Foundational Model Tutorial#

This tutorial demonstrates how to load a pretrained foundational model and attach it as an ASE calculator.

For more advanced topics—such as fine-tuning, or alternative interfaces, please refer to the corresponding dedicated tutorials.

Model Download and Cache#

When loading a model through from tace.foundations import tace_foundations, the pretrained weights will be downloaded automatically and cached locally.

By default, all models are stored under:

~/.cache/tace/

If your network connection is unstable or restricted, you may manually download the pretrained models from:

xvzemin/tace-foundations

After downloading, please keep the original directory and file names unchanged and place them directly under ~/.cache/tace/ so that TACE can locate them correctly.

Minimal ASE Example#

Below is a minimal working example showing how to use a TACE Foundational Model as an ASE calculator:

import torch
from ase.io import read
from tace.foundations import tace_foundations
from tace.interface.ase import TACEAseCalc, add_dispersion

# Load a pretrained foundational model
# The model will be auto-downloaded to ~/.cache/tace if not present
model = tace_foundations["TACE-v1-OAM-M"]

dtype = "float32"
device = "cuda" if torch.cuda.is_available() else "cpu"

# Fidelity level (0 corresponds to the first fidelity)
level = 0

atoms = read("../unrelaxed.xyz", index=0)
calc = TACEAseCalc(
    model=model,
    dtype=dtype,
    device=device,
    level=level,
)
atoms.calc = calc

Dispersion Correction (Optional)#

Dispersion interactions can also be supported by calling third-party libraries. For detailed instructions, see ase guide.