10.1. Model Composer Overview

Edge AI Studio Model Composer provides a graphical user interface (GUI) for Tiny ML Tensorlab, making it easier to train and deploy models without using the command line.

10.1.1. What is Model Composer?

Model Composer is TI’s cloud-based or desktop GUI for:

  • Uploading and managing datasets

  • Configuring training parameters visually

  • Training models with progress visualization

  • Exporting trained models for deployment

It provides the same functionality as the CLI tools but through a user-friendly web interface.

10.1.2. Model Composer vs CLI

Feature

Model Composer (GUI)

CLI Tools

Ease of use

Beginner-friendly

Requires CLI knowledge

Flexibility

Guided workflows

Full configuration control

Automation

Manual only

Scriptable

Visualization

Built-in

Requires external tools

Remote access

Cloud-accessible

Local only

When to use Model Composer:

  • First-time users

  • Quick experiments

  • Visual dataset inspection

  • Non-technical stakeholders

When to use CLI:

  • Automation and scripting

  • Advanced configuration

  • Custom workflows

  • Integration with CI/CD

10.1.3. Accessing Model Composer

Cloud Version:

Access via Edge AI Studio:

  1. Visit https://dev.ti.com/edgeaistudio/

  2. Create or log into your TI account

  3. Select “Model Composer” from the dashboard

Local Version (if available):

Some installations include a local GUI:

# Start local Model Composer
cd tinyml-mlbackend
python app.py
# Open http://localhost:5000 in browser

10.1.4. Key Features

Dataset Management:

  • Upload CSV or ZIP datasets

  • Preview data in browser

  • Visualize time series plots

  • Automatic format validation

Model Configuration:

  • Visual parameter selection

  • Device dropdown menus

  • Model size sliders

  • Feature extraction presets

Training:

  • One-click training start

  • Real-time loss graphs

  • Progress indicators

  • Training history

Results Analysis:

  • Confusion matrices

  • ROC curves

  • Accuracy metrics

  • Model comparison

Export:

  • Download trained models

  • Get compilation artifacts

  • Generate CCS projects

10.1.5. Supported Workflows

Model Composer supports all Tiny ML Tensorlab workflows:

Time Series Classification:

  • Upload labeled time series data

  • Configure feature extraction

  • Train classification models

  • Evaluate with confusion matrix

Time Series Regression:

  • Upload regression datasets

  • Set target variables

  • Train regression models

  • Evaluate with scatter plots

Anomaly Detection:

  • Upload normal-only data

  • Train autoencoder models

  • Set detection thresholds

  • Evaluate reconstruction error

Forecasting:

  • Upload sequential data

  • Configure forecast horizon

  • Train forecasting models

  • Evaluate prediction accuracy

10.1.6. System Requirements

For Cloud Version:

  • Modern web browser (Chrome, Firefox, Edge)

  • Internet connection

  • TI account

For Local Version:

  • Python 3.10 environment

  • Tiny ML Tensorlab installed

  • 8 GB RAM minimum

  • GPU recommended for large models

10.1.7. Limitations

Model Composer may have some limitations compared to CLI:

  • Fewer advanced configuration options

  • May not support all presets

  • File size limits for uploads

  • Processing time limits

For advanced use cases, consider using the CLI tools directly.

10.1.8. Integration with CLI

You can combine Model Composer and CLI workflows:

Export Configuration:

Model Composer can export YAML config files:

  1. Configure your project in Model Composer

  2. Click “Export Configuration”

  3. Download the YAML file

  4. Use with CLI for further customization

Import Models:

CLI-trained models can be analyzed in Model Composer:

  1. Upload trained model files

  2. View analysis results

  3. Export for deployment

10.1.9. Getting Started

Ready to try Model Composer? See:

For CLI-based workflow, see Quickstart.