Train DeePMD-kit machine learning potentials with DPA-3 and se_e2_a descriptors.
npx skills add https://github.com/Hello-QM/catgo-LRG --skill deepmd-train
Please help me install this Agent Skill. Skill: deepmd-train Source: https://github.com/Hello-QM/catgo-LRG/tree/main/.claude/skills/deepmd-train Command: npx skills add https://github.com/Hello-QM/catgo-LRG --skill deepmd-train
Train DeePMD-kit machine learning potentials across DPA-3, se_e2_a, and fine-tuning from pretrained models.
Provide your dpdata-formatted training dataset and execute the training workflow to produce a production-ready DeePMD model.
High-intent search queries and answers about installing and using this skill.
Train DeePMD-kit machine learning potentials by deploying a production workflow on GPU-enabled HPC environments. The workflow processes dpdata-formatted datasets and manages end-to-end training, from data preparation to final model freezing.
Fine-tuning pretrained DeePMD models is fully supported for materials science simulations. You can apply pretrained models to new bulk systems or adsorbates, leveraging configurable descriptors and network architectures to produce production-ready potentials.
Yes, dpdata-formatted training data is a required input to start training DeePMD potentials. The workflow explicitly enforces proper data preparation alongside GPU availability and correct configuration of training and freezing steps before execution.
Supported network architectures include DPA-3 and se_e2_a for training machine learning potentials. These configurable descriptors allow you to tailor the training process to specific materials science contexts and scale across HPC clusters.
Model freezing is the final step in the DeePMD-kit training workflow, producing a production-ready model. The workflow enforces proper configuration of both training and freezing steps to ensure the output machine learning potentials are fully optimized for simulations.
jinzhezenggroupConvert atomic simulation data between VASP, LAMMPS, QE, CP2K, and ABACUS formats.
Hello-QMRun LAMMPS molecular dynamics with DeePMD-kit potentials and frozen .pb models.
jinzhezenggroupTrain DeePMD-kit models with the DPA3 descriptor using dp --pt train.
learningmatter-mitBenchmark machine learning interatomic potentials against labeled datasets using MAE/RMSE metrics.
