Install & Compatibility
Where this runs
No compatibility data collected yet for this library.
Code
Verified usage
Verified import paths — ran on the pinned version, not inferred.
TorchSim
✓ from torch_sim_atomistic import TorchSim
✗ from torch_sim import TorchSim
Old package name was torch-sim; now torch-sim-atomistic
MACE
✓ from torch_sim_atomistic.models import MACE
Models submodule under torch_sim_atomistic.models
Compute potential energy of bulk silicon using a MACE model.
import torch
from torch_sim_atomistic import TorchSim
from ase.build import bulk
# Create a silicon crystal
atoms = bulk('Si', 'diamond', a=5.431)
# Initialize MLIP calculator
calc = TorchSim(model='MACE', device='cpu')
atoms.calc = calc
# Compute energy
energy = atoms.get_potential_energy()
print(f"Energy: {energy:.3f} eV")
Debug
Known issues
breakingIn v0.6.0, the package name changed from 'torch-sim' to 'torch-sim-atomistic'. Import paths updated accordingly.fixUpdate imports: from torch_sim_atomistic import ... instead of from torch_sim import ...
affects: <0.6.0 -> 0.6.0+
deprecatedModels like CHGNet and M3GNet are deprecated in v0.6.0 and will be removed in v0.7.0. Use MACE or SIMPL instead.fixSwitch to supported models: TorchSim(model='MACE') or TorchSim(model='SIMPL')
affects: 0.6.0
gotchaThe default device is 'cuda' if available; this can cause issues on CPU-only systems without explicit device setting.fixAlways set device explicitly: TorchSim(..., device='cpu') or device='cuda'.
affects: all
Upgrade
Version history
0.6.0latest on PyPI · released Apr 24, 2026
Audit
Dependencies
torchrequiredCore tensor operations and autograd
aserequiredAtomic simulation environment for structure handling