TI Tiny ML Tensorlab
v1.3.0

Contents

  • 1. Introduction
    • 1.1. What is Tiny ML Tensorlab?
      • 1.1.1. Overview
      • 1.1.2. Target Applications
      • 1.1.3. Key Capabilities
      • 1.1.4. Repository Structure
      • 1.1.5. Workflow Summary
      • 1.1.6. Next Steps
    • 1.2. System Architecture
      • 1.2.1. High-Level Architecture
      • 1.2.2. Component Details
        • 1.2.2.1. tinyml-modelzoo
        • 1.2.2.2. tinyml-modelmaker
        • 1.2.2.3. tinyml-tinyverse
        • 1.2.2.4. tinyml-modeloptimization
      • 1.2.3. Which Repository Do I Need?
      • 1.2.4. Data Flow
      • 1.2.5. Configuration System
      • 1.2.6. Integration Points
    • 1.3. Terminology
      • 1.3.1. General ML Terms
      • 1.3.2. Tiny ML Tensorlab Terms
      • 1.3.3. Device & Hardware Terms
      • 1.3.4. Configuration Terms
      • 1.3.5. Data Terms
      • 1.3.6. Model Size Conventions
      • 1.3.7. Abbreviations
    • 1.4. Overview
  • 2. Installation
    • 2.1. Prerequisites
      • 2.1.1. System Requirements
      • 2.1.2. Software Requirements
      • 2.1.3. For Compilation (Optional)
      • 2.1.4. For Device Deployment (Optional)
      • 2.1.5. CUDA (Optional)
      • 2.1.6. Verification Checklist
    • 2.2. User Installation
      • 2.2.1. Quick Install
      • 2.2.2. Running Your First Example
      • 2.2.3. Verifying Installation
      • 2.2.4. Updating
      • 2.2.5. Uninstalling
      • 2.2.6. Limitations of User Install
      • 2.2.7. Troubleshooting
      • 2.2.8. Next Steps
    • 2.3. Developer Installation
      • 2.3.1. Overview
      • 2.3.2. Step 1: Clone the Repository
      • 2.3.3. Step 2: Set Up Python Environment
      • 2.3.4. Step 3: Verify Installation
      • 2.3.5. Directory Structure After Installation
      • 2.3.6. Updating
      • 2.3.7. Common Developer Tasks
      • 2.3.8. Troubleshooting
      • 2.3.9. Next Steps
    • 2.4. Windows Setup
      • 2.4.1. Option 1: Native Windows Installation
      • 2.4.2. Option 2: WSL2 (Recommended for Full Compatibility)
        • 2.4.2.1. Network and Proxy Configuration in WSL (Optional)
        • 2.4.2.2. Docker in WSL (Optional)
        • 2.4.2.3. USB Device Access from WSL
        • 2.4.2.4. VS Code with WSL
      • 2.4.3. Path Configuration
      • 2.4.4. Common Windows Issues
      • 2.4.5. GPU Support on Windows
      • 2.4.6. WSL2 vs Native Windows Comparison
      • 2.4.7. Next Steps
    • 2.5. Linux Setup
      • 2.5.1. System Preparation
      • 2.5.2. Installing Python 3.10
      • 2.5.3. Installation
      • 2.5.4. Verification
      • 2.5.5. GPU Setup (Optional)
      • 2.5.6. Shell Configuration
      • 2.5.7. Permission Issues
      • 2.5.8. Multiple Python Versions
      • 2.5.9. System Service (Optional)
      • 2.5.10. Troubleshooting
      • 2.5.11. Next Steps
    • 2.6. Environment Variables
      • 2.6.1. Required Variables by Device Family
        • 2.6.1.1. C2000 Devices (F28P55, F28P65, F2837, etc.)
        • 2.6.1.2. MSPM0 Devices (MSPM0G3507, MSPM0G5187, etc.)
        • 2.6.1.3. AM26x Devices (AM263, AM261, etc.)
      • 2.6.2. Setting Environment Variables
      • 2.6.3. Installing TI Tools
      • 2.6.4. Verifying Configuration
      • 2.6.5. Troubleshooting
      • 2.6.6. Next Steps
    • 2.7. Quick Start
  • 3. Getting Started
    • 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
  • 4. Supported Task Types
    • 4.1. Time Series Classification
      • 4.1.1. Overview
      • 4.1.2. Configuration
      • 4.1.3. Dataset Format
      • 4.1.4. Available Models
      • 4.1.5. Feature Extraction
      • 4.1.6. Metrics
      • 4.1.7. Example: Arc Fault Detection
      • 4.1.8. Tips
      • 4.1.9. See Also
    • 4.2. Time Series Regression
      • 4.2.1. Overview
      • 4.2.2. Configuration
      • 4.2.3. Dataset Format
      • 4.2.4. Available Models
      • 4.2.5. Key Configuration
      • 4.2.6. Metrics
      • 4.2.7. Example: Torque Measurement
      • 4.2.8. Tips
      • 4.2.9. See Also
    • 4.3. Time Series Forecasting
      • 4.3.1. Overview
