SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2209.12892 · 2022

Learning to Learn with Generative Models of Neural Network Checkpoints

arXiv · PDF · Open in the Atlas

Code that ran

16 of 20 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; 2 failures of our instrument; 4 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/2209.12892.json · badge: /api/ran/2209.12892.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
wpeebles/g.pt canonical (classifier label)
classifier basis not recorded on the link
16 ran · 4 no recorded run
FunctionStatusWhere it lives
all_gather Ran
ran on a synthesized input
wpeebles/g.pt/Gpt/distributed.py
code served (permissive licence) · get_code("250494257e1f8188")
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 Tensor, 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 torch

torch.manual_seed(0)

# The function all_gather expects:
# - input: a tensor
# - cat: boolean (default True)
# - dim: integer (default 0)

# Since get_world_size() will return 1 (distributed not initialized),
# the function will take the early return path and just return the input
# (or input.unsqueeze(dim) if cat=False)

input_tensor = torch.randn(2, 4)

ARGS = {
    "input": input_tensor,
    "cat": True,
    "dim": 0,
}
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 Tensor, 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
color_norm Ran
ran, violated the contract we drafted
wpeebles/g.pt/data_gen/train_cifar10.py
code served (permissive licence) · get_code("d160a205c24a70f9")
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 Tensor, ndim 3, dtype family float, shape (3, 4, 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 torch

torch.manual_seed(0)

# The function performs per-channel normalization on a CHW format image
# It iterates over the first dimension (channels) and normalizes each channel
# by subtracting mean and dividing by std

# Create a small image tensor in CHW format (Channels, Height, Width)
im = torch.randn(3, 4, 5, dtype=torch.float32)

# Create mean and std tensors with one value per channel
mean = torch.tensor([0.5, 0.4, 0.3], dtype=torch.float32)
std = torch.tensor([0.2, 0.15, 0.1], dtype=torch.float32)

ARGS = {
    "im": im,
    "mean": mean,
    "std": std
}
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 Tensor, ndim 3, dtype family float, shape (3, 4, 5) — shape (3, 4, 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
construct_loader Ran
ran on a synthesized input
wpeebles/g.pt/data_gen/train_mnist.py
code served (permissive licence) · get_code("337c2966cca20001")
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 DataLoader. 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
from torch.utils.data import TensorDataset

torch.manual_seed(0)

# Create a small dataset with 2 samples, each containing an input tensor (dim 4) and a label
x = torch.randn(2, 4)
y = torch.randint(0, 2, (2,))
dataset = TensorDataset(x, y)

ARGS = {
    "dataset": dataset,
    "mb_size": 2,
    "shuffle": True,
    "drop_last": False
}
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 DataLoader — a DataLoader output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a DataLoader 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
create_thresholding_fn Ran
ran on a synthesized input
wpeebles/g.pt/Gpt/vis.py
code served (permissive licence) · get_code("9759670b96772cd5")
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 torch

torch.manual_seed(0)

# Define the function
def create_thresholding_fn(thresholding, param_range):
    """
    Creates a function that thresholds after each diffusion sampling step.

    thresholding = "none": No thresholding.
    thresholding = "static": Clamp the sample to param_range.
    """
    if thresholding == 'none':

        def denoised_fn(x):
            return x
    elif thresholding == 'static':

        def denoised_fn(x):
            return torch.clamp(x, param_range[0], param_range[1])
    else:
        raise NotImplementedError
    return denoised_fn

# Test case 1: none thresholding
ARGS = {
    "thresholding": "none",
    "param_range": [-1.0, 1.0]
}
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 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
ddim_synth Ran
ran on a synthesized input
wpeebles/g.pt/playground.py
code served (permissive licence) · get_code("06ebeb225cfdc4a0")
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 Tensor, ndim 3, dtype family float, shape (2, 4, 5) as a later run recorded it. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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
import torch.nn as nn

torch.manual_seed(0)

class MockDiffusion(nn.Module):
    def __init__(self):
        super().__init__()
    
    def ddim_sample_loop(self, G, shape, clip_denoised=False, model_kwargs=None, device='cpu', **kwargs):
        # Return a tensor of the specified shape
        batch_size = shape[0]
        return torch.randn(shape)

class MockG(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 4)
    
    def forward(self, x, t, **kwargs):
        return self.linear(x)

# Create instances
diffusion = MockDiffusion()
G = MockG()

# Create tensor arguments with distinct dimensions
batch_size = 2
loss_target = torch.randn(batch_size, 3)  # batch_size x 3
loss_prev = torch.randn(batch_size, 3)    # batch_size x 3
w_prev = torch.randn(batch_size, 4, 5)    # batch_size x 4 x 5

# Additional kwargs for ddim_sample_loop
ddim_sample_loop_kwargs = {
    'steps': 50,
    'eta': 0.0,
}

ARGS = {
    'diffusion': diffusion,
    'G': G,
    'loss_target': loss_target,
    'loss_prev': loss_prev,
    'w_prev': w_prev,
    'ddim_sample_loop_kwargs': ddim_sample_loop_kwargs,
}
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 Tensor, ndim 3, dtype family float, shape (2, 4, 5) — shape (2, 4, 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
dump_json_stats Ran
ran on a synthesized input
wpeebles/g.pt/Gpt/meters.py
code served (permissive licence) · get_code("9817cd76b1cffb0c")
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 decimal
import simplejson

def float_to_decimal(data, prec=5):
    """Converts floats to decimals which allows for fixed width json."""
    if prec and isinstance(data, dict):
        return {k: float_to_decimal(v, prec) for k, v in data.items()}
    if prec and isinstance(data, float):
        return decimal.Decimal(('{:.' + str(prec) + 'f}').format(data))
    else:
        return data

def dump_json_stats(stats):
    """Converts stats dict into json string for logging."""
    stats = float_to_decimal(stats)
    stats_json = simplejson.dumps(stats, sort_keys=True, use_decimal=True)
    return stats_json

