HPC Lab — Apptainer Container Environment

Hands-on HPC portfolio project: building and running GPU-accelerated containers on Rocky Linux 9.

What's in this repo

• hpc_pytorch.def — Apptainer definition file for a custom PyTorch container
• matrix_benchmark.py — CPU/GPU matrix multiply benchmark (GFLOPS output)
• docs/as-built.md — As-built documentation and performance results

Environment

• OS: Rocky Linux 9.8 (RHEL-compatible)
• Container runtime: Apptainer 1.5.2
• Base image: pytorch/pytorch:2.3.0-cuda11.8-cudnn8-runtime
• Added packages: numpy, scipy, matplotlib, pandas

Benchmark Results (CPU baseline)

Matrix Size	Time	GFLOPS
1024x1024	45.9 ms	46.8
2048x2048	225.3 ms	76.3
4096x4096	1669.8 ms	82.3

GPU node with NVIDIA A100 expected to deliver 300,000+ GFLOPS.