Menu
Prepare and submit HamGNN inference jobs through the canonical HamGNN Inspire submit flow.
Instalar com Codex ou Claude Copie este prompt, cole no Codex, Claude ou outro assistente e deixe que ele revise a página da skill e instale para você.
Baseado na classificação ocupacional SOC
| name | hamgnn-inference |
| description | Prepare and submit HamGNN inference jobs through the canonical HamGNN Inspire submit flow. |
<skill_root>/templates/HamGNN/test_config.yaml.template, then render a run-local config under <target_dir>/HamGNN/.inspire-hamgnn-submit for the actual remote submission. inspire-hamgnn-submit is the sole authority for HamGNN workspace, GPU resource, image, and helper selection.<skill_root>/scripts/inference.sh --workdir <hamgnn_workdir> --config <rendered_test_config> as the required runnable entrypoint for this stage.HamGNN/test_config.yaml in place..npz generated in the OpenMX data-processing stagetest_config.yaml.npz, render the config so stage test consumes that entire dataset as the inference set. Do not ask HamGNN to split the external test dataset again.uni_model_pkl_path: null by default. Only populate uni_model_pkl_path when the user explicitly asks to load a Uni model.Prepare ABACUS calculation inputs from CIF/POSCAR with default settings and generate per-structure INPUT/KPT/STRU run directories.
Prepare HamGNN training configs and run locally or submit through the canonical HamGNN Inspire flow.
Prepare split-local OpenMX preprocessing inputs and submit recognized OpenMX processing jobs through `inspire-openmx-submit`.
Legacy compatibility wrapper for Inspire remote job submission. Redirect to `inspire-cli-training-submit` for the canonical generic submission flow.
Structure-file-driven OpenMX/ABACUS-to-HamGNN training pipeline. Use when users provide a crystal structure file such as CIF or POSCAR and want to build a perturbed CIF dataset, run DFT preprocessing (OpenMX by default, ABACUS when explicitly requested), and train a HamGNN model, stopping at the trained checkpoint.