Skip to content

Commit 79a66c0

Browse files
committed
Update torch-sim engine page to reference metatomic-torchsim
Points to the standalone metatomic-torchsim package with install instructions and a minimal usage snippet.
1 parent f28d191 commit 79a66c0

1 file changed

Lines changed: 33 additions & 10 deletions

File tree

docs/src/engines/torch-sim.rst

Lines changed: 33 additions & 10 deletions
Original file line numberDiff line numberDiff line change
@@ -9,23 +9,46 @@ torch-sim
99
* - Official website
1010
- How is metatomic supported?
1111
* - https://radical-ai.github.io/torch-sim/
12-
- In the official version
12+
- Via the ``metatomic-torchsim`` package
1313

14-
Supported model outputs
14+
How to install the code
1515
^^^^^^^^^^^^^^^^^^^^^^^
1616

17-
Only the :ref:`energy <energy-output>` output is supported.
17+
Install the integration package from PyPI:
1818

19-
How to install the code
19+
.. code-block:: bash
20+
21+
pip install metatomic-torchsim
22+
23+
This pulls in ``torch-sim-atomistic`` and ``metatomic-torch`` as dependencies.
24+
25+
For the full TorchSim documentation, see
26+
https://radical-ai.github.io/torch-sim/.
27+
28+
Supported model outputs
2029
^^^^^^^^^^^^^^^^^^^^^^^
2130

22-
The code is available in the ``torch-sim`` package, see the corresponding
23-
`installation instructions <https://radical-ai.github.io/torch-sim/user/introduction.html#installation>`_.
31+
Only the :ref:`energy <energy-output>` output is supported. Forces and stresses
32+
are derived via autograd.
2433

2534
How to use the code
2635
^^^^^^^^^^^^^^^^^^^
2736

28-
You can find the documentation for metatomic models in torch-sim `here
29-
<https://radical-ai.github.io/torch-sim/tutorials/metatomic_tutorial.html>`_,
30-
and generic documentation on torch-sim `there
31-
<radical-ai.github.io/torch-sim/>`_.
37+
.. code-block:: python
38+
39+
import ase.build
40+
import torch_sim as ts
41+
from metatomic.torchsim import MetatomicModel
42+
43+
model = MetatomicModel("model.pt", device="cpu")
44+
45+
atoms = ase.build.bulk("Si", "diamond", a=5.43, cubic=True)
46+
sim_state = ts.io.atoms_to_state([atoms], model.device, model.dtype)
47+
48+
results = model(sim_state)
49+
print(results["energy"]) # shape [1]
50+
print(results["forces"]) # shape [n_atoms, 3]
51+
print(results["stress"]) # shape [1, 3, 3]
52+
53+
For more details, see the `metatomic-torchsim documentation
54+
<https://docs.metatensor.org/metatomic/latest/torchsim/>`_.

0 commit comments

Comments
 (0)