Installation Guide

Installation Guide#

This guide covers how to install Iris on your system using various methods.

Overview#

Iris has minimal dependencies including Python, PyTorch, ROCm HIP runtime, MPI, and Triton. This guide will walk you through the installation process using different approaches.

Prerequisites#

System Requirements#

Linux operating system (Ubuntu 22.04+)
AMD GPU with ROCm 6.3.1+ support (MI300X, MI350X, MI355X, or other ROCm-compatible GPUs)
Python 3.10+
ROCm 6.3.1+ HIP runtime

Required Software#

Minimum working requirements based on the Docker setup:

Python 3.10+
PyTorch 2.0+ (ROCm version)
ROCm 6.3.1+ HIP runtime
OpenMPI
Git
CMake, Ninja, build-essential
Triton (specific commit: dd5823453bcc7973eabadb65f9d827c43281c434)

Note: These versions represent the minimum working configuration. Using different versions may cause compatibility issues.

Installation Methods#

1. Using Docker Compose (Recommended)#

The easiest way to get started is using Docker Compose:

# Clone the repository
git clone https://github.com/ROCm/iris.git
cd iris

# Start the development container
docker compose up --build -d

# Attach to the running container
docker attach iris-dev

# Install Iris in development mode
cd iris && pip install -e .

2. Manual Docker Setup#

If you prefer to build and run Docker containers manually:

# Build the Docker image
./docker/build.sh

# Run the container
./docker/run.sh

# Install Iris in development mode
pip install -e .

3. Apptainer/Singularity#

For HPC environments or systems where Docker is not available:

# Build the Apptainer image
./apptainer/build.sh

# Run the container
./apptainer/run.sh

# Install Iris in development mode
pip install -e .

Next Steps#

Once you have Iris running with any of these methods:

Explore the Examples directory
Learn about the Programming Model