      • 4.3.2. Configuration
      • 4.3.3. Key Parameters
      • 4.3.4. Dataset Format
      • 4.3.5. Available Models
      • 4.3.6. Windowing Example
      • 4.3.7. Metrics
      • 4.3.8. Example: PMSM Temperature Forecasting
      • 4.3.9. Important Notes
      • 4.3.10. Tips
      • 4.3.11. See Also
    • 4.4. Anomaly Detection
      • 4.4.1. Overview
      • 4.4.2. How It Works
      • 4.4.3. Autoencoder Architecture
        • 4.4.3.1. What is an Autoencoder?
        • 4.4.3.2. Architecture Diagram
        • 4.4.3.3. Dimensionality Flow
        • 4.4.3.4. Key Components
        • 4.4.3.5. Training with MSE Loss
        • 4.4.3.6. Semi-Supervised Learning: Normal Data Only
        • 4.4.3.7. Advantages and Limitations
      • 4.4.4. Configuration
      • 4.4.5. Dataset Format for Anomaly Detection
        • 4.4.5.1. Folder Structure
        • 4.4.5.2. Concrete Example
        • 4.4.5.3. Data Splitting Strategy
        • 4.4.5.4. Datafile Format (CSV)
      • 4.4.6. Dataset Format
      • 4.4.7. Available Models
      • 4.4.8. Training Workflow
        • 4.4.8.1. Overview
        • 4.4.8.2. Pipeline Steps in Detail
        • 4.4.8.3. Running the Pipeline
        • 4.4.8.4. Output Files
        • 4.4.8.5. Understanding Training Logs
      • 4.4.9. Threshold Selection
        • 4.4.9.1. Formula
        • 4.4.9.2. k-Value Impact
        • 4.4.9.3. threshold_performance.csv
        • 4.4.9.4. How to Choose k Based on Application
      • 4.4.10. Evaluation Metrics
        • 4.4.10.1. Confusion Matrix
        • 4.4.10.2. Metric Definitions
        • 4.4.10.3. Metric Summary
      • 4.4.11. What If You Don’t Have Anomaly Data?
        • 4.4.11.1. What You CAN Do
        • 4.4.11.2. What You CANNOT Do
      • 4.4.12. Semi-Supervised vs Supervised
      • 4.4.13. Example: Motor Bearing Anomaly
      • 4.4.14. Tips
      • 4.4.15. See Also
    • 4.5. Image Classification
      • 4.5.1. Overview
      • 4.5.2. Configuration
      • 4.5.3. Dataset Format
      • 4.5.4. Available Models
      • 4.5.5. Current Limitations
      • 4.5.6. Example: MNIST Digit Recognition
      • 4.5.7. Preparing Image Data
      • 4.5.8. Tips
      • 4.5.9. See Also
    • 4.6. Task Overview
    • 4.7. Choosing the Right Task
  • 5. Bring Your Own Data
    • 5.1. Classification Dataset Format
      • 5.1.1. Directory Structure
      • 5.1.2. Data File Format
      • 5.1.3. Supported File Types
      • 5.1.4. Annotations (Optional)
      • 5.1.5. Configuration
      • 5.1.6. Dataset Splitting Modes
      • 5.1.7. Example: 3-Class Vibration Data
      • 5.1.8. Class Balancing
      • 5.1.9. Common Issues
      • 5.1.10. Key Differences: Classification vs Regression vs Forecasting
    • 5.2. Regression Dataset Format
      • 5.2.1. Directory Structure
      • 5.2.2. Data File Format
      • 5.2.3. Time Column Handling
      • 5.2.4. Annotation Files (Required)
      • 5.2.5. Configuration
      • 5.2.6. Target Processing
      • 5.2.7. Complete Example
      • 5.2.8. Troubleshooting
      • 5.2.9. Best Practices
      • 5.2.10. Common Issues
    • 5.3. Forecasting Dataset Format
      • 5.3.1. Directory Structure
      • 5.3.2. Data File Format
      • 5.3.3. Key Difference from Regression
      • 5.3.4. Configuration
      • 5.3.5. Variable Specification Options
      • 5.3.6. Windowing Behavior
      • 5.3.7. Complete Example
      • 5.3.8. Important Notes
      • 5.3.9. Minimum Data Requirements
      • 5.3.10. Common Issues
    • 5.4. Anomaly Detection Dataset Format
      • 5.4.1. Folder Structure
      • 5.4.2. Concrete Example
      • 5.4.3. Data Splitting Strategy
      • 5.4.4. What If You Don’t Have Anomaly Data?
      • 5.4.5. Datafile Format (CSV)
      • 5.4.6. See Also
    • 5.5. Data Splitting
      • 5.5.1. Split Methods
      • 5.5.2. Configuration
      • 5.5.3. Annotation File Format
      • 5.5.4. Split Examples
      • 5.5.5. When to Use Each Method
      • 5.5.6. Best Practices
      • 5.5.7. Creating Annotation Files
      • 5.5.8. Verifying Splits
    • 5.6. Dataset Format Overview
    • 5.7. Supported File Formats
    • 5.8. Data Sources
  • 6. Supported Devices
    • 6.1. Device Overview
      • 6.1.1. Supported Device Families
      • 6.1.2. Complete Device List
      • 6.1.3. Target Device Configuration
      • 6.1.4. NPU vs Non-NPU Devices
      • 6.1.5. Choosing a Device
    • 6.2. NPU Guidelines
      • 6.2.1. NPU-Enabled Devices
      • 6.2.2. Layer Constraints
      • 6.2.3. Using NPU-Compatible Models
      • 6.2.4. Channel Multiples of 4
      • 6.2.5. Kernel Size Restrictions
      • 6.2.6. Compilation Preset
      • 6.2.7. Custom NPU-Compatible Models
      • 6.2.8. Troubleshooting NPU Compilation
      • 6.2.9. Performance Comparison
    • 6.3. C2000 Family
      • 6.3.1. Overview
      • 6.3.2. Supported Devices
      • 6.3.3. F28P55 (Recommended)
      • 6.3.4. F28P65
      • 6.3.5. F2837
      • 6.3.6. C29x Family (F29H85, F29P58, F29P32)
      • 6.3.7. Typical Applications
      • 6.3.8. Development Tools
      • 6.3.9. Memory Considerations
      • 6.3.10. Next Steps
    • 6.4. MSPM0 Family
      • 6.4.1. Overview
      • 6.4.2. Supported Devices
      • 6.4.3. MSPM0G5187 (NPU-Enabled)
      • 6.4.4. MSPM0G3507
      • 6.4.5. MSPM0G3519
      • 6.4.6. AM13 Family
      • 6.4.7. Power Considerations
      • 6.4.8. Memory Constraints
      • 6.4.9. Typical Applications
      • 6.4.10. Development Tools
      • 6.4.11. Getting Started
      • 6.4.12. Next Steps
    • 6.5. Connectivity Devices
      • 6.5.1. Overview
      • 6.5.2. Supported Devices
      • 6.5.3. CC2755
      • 6.5.4. CC1352
      • 6.5.5. CC1354
      • 6.5.6. CC35X1
      • 6.5.7. AM26x Family
      • 6.5.8. Typical Applications
      • 6.5.9. Power Optimization
      • 6.5.10. Memory Constraints
      • 6.5.11. Development Tools
      • 6.5.12. Wireless Protocol Considerations
      • 6.5.13. Getting Started
      • 6.5.14. Next Steps
    • 6.6. Device Families at a Glance
    • 6.7. Complete Device List
    • 6.8. NPU vs Non-NPU Devices
  • 7. Examples & Applications
    • 7.1. Running an Example
    • 7.2. Generic Examples
    • 7.3. Classification Examples
    • 7.4. Regression Examples
    • 7.5. Forecasting Examples
    • 7.6. Anomaly Detection Examples
    • 7.7. Image Classification Examples
      • 7.7.1. Generic Time Series Classification
        • 7.7.1.1. Overview
        • 7.7.1.2. Running the Example
        • 7.7.1.3. Understanding the Dataset
        • 7.7.1.4. Dataset Format
        • 7.7.1.5. Configuration
        • 7.7.1.6. Feature Extraction
        • 7.7.1.7. Evaluation Metrics
        • 7.7.1.8. Expected Results
        • 7.7.1.9. Output Location
        • 7.7.1.10. Variations to Try
        • 7.7.1.11. Next Steps
      • 7.7.2. Generic Time Series Regression
        • 7.7.2.1. Overview
        • 7.7.2.2. Running the Example
        • 7.7.2.3. Understanding the Dataset
        • 7.7.2.4. Dataset Format
        • 7.7.2.5. Configuration
        • 7.7.2.6. Evaluation Metrics
        • 7.7.2.7. Expected Results
        • 7.7.2.8. Output Location
        • 7.7.2.9. Next Steps
      • 7.7.3. Generic Time Series Forecasting
        • 7.7.3.1. Overview
        • 7.7.3.2. Running the Example
        • 7.7.3.3. Understanding the Dataset
        • 7.7.3.4. Dataset Format
        • 7.7.3.5. Configuration
        • 7.7.3.6. Target Variables
        • 7.7.3.7. Evaluation Metrics
        • 7.7.3.8. Expected Results
        • 7.7.3.9. Output Location
        • 7.7.3.10. Next Steps
      • 7.7.4. Generic Time Series Anomaly Detection
        • 7.7.4.1. Overview
        • 7.7.4.2. Running the Example
        • 7.7.4.3. Understanding the Dataset
        • 7.7.4.4. Dataset Format
        • 7.7.4.5. Configuration
        • 7.7.4.6. How Autoencoder Detection Works
        • 7.7.4.7. Why Frame Size Matters
        • 7.7.4.8. Expected Results
        • 7.7.4.9. Output Location
        • 7.7.4.10. Next Steps
      • 7.7.5. Arc Fault Detection
        • 7.7.5.1. Overview
        • 7.7.5.2. Why Arc Fault Detection?
        • 7.7.5.3. Running the Example
