SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2207.09453 · 2022

e3nn: Euclidean Neural Networks

arXiv · PDF · Open in the Atlas

Code that ran

14 of 17 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 1 violated it and 13 have no contract verdict; 0 failures of our instrument; 3 no recorded run), measured 2026-11-22; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2207.09453.json · badge: /api/ran/2207.09453.svg

Each repository's role is a label: a classifier's output recorded on the link, or, where no role is recorded, what the Papers with Code archive marked. The line under it names its basis. Neither says who wrote the code.

RepositoryRoleRan
atomicarchitects/phonax community (classifier label)
classifier basis not recorded on the link
9 ran · 3 no recorded run
e3nn/e3nn-jax community (classifier label)
classifier basis not recorded on the link
4 ran · 0 no recorded run
copy not recorded community
not in the paper's declared repositories
1 ran · 0 no recorded run
FunctionStatusWhere it lives
bias_correction Ran
ran on a synthesized input
atomicarchitects/phonax/phonax/amsgrad.py
code served (permissive licence) · get_code("cfc6dd18efbdcf48")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import jax
import jax.numpy as jnp

# Construct arguments for bias_correction function
# The function applies bias correction to moments using a decay factor and count

# moment: a PyTree (can be dict, list, or array of JAX arrays)
# We'll use a dict containing JAX arrays of different shapes to test tree_map
moment = {
    'first': jnp.array([[1.0, 2.0, 3.0, 4.0],
                        [5.0, 6.0, 7.0, 8.0]]),  # shape (2, 4)
    'second': jnp.array([[9.0, 10.0, 11.0],
                         [12.0, 13.0, 14.0]]),  # shape (2, 3)
}

# decay: scalar value between 0 and 1 (like momentum/exponential decay rate)
decay = jnp.array(0.9)

# count: integer count (number of steps/iterations)
count = jnp.array(5)

ARGS = {
    'moment': moment,
    'decay': decay,
    'count': count,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
config Ran
ran on a synthesized input
e3nn/e3nn-jax/e3nn_jax/_src/config.py
code served (permissive licence) · get_code("598dbf45f78dc887")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned str. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

# The function 'config' takes two parameters:
# 1. name: a string key to look up in the __conf dictionary
# 2. value: optional, if provided sets the config value, if None retrieves it

# Test case 1: Getting a value (value=None)
test_name_1 = 'irrep_normalization'
test_value_1 = None

# Test case 2: Setting a value
test_name_2 = 'fused'
test_value_2 = True

# We'll use the first test case as the primary ARGS
ARGS = {
    'name': 'irrep_normalization',
    'value': None
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
load_config Ran
ran on a synthesized input
atomicarchitects/phonax/phonax/data_utils.py
code served (permissive licence) · get_code("590a74006285068b")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned float, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

config_dict = {
    "model_name": "bert-base",
    "learning_rate": 0.001,
    "batch_size": 32,
    "num_epochs": 10
}

config_key = "learning_rate"
key_default = 0.01

ARGS = {
    "config_dict": config_dict,
    "config_key": config_key,
    "key_default": key_default
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type float, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
mean_squared_error_stress Ran
ran on a synthesized input
atomicarchitects/phonax/phonax/loss.py
code served (permissive licence) · get_code("d7f12a4b2f0cf768")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ArrayImpl, ndim 1, dtype family float, shape (2,) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import jax.numpy as jnp
import numpy as np

class MockGlobals:
    def __init__(self, stress, weight):
        self.stress = stress
        self.weight = weight

class MockGraph:
    def __init__(self, globals_obj):
        self.globals = globals_obj

# Create small tensors with distinct dimensions
# stress_ref shape: (2, 4, 3) - batch=2, dim1=4, dim2=3
stress_ref = jnp.array(np.random.randn(2, 4, 3), dtype=jnp.float32)

