HPC execution

Start with: prepared and checked simulation inputs. This workflow handles planning and execution; it does not replace topology preparation or validate equilibration automatically.

1. Review the configuration

Download the existing workflow YAML and SLURM example. Work from the repository root, and edit:

Target monomers and repeat counts; use small chains first.
Which stages are enabled (build, gaff2, openmm).
Charge model and engine settings.
Cluster partition, time, CPUs, memory and Conda environment.

Unlike the contact-analysis YAML, the configured workflow’s output paths are used relative to the process working directory. Launch it from the repository root.

2. Print a plan and generate a script

From the active installed environment:

python examples/run_configured_workflow.py --config examples/hpc_validation_workflow.yaml --dry-run
python examples/run_configured_workflow.py --config examples/hpc_validation_workflow.yaml --write-slurm examples/output/run_validation.slurm

Read the generated script and adapt the environment block to your cluster. The workflow plan does not submit a job. When the inputs, allocations and script are ready, submission is a separate user action:

sbatch examples/output/run_validation.slurm

3. Monitor and restart

Keep checkpoints, topology, coordinates, parameter files, configuration and logs together. Review stage completion and numerical behavior before continuation. For GROMACS restarts, use the appropriate checkpoint with mdrun -cpi; for OpenMM, use the restart format supported by the chosen runner. Benchmark a small segment on the intended hardware before allocating a long production run.

The 07 notebook covers GROMACS SLURM execution and restart guidance; notebook 10 is a six-system benchmark launcher with fixed workflow assumptions. Batch benchmarks are separate from the current single-system enzyme-contact analysis.

API: iphasimulator.workflows.hpc.workflow_plan(), iphasimulator.workflows.hpc.render_slurm_script().

Prepared 06A/06B/06C folders

The configuration runner above is an additional interface, separate from 07’s prepared-folder hand-off. 06A/06B need local step6.0_minimization.gro before their generated NVT/NPT/production script. 06C needs step6.0_minimization.xml, its original PRMTOP/INPCRD and protocol.json; its standalone runner resumes each MD stage from a compatible checkpoint and takes no additional steps in completed stages. Review output/checkpoint consistency after interruption and keep physical inputs unchanged. See the 06C restart guide.