        • 7.7.5.4. Configuration
        • 7.7.5.5. Dataset Description
        • 7.7.5.6. Feature Extraction
        • 7.7.5.7. Available Models
        • 7.7.5.8. Expected Results
        • 7.7.5.9. Interpreting Results
        • 7.7.5.10. Deployment Considerations
        • 7.7.5.11. AC Arc Fault Detection
        • 7.7.5.12. Next Steps
      • 7.7.6. AC Arc Fault
        • 7.7.6.1. Overview
        • 7.7.6.2. Configuration
        • 7.7.6.3. Running the Example
        • 7.7.6.4. Dataset Details
        • 7.7.6.5. Recommended Models
        • 7.7.6.6. See Also
      • 7.7.7. Motor Bearing Fault
        • 7.7.7.1. Overview
        • 7.7.7.2. Fault Classes
        • 7.7.7.3. Running the Example
        • 7.7.7.4. Configuration
        • 7.7.7.5. Dataset Description
        • 7.7.7.6. Feature Extraction Presets
        • 7.7.7.7. Available Models
        • 7.7.7.8. Expected Results
        • 7.7.7.9. Multi-Class Evaluation
        • 7.7.7.10. Dataset Quality Analysis
        • 7.7.7.11. Practical Considerations
        • 7.7.7.12. Anomaly Detection Alternative
        • 7.7.7.13. Next Steps
      • 7.7.8. Blower Imbalance
        • 7.7.8.1. Overview
        • 7.7.8.2. Configuration
        • 7.7.8.3. Running the Example
        • 7.7.8.4. Dataset Details
        • 7.7.8.5. Recommended Models
        • 7.7.8.6. See Also
      • 7.7.9. Fan Blade Fault Classification
        • 7.7.9.1. Overview
        • 7.7.9.2. Demo Setup
        • 7.7.9.3. Fault Types
        • 7.7.9.4. Configuration
        • 7.7.9.5. Running the Example
        • 7.7.9.6. Dataset Details
        • 7.7.9.7. Results and Analysis
        • 7.7.9.8. Anomaly Detection Variant
        • 7.7.9.9. See Also
      • 7.7.10. Electrical Fault
        • 7.7.10.1. Overview
        • 7.7.10.2. Configuration
        • 7.7.10.3. Running the Example
        • 7.7.10.4. Dataset Details
        • 7.7.10.5. See Also
      • 7.7.11. Grid Stability
        • 7.7.11.1. Overview
        • 7.7.11.2. Configuration
        • 7.7.11.3. Running the Example
        • 7.7.11.4. Dataset Details
        • 7.7.11.5. See Also
      • 7.7.12. Gas Sensor
        • 7.7.12.1. Overview
        • 7.7.12.2. Configuration
        • 7.7.12.3. Running the Example
        • 7.7.12.4. Dataset Details
        • 7.7.12.5. Quantization Analysis
        • 7.7.12.6. See Also
      • 7.7.13. Human Activity Recognition
        • 7.7.13.1. Overview
        • 7.7.13.2. Configuration
        • 7.7.13.3. Running the Example
        • 7.7.13.4. Dataset Details
        • 7.7.13.5. See Also
      • 7.7.14. ECG Classification
        • 7.7.14.1. Overview
        • 7.7.14.2. Key Performance Targets
        • 7.7.14.3. System Components
        • 7.7.14.4. Dataset Details
        • 7.7.14.5. Feature Extraction
        • 7.7.14.6. Model
        • 7.7.14.7. Training Configuration
        • 7.7.14.8. Configuration (MSPM0)
        • 7.7.14.9. Running the Example
        • 7.7.14.10. Available Configurations
        • 7.7.14.11. Anomaly Detection Mode
        • 7.7.14.12. References
        • 7.7.14.13. See Also
      • 7.7.15. NILM Appliance Usage Classification
        • 7.7.15.1. Overview
        • 7.7.15.2. Configuration
        • 7.7.15.3. Running the Example
        • 7.7.15.4. Dataset Details
        • 7.7.15.5. On-Device Results
        • 7.7.15.6. PLAID Dataset Variant
        • 7.7.15.7. See Also
      • 7.7.16. PIR Detection
        • 7.7.16.1. Overview
        • 7.7.16.2. Device Support
        • 7.7.16.3. System Components
        • 7.7.16.4. Running the Example
        • 7.7.16.5. Dataset Description
        • 7.7.16.6. Feature Extraction Pipeline
        • 7.7.16.7. Model
        • 7.7.16.8. Expected Results
        • 7.7.16.9. Training Configuration
        • 7.7.16.10. References
      • 7.7.17. Grid Fault Detection
        • 7.7.17.1. Overview
        • 7.7.17.2. Configuration
        • 7.7.17.3. Running the Example
        • 7.7.17.4. Dataset Details
        • 7.7.17.5. See Also
      • 7.7.18. Torque Measurement Regression
        • 7.7.18.1. Overview
