Settings

Personalize behavior and performance trade‑offs. Changes take effect immediately and persist to your user config.

Personalization

Force data format: Choose how forces are plotted
- Raw: Use 3N vector format grouped by atoms
- Norm: Use per‑structure |F| (faster plotting and simpler visuals)
Canvas Engine: Choose rendering backend
- PyQtGraph (CPU): Default, broad compatibility
- Vispy (GPU): Uses OpenGL, faster for large scenes (requires working GPU/driver)
Auto loading: On startup, if ./train.xyz and ./nep.txt exist, auto‑load the dataset
Cache output files: When enabled (default), NepTrainKit writes cached *.out, descriptor.out, and nep.in files to speed up the next load. Disable for temporary viewing to avoid creating/updating cache files (applies to both NEP and DeepMD loaders).
Covalent radius coefficient: Threshold for “non‑physical bond” detection in structure view
Sort atoms: Normalize atom ordering when processing structures in Make Dataset
Use card group menu: Group console cards by category in the add‑card menu

NEP Backend: Select CPU, GPU, or Auto
- Auto: Tries GPU first; falls back to CPU if GPU is unavailable
- GPU: Requires a supported NVIDIA driver and CUDA 12.4 runtime; otherwise an in‑app warning will appear and the app switches to CPU
- CPU: Always use the CPU backend
GPU Batch Size: Number of frames per GPU slice when running NEP calculations
- Larger batches improve throughput but use more GPU memory
- Reduce if you encounter out‑of‑memory errors

Switching Canvas to Vispy requires a functional OpenGL environment.
“Norm” force mode reduces data size for plotting and improves responsiveness.
Auto loading aims to streamline repetitive workflows when working in a project directory.
GPU backend availability depends on your platform and package build. On systems without the GPU binary or compatible drivers, NepTrainKit automatically uses the CPU backend.