Rootstock

Select a cluster from the sidebar

della

Rootstock v0.5.0

4 envs
Root
/scratch/gpfs/ROSENGROUP/common/rootstock
Python
3.12.12
Maintainer
Will Engler
willengler@uchicago.edu
Last Updated
Mar 16, 2026

Environments

chgnet_env Ready
Built: 2026-03-16T16:13:58.992562+00:00
Python: >=3.10

Source Code

 
# /// script
# requires-python = ">=3.10"
# dependencies = [
#     "chgnet>=0.3.0",
#     "ase>=3.22",
#     "torch>=2.0",
# ]
# ///
"""
CHGNet environment for Rootstock.

This environment provides access to CHGNet, a pretrained universal neural
network potential for charge-informed atomistic modeling.
"""


def setup(model: str | None = None, device: str = "cuda"):
    """
    Load a CHGNet calculator.

    Args:
        model: Optional path to a fine-tuned model. If None, uses the
               default pre-trained CHGNet model.
        device: PyTorch device string (e.g., "cuda", "cuda:0", "cpu")

    Returns:
        ASE-compatible calculator
    """
    from chgnet.model import CHGNetCalculator

    if model:
        return CHGNetCalculator(model_path=model, use_device=device)
    return CHGNetCalculator(use_device=device)

Dependencies (4)

ase 3.27.0
chgnet 0.4.2
rootstock 0.5.1
torch 2.10.0
mace_env Ready
Built: 2026-03-16T16:13:59.115604+00:00
Python: >=3.10

Source Code

 
# /// script
# requires-python = ">=3.10"
# dependencies = [
#     "mace-torch>=0.3.0",
#     "ase>=3.22",
#     "torch>=2.4.0,<2.10",
# ]
# ///
"""
MACE environment for Rootstock.

This environment provides access to MACE foundation models for
machine learning interatomic potentials.

Models:
    - "small", "medium", "large": Pre-trained MACE-MP-0 models
    - Path to a .pt file: Custom fine-tuned model
"""


def setup(model: str, device: str = "cuda"):
    """
    Load a MACE calculator.

    Args:
        model: Model identifier. Can be:
            - "small", "medium", "large" for MACE-MP-0 foundation models
            - Path to a .pt file for custom models
        device: PyTorch device string (e.g., "cuda", "cuda:0", "cpu")

    Returns:
        ASE-compatible calculator
    """
    from mace.calculators import mace_mp

    return mace_mp(model=model, device=device, default_dtype="float32")

Dependencies (4)

ase 3.27.0
mace-torch 0.3.15
rootstock 0.5.1
torch 2.9.1
tensornet_env Ready
Built: 2026-03-16T16:13:59.289290+00:00
Python: >=3.12

Source Code

 
# /// script
# requires-python = ">=3.12"
# dependencies = [
#     "torch>=2.4.0",
#     "ase>=3.22",
#     "matgl",
#     "nvalchemi-toolkit-ops",
#     "torch-geometric",
#     "torch-scatter",
#     "torch-sparse",
#     "torch-cluster",
#     "torch-spline-conv",
# ]
#
# [tool.uv]
# find-links = ["https://data.pyg.org/whl/torch-2.4.0+cu121.html"]
# ///
"""
TensorNet environment for Rootstock.

This environment provides access to TensorNet models via the MatGL library
from the Materials Virtual Lab.

Models:
    - "TensorNet-MatPES-PBE-v2025.1-PES": MatPES PBE functional (default)
    - Other MatGL models as available
"""


def setup(model: str = "TensorNet-MatPES-PBE-v2025.1-PES", device: str = "cuda"):
    """
    Load a TensorNet/MatGL calculator.

    Args:
        model: Model identifier (e.g., "TensorNet-MatPES-PBE-v2025.1-PES").
               Passed directly to matgl.load_model().
        device: PyTorch device string (currently MatGL handles device internally)

    Returns:
        ASE-compatible calculator
    """
    import torch
    torch.set_default_device(device)

    import matgl
    from matgl.ext.ase import PESCalculator

    pot = matgl.load_model(model)
    return PESCalculator(potential=pot)

Dependencies (10)

ase 3.27.0
matgl 2.0.6
nvalchemi-toolkit-ops 0.2.0
rootstock 0.5.1
torch 2.8.0
torch-cluster 1.6.3+pt24cu121
torch-geometric 2.7.0
torch-scatter 2.1.2+pt24cu121
torch-sparse 0.6.18+pt24cu121
torch-spline-conv 1.2.2+pt24cu121
uma_env Ready
Built: 2026-03-16T16:13:59.427557+00:00
Python: >=3.10,<3.11

Source Code

 
# /// script
# requires-python = ">=3.10,<3.11"
# dependencies = [
#     "torch>=2.4.0",
#     "fairchem-core>=2.0.0",
#     "ase>=3.22",
#     "torch-geometric",
# ]
#
# [tool.uv]
# find-links = ["https://data.pyg.org/whl/torch-2.4.0+cu121.html"]
# ///
"""
UMA (Universal Atomistic Model) environment for Rootstock.