        • 7.7.18.2. Configuration
        • 7.7.18.3. Running the Example
        • 7.7.18.4. Dataset Details
        • 7.7.18.5. Recommended Models
        • 7.7.18.6. See Also
      • 7.7.19. Induction Motor Speed Prediction
        • 7.7.19.1. Overview
        • 7.7.19.2. Configuration
        • 7.7.19.3. Running the Example
        • 7.7.19.4. Dataset Details
        • 7.7.19.5. See Also
      • 7.7.20. Washing Machine Regression
        • 7.7.20.1. Overview
        • 7.7.20.2. Configuration
        • 7.7.20.3. Running the Example
        • 7.7.20.4. Dataset Details
        • 7.7.20.5. Results
        • 7.7.20.6. See Also
      • 7.7.21. MOSFET Junction Temperature Prediction
        • 7.7.21.1. Overview
        • 7.7.21.2. Configuration
        • 7.7.21.3. Running the Example
        • 7.7.21.4. Dataset Details
        • 7.7.21.5. On-Device Deployment
        • 7.7.21.6. See Also
      • 7.7.22. PMSM Rotor Forecasting
        • 7.7.22.1. Overview
        • 7.7.22.2. Configuration
        • 7.7.22.3. Running the Example
        • 7.7.22.4. Dataset Details
        • 7.7.22.5. Recommended Models
        • 7.7.22.6. Results
        • 7.7.22.7. See Also
      • 7.7.23. HVAC Indoor Temp Forecast
        • 7.7.23.1. Overview
        • 7.7.23.2. Configuration
        • 7.7.23.3. Running the Example
        • 7.7.23.4. Dataset Details
        • 7.7.23.5. Results
        • 7.7.23.6. See Also
      • 7.7.24. Arc Fault Anomaly Detection Example
        • 7.7.24.1. Overview
        • 7.7.24.2. Running the Example
        • 7.7.24.3. Configuration
        • 7.7.24.4. Dataset Format
        • 7.7.24.5. Available Models
        • 7.7.24.6. Expected Results
        • 7.7.24.7. Threshold Selection
        • 7.7.24.8. Interpreting Outputs
        • 7.7.24.9. Advanced Configuration
        • 7.7.24.10. Practical Applications
        • 7.7.24.11. Comparison with Classification
        • 7.7.24.12. Troubleshooting
        • 7.7.24.13. Next Steps
      • 7.7.25. Forecasting Example
        • 7.7.25.1. Overview
        • 7.7.25.2. When to Use Forecasting
        • 7.7.25.3. Running the Example
        • 7.7.25.4. Configuration
        • 7.7.25.5. How Forecasting Works
        • 7.7.25.6. Dataset Format
        • 7.7.25.7. Available Models
        • 7.7.25.8. Expected Results
        • 7.7.25.9. Key Metrics
        • 7.7.25.10. Forecast Horizon Trade-offs
        • 7.7.25.11. Multi-Variable Forecasting
        • 7.7.25.12. Feature Extraction Options
        • 7.7.25.13. Practical Applications
        • 7.7.25.14. Deployment Considerations
        • 7.7.25.15. Troubleshooting
        • 7.7.25.16. Comparison with Regression
        • 7.7.25.17. Next Steps
      • 7.7.26. Image Classification Example
        • 7.7.26.1. Overview
        • 7.7.26.2. When to Use Image Classification
        • 7.7.26.3. Running the Example
        • 7.7.26.4. Configuration
        • 7.7.26.5. Dataset Format
        • 7.7.26.6. Image Size Considerations
        • 7.7.26.7. Available Models
        • 7.7.26.8. Expected Results
        • 7.7.26.9. Grayscale vs RGB
        • 7.7.26.10. Data Augmentation
        • 7.7.26.11. Practical Applications
        • 7.7.26.12. Memory Constraints
        • 7.7.26.13. Inference Performance
        • 7.7.26.14. Transfer Learning
        • 7.7.26.15. Camera Integration
        • 7.7.26.16. Troubleshooting
        • 7.7.26.17. Limitations
        • 7.7.26.18. Next Steps
      • 7.7.27. MNIST Image Classification
        • 7.7.27.1. Overview
        • 7.7.27.2. Key Targets
        • 7.7.27.3. System Components
        • 7.7.27.4. Running the Example
        • 7.7.27.5. Dataset
        • 7.7.27.6. Feature Extraction
        • 7.7.27.7. Model Architecture
        • 7.7.27.8. Training Configuration
        • 7.7.27.9. Expected Results
        • 7.7.27.10. Supported Devices
        • 7.7.27.11. References
  • 8. Advanced Features
    • 8.1. Neural Architecture Search
      • 8.1.1. Overview
      • 8.1.2. When to Use NAS
      • 8.1.3. Code Flow
      • 8.1.4. Configuration
      • 8.1.5. Model Size Presets
      • 8.1.6. Usage
      • 8.1.7. Running NAS
      • 8.1.8. Example: Full NAS Configuration
      • 8.1.9. Tips
      • 8.1.10. Best Practices
      • 8.1.11. Search Algorithm
      • 8.1.12. Search Space
      • 8.1.13. NAS Framework Internals
      • 8.1.14. References
      • 8.1.15. Next Steps
    • 8.2. Quantization
      • 8.2.1. Overview
      • 8.2.2. Configuration Parameters
      • 8.2.3. Quantization Modes
      • 8.2.4. Quantization Methods
      • 8.2.5. Bit Widths
      • 8.2.6. NPU Quantization Requirements
      • 8.2.7. Output Files
      • 8.2.8. Accuracy Comparison
      • 8.2.9. Troubleshooting Accuracy Loss
      • 8.2.10. Best Practices
      • 8.2.11. Example: Full Quantization Workflow
      • 8.2.12. Memory Savings
      • 8.2.13. Performance Impact
      • 8.2.14. Quantization Wrapper Architecture
      • 8.2.15. NPU Hardware Constraints
      • 8.2.16. Using Quantization Wrappers Directly
      • 8.2.17. Wrapper API Reference
      • 8.2.18. Model Surgery
      • 8.2.19. Next Steps
    • 8.3. Standalone Quantization Examples
      • 8.3.1. Overview
      • 8.3.2. FMNIST Image Classification
      • 8.3.3. Audio Keyword Spotting
      • 8.3.4. Motor Fault Time Series Classification
      • 8.3.5. MNIST Digit Classification
      • 8.3.6. Torque Time Series Regression
      • 8.3.7. Quantization Guidance
      • 8.3.8. Next Steps
    • 8.4. Feature Extraction
      • 8.4.1. Overview
      • 8.4.2. Feature Extraction Pipeline
      • 8.4.3. Configuration Parameters
      • 8.4.4. Preset System
      • 8.4.5. Available Presets
      • 8.4.6. Data Processing Transforms
      • 8.4.7. Feature Extraction Transforms
      • 8.4.8. Custom Feature Extraction
      • 8.4.9. Multi-Channel Data
      • 8.4.10. Forecasting Configuration
      • 8.4.11. Data Augmentation
      • 8.4.12. Choosing the Right Preset
      • 8.4.13. Performance Impact
      • 8.4.14. On-Device Feature Extraction
      • 8.4.15. Example Configurations
      • 8.4.16. Stacking Modes
      • 8.4.17. Gain Variation Augmentation
      • 8.4.18. Q15 Fixed-Point Transforms
      • 8.4.19. Frame Offset (Overlap Control)
      • 8.4.20. Analysis Bandwidth
      • 8.4.21. Feature Extraction Only Mode
      • 8.4.22. Evaluating Feature Extraction Quality
      • 8.4.23. Best Practices
      • 8.4.24. Next Steps
    • 8.5. Goodness of Fit
      • 8.5.1. Overview
      • 8.5.2. Enabling GoF Test
      • 8.5.3. Running the Test
      • 8.5.4. Output Files
      • 8.5.5. Understanding the Visualizations
      • 8.5.6. Interpreting Results
      • 8.5.7. 8-Plot Analysis
      • 8.5.8. Common Patterns
      • 8.5.9. Actionable Insights
      • 8.5.10. Frame Size Sweeping
      • 8.5.11. Multi-Cluster Analysis
      • 8.5.12. Example: Motor Fault GoF Analysis
      • 8.5.13. GoF Without Training
      • 8.5.14. Comparing Feature Extraction
      • 8.5.15. Best Practices
      • 8.5.16. Limitations
      • 8.5.17. Next Steps
    • 8.6. Post-Training Analysis
      • 8.6.1. Overview
      • 8.6.2. Enabling Analysis
      • 8.6.3. Output Files
      • 8.6.4. Confusion Matrix
      • 8.6.5. ROC Curves
      • 8.6.6. Class Score Histograms
      • 8.6.7. FPR/TPR Thresholds
      • 8.6.8. Classification Report
      • 8.6.9. Error Analysis
      • 8.6.10. Quantized vs Float Comparison
      • 8.6.11. File-Level Classification Summary
      • 8.6.12. Regression Analysis
      • 8.6.13. Anomaly Detection Analysis
      • 8.6.14. Custom Analysis Scripts
      • 8.6.15. Generating Reports
      • 8.6.16. Example: Complete Analysis Configuration
      • 8.6.17. Best Practices
      • 8.6.18. Troubleshooting Low Accuracy
      • 8.6.19. Next Steps
    • 8.7. Feature Overview
  • 9. Device Deployment
    • 9.1. CCS Integration Guide
      • 9.1.1. Prerequisites
      • 9.1.2. Compilation Output
      • 9.1.3. Creating a CCS Project
      • 9.1.4. Integration Code
      • 9.1.5. Memory Placement
      • 9.1.6. Linker Command File
      • 9.1.7. Interrupt-Based Inference
      • 9.1.8. Timing and Profiling
