installation.md

Installation Guide

This guide covers the installation and setup of InfiniMetrics.

Prerequisites

• Python: 3.8 or higher
• Compiler: GCC 11.3
• CMake: 3.20+

Dependencies Installation

Core Python Dependencies

Install the core dependencies based on your needs:

# For InfiniLM adapter
pip install numpy torch

# For vLLM adapter
pip install vllm

# For data processing
pip install pandas

Build Hardware Benchmarks (Optional)

If you plan to use the hardware testing modules, you need to build the CUDA memory benchmark suite:

cd infinimetrics/hardware/cuda-memory-benchmark
bash build.sh

Note: This requires:

• CUDA toolkit (compatible with your GPU driver)
• C++ compiler with CUDA support
• CMake 3.20 or higher

Verification

After installation, verify your setup:

# Run a simple hardware test
python main.py format_input_comprehensive_hardware.json

Platform-Specific Notes

NVIDIA GPUs

Ensure CUDA_HOME is set:

export CUDA_HOME=/usr/local/cuda
export PATH=$CUDA_HOME/bin:$PATH
export LD_LIBRARY_PATH=$CUDA_HOME/lib64:$LD_LIBRARY_PATH

Other Accelerators

Refer to your accelerator vendor's documentation for driver and toolkit installation.