcnName="adder")
inSpecs = inputSpecifications(model, FcnName="adder");

% Invoke a specific function
out = invoke(model, in, FcnName="adder");

If the model contains multiple functions, you must specify FcnName when calling invoke, inputSpecifications, outputSpecifications, or summary. Single-signature models have FcnNames = {'main'} — you do not need to pass FcnName for these; invoke(model, input) works directly.

Entry-point for multi-signature codegen:

function out = mInvokeAdder(modelPath, input) %#codegen
    model = loadLiteRTModel(modelPath);
    out = invoke(model, input, FcnName="adder");
end

ExecutionMode: Controls how the LiteRTModel executes inference in MATLAB. Options are "JIT" (default — just-in-time compiled) and "MEX" (compiles a MEX for faster repeated inference in MATLAB). This is a MATLAB-side execution setting and is separate from generating a MEX via codegen for deployment.

model = loadLiteRTModel("model.tflite", ExecutionMode="MEX");
model = loadLiteRTModel("model.tflite", ExecutionMode="JIT");

Entry-Point Pattern

The standard entry-point for LiteRT model codegen:

function out = mInvoke(modelPath, input) %#codegen
    model = loadLiteRTModel(modelPath);
    out = invoke(model, input);
end

Entry-point for multi-signature codegen:

function out = mInvokeAdder(modelPath, input) %#codegen
    model = loadLiteRTModel(modelPath);
    out = invoke(model, input, FcnName="adder");
end

Code Generation (MEX, Library, Executable)

For the full codegen workflow (MEX generation, verification, library/executable targets, coder.Constant usage, and coder.DeepLearningConfig notes), see codegen-workflow.md.

Quick reference — CPU MEX:

cfg = coder.config("mex");
input = ones(1, 224, 224, 3, "single");  % NHWC — match inputSpecifications
codegen -config cfg -args {coder.Constant("model.tflite"), input} mInvoke

Common Mistakes

Mistake	Why It's Wrong	Correct Approach
`importNetworkFromTensorFlow("model")`	Returns `dlnetwork` (DLT path), cannot use `invoke`	`loadLiteRTModel("model.tflite")`
`importNetworkFromKeras("model.keras")`	Returns `dlnetwork` (DLT path), unrelated to LiteRT codegen	Convert to `.tflite` first, then `loadLiteRTModel`
`loadTFLiteModel(...)` for standalone codegen	Generates code that calls LiteRT C API at runtime — not standalone C/CUDA	`loadLiteRTModel("model.tflite")` for standalone code generation
`predict(model, input)`	Wrong inference function for LiteRTModel	`invoke(model, input)`
`classify(model, input)`	Wrong inference function for LiteRTModel	`invoke(model, input)`
`coder.loadDeepLearningNetwork(...)`	DLT codegen API, not compatible with LiteRTModel	Use `loadLiteRTModel` in the entry-point directly
`coder.DeepLearningConfig('tensorrt')`	Ignored by the MATLAB Coder Support Package path	Optional — has no effect on LiteRTModel codegen
Input shape `[224, 224, 3]` (HWC) from `imread`	LiteRTModel expects NHWC with batch dim	Use `[1, 224, 224, 3]` (check `inputSpecifications`)
Input shape `[1, 3, 224, 224]` (NCHW)	That's PyTorch convention, not LiteRT	LiteRT uses NHWC: `[1, 224, 224, 3]`
Passing LiteRTModel object to codegen	Rare pattern, not recommended	Pass model path as `coder.Constant` string
Confusing `ExecutionMode` with codegen MEX target	`ExecutionMode` controls in-MATLAB execution, not code generation	Use `coder.config("mex")` + `codegen` for the MEX-first verification workflow
Omitting `coder.Constant` for model path	Codegen needs model path at compile time	Always wrap model path: `coder.Constant("model.tflite")`
Calling `invoke` without `FcnName` on a multi-signature model	Error — must specify which function to invoke	`invoke(model, input, FcnName="adder")`

