القائمة
Evaluate MatterGen-generated crystal structures for novelty, uniqueness, stability, and S.U.N. criteria, and produce filtered relaxed structures and evaluation reports.
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
استنادا إلى تصنيف SOC المهني
| name | mattergen-evaluation |
| description | Evaluate MatterGen-generated crystal structures for novelty, uniqueness, stability, and S.U.N. criteria, and produce filtered relaxed structures and evaluation reports. |
| tags | ["MatterGen","evaluation"] |
| tools | ["run_bash"] |
| dependent_skills | [] |
Use this skill to evaluate structures generated by a MatterGen model, compute quality metrics such as novelty, uniqueness, stability, and S.U.N.-related results, and filter the relaxed structures that satisfy the S.U.N. criteria when detailed evaluation outputs are available.
The official MatterGen README also documents mattergen-evaluate, which can:
Reference:
Important:
First download the reference dataset, then evaluate the generated structures:
git lfs pull -I data-release/alex-mp/reference_MP2020correction.gz --exclude=""
ts=$(date +"%Y%m%d%H%M%S")
workdir="matg/${ts}.mattergen_evaluation"
mkdir -p "$workdir"
cd "$workdir"
"${MATTERGEN_ENV}/bin/mattergen-evaluate" \
--structures_path=/abs/path/to/generated_results \
--save_as=/abs/path/to/generated_results/metrics.json \
--save_detailed_as=/abs/path/to/generated_results/detailed_metrics.json \
--structures_output_path=/abs/path/to/generated_results/relaxed_structures.extxyz
When submitting MatterGen jobs to Bohrium through the bohrium skill, Bohrium-specific submission settings can be read from environment variables such as BOHRIUM_MAT_IMAGE and BOHRIUM_MAT_MACHINE.
Prepare a writable local job directory first:
ts=$(date +"%Y%m%d%H%M%S")
workdir="matg/${ts}.mattergen_evaluation"
mkdir -p "$workdir"
cd "$workdir"
Then copy the reference dataset into the workdir, using the local paths from the environment variables eval_reference, unless the user explicitly provides different paths.
Then copy the structures to be evaluated into the workdir, preferably in extxyz format.
In submission.json:
command to a generation command such as:mkdir -p results && mattergen-evaluate --structures_path=generated_results --reference_dataset_path=dataset.gz --save_as=results/metrics.json --save_detailed_as=results/detailed_metrics.json --structures_output_path=results/relaxed_structures.extxyz
structures_path and reference_dataset_path must point to the actual input files you are using, with their real filenames, for example generated.extxyz and MP_2020.gz.
forward_files should include the reference dataset, structures, such as generated.extxyz, MP_2020.gz.
backward_files can be results.
After evaluation, when both relaxed_structures.extxyz and detailed_metrics.json are available, filter the relaxed structures using the novel_unique_stable list in detailed_metrics.json.
Filtering procedure:
novel_unique_stable from detailed_metrics.json.true entry in novel_unique_stable as one structure that satisfies the S.U.N. criteria.novel_unique_stable to the frame index in relaxed_structures.extxyz.novel_unique_stable value is true./abs/path/to/generated_results/sun_structures.extxyz.extxyz file and the number of retained structures.After evaluation, report at minimum:
detailed_metrics.json, if generated,relaxed_structures.extxyz, if generated,sun_structures.extxyz, if generated,metrics.json, especially stability, uniqueness, novelty, and S.U.N.-related results when present.Generate VASP inputs (INCAR/POSCAR/POTCAR/KPOINTS) via pymatgen.io.vasp.sets for DFT calculations: MPRelaxSet (geometry relaxation), MPStaticSet (SCF), MPNonSCFSet (band structure), MatPESStaticSet (MLFF energy/force labeling). INCAR is driven by pymatgen defaults; the agent only supplies user_incar_settings overrides. Use when the user asks to prepare VASP calculations, run DFT, or generate MLFF training data from structures. Do NOT use for VASP result post-processing / analysis, non-VASP DFT codes, or molecular dynamics — this skill only generates input files and submits jobs.
Deep potential models finetuning and testing using the DeePMD-kit. Use this skill whenever finetuning a Deep Potential (DPA-1 / DPA-2 / DPA-3 / DPA-4) model or running model tests on a dataset. The oldest DP descriptors such as se_e2_a, se_e2_r, and se_e3 are no longer supported. Training from scratch is NEVER advised unless distilling a student model from a teacher model. Multitask fine-tuning is NOT supported.
Skill for running ASE calculations, including energy/force/stress evaluation, molecular dynamics (MD) and structure optimization. Now only use machine-learned force fields (MLFFs) as calculators.
Concept skill for Machine Learning Force Fields (MLFFs). Describes what MLFFs are, the distinction between fine-tuning and distillation, and which tool skills to use. Load this before selecting a specific MLFF framework (DeePMD, MatterSim, etc.).
Concept skill for Molecular Dynamics (MD) simulation. Describes MD ensembles, key parameters, and which tool skills to invoke for MD runs. Use this to understand MD concepts before selecting a specific simulation tool.
Equation-of-State (EOS) benchmark skill — compare DFT, pretrained, and finetuned model E(V) curves to evaluate force-field quality for bulk crystals and simple systems.