xnn.common.finetune.reference.predict_energies

xnn.common.finetune.reference.predict_energies(model, structures, batch_size=32, device=None)

Energies model predicts for structures, float64 (n,).

Parameters:

• model (torch.nn.Module) – A potential (bare or wrapped in ForceStressOutput); its cutoff builds the neighbor lists.

• structures (iterable of dict or AtomicDataset) – The structures.

• batch_size (int, optional) – Structures per forward pass, by default 32.

• device (torch.device, optional) – Where to evaluate; by default where the model’s parameters are.

Return type:

Tensor