jinzhezenggroup/computational-chemistry-agent-skills/dpgen-simplify

DP-GEN Simplify

Use this skill when the user wants to prepare, explain, validate, or execute the dpgen simplify workflow.

This skill is for dataset simplification workflows where the user already has candidate data in DeepMD-compatible format and wants to reduce repeated or redundant structures through iterative selection.

Core Rule (Critical)

DP-GEN simplify always uses two parameter classes and therefore two JSON files:

• Workflow parameters -> param.json
• Execution / machine parameters -> machine.json

Run exactly:

dpgen simplify param.json machine.json

Environment boundary rule:

• Outer layer: run dpgen simplify param.json machine.json in an activated environment where dpgen -h works.
• Inner layer: for scheduler stages, explicitly activate runtime in resources.source_list on the server side.

Agent responsibilities

When using this skill, the agent should:

1. confirm that the task is a simplify workflow
2. check whether existing configs or templates are already available
3. collect only the missing dataset, training, FP, and machine inputs
4. generate or patch param.json
5. generate or patch machine.json
6. explain important simplify parameters in plain language when asked
7. validate the workflow before execution
8. provide the exact command for running simplify
9. if requested, help structure repeated experiments
10. after execution, summarize outputs and next inspection targets

Working policy

1. Ask only for missing inputs

Do not ask the user for everything if part of the configuration is already available.

If the user already provides:

• a partial param.json
• a partial machine.json
• a known training template
• a known cluster template

then patch those files instead of rebuilding everything from scratch.

2. Preserve the user's scientific choices

Do not silently change:

• descriptor family
• fitting net structure
• fp backend
• trust thresholds
• type_map ordering

If a value looks scientifically questionable, explain the concern instead of silently replacing it.

3. Keep local and scheduler execution explicit

If the user wants local execution, produce local-friendly commands.

If the user wants scheduler execution, produce scheduler-friendly commands and keep queue, partition, and resource requests explicit.

Do not invent scheduler module names or executable paths.

4. Do not invent environment activation commands

If the user already has a working activation command such as:

• conda activate ...
• module load ...
• source ...

reuse it exactly.

If execution is requested and the activation method is unknown, ask the user for the precise activation command.

Do not guess conda environment names, module names, or site-specific paths.

4.1 Outer launcher policy

Use an activated DP-GEN environment and verify with:

dpgen -h

Do not start simplify from a shell where dpgen is unavailable.

4.2 Outer vs inner runtime boundaries (critical)

Treat simplify execution as two separate environment layers:

1. Outer layer: the shell that launches dpgen simplify param.json machine.json (must have dpgen in PATH)
2. Inner layer: stage tasks dispatched by DP-GEN (train / model_devi / fp) on server/runtime side

Even if the outer layer is correct, inner stage tasks still need explicit runtime setup in machine.json. Do not assume the outer shell environment will be inherited by dispatched stage jobs. For scheduler-style execution, resources.source_list must explicitly activate the required runtime environment.

5. Prefer reproducible output layout

When generating a simplify workflow, keep files organized and predictable.

Recommended structure:

project/
├── param.json
├── machine.json
├── run.sh
├── logs/
└── summary/

For repeated experiments:

project/
├── base/
├── exp_01/
├── exp_02/
├── exp_03/
└── summary/

Minimum required inputs

Collect the following information before generating files.

Dataset information

• pick_data
• sys_configs
• init_data_prefix
• init_data_sys
• sys_batch_size
• dataset format
• type_map
• mass_map if needed
• labeled

Simplify controls

• init_pick_number
• iter_pick_number
• model_devi_f_trust_lo
• model_devi_f_trust_hi
• model_devi_e_trust_lo / model_devi_e_trust_hi if energy trust is used
• numb_models if not already specified

Training setup

• train_backend if required by environment (for example pytorch)
• default_training_param
• descriptor settings
• fitting network settings
• learning rate settings
• loss settings
• training step settings

FP setup

• fp_style
• If data is already labeled (energy/force/virial available) and no re-labeling is requested, set fp_style to none.
• if fp_style != "none", collect matching FP runtime settings such as:
• fp_task_max
• fp_task_min
• fp_params
• pseudopotential or backend file paths if required

