nvalchemi-data-structures

What problem does it solve?

Graph-based representations of atomic systems and efficient batching for GPU computation are essential for scalable simulations and analytics. This skill provides core data structures to manage atomic graphs (AtomicData) and batched multiple systems (Batch).

Core Features & Use Cases

• Core classes: AtomicData and Batch (Pydantic BaseModel with DataMixin) for GPU-ready data handling.
• Flexible fields: positions, atomic_numbers, neighbor_list, energy, cell, pbc, and more; data can be extended via info dict.
• Batch operations: create batches from data lists, inspect batch sizes, and transfer to devices for high-throughput computations.
• ASE integration: construct AtomicData from ASE Atoms objects for easy ingestion of external structures.
• High-performance workflow: includes graph-based representations and preparation for GPU-accelerated pipelines.

Quick Start

Import AtomicData and Batch from nvalchemi.data and start by creating a small AtomicData instance, then batch it with Batch.from_data_list.