This environment provides access to Meta's UMA foundation model
via the FAIRChem library.

Models:
    - "uma-s-1p1": UMA small model (default)
    - Other UMA variants as released by FAIRChem
"""


def setup(model: str = "uma-s-1p1", device: str = "cuda"):
    """
    Load a UMA calculator.

    Args:
        model: Model identifier (e.g., "uma-s-1p1"). Passed directly to
               pretrained_mlip.get_predict_unit().
        device: PyTorch device string (e.g., "cuda", "cuda:0", "cpu")

    Returns:
        ASE-compatible calculator
    """
    from fairchem.core import FAIRChemCalculator, pretrained_mlip

    predictor = pretrained_mlip.get_predict_unit(model, device=device)
    return FAIRChemCalculator(predictor, task_name="omat")

Dependencies (5)

ase 3.27.0
fairchem-core 2.14.0
rootstock 0.5.1
torch 2.8.0
torch-geometric 2.7.0

sophia

Rootstock v0.5.0

6 envs
Root
/lus/eagle/projects/Garden-Ai/rootstock
Python
3.10.19
Maintainer
Hayden Holbrook
hholbrook@uchicago.edu
Last Updated
Mar 19, 2026

Environments

chgnet_env Ready
Built: 2026-03-16T15:59:22.114091+00:00
Python: >=3.10

Source Code

 
# /// script
# requires-python = ">=3.10"
# dependencies = [
#     "chgnet>=0.3.0",
#     "ase>=3.22",
#     "torch>=2.0",
# ]
# ///
"""
CHGNet environment for Rootstock.

This environment provides access to CHGNet, a pretrained universal neural
network potential for charge-informed atomistic modeling.
"""


def setup(model: str | None = None, device: str = "cuda"):
    """
    Load a CHGNet calculator.

    Args:
        model: Optional path to a fine-tuned model. If None, uses the
               default pre-trained CHGNet model.
        device: PyTorch device string (e.g., "cuda", "cuda:0", "cpu")

    Returns:
        ASE-compatible calculator
    """
    from chgnet.model import CHGNetCalculator

    if model:
        return CHGNetCalculator(model_path=model, use_device=device)
    return CHGNetCalculator(use_device=device)

Dependencies (4)

ase 3.27.0
chgnet 0.4.2
rootstock 0.6.1
torch 2.10.0
tensornet_env Ready
Built: 2026-03-16T15:59:22.153004+00:00
Python: ==3.11

Source Code

 
# /// script
# requires-python = "==3.11"
# dependencies = [
#     "torch>=2.4.0",
#     "ase>=3.22",
#     "matgl",
#     "nvalchemi-toolkit-ops",
#     "torch-geometric",
#     "torch-scatter",
#     "torch-sparse",
#     "torch-cluster",
#     "torch-spline-conv",
# ]
#
# [tool.uv]
# find-links = ["https://data.pyg.org/whl/torch-2.4.0+cu124.html"]
# 
# ///
"""
TensorNet environment for Rootstock.

This environment provides access to TensorNet models via the MatGL library
from the Materials Virtual Lab.

Models:
    - "TensorNet-MatPES-PBE-v2025.1-PES": MatPES PBE functional (default)
    - Other MatGL models as available
"""


def setup(model: str | None = "TensorNet-MatPES-PBE-v2025.1-PES", device: str = "cuda"):
    """
    Load a TensorNet/MatGL calculator.