Execution setup

For each stage train, model_devi, and fp, collect or preserve:

• command
• machine.batch_type
• machine.context_type
• machine.local_root
• machine.remote_root
• resources.batch_type (use the same backend as machine.batch_type)
• resources.number_node
• resources.cpu_per_node
• resources.gpu_per_node
• resources.group_size
• resources.source_list (required for scheduler jobs; use it to activate environment explicitly)
• any explicit queue / partition / custom scheduler flags if the user already uses them

Choose a runtime profile first, then fill the matching template:

• server-local Slurm: assets/machine.template.server-local-slurm.json
• local machine -> remote Slurm via SSH: assets/machine.template.ssh-remote-slurm.json
• pure local shell testing: assets/machine.template.local-shell.json

How to build param.json

Construct param.json around these logical blocks:

1. element and mass definitions
2. data source and batch settings
3. model ensemble count
4. default DeePMD training parameters
5. FP backend settings
6. simplify pick settings
7. trust thresholds

Key fields usually include:

• type_map
• mass_map
• pick_data
• init_data_prefix
• init_data_sys
• sys_batch_size
• numb_models
• default_training_param
• fp_style
• fp_task_max
• fp_task_min
• fp_params
• backend support fields such as fp_pp_path and fp_pp_files only when the

selected FP schema requires them

• init_pick_number
• iter_pick_number
• model_devi_f_trust_lo
• model_devi_f_trust_hi

If the user is doing grid experiments, keep a base template and derive variants from it.

Official reference example (QM7-style, adapted with path placeholders):

• assets/param.example.qm7.from-official-docs.json

How to build machine.json

Construct machine.json with separate stage blocks for:

• train
• model_devi
• fp

For each stage, keep the following explicit:

• command
• machine or context configuration
• resources, including the same batch_type used by the machine block
• queue or partition if needed
• cpu and gpu counts
• custom scheduler flags
• environment activation commands

Do not merge all stages into one vague machine block.

Validation before run

Before execution, validate the workflow in this order:

1. confirm outer-layer dpgen is available:

dpgen -h

1. validate JSON syntax:

python -m json.tool param.json
python -m json.tool machine.json

1. verify required dataset paths exist
2. verify stage commands match the selected software stack
3. if fp_style is none, do not require FP-specific backend settings
4. only then run:

dpgen simplify param.json machine.json

Output contract

Always provide:

1. final absolute paths to param.json and machine.json
2. the exact simplify command to run (dpgen simplify param.json machine.json)
3. a short pre-run checklist
4. any unresolved required fields
5. if execution was performed, the main output locations and next files to inspect

Guardrails

• Never merge workflow and machine parameters into one file.
• Never run dpgen simplify before both JSON files are present.
• Never hardcode personal cluster, account, queue, or path settings as universal defaults.
• Never silently change the user's scientific choices.
• Keep type_map ordering consistent with dataset typing.
• If required inputs are missing, stop and ask instead of guessing.
• If fp_style is none, skip FP-specific prompts and keep FP-specific settings disabled or unset.
• The machine schema still requires an fp stage and a string command. Use the

no-op command true when fp_style is none.

• If data is already labeled and the user does not request new labels, enforce fp_style = "none" and do not require active FP runtime fields.
• Do not assume outer-shell activation is inherited by stage jobs; for scheduler execution, require explicit source_list per stage.
• If the user already has working templates, patch them rather than overwriting them blindly.

References and bundled files

Use these bundled files:

• assets/param.template.json
• assets/param.example.qm7.from-official-docs.json
• assets/machine.template.json
• assets/machine.template.server-local-slurm.json
• assets/machine.template.ssh-remote-slurm.json
• assets/machine.template.local-shell.json
• references/param-fields.md
• references/machine-fields.md
• references/workflow-notes.md

External references:

• DP-GEN simplify overview: https://docs.deepmodeling.com/projects/dpgen/en/latest/simplify/simplify.html
• simplify parameter definitions: https://docs.deepmodeling.com/projects/dpgen/en/latest/simplify/simplify-jdata.html
• simplify machine definitions: https://docs.deepmodeling.com/projects/dpgen/en/latest/simplify/simplify-mdata.html