      • 9.1.9. Debugging
      • 9.1.10. Build Configurations
      • 9.1.11. Example Project Structure
      • 9.1.12. Common Issues
      • 9.1.13. Testing on Hardware
      • 9.1.14. Next Steps
    • 9.2. NPU Device Deployment
      • 9.2.1. NPU-Enabled Devices
      • 9.2.2. NPU Compilation
      • 9.2.3. NPU Model Requirements
      • 9.2.4. NPU Compilation Artifacts
      • 9.2.5. NPU Initialization
      • 9.2.6. NPU Inference Code
      • 9.2.7. NPU Memory Management
      • 9.2.8. NPU Performance
      • 9.2.9. NPU Power Considerations
      • 9.2.10. NPU Debugging
      • 9.2.11. NPU Error Handling
      • 9.2.12. CCS Project Setup for NPU
      • 9.2.13. Example: Arc Fault on F28P55 NPU
      • 9.2.14. Troubleshooting NPU Issues
      • 9.2.15. CCS Studio Walkthrough: F28P55x
        • 9.2.15.1. Step 1 – Load the Example from Resource Explorer
        • 9.2.15.2. Step 2 – Build the Project
        • 9.2.15.3. Step 3 – Set Target Configuration
        • 9.2.15.4. Step 4 – Flash the Device
        • 9.2.15.5. Step 5 – Debug and Verify
      • 9.2.16. Required Files from ModelMaker
      • 9.2.17. Model Performance Profiling
      • 9.2.18. Next Steps
    • 9.3. Non-NPU Deployment
      • 9.3.1. Non-NPU Devices
      • 9.3.2. Configuration
      • 9.3.3. Model Selection
      • 9.3.4. CPU Inference Performance
      • 9.3.5. Compilation Artifacts
      • 9.3.6. CCS Project Setup
      • 9.3.7. Basic Integration
      • 9.3.8. Optimizing CPU Inference
      • 9.3.9. Memory Optimization
      • 9.3.10. Power Optimization
      • 9.3.11. Real-Time Considerations
      • 9.3.12. Device-Specific Notes
      • 9.3.13. Example: Vibration Monitoring on MSPM0G3507
      • 9.3.14. CCS Studio Walkthrough: F28004x
        • 9.3.14.1. Step 1 – Import the Project Manually
        • 9.3.14.2. Step 2 – Build the Project
        • 9.3.14.3. Step 3 – Set Target Configuration
        • 9.3.14.4. Step 4 – Flash the Device
        • 9.3.14.5. Step 5 – Debug and Verify
      • 9.3.15. Required Files from ModelMaker
      • 9.3.16. Comparison: NPU vs Non-NPU
      • 9.3.17. Next Steps
    • 9.4. Supported Devices
    • 9.5. Deployment Overview
    • 9.6. Prerequisites
    • 9.7. Output File Locations
    • 9.8. CCS Example Project Locations
    • 9.9. Task-Type-Specific Deployment Notes
    • 9.10. Testing Multiple Cases
    • 9.11. Model Compilation Details
  • 10. Edge AI Studio Model Composer
    • 10.1. Model Composer Overview
      • 10.1.1. What is Model Composer?
      • 10.1.2. Model Composer vs CLI
      • 10.1.3. Accessing Model Composer
      • 10.1.4. Key Features
      • 10.1.5. Supported Workflows
      • 10.1.6. System Requirements
      • 10.1.7. Limitations
      • 10.1.8. Integration with CLI
      • 10.1.9. Getting Started
    • 10.2. Getting Started (GUI)
      • 10.2.1. Prerequisites
      • 10.2.2. Step 1: Access Model Composer
      • 10.2.3. Step 2: Create a Project
      • 10.2.4. Step 3: Upload Dataset
      • 10.2.5. Step 4: Configure Feature Extraction
      • 10.2.6. Step 5: Configure Training
      • 10.2.7. Step 6: Start Training
      • 10.2.8. Step 7: Analyze Results
      • 10.2.9. Step 8: Evaluate on Test Set
      • 10.2.10. Step 9: Export Model
      • 10.2.11. Using Exported Model
      • 10.2.12. GUI Tips and Tricks
      • 10.2.13. Troubleshooting
      • 10.2.14. What’s Next?
      • 10.2.15. Additional Resources
    • 10.3. Exporting Models
      • 10.3.1. Export Options
      • 10.3.2. Exporting as CCS Project
      • 10.3.3. Exporting Artifacts Only
      • 10.3.4. Exporting ONNX Model
      • 10.3.5. Exporting Configuration
      • 10.3.6. Batch Export
      • 10.3.7. Export History
      • 10.3.8. Validating Exports
      • 10.3.9. Export Troubleshooting
      • 10.3.10. Best Practices
      • 10.3.11. Next Steps
    • 10.4. What is Model Composer?
    • 10.5. GUI vs CLI Comparison
  • 11. Bring Your Own Model
    • 11.1. Adding Custom Models
      • 11.1.1. Overview
      • 11.1.2. Quick Summary
      • 11.1.3. Step 1: Choose the Right Model File