    Args:
        device: PyTorch device string (currently MatGL handles device internally)

    Returns:
        ASE-compatible calculator
    """
    import matgl
    from matgl.ext.ase import PESCalculator
    from pathlib import Path
    pot = matgl.load_model(Path("/eagle/Garden-Ai/rootstock/cache/tensornet") / model)
    calc = PESCalculator(pot, device=device)
    return calc

Dependencies (10)

ase 3.28.0
matgl 2.1.1
nvalchemi-toolkit-ops 0.3.0
rootstock 0.7.0
torch 2.10.0
torch-cluster 1.6.3+pt24cu124
torch-geometric 2.7.0
torch-scatter 2.1.2+pt24cu124
torch-sparse 0.6.18+pt24cu124
torch-spline-conv 1.2.2+pt24cu124
uma_env Ready
Built: 2026-03-16T15:59:22.197145+00:00
Python: >=3.10,<3.11

Source Code

 
# /// script
# requires-python = ">=3.10,<3.11"
# dependencies = [
#     "torch>=2.4.0",
#     "fairchem-core>=2.0.0",
#     "ase>=3.22",
#     "torch-geometric",
# ]
#
# [tool.uv]
# find-links = ["https://data.pyg.org/whl/torch-2.4.0+cu121.html"]
# ///
"""
UMA (Universal Atomistic Model) environment for Rootstock.

This environment provides access to Meta's UMA foundation model
via the FAIRChem library.

Models:
    - "uma-s-1p1": UMA small model (default)
    - Other UMA variants as released by FAIRChem
"""


def setup(model: str = "uma-s-1p1", device: str = "cuda"):
    """
    Load a UMA calculator.

    Args:
        model: Model identifier (e.g., "uma-s-1p1"). Passed directly to
               pretrained_mlip.get_predict_unit().
        device: PyTorch device string (e.g., "cuda", "cuda:0", "cpu")

    Returns:
        ASE-compatible calculator
    """
    from fairchem.core import FAIRChemCalculator, pretrained_mlip

    predictor = pretrained_mlip.get_predict_unit(model, device=device)
    return FAIRChemCalculator(predictor, task_name="omat")

Dependencies (5)

ase 3.27.0
fairchem-core 2.14.0
rootstock 0.6.1
torch 2.8.0
torch-geometric 2.7.0
mace_env Ready
Built: 2026-03-16T16:01:51.169812+00:00
Python: >=3.10

Source Code

 
# /// script
# requires-python = ">=3.10"
# dependencies = [
#     "mace-torch>=0.3.0",
#     "ase>=3.22",
#     "torch>=2.4.0,<2.10",
# ]
# ///
"""
MACE environment for Rootstock.

This environment provides access to MACE foundation models for
machine learning interatomic potentials.

Models:
    - "small", "medium", "large": Pre-trained MACE-MP-0 models
    - Path to a .pt file: Custom fine-tuned model
"""


def setup(model: str, device: str = "cuda"):
    """
    Load a MACE calculator.

    Args:
        model: Model identifier. Can be:
            - "small", "medium", "large" for MACE-MP-0 foundation models
            - Path to a .pt file for custom models
        device: PyTorch device string (e.g., "cuda", "cuda:0", "cpu")

    Returns:
        ASE-compatible calculator
    """
    from mace.calculators import mace_mp

    return mace_mp(model=model, device=device, default_dtype="float32")

Dependencies (4)

ase 3.27.0
mace-torch 0.3.15
rootstock 0.6.1
torch 2.9.1
orb_env Ready
Built: 2026-03-19T16:59:19.555708+00:00
Python: >=3.12

Source Code

 
# /// script
# requires-python = ">=3.12"
# dependencies = [
#     "orb-models>=0.6.2", 
#     "ase>=3.22",
#     "torch>=2.0",
# ]
# ///

def setup(model: str, device: str = "cuda"):
    """
    Load an ORBCalculator
    """

    from orb_models.forcefield import pretrained
    from orb_models.forcefield.inference.calculator import ORBCalculator

    orbff, atoms_adapter = pretrained.ORB_PRETRAINED_MODELS[model]()
    calc = ORBCalculator(orbff, atoms_adapter=atoms_adapter, device=device)
    return calc

Dependencies (4)

ase 3.28.0
orb-models 0.6.2
rootstock 0.7.0
torch 2.10.0
sevennet_env Ready
Built: 2026-03-19T17:32:28.231635+00:00
Python: >=3.12

Source Code

 
# /// script
# requires-python = ">=3.12"
# dependencies = [
#   "sevenn[queq12]>=0.12.1",
# ]
# ///

def setup(model: str | None = None, device: str = "cuda"):
    """
    Load a SevenNetCalculator
    """
    from sevenn.calculator import SevenNetCalculator
    if model is None:
        model = "7net-omat"
    calc = SevenNetCalculator(model=model, enable_queq=True, device=device)
    return calc

Dependencies (2)

rootstock 0.7.0
sevenn 0.12.1