# gpumd-workflow

> End-to-end guidance for compiling, running, and debugging the GPUMD/nep executables. Use when handling GPU molecular dynamics with GPUMD, preparing model.xyz/run.in, training or validating NEP potentials (nep.in, train.xyz, test.xyz), analyzing outputs (thermo.out, loss.out, nep.txt), or resolving CUDA/build/runtime issues.

- Author: Siyu Liu
- Repository: IntelligentMat/Materials-Science-Skills-For-LLM
- Version: 20251221161123
- Stars: 0
- Forks: 0
- Last Updated: 2026-02-07
- Source: https://github.com/IntelligentMat/Materials-Science-Skills-For-LLM
- Web: https://mule.run/skillshub/@@IntelligentMat/Materials-Science-Skills-For-LLM~gpumd-workflow:20251221161123

---

---
name: gpumd-workflow
description: End-to-end guidance for compiling, running, and debugging the GPUMD/nep executables. Use when handling GPU molecular dynamics with GPUMD, preparing model.xyz/run.in, training or validating NEP potentials (nep.in, train.xyz, test.xyz), analyzing outputs (thermo.out, loss.out, nep.txt), or resolving CUDA/build/runtime issues.
---

# GPUMD Workflow

## Overview
- Build and operate the `gpumd` (molecular dynamics) and `nep` (NEP training/prediction) executables efficiently.
- Prepare the required input decks, interpret outputs, and apply quick fixes for common CUDA or configuration problems.

## Quick Start
1) Open `references/quickstart.md` for the build + first-run checklist.  
2) Choose a path: **MD simulation** → `references/gpumd-run.md`; **NEP training/prediction** → `references/nep-training.md`.  
3) Keep `references/outputs.md` nearby to map outputs; use `references/troubleshooting.md` if something fails.

## Task Playbook
- **Compile**: `cd src && make` (CUDA ≥9.0). Expect `gpumd` and `nep` in `src/`.
- **Run an MD case**: Place `model.xyz` and `run.in` together; run `path/to/gpumd`. Set `CUDA_VISIBLE_DEVICES` to pick GPUs; add `x|y|z` after `potential` to force partition axis.
- **Train a NEP model**: Ensure `nep.in`, `train.xyz` (and optionally `test.xyz`) are present; run `path/to/nep`. Monitor `loss.out`; final model is `nep.txt`.
- **Use a NEP in MD**: Copy `nep.txt` next to `run.in`; add `potential nep.txt`. Do a short stability test before production.
- **Restart / iterate**: Resume NEP via `nep.restart`; restart MD via `dump_restart` → `restart.xyz`. Adjust `time_step`/thermostat if energy drifts.

## References in This Skill
- `references/quickstart.md` — build + example run checklist.
- `references/gpumd-run.md` — `model.xyz`/`run.in` essentials and common keywords.
- `references/nep-training.md` — `nep.in` basics, dataset format, outputs to monitor.
- `references/outputs.md` — meaning of key output files for `gpumd` and `nep`.
- `references/troubleshooting.md` — quick fixes for build/runtime/accuracy issues.