3. Getting Started

This section helps you get up and running with Tiny ML Tensorlab quickly. Follow these guides to train your first model and understand the toolchain workflow.

Contents

• 3.1. Quickstart
  • 3.1.1. Prerequisites
  • 3.1.2. Step 1: Navigate to ModelZoo
  • 3.1.3. Step 2: Run Hello World Example
  • 3.1.4. Step 3: View Results
  • 3.1.5. Step 4: Understand the Config
  • 3.1.6. Step 5: Try a Different Example
  • 3.1.7. Expected Results
  • 3.1.8. Next Steps
• 3.2. First Example
  • 3.2.1. Overview
  • 3.2.2. Step 1: Examine the Configuration
  • 3.2.3. Step 2: Run Training
  • 3.2.4. Step 3: Understand Training Output
  • 3.2.5. Step 4: Examine Output Files
  • 3.2.6. Step 5: Analyze Results
  • 3.2.7. Step 6: Customize the Example
  • 3.2.8. Next Steps
• 3.3. Understanding Config
  • 3.3.1. Configuration Overview
  • 3.3.2. Common Section
  • 3.3.3. Dataset Section
  • 3.3.4. Data Processing & Feature Extraction Section
  • 3.3.5. Training Section
  • 3.3.6. Testing Section
  • 3.3.7. Compilation Section
  • 3.3.8. Complete Example
  • 3.3.9. Tips
  • 3.3.10. See Also
• 3.4. Running Examples
  • 3.4.1. Finding Examples
  • 3.4.2. Running an Example
  • 3.4.3. Example Directory Structure
  • 3.4.4. Understanding Example Configs
  • 3.4.5. Customizing Examples
  • 3.4.6. Output Location
  • 3.4.7. Example Categories
  • 3.4.8. Running Multiple Examples
  • 3.4.9. Troubleshooting Examples
  • 3.4.10. Next Steps

3.5. 5-Minute Quickstart

After completing the installation, you can train your first model in just a few commands:

LinuxWindows

cd tinyml-modelzoo
./run_tinyml_modelzoo.sh examples/generic_timeseries_classification/config.yaml

This runs the “Hello World” example, which:

1. Downloads a simple waveform classification dataset (sine, square, sawtooth)

2. Applies feature extraction (FFT)

3. Trains a small neural network (~1K parameters)

4. Quantizes the model for MCU deployment

5. Compiles for the target device (F28P55 by default)

Output artifacts are saved to ../tinyml-modelmaker/data/projects/generic_timeseries_classification/.