# weight shape: (2,) - one weight per batch element
weight = jnp.array(np.array([1.0, 1.5]), dtype=jnp.float32)

# stress_pred shape: (2, 4, 3) - same shape as stress_ref
stress_pred = jnp.array(np.random.randn(2, 4, 3), dtype=jnp.float32)

# Construct graph with globals
globals_obj = MockGlobals(stress=stress_ref, weight=weight)
graph = MockGraph(globals_obj=globals_obj)

ARGS = {
    "graph": graph,
    "stress_pred": stress_pred
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ArrayImpl, ndim 1, dtype family float, shape (2,) — shape (2,) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
naive_broadcast_decorator Ran
ran on a synthesized input
atomicarchitects/phonax/phonax/nequip_model.py
code served (permissive licence) · get_code("3b1c22e06d1c7364")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned function. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import jax
import jax.nn
import jax.numpy as jnp

def naive_broadcast_decorator(func):
    def wrapper(*args):
        leading_shape = jnp.broadcast_shapes(*(arg.shape[:-1] for arg in args))
        args = [jnp.broadcast_to(arg, leading_shape + (-1,)) for arg in args]
        f = func
        for _ in range(len(leading_shape)):
            f = jax.vmap(f)
        return f(*args)
    return wrapper

# Create a simple function that the decorator will wrap
def simple_func(x, y):
    return x + y

# Test function with appropriate signature
@naive_broadcast_decorator
def decorated_func(x, y):
    return x + y

ARGS = {
    'func': simple_func
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
normalize_function Ran
ran on a synthesized input
e3nn/e3nn-jax/e3nn_jax/_src/activation.py
code served (permissive licence) · get_code("70b416517c39f1d7")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned function. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import jax
import jax.numpy as jnp
import jax.scipy.special as jsp

def normalspace(n: int) -> jax.Array:
    """Sequence of normally distributed numbers :math:`x_i` for :math:`i=1, \\ldots, n` such that

    .. math::

        \\int_{-\\infty}^{x_i} \\phi(x) dx = \\frac{i}{n+1}

    where :math:`\\phi(x) = \\frac{1}{\\sqrt{2\\pi}} e^{-x^2/2}` is the normal distribution.

    Args:
        n (int): Number of points

    Returns:
        jax.Array: Sequence of normally distributed numbers

    Examples:
        >>> normalspace(5)
        Array([-0.96742135, -0.4307273 ,  0.        ,  0.43072742,  0.96742165],      dtype=float32)
    """
    return jnp.sqrt(2) * jsp.erfinv(jnp.linspace(-1.0, 1.0, n + 2)[1:-1])

def normalize_function(phi):
    """Normalize a function, :math:`\\psi(x)=\\phi(x)/c` where :math:`c` is the normalization constant such that

    .. math::

        \\int_{-\\infty}^{\\infty} \\psi(x)^2 \\frac{e^{-x^2/2}}{\\sqrt{2\\pi}} dx = 1
    """
    with jax.ensure_compile_time_eval():
        x = normalspace(1000001)
        c = jnp.mean(phi(x) ** 2) ** 0.5
        c = c.item()
        if jnp.allclose(c, 1.0):
            return phi
        else:

            def rho(x):
                return phi(x) / c
            return rho


def simple_phi(x):
    """A simple test function that takes a jax array and returns a jax array."""
    return jnp.sin(x)


ARGS = {
    "phi": simple_phi
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
normalspace Ran
ran on a synthesized input
e3nn/e3nn-jax/e3nn_jax/_src/activation.py
code served (permissive licence) · get_code("e0a387be0fb8229f")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ArrayImpl, ndim 1, dtype family float, shape (5,) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import jax
import jax.numpy as jnp
import jax.scipy.special as jsp

# The function normalspace takes a single parameter: n (int)
# It returns a jax.Array of normally distributed numbers
# The parameter n specifies how many points to generate