Conventions

Always use loadLiteRTModel for .tflite code generation workflows
LiteRT models use NHWC layout — do not confuse with PyTorch's NCHW
For multi-signature models, always specify FcnName in invoke
Do NOT use importNetworkFromTensorFlow, importNetworkFromKeras, coder.loadDeepLearningNetwork, dlarray, predict, or classify on this skill's .tflite codegen path. If the user needs a dlnetwork for quantization/projection/pruning or exportNetworkToSimulink prior to codegen, route to the matlab-deploy-embedded-ai skill
Do NOT use loadTFLiteModel on this path — it produces code that calls the LiteRT C API at runtime (bit-true interpreter path), not standalone C/CUDA. Use loadLiteRTModel for standalone code generation
For shared conventions (MEX-first, coder.Constant, verify before lib), see codegen-workflow.md

Use in Simulink

The dlosslib library provides a dedicated LiteRT block for .tflite models. Usage is identical to the PyTorch ExportedProgram block:

modelName = "my_litert_model";
new_system(modelName);
add_block("dlosslib/LiteRT", [modelName "/MyModel"]);
set_param([modelName "/MyModel"], "ModelFilePath", "/path/to/model.tflite");

The block auto-detects input/output shapes from the .tflite file.

Multi-signature models in Simulink: For .tflite models with multiple functions, the LiteRT block provides a FcnName drop-down parameter to select which function to invoke. Set it programmatically with:

set_param([modelName "/MyModel"], "FcnName", "detect");

For full Simulink integration details (simulation modes, code generation config, MATLAB Function block alternative), see simulink-workflow.md.

Detailed References

codegen-workflow.md — Shared codegen steps: MEX generation, verification, library/executable targets, coder.Constant, coder.DeepLearningConfig.
litert-numeric-verification.md — pyrun-based pattern for comparing Python LiteRT Interpreter vs MATLAB outputs.
simulink-workflow.md — Simulink integration for both PyTorch and LiteRT models (dedicated blocks, MATLAB Function block path, code generation config).
"Generate CUDA Code for YOLO v11 (LiteRT)" (MATLAB documentation) — End-to-end example with preprocessing, postprocessing (NMS), and GPU Coder CUDA code generation from a LiteRT object detection model.

Requirements and compatibility

Works with: Claude Code, Claude apps, OpenAI Codex, GitHub Copilot, Cursor, Gemini CLI, Kiro, OpenCode, Goose, Amp, Roo Code, JetBrains Junie, Factory Droid, Devin (ex-Windsurf), Mistral Vibe, OpenHands, Letta

“Works with” means the tool reads the open Agent Skills format; a skill written for one tool may use features others lack.

Security

What our scanner noticed· Low risk

Instructions that use the network

A static scan is not a guarantee. Skills are third-party content: review a skill before you let an agent use it.

Attribution and licence

Created by

MathWorks

Publisher

MathWorks (Official)

Source

matlab/matlab-agentic-toolkit · skills-catalog/code-generation/matlab-deploy-ai-model · commit 2a7f215

Licence

BSD 2-Clause License (redistribution allowed)

LICENSE (added by AutoGineer from LICENSE.md)
ATTRIBUTION.md (written by AutoGineer)

Synced

2026-11-09

AutoGineer is not affiliated with or endorsed by MathWorks or MathWorks.

Download ZIP

The ZIP holds the skill’s files, its licence files and an ATTRIBUTION.md. Download matlab-deploy-ai-model.zip

From the author

From the skill’s files

What it does

When to use it

Advantages

Usage examples

Install

Documentation by MathWorks

Generate C/C++/CUDA Code from an AI Model

When to Use

When NOT to Use

Supported Frameworks

Generic Workflow

1. Load and Inspect

2. Write Entry-Point Function

3. Verify Numerics

4. Generate MEX (First!)

5. Verify MEX Output

6. Generate Library/Executable

7. Use in Simulink

8. Deploy to Hardware (Optional — requires Embedded Coder)

Key Functions

Conventions

References

See Also

Files

references/litert-workflow.md

Generate C/C++/CUDA Code from a LiteRT Model

Critical: API Routing

Obtaining a .tflite Model

Workaround: Stateful Models (R2026a)

Load and Inspect

Entry-Point Pattern

Code Generation (MEX, Library, Executable)

Common Mistakes

Conventions

Use in Simulink

Detailed References

Requirements and compatibility

Security

Attribution and licence

Related skills

Implement Spec

MCP Builder

Schema

Postgresql Table Design

SQL Optimization Patterns

API Scaffolding Fastapi Pro

Download ZIP