Features

Think of NepTrainKit as a data flow rather than a list of menu names:

What do you need to do?	Where to go	Typical output
Generate candidate training structures	Make Dataset	candidate_pool.xyz
Clean candidates and remove problematic samples	NEP Dataset Display	candidate_pool_clean.xyz
Inspect training-data quality and distribution signals	Training Set Audit	Traceable findings and structure subsets
Look up the purpose and parameters of a generation card	Make Dataset Card Guide	Reusable card configurations
Manage projects, model versions, and result paths	Data Management	Searchable experiment records
Adjust display, backend, and performance behavior	Settings	Defaults suited to the local machine

If you are not sure where to begin, start with Quick start. If you already have a candidate pool, go directly to Clean Candidate Structures Before DFT.

• NEP Dataset Display
• Training Set Audit
• NEP Dataset Display reference
• Generate data (Make Dataset)
• Make Dataset card manual
• Develop custom cards
• Data management
• Settings