ARGS = {
    "n": 5
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ArrayImpl, ndim 1, dtype family float, shape (5,) — shape (5,) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
prod Ran
ran on a synthesized input
atomicarchitects/phonax/phonax/nequip_model.py
code served (permissive licence) · get_code("3f10eb76dacbfed5")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned int, ndim 0, dtype family int, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import functools
import operator

def prod(xs):
    """From e3nn_jax/util/__init__.py."""
    return functools.reduce(operator.mul, xs, 1)

# The function takes a single parameter 'xs' which is an iterable
# It uses functools.reduce with operator.mul to compute the product of all elements
# Starting with an initial value of 1

# Test with various iterables of numbers
ARGS = {
    "xs": [2, 3, 4, 5]
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type int, ndim 0, dtype family int, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
random_train_valid_split Ran
ran on a synthesized input
atomicarchitects/phonax/phonax/data_utils.py
code served (permissive licence) · get_code("67cb9fe4747043a7")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# The function takes a Sequence of items, a valid_num int, and a seed int.
# It shuffles indices and splits the items into train and validation sets.
# No special object requirements - just basic Python types.

items = ['item_0', 'item_1', 'item_2', 'item_3', 'item_4', 'item_5', 'item_6', 'item_7', 'item_8', 'item_9']
valid_num = 3
seed = 42

ARGS = {
    'items': items,
    'valid_num': valid_num,
    'seed': seed
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
soft_odd Ran
ran on a synthesized input
e3nn/e3nn-jax/e3nn_jax/_src/activation.py
code served (permissive licence) · get_code("1bd170153085e595")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ArrayImpl, ndim 1, dtype family float, shape (6,) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import jax.numpy as jnp
import numpy as np

def soft_odd(x):
    """Smooth odd function that can be used as activation function for odd scalars.

    .. math::

        x (1 - e^{-x^2})

    Note:
        Odd scalars (l=0 and p=-1) has to be activated by functions with well defined parity:

        * even (:math:`f(-x)=f(x)`)
        * odd (:math:`f(-x)=-f(x)`).
    """
    return (1 - jnp.exp(-x ** 2)) * x


# Create test input - a small JAX array
x = jnp.array([-2.5, -1.0, 0.0, 1.5, 2.0, -0.5])

ARGS = {
    "x": x
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ArrayImpl, ndim 1, dtype family float, shape (6,) — shape (6,) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
sqrt Ran
ran on a synthesized input
atomicarchitects/phonax/phonax/gradio.py
code served (permissive licence) · get_code("1c715fe36ed81d4d")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ArrayImpl, ndim 1, dtype family float, shape (6,) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import jax.numpy as jnp
import numpy as np

# The function sqrt takes a single parameter x
# It uses jnp.sign, jnp.sqrt, and jnp.abs on x
# So x should be a JAX array or something that can be converted to one

# Create a small array with both positive and negative values
x = jnp.array([-4.0, -1.0, 1.0, 4.0, 9.0, 16.0])

ARGS = {
    'x': x
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ArrayImpl, ndim 1, dtype family float, shape (6,) — shape (6,) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
t_or_f Ran
ran, violated the contract we drafted
this paper's copy was not recorded; identical code first harvested from e3nn/e3nn
pointer only; the stored copy get_code serves, from e3nn/e3nn, is pointer only (licence: NOASSERTION) · get_code("440a098dda4f24f0")
How it ran

The body of this sample is withheld: this paper's copy of the code was not recorded (its link names no repository or source path), so no licence was read for it. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

update_moment Ran
ran on a synthesized input
atomicarchitects/phonax/phonax/amsgrad.py
code served (permissive licence) · get_code("643f80b3495d96e8")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ArrayImpl, ndim 2, dtype family float, shape (2, 4) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import jax
import jax.numpy as jnp

# Create mock updates and moments using JAX arrays
# The function applies tree_map over updates and moments, so they need to be
# nested structures (can be dicts, lists, or just single arrays)

