Allegro-FM: Model and LAMMPS input file

2025/02/09 by Nomura, Ken-ichi, Shinnosuke Hattori, Hattori, Shinnosuke +13 · 5 citations
Environmental Science · Physics and Astronomy · #Advanced Chemical Physics Studies #FOS: Physical sciences #Materials Science (cond-mat.mtrl-sci) #Methane Hydrates and Related Phenomena #Quantum, superfluid, helium dynamics

paper · pdf · doi:10.48550/arxiv.2502.06073

openalex publication_date 2025/02/09 · openalex created_date 2025/02/12 · openalex updated_date 2026/07/28

Abstract

Two models used in the manuscript and an example LAMMPS input file. Model Details afm256₀1_HL.pt: Trained for 109 epochs over approximately 3 days afm512₀1_HL.pt: Trained for 93 epochs over approximately 6 days Optimizer: torch.optim.Adam with a learning rate of 0.001 Learning Rate Scheduler: torch.optim.lr_scheduler.ReduceLROnPlateau Factor: 0.6 Patience: 25 epochs Minimum learning rate: 1e-4. Threshold: 0.02 in relative mode, Early Stopping: Callback: lightning.pytorch.callbacks.EarlyStopping. Minimum delta: 1e-3 Patience: 30 epochs before early stopping is triggered. Environment and Version Information torch: 2.5.1+cu124 e3nn: 0.5.4 nequip: 0.6.1 allegro: 0.3.0

Cited by

• Molecular Dynamics Simulations of γ-Belite(010)-Water Interfaces with High-Dimensional Neural Network Potentials
• QuantumCanvas: A Multimodal Benchmark for Visual Learning of Atomic Interactions
• Node-Equivariant Message Passing for Efficient and Accurate Machine Learning Interatomic Potentials
• Multiscale light-matter dynamics in quantum materials: from electrons to topological superlattices
• Beyond Scaling: Chemical Intuition as Emergent Ability of Universal Machine Learning Interatomic Potentials

Related

• Taming Multi-Domain, -Fidelity Data: Towards Foundation Models for Atomistic Scale Simulations
• AlphaNet: Scaling Up Local-frame-based Atomistic Interatomic Potential
• Fast and Accurate Foundation Models for Equivariant Machine-Learned Interatomic Potentials
• A foundation model for atomistic materials chemistry
• Learning Local Equivariant Representations for Large-Scale Atomistic Dynamics