zrwangems/dsh-zrwang-plugin / skills/engineering/matlab/ai-and-statistics/matlab-import-external-ai-model
matlab-import-external-ai-model
VerifiedImport PyTorch, ONNX, or Keras 3 / TensorFlow 2.16+ deep learning models into MATLAB as dlnetwork objects. Use when importing .pt2 exported programs, traced .pt files, .onnx models, or Keras 3 models via matlabsaver. Covers importNetworkFromPyTorch, importNetworkFromONNX, importNetworkFromKeras, importNetworkFromTensorFlow, torch.export.export, PyTorchInputSizes, InputDataFormats, matlabsaver, tf_keras downgrade, numeric validation against PyTorch or ONNX Runtime, and placeholder/custom layer implementation. Applies when user mentions any of these functions, file formats, or encounters import errors, unsupported operator warnings, 0 learnables, or uninitialized networks.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 11.6 KB |
| manifest.yaml | config | 626 B |
| references/keras-matlabsaver-workflow.md | markdown | 4.0 KB |
| references/keras-tf-keras-downgrade.md | markdown | 3.8 KB |
| references/onnx-validation-workflow.md | markdown | 6.4 KB |
| references/pytorch-export-guidance.md | markdown | 3.6 KB |
| references/pytorch-import-guidance.md | markdown | 3.9 KB |
| references/pytorch-numeric-validation.md | markdown | 4.3 KB |
| references/pytorch-placeholder-guidance.md | markdown | 4.3 KB |
| scripts/validateImportedNetwork.m | other | 4.3 KB |
references/keras-matlabsaver-workflow.md
View on GitHub# matlabsaver + importNetworkFromKeras Workflow
Complete step-by-step for importing Keras 3 models into MATLAB R2026a+ with full
layer structure preserved.
## Prerequisites
- MATLAB R2026a or newer
- Deep Learning Toolbox Converter for TensorFlow Models support package installed
- Python environment with TensorFlow 2.16+ and Keras 3.x
Verify the support package is installed:
```matlab
matlabsaverPath = which("matlabsaver.py");
assert(~isempty(matlabsaverPath), ...
"matlabsaver.py not found. Install the Deep Learning Toolbox " + ...
"Converter for TensorFlow Models support package.")
disp("matlabsaver.py location: " + matlabsaverPath)
```
## Step 1: Locate and Deploy matlabsaver.py
From MATLAB, find the utility script:
```matlab
matlabsaverPath = which("matlabsaver.py");
disp(matlabsaverPath)
```
Copy `matlabsaver.py` to the Python project directory, or add its parent folder
to `sys.path` in the Python script.
## Step 2: Save the Model (Python)
In the Python training script:
```python
import sys
sys.path.insert(0, "/path/to/folder/containing/matlabsaver")
import matlabsaver
# model is your trained keras.Model (Functional, Sequential, or subclass)
matlabsaver.save_for_matlab(model, "exportedModelFolder")
```
### What save_for_matlab Does
For Functional and Sequential models:
1. Saves the model as `.keras` format, then unzips it into the output folder
2. Iterates through layers and creates `ExportArchive` entries for custom layers
3. Writes a `layerClassesAndNames.json` mapping custom layer classes to instance names
For subclassed (custom) models:
1. Calls `model.export(modelFolder)` to create a SavedModel with full tracing
### Requirements
- The model must be built (`model.built == True`)
- The model must be a `keras.Model` instance (Functional, Sequential, or subclass)
- Custom layers must have deterministic `call()` methods for tracing to succeed
## Step 3: Import into MATLAB
```matlab
net = importNetworkFromKeras("exportedModelFolder");
```
The function:
- Reads the Keras 3 architecture from the saved folder
- Maps Keras 3 layers to MATLAB equivalents
- Initializes weights from the saved variables
- Returns an initialized `dlnetwork`
## Step 4: Verify the Import
The code below is illustrative — adapt the input shape and dlarray format labels to match the imported network's expected inputs.
```matlab
summary(net)
analyzeNetwork(net)
numLearnables = numel(net.Learnables.Value);
assert(numLearnables > 0, "Import produced 0 learnables — check the export step")
inputSize = net.Layers(1).InputSize;
testInput = dlarray(randn([inputSize 1], "single"), "CB");
output = predict(net, testInput);
disp("Inference OK. Output size: " + join(string(size(output)), "x"))
```
## Handling Custom Layers
When the model contains custom layers (subclasses of `keras.layers.Layer`), the
importer creates auto-generated custom layers in MATLAB. These layers are
functional but may appear as opaque blocks in `analyzeNetwork`.
To check which layers were imported as custom:
```matlab
layerNames = {net.Layers.Name}';
layerTypes = arrayfun(@(l) class(l), net.Layers, UniformOutput=false)';
customIdx = contains(layerTypes, "nnet.keras");
if any(customIdx)
disp("Custom layers imported:")
disp(layerNames(customIdx))
end
```
## Troubleshooting
| Symptom | Cause | Fix |
|---------|-------|-----|
| "Brace indexing is not supported for variables of type char" | Keras 3.10+ changed `config.json` — `input_layers`/`output_layers` are flat instead of nested | Apply the config.json patch (see Step 2 note in SKILL.md) to wrap them in a list |
| 0 learnables after import | Used `model.export` instead of `matlabsaver.save_for_matlab` | Re-export using `matlabsaver.save_for_matlab` |
| "model must be a keras.Model" error in Python | Passed a layer or non-Model object | Wrap in `keras.Model(inputs, outputs)` first |
| Import warning about untested Keras version | Keras is newer than the support package was tested against | Usually works despite warning. Update support package if errors occur |
----
Copyright 2026 The MathWorks, Inc.
----
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack." | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |