Installationο
System Requirementsο
Before proceeding, ensure your system meets these hardware and software specifications:
Operating System: Linux or Windows Subsystem for Linux (WSL2).
Hardware: An NVIDIA GPU is required for model training and inference.
Driver: NVIDIA Driver supporting CUDA 12.0 or higher.
Step 1: Install Dependenciesο
Lightning Pose requires specific system-level tools to handle video processing and GPU acceleration.
FFmpegο
Lightning Pose uses FFmpeg for high-performance video decoding.
# Ubuntu/WSL
sudo apt update && sudo apt install ffmpeg
# Verify installation
ffmpeg -version
NVIDIA GPU Setupο
Ensure your GPU drivers are correctly installed and recognized by the system.
nvidia-smi
The output should display a table showing your GPU model, Driver Version, and CUDA Version.
Step 2: Create Conda Environmentο
We recommend using Conda (or another python environment management tool) to create an isolated python environment.
If you donβt already have conda installed, download and install it from here.
# Create a new environment named 'lp' with Python 3.12
conda create -n lp python=3.12
# Activate the environment
conda activate lp
Tip
Check your environment: Always ensure your terminal prompt is prefixed with (lp) before running installation or training commands. This confirms you are working within the isolated environment and not your systemβs global Python.
Step 3: Install Lightning Poseο
Choose the method that best fits your needs.
Option A: Standard Installation (Standard)ο
Use this method if you donβt know which method to use.
pip install lightning-pose lightning-pose-app
Option B: Installation from Source (Development)ο
Use this if you plan to modify the source code. This requires cloning the repositories for an editable install.
If you want to edit the core training/inference logic:
# Clone the core repo and install
git clone https://github.com/paninski-lab/lightning-pose.git
cd lightning-pose
pip install -e ".[dev]"
cd ..
If you want to edit the app/UI code:
# Clone the app repo and install
git clone https://github.com/paninski-lab/lightning-pose-app.git
cd lightning-pose-app/app_server
pip install -e .
cd ..
# Install node.js via nvm to be able to compile and run the web UI.
# Follow the directions here: https://github.com/nvm-sh/nvm
# Then:
npm install -g @angular/cli
cd web_ui
npm install
cd ..
honcho -f Procfile.dev
Note
In case of a PyTorch installation issue in either Option A or B, You may need to install the PyTorch matching your CUDA version by following the PyTorch Installation Guide.
How to Choose?ο
Goal |
Method |
Benefit |
|---|---|---|
Standard Usage |
PyPI (Pip) |
Simple setup; stable release versions. |
Development/Research |
Source (Git) |
Access to unreleased features; ability to edit code. |
You can migrate from one to the other, just pip uninstall and then re-install using the new method.
Verificationο
Verify the installation by checking the Command Line Interface (CLI):
litpose --help
If successful, you will see a list of available commands for the Lightning Pose suite.