--- name: lammps description: > LAMMPS molecular dynamics with DeepMD/DPA machine-learning potentials ONLY. This skill generates LAMMPS input files (in.lammps) using pair_style deepmd. Classical force fields (Stillinger-Weber, Tersoff, EAM, etc.) are NOT supported and must NEVER be suggested. The default model is the DPA3 pretrained checkpoint frozen with the Omat24 head (from DEEPMD_MODEL_PATH env var), identical to the ase-deepmd skill. A model file is always required — do NOT ask the user whether they want to use a classical potential instead. Bohrium's default DeepMD image (registry.dp.tech/dptech/deepmd-kit) already includes LAMMPS — no separate LAMMPS installation is needed. Covers structure conversion, input generation, Bohrium submission via bohrium skill, and result collection. metadata: tools: - run_bash - run_python_file - load_skill_resource dependent_skills: - bohrium - dpdisp tags: - lammps - deepmd - dpa - dpa3 - deep-potential - machine-learning-potential - md - molecular-dynamics - npt - nve - bohrium --- # LAMMPS / DeepMD Skill ## CRITICAL CONSTRAINTS — READ FIRST 1. **This skill ONLY supports `pair_style deepmd`.** Do NOT suggest, mention, or offer classical force fields (Stillinger-Weber, Tersoff, EAM, ReaxFF, etc.) as alternatives. There is no "choice" — DeepMD is the only option. 2. **A model is always available.** The environment variable `DEEPMD_MODEL_PATH` points to a pretrained DPA3 model. The tool automatically freezes it with the Omat24 head (`dp --pt freeze --head Omat24`). You do NOT need to ask the user for a model file — just run `generate_input` and it will resolve the model automatically. Only ask if `DEEPMD_MODEL_PATH` is unset AND the user did not pass `--model_path`. 3. **LAMMPS is already installed in the Bohrium DeepMD image.** The default image `registry.dp.tech/dptech/deepmd-kit` provides both LAMMPS and DeePMD-kit. Do NOT ask whether LAMMPS is installed — it is. Use `BOHRIUM_DEEPMD_IMAGE` (the same env var as the `deepmd` skill) as the container image. 4. **Never present this as a plan to be confirmed.** If the user asks for a LAMMPS MD simulation, execute immediately: generate input → submit → collect. --- ## How it works | Component | Role | |---|---| | `lammps_tools.py` | Generate `in.lammps` + `conf.lmp`; collect results | | `bohrium` skill | Submit job directories to Bohrium (recommended) | | `dpdisp` skill | Submit to Slurm/HPC clusters (alternative) | Every command prints JSON to stdout and exits 0 on success, 1 on error. --- ## Mandatory workflow sequence 1. **Obtain a structure** — supply an extxyz, POSCAR, CIF, or any ASE-readable file. If the user has no structure, generate one first using the `atomic-structure` skill. 2. **Prepare job directory** — run `lammps_tools.py generate_input`. The model is resolved automatically from `DEEPMD_MODEL_PATH` and frozen with Omat24 head. 3. **Submit jobs** — use the `bohrium` skill for Bohrium platform submission. For full submission details, see: ``` load_skill_resource(skill_name="lammps", path="references/bohrium-submission.md") ``` 4. **Collect results** — run `collect_results` after jobs finish. --- ## 1. Generate LAMMPS input ### Simplest usage (model resolved from DEEPMD_MODEL_PATH automatically) ```bash python lammps_tools.py generate_input --structures POSCAR ``` This uses all defaults: DPA3 model, Omat24 head, NPT, 300 K, 1 bar, 1M steps. ### NVE ensemble at 600 K ```bash python lammps_tools.py generate_input \ --structures structures.extxyz \ --ensemble nve --temperature 600 --runtime_steps 500000 ``` ### NPT at zero pressure ```bash python lammps_tools.py generate_input \ --structures POSCAR \ --temperature 300 --pressure 0.0 ``` ### All frames from a multi-frame file (one job per frame) ```bash python lammps_tools.py generate_input \ --structures traj.extxyz --frame -1 ``` **Key flags** | Flag | Default | Description | |---|---|---| | `--structures` | required | Any ASE-readable structure file (extxyz, POSCAR, CIF, …) | | `--model_path` | `DEEPMD_MODEL_PATH` | DeePMD model file. Auto-resolved from env var — usually不需要指定. | | `--head` | `Omat24` | Multi-task head to freeze. Pass `none` to skip freezing. | | `--frame` | 0 | Frame index. `-1` = all frames. | | `--ensemble` | `npt` | `npt` or `nve` | | `--temperature` | 300.0 | Target temperature (K) | | `--pressure` | 1 | Target pressure in bar (NPT only). LAMMPS units metal uses bar. | | `--timestep` | 0.001 | Timestep in ps | | `--runtime_steps` | 1000000 | Number of MD steps | | `--dump_interval` | 10000 | Trajectory write frequency (steps) | | `--thermo_interval` | 100 | Thermo output frequency (steps) | ### Model handling (same as ase-deepmd) The pretrained DPA3 model (from `DEEPMD_MODEL_PATH`) is frozen with `dp --pt freeze --head Omat24` before being copied into the job directory. The frozen single-task model is used as `pair_style deepmd frozen_model.pth`. Pass `--head none` to skip freezing. ### Generated in.lammps example ```lammps units metal boundary p p p atom_style atomic neighbor 1.0 bin read_data conf.lmp pair_style deepmd frozen_model.pth pair_coeff * * Si # elements sorted alphabetically velocity all create 300.0 23456789 timestep 0.001 thermo 100 thermo_style custom step pe ke etotal temp press vol fix 1 all npt temp 300.0 300.0 0.1 aniso 1.0 1.0 0.5 # pressure in bar dump 1 all custom 10000 traj.dump id type xu yu zu dump_modify 1 sort id run 1000000 undump 1 unfix 1 ``` --- ## 2. Collect results ```bash python lammps_tools.py collect_results \ --calc_dirs /tmp/lammps_jobs/lammps_npt_001 /tmp/lammps_jobs/lammps_npt_002 ``` Returns per-job summary: average temperature, pressure, potential energy, volume. --- ## 3. Submit to Bohrium For Bohrium platform submission, use the `bohrium` skill. Full submission details including environment variables, JSON template, and submission commands are in: ``` load_skill_resource(skill_name="lammps", path="references/bohrium-submission.md") ``` --- ## config.yaml ```yaml work_dir: lammps ```