      • 11.1.4. Step 2: Create Your Model Class
        • 11.1.4.1. Option A: Spec-Based Model (Recommended)
        • 11.1.4.2. Option B: Custom PyTorch Model
        • 11.1.4.3. NPU-Compatible Model
      • 11.1.5. Available Layer Types
      • 11.1.6. Step 3: Add to __all__
      • 11.1.7. Step 4: Verify Your Model
      • 11.1.8. Step 5 (Optional): Add Device Performance Info
      • 11.1.9. Step 6 (Optional): Add GUI Model Description
      • 11.1.10. Using Your Custom Model
      • 11.1.11. Naming Conventions
      • 11.1.12. Configuring Model Layer Parameters
      • 11.1.13. Troubleshooting
      • 11.1.14. Summary Checklist
      • 11.1.15. Next Steps
    • 11.2. Compilation Only
      • 11.2.1. Overview
      • 11.2.2. ONNX Model Requirements
      • 11.2.3. Compilation Configuration
      • 11.2.4. Running Compilation
      • 11.2.5. Model Formats
      • 11.2.6. Output Artifacts
      • 11.2.7. NPU Compilation
      • 11.2.8. Example: External PyTorch Model
      • 11.2.9. Example: TensorFlow Model
      • 11.2.10. Troubleshooting
      • 11.2.11. Best Practices
      • 11.2.12. Next Steps
    • 11.3. Two Approaches
    • 11.4. Model Requirements
  • 12. Troubleshooting
    • 12.1. Common Errors
      • 12.1.1. Installation Errors
      • 12.1.2. Dataset Errors
      • 12.1.3. Training Errors
      • 12.1.4. Compilation Errors
      • 12.1.5. Quantization Errors
      • 12.1.6. Deployment Errors
      • 12.1.7. Configuration Errors
      • 12.1.8. Getting Help
    • 12.2. FAQ
      • 12.2.1. General Questions
      • 12.2.2. Installation Questions
      • 12.2.3. Dataset Questions
      • 12.2.4. Model Questions
      • 12.2.5. Training Questions
      • 12.2.6. Deployment Questions
      • 12.2.7. Advanced Questions
      • 12.2.8. Support Questions
      • 12.2.9. Still Have Questions?
    • 12.3. Quick Fixes
    • 12.4. Getting Help
  • 13. Appendix
    • 13.1. Configuration Reference
      • 13.1.1. Configuration File Structure
      • 13.1.2. Common Section
      • 13.1.3. Dataset Section
      • 13.1.4. Feature Extraction Section
      • 13.1.5. Training Section
      • 13.1.6. Testing Section
      • 13.1.7. NAS Options (under training section)
      • 13.1.8. Compilation Section
      • 13.1.9. BYOM Section
      • 13.1.10. Complete Example
    • 13.2. Model Zoo Reference
      • 13.2.1. Classification Models
        • 13.2.1.1. Standard Classification
        • 13.2.1.2. NPU Classification
        • 13.2.1.3. Application-Specific Classification (Edge AI Studio Only)
      • 13.2.2. Regression Models
        • 13.2.2.1. Standard Regression
        • 13.2.2.2. NPU Regression
      • 13.2.3. Anomaly Detection Models
        • 13.2.3.1. Standard AD
        • 13.2.3.2. NPU AD
      • 13.2.4. Forecasting Models
        • 13.2.4.1. Standard Forecasting
        • 13.2.4.2. NPU Forecasting
      • 13.2.5. Image Classification Models
      • 13.2.6. Model Selection Guide
      • 13.2.7. Model Architecture Details
      • 13.2.8. Using Models
    • 13.3. Changelog
      • 13.3.1. Version 1.3.0
      • 13.3.2. Version 1.1.0
      • 13.3.3. Version 1.0.0
      • 13.3.4. Migration Guides
        • 13.3.4.1. Migrating from 1.1.x to 1.2.x
      • 13.3.5. Deprecation Notices
      • 13.3.6. Known Issues
      • 13.3.7. Roadmap
      • 13.3.8. Contributing
    • 13.4. Quick Reference
    • 13.5. External Links
TI Tiny ML Tensorlab
  • Index

Index

A | B | C | F | G | I | L | M | N | O | P | Q | S | T | V

A

  • AM13
  • Autoencoder

B

  • BYOD
  • BYOM

C

  • C2000
  • C2000Ware
  • CCS

F

  • Feature Extraction
  • feature_extraction_name
  • frame_size

G

  • GoF Test

I

  • Inference

L

  • LaunchPad

M

  • MACs
  • model_name
  • ModelMaker
  • ModelZoo
  • MSPM0
  • MSPM33C

N

  • NAS
  • NNC
  • NPU

O

  • ONNX

P

  • PTQ

Q

  • QAT
  • Quantization

S

  • stride_size

T

  • target_device
  • task_type
  • TINPU
  • TinyVerse

V

  • variables

© Copyright 2026, Texas Instruments Incorporated. Last updated on Feb 26, 2026.