Auto3D.model_factory.create_model

Auto3D.model_factory.create_model(name: str, device: device | None = None, compile_model: bool | None = None, use_ensemble: bool | None = None, use_cache: bool = True, **kwargs: Any) → BaseModelAdapter

Convenience function to create a model adapter.

Parameters:

  • name – Model name or path to custom model.

  • device – Target device (default: CPU).

  • compile_model – Whether to use torch.compile() for optimization. If None, checks AUTO3D_COMPILE_MODEL environment variable.

  • use_ensemble – For AIMNET: use ensemble model (default False). Single model is ~35x faster. Set True for highest accuracy.

  • use_cache – Whether to cache and reuse model instances. Default True.

  • **kwargs – Additional model arguments.

Returns:

Initialized model adapter.

Example

>>> model = create_model("AIMNET", device=torch.device("cuda:0"))  # Fast single model (default)
>>> # Use ensemble for highest accuracy (slower)
>>> model = create_model("AIMNET", device=torch.device("cuda:0"), use_ensemble=True)
>>> # Enable torch.compile for ANI models
>>> model = create_model("ANI2xt", device=torch.device("cuda:0"), compile_model=True)
>>> # Clear cache when done
>>> ModelFactory.clear_cache()