2.1. Prerequisites

Before installing Tiny ML Tensorlab, ensure your system meets these requirements.

2.1.1. System Requirements

Operating System

Linux	Ubuntu 20.04 or later (recommended)
Windows	Windows 10/11 with WSL2 or native support
macOS	macOS 12 or later (limited testing)

Hardware

CPU	Modern multi-core processor (Intel/AMD x86_64 or Apple Silicon)
RAM	Minimum 8 GB (16 GB recommended)
Storage	10 GB free space for installation and datasets
GPU	Optional but recommended for Neural Architecture Search (NAS)

2.1.2. Software Requirements

Python 3.10

Tiny ML Tensorlab requires Python 3.10.x specifically. Other versions are not supported.

LinuxWindows

We recommend using pyenv:

# Install pyenv
curl https://pyenv.run | bash

# Add to ~/.bashrc
export PYENV_ROOT="$HOME/.pyenv"
export PATH="$PYENV_ROOT/bin:$PATH"
eval "$(pyenv init -)"

# Install Python 3.10
pyenv install 3.10.14
pyenv global 3.10.14

# Verify
python --version  # Should show Python 3.10.14

Git

Git is required to clone the repository.

LinuxWindows

sudo apt-get install git

pip

pip should be included with Python. Ensure it’s up to date:

python -m pip install --upgrade pip

2.1.3. For Compilation (Optional)

To compile models for TI MCUs, you’ll need the TI toolchain. These are only required if you want to compile models for device deployment.

For C2000 Devices

• TI C2000 Code Generation Tools (CGT)

• C2000Ware SDK

Download from:

• CGT: https://www.ti.com/tool/C2000-CGT

• C2000Ware: https://www.ti.com/tool/C2000WARE

For MSPM0 Devices

• TI Arm Clang Compiler

• MSPM0 SDK

Download from:

• Arm Clang: https://www.ti.com/tool/ARM-CGT

• MSPM0 SDK: https://www.ti.com/tool/MSPM0-SDK

2.1.4. For Device Deployment (Optional)

To flash and debug models on hardware:

Code Composer Studio (CCS)

TI’s IDE for MCU development. Download from: https://www.ti.com/tool/CCSTUDIO

Recommended version: 20.x or later

TI LaunchPad or EVM

A development board for your target device.

2.1.5. CUDA (Optional)

For Neural Architecture Search (NAS), a CUDA-capable GPU significantly speeds up training:

• NVIDIA GPU with CUDA support

• CUDA Toolkit 11.x or 12.x

• cuDNN library

NAS can run on CPU but will be much slower.

2.1.6. Verification Checklist

Before proceeding with installation, verify:

# Python version (must be 3.10.x)
python --version

# pip is available
pip --version

# Git is available
git --version

If all checks pass, proceed to User Installation or Developer Installation.