How adamlaho/molcryst works
MolCryst-MLIPs is a published collection of fine-tuned interatomic-potential models and datasets for molecular crystals, rather than a software runtime. The documented use path is to obtain a compound-specific model and run simulations through AMLP-Analysis or the MACE calculator. Training and validation are also important product stages. The tree contains only README.md, so paths and relationships below reflect documented capabilities; implementation details and internal wiring are unavailable.
Main entry points: README.md Crystal models
Components 11
Model resources
- Crystal models
README.md - Train/valid data
README.md
Model development
- DFT reference data
README.md - MACE-MH1
README.md - Two-stage training
README.md - Model validation
README.md
Simulation use
- AMLP-Analysis
README.md - MACE calculator
README.md - Crystal simulations
README.md
Other components
- Researcher
- Hugging Face
Connections 14
- Hugging Face → Crystal models: hosts
- Hugging Face → Train/valid data: hosts
- DFT reference data → Train/valid data: produces
- MACE-MH1 → Two-stage training: fine-tuned by
- Train/valid data → Two-stage training: trains
- Two-stage training → Crystal models: produces
- Crystal models → Model validation: evaluated by
- Researcher → Hugging Face: downloads
- Researcher → AMLP-Analysis: runs
- Crystal models → AMLP-Analysis: configured in
- AMLP-Analysis → Crystal simulations: runs
- Researcher → MACE calculator: configures
- Crystal models → MACE calculator: loaded by
- MACE calculator → Crystal simulations: calculates