SLURM on Rachid Youven Zeghlache

Apptainer & Conda: Reproducible HPC Python Environments

Thu, 05 Mar 2026 00:00:00 +0000

Build Reproducible Python Environments with Apptainer & Conda — For HPC Researchers

Your Python script runs perfectly on your laptop. You transfer it to the cluster. Runtime fails. Missing dependencies. Version conflicts. Different Python builds. You spend hours debugging environment mismatches instead of running science.

This is the researcher’s tax: environment reproduction across machines is broken by default.

What This Solves

Apptainer (formerly Singularity) is a containerization tool built for HPC clusters. Unlike Docker, it doesn’t require root privileges and integrates seamlessly with cluster job schedulers like SLURM and PBS. Combined with Conda, it lets you package a complete, reproducible Python environment—dependencies, versions, and all—into a single .sif file that runs identically on your laptop, your colleague’s machine, and any HPC node.

SLURM Job Scheduling: Submit & Monitor HPC Jobs

Thu, 05 Mar 2026 00:00:00 +0000

Submit & Monitor HPC Jobs Using SLURM — Without Losing Your Data to Path Errors

You’ve just logged into your university’s HPC cluster for the first time. You write what you think is a perfect job script, submit it with sbatch, and 10 minutes later you find a cryptic error in the log file: “file does not exist.” Your absolute path was wrong. Or you forgot to specify GPU resources and your job waited in the queue for 3 hours doing nothing.

SSH & SLURM: Transfer Code to HPC Clusters

Thu, 05 Mar 2026 00:00:00 +0000

Transfer Code to HPC Clusters Using SSH, Singularity/Apptainer & SLURM

You’ve written working code on your laptop. Now you need to run it on your institution’s HPC cluster with 100+ GPUs—but you’re staring at a terminal with no idea how to get your files there, containerize your environment, or submit a job without breaking it.

This is the moment most researchers feel lost.

The HPC Stack in 90 Seconds

High-performance computing (HPC) clusters are shared servers with massive CPU and GPU resources. To use them safely and reproducibly, you need three tools working together: