Stan31/quantumflow-prototypes

license: mit tags: - nvidia - cuda - gpu - quantum - prototypes - benchmarks - h100 - cuquantum - tensor-cores datasets: - custom metrics: - gpu-utilization - tflops - memory-usage

QuantumFlow — GPU-Accelerated Prototypes Ecosystem

Model Summary

QuantumFlow is a collection of three GPU-first prototypes designed to run together in one environment, demonstrating reproducible GPU validation, cuQuantum acceleration, and ecosystem compatibility.

Achieving 95%+ GPU utilization on NVIDIA H100 with reproducible benchmarks.

Key Results (tested on NVIDIA H100 PCIe)

Component	What it proves	Key metric	Evidence
Team1 Quantum	Tensor Core-heavy screening workload (CPU fallback)	NVML GPU util avg 95.19%	prototypes/team1_quantum/benchmarks/results/latest.json
Team2 Energy	Differentiable thermo + grid optimization (CPU fallback)	NVML GPU util avg 95.44%	prototypes/team2_energy/benchmarks/results/latest.json
Team3 Innovation	cuQuantum contraction + sustained soak	NVML GPU util avg 95.47% + cuquantum_used=true	prototypes/team3_innovation/benchmarks/results/latest.json

Quick Start

Installation

# CPU (no CUDA required)
python -m pip install -U pip
python -m pip install -r prototypes/requirements.cpu.txt

# NVIDIA GPU (CUDA 12, without cuQuantum)
python -m pip install -U pip
python -m pip install -r prototypes/requirements.gpu-cu12.txt

# NVIDIA GPU (CUDA 12) + cuQuantum (Team3 acceleration)
python -m pip install -U pip
python -m pip install -r prototypes/requirements.gpu-cu12-cuquantum.txt

Run Ecosystem Smoke Test

python prototypes/ecosystem_smoke.py

Run Individual Prototypes

# Team1 Quantum
python prototypes/team1_quantum/demo/demo.py
python prototypes/team1_quantum/benchmarks/run_benchmarks.py

# Team2 Energy
python prototypes/team2_energy/demo/demo.py
python prototypes/team2_energy/benchmarks/run_benchmarks.py

# Team3 Innovation
python prototypes/team3_innovation/demo/demo.py
python prototypes/team3_innovation/benchmarks/run_benchmarks.py

NVIDIA Technologies Used

• CUDA 12.x (PyTorch CUDA)
• Tensor Cores (BF16/FP16 matmul soak)
• NVML (nvidia-ml-py) for utilization/memory metrics
• cuQuantum (Team3 only): cutensornet / tensornet contractions + custatevec

Links

• GitHub Repository: https://github.com/Corusant-world/quantumflow-prototypes
• Documentation: https://github.com/Corusant-world/quantumflow-prototypes/blob/main/README.md
• Docker Images: https://github.com/Corusant-world/quantumflow-prototypes/pkgs/container/quantumflow
• Release Notes: https://github.com/Corusant-world/quantumflow-prototypes/releases/tag/v0.1.1

Citation

@software{quantumflow2025,
  author = {Corusant-world},
  title = {QuantumFlow: GPU-Accelerated Prototypes Ecosystem},
  year = {2025},
  url = {https://github.com/Corusant-world/quantumflow-prototypes}
}
---