Skip to content

installation.md

Latest commit

 

History

History
106 lines (73 loc) · 2.45 KB

installation.md

File metadata and controls

106 lines (73 loc) · 2.45 KB

Installation Guide

This guide walks you through installing HPC-AI-Optimization-Lab on your system.

Prerequisites

Before installing, ensure your system meets the following requirements:

Hardware Requirements

  • GPU: NVIDIA CUDA-capable GPU (Compute Capability 7.0+)
  • Recommended: Volta (7.0), Ampere (8.0), or Hopper (9.0) architecture
  • Memory: 8GB+ GPU memory recommended for large kernels

Software Requirements

  • Operating System: Linux (Ubuntu 20.04+, CentOS 8+, or similar)
  • CUDA Toolkit: 12.4 or later
  • CMake: 3.24 or later
  • Compiler: GCC 11+ or Clang 14+
  • Python (optional): 3.9+ for Python bindings

Installation Methods

Method 1: Clone and Build (Recommended)

# Clone the repository
git clone https://github.com/AICL-Lab/hpc-ai-optimization-lab.git
cd hpc-ai-optimization-lab

# Create build directory
mkdir build && cd build

# Configure with CMake
cmake .. -DCMAKE_BUILD_TYPE=Release

# Build the project
cmake --build . -j$(nproc)

# Run tests (optional)
ctest --output-on-failure

Method 2: Build with Examples

# Configure with examples enabled
cmake .. \
  -DCMAKE_BUILD_TYPE=Release \
  -DBUILD_EXAMPLES=ON

# Build
cmake --build . -j$(nproc)

# Run an example
./bin/examples/relu_example

Method 3: Build with Python Bindings

# Install Python dependencies first
pip install nanobind numpy

# Configure with Python bindings
cmake .. \
  -DCMAKE_BUILD_TYPE=Release \
  -DBUILD_PYTHON_BINDINGS=ON \
  -DPython_EXECUTABLE=$(which python3)

# Build
cmake --build . -j$(nproc)

Verifying Installation

After building, verify your installation:

# Run all tests
ctest --output-on-failure

# Run a specific test
./tests/elementwise/test_relu

# Check Python binding (if enabled)
python3 -c "import hpc_ai_opt; print(hpc_ai_opt.__version__)"

Troubleshooting

If you encounter issues:

  1. CUDA not found: Ensure CUDA_HOME or CUDA_PATH is set
  2. CMake version too old: Install newer CMake from https://cmake.org/download/
  3. Compilation errors: Verify GCC 11+ or Clang 14+ is installed
  4. Test failures: Check GPU availability and CUDA driver version

For more help, see the Troubleshooting Guide.

Next Steps