# Simple case: single arrays
updates = jnp.array([[1.0, 2.0, 3.0, 4.0],
                     [5.0, 6.0, 7.0, 8.0]])

moments = jnp.array([[0.1, 0.2, 0.3, 0.4],
                     [0.5, 0.6, 0.7, 0.8]])

decay = 0.9

order = 2

ARGS = {
    "updates": updates,
    "moments": moments,
    "decay": decay,
    "order": order,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ArrayImpl, ndim 2, dtype family float, shape (2, 4) — shape (2, 4) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
update_moment_per_elem_norm Ran
ran on a synthesized input
atomicarchitects/phonax/phonax/amsgrad.py
code served (permissive licence) · get_code("50ec154b3d6e92bb")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import jax
import jax.numpy as jnp
import numpy as np

# Set seeds for reproducibility
np.random.seed(0)

# Create small sample data for the function
# The function expects:
# - updates: a pytree of arrays (the model updates/gradients)
# - moments: a pytree of arrays (the running moments, same structure as updates)
# - decay: a float (decay rate for EMA)
# - order: a float or int (the order of the moment to compute)

# Create sample updates (pytree of JAX arrays)
updates = {
    'layer1': jnp.array([[1.0, 2.0, 3.0, 4.0],
                         [5.0, 6.0, 7.0, 8.0]]),
    'layer2': jnp.array([[0.5, 1.5, 2.5, 3.5],
                         [4.5, 5.5, 6.5, 7.5]])
}

# Create sample moments (same structure as updates)
moments = {
    'layer1': jnp.array([[0.1, 0.2, 0.3, 0.4],
                         [0.5, 0.6, 0.7, 0.8]]),
    'layer2': jnp.array([[0.05, 0.15, 0.25, 0.35],
                         [0.45, 0.55, 0.65, 0.75]])
}

# Decay rate for exponential moving average
decay = 0.9

# Order of the moment
order = 2

ARGS = {
    'updates': updates,
    'moments': moments,
    'decay': decay,
    'order': order
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax, numpy, optax — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
datasets_groups Not yet run
unverified (harvested, no recorded run)
atomicarchitects/phonax/phonax/datasets.py
code served (permissive licence) · get_code("9acae8f41c549060")
How it ran

no invocation recorded.

mean_squared_error_energy Not yet run
unverified (harvested, no recorded run)
atomicarchitects/phonax/phonax/loss.py
code served (permissive licence) · get_code("dce83cf2c8205188")
How it ran

no invocation recorded.

mean_squared_error_forces Not yet run
unverified (harvested, no recorded run)
atomicarchitects/phonax/phonax/loss.py
code served (permissive licence) · get_code("bf974b17cbbd7b86")
How it ran

no invocation recorded.

Also present in other papers' code

Of 17 samples lifted from this paper's linked repositories, 17 carry a function identity and 0 do not. Of the 17 distinct functions with an identity, 1 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 1 later) and 16 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

FunctionAlso present inEarliest-datedOthers named
t_or_f 1 other paper
1 later
2401.10216 (2024-01)
Enabling Efficient Equivariant Operations in the Fourier Basis via Gaunt Tensor Products

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

We present e3nn, a generalized framework for creating E(3) equivariant trainable functions, also known as Euclidean neural networks. e3nn naturally operates on geometry and geometric tensors that describe systems in 3D and transform predictably under a change of coordinate system. The core of e3nn are equivariant operations such as the TensorProduct class or the spherical harmonics functions that can be composed to create more complex modules such as convolutions and attention mechanisms. These core operations of e3nn can be used to efficiently articulate Tensor Field Networks, 3D Steerable CNNs, Clebsch-Gordan Networks, SE(3) Transformers and other E(3) equivariant networks.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2207.09453")
get_code_for_paper("2207.09453")
have("2207.09453")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2207.09453.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2207.09453.svg)](https://syntology.ai/paper/2207.09453)

Connect an agent — have() is free.