wedgebox-hpc 1.0.0

GPU Performance Monitor for HPC Clusters

WedgeBox-HPC

Simple GPU Performance Monitoring for HPC Clusters

Overview

WedgeBox-HPC provides detailed GPU telemetry for debugging and optimizing machine learning workloads on HPC infrastructure.

Features:

• Multi-vendor GPU support (NVIDIA, AMD)
• Process-level GPU resource tracking
• Zero-configuration monitoring
• Human-readable console output + detailed JSON reports

Quick Start

# Download
wget https://raw.githubusercontent.com/wedgebox/wedgebox-hpc/main/wedgebox-hpc
chmod +x wedgebox-hpc

# Run
./wedgebox-hpc

Requirements

• Python 3.8+
• psutil (pip install psutil --user)
• GPU drivers (nvidia-smi for NVIDIA, rocm-smi for AMD)

For HPC Administrators

See docs/installation.md for cluster-wide deployment instructions.

Output

Generates:

• Console summary (human-readable)
• wedgebox-report-[timestamp].json (machine-readable)

Use Cases

• Debug slow training jobs
• Monitor GPU utilization
• Profile memory usage
• Identify resource bottlenecks
• Track multi-GPU workload distribution

License

MIT License - See LICENSE file