# Construct valid arguments for dump_json_stats
ARGS = {
    "stats": {
        "loss": 0.12345,
        "accuracy": 0.98765,
        "epoch": 5,
        "learning_rate": 0.001234,
        "nested": {
            "train_loss": 0.05678,
            "val_loss": 0.09876
        },
        "string_value": "test",
        "int_value": 42
    }
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
simplejson — 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
eta_str Ran
ran on a synthesized input
wpeebles/g.pt/data_gen/train_mnist.py
code served (permissive licence) · get_code("5c827254b86e623d")
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.

from datetime import timedelta

# The function eta_str takes a timedelta object and formats it as a string
# It accesses: eta_td.days and eta_td.seconds attributes
# We need to construct a valid timedelta object

eta_td = timedelta(days=5, hours=3, minutes=42, seconds=17)

ARGS = {
    'eta_td': eta_td
}
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 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
float_to_decimal Ran
ran on a synthesized input
wpeebles/g.pt/Gpt/meters.py
code served (permissive licence) · get_code("ed27fb884352e87a")
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 decimal

def float_to_decimal(data, prec=5):
    """Converts floats to decimals which allows for fixed width json."""
    if prec and isinstance(data, dict):
        return {k: float_to_decimal(v, prec) for k, v in data.items()}
    if prec and isinstance(data, float):
        return decimal.Decimal(('{:.' + str(prec) + 'f}').format(data))
    else:
        return data

# Test case 1: dict with floats
test_data = {
    "loss": 0.123456,
    "accuracy": 0.987654,
    "learning_rate": 0.0001234
}

ARGS = {
    "data": test_data,
    "prec": 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 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
horizontal_flip Ran
ran on a synthesized input
wpeebles/g.pt/data_gen/train_cifar10.py
code served (permissive licence) · get_code("1cd4047c59310f59")
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 ndarray, ndim 3, dtype family int, shape (3, 4, 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 numpy as np

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

# Create a small image in CHW (Channels, Height, Width) format
# Using distinct dimensions: 3 channels, 4 height, 5 width
im = np.random.randint(0, 256, size=(3, 4, 5), dtype=np.uint8)

# Probability of horizontal flip
p = 0.5

ARGS = {
    "im": im,
    "p": p,
}
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 ndarray, ndim 3, dtype family int, shape (3, 4, 5) — shape (3, 4, 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
is_main_proc Ran
ran on a synthesized input
wpeebles/g.pt/Gpt/distributed.py
code served (permissive licence) · get_code("13bc2d293163e97b")
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 bool, 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)

ARGS = {
    "local": False,
}
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 bool, 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
lr_fun_cos Ran
ran on a synthesized input
wpeebles/g.pt/Gpt/utils.py
code served (permissive licence) · get_code("3beb27a8d68b5a5e")
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 float64, ndim 0, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim 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

class Config:
    def __init__(self):
        self.train = TrainConfig()

class TrainConfig:
    def __init__(self):
        self.base_lr = 0.1
        self.num_ep = 10

cfg = Config()
cur_epoch = 5

ARGS = {
    "cfg": cfg,
    "cur_epoch": cur_epoch
}
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 float64, ndim 0, dtype family float — no output shape is stated: observe_signature.py recorded float ndim 0, which disagrees with the fixture run
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
scaled_all_reduce Ran
ran on a synthesized input
wpeebles/g.pt/Gpt/distributed.py
code served (permissive licence) · get_code("a45c474751d02fe9")
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 list. 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

torch.manual_seed(0)

class Config:
    def __init__(self, num_gpus):
        self.num_gpus = num_gpus

cfg = Config(num_gpus=1)

tensors = [
    torch.randn(2, 4),
    torch.randn(3, 5),
]

ARGS = {
    "cfg": cfg,
    "tensors": tensors,
}
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 list — a list 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
slerp Ran
ran on a synthesized input
wpeebles/g.pt/Gpt/latent_walk_helpers.py
code served (permissive licence) · get_code("c2e26f19600bfc6a")
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 Tensor, ndim 3, dtype family float, shape (2, 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 torch

torch.manual_seed(0)

# Construct valid arguments for slerp function
# The function expects:
# - low: tensor of shape (batch, dim)
# - high: tensor of shape (batch, dim)
# - weight: tensor of shape (batch,)

batch_size = 2
dim = 4

# Create random tensors for low and high points in latent space
low = torch.randn(batch_size, dim)
high = torch.randn(batch_size, dim)

# Create weight tensor for interpolation (values between 0 and 1)
weight = torch.rand(batch_size)

ARGS = {
    "low": low,
    "high": high,
    "weight": weight,
}
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 Tensor, ndim 3, dtype family float, shape (2, 2, 4) — shape (2, 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
slerpify Ran
ran on a synthesized input
wpeebles/g.pt/Gpt/latent_walk_helpers.py
code served (permissive licence) · get_code("d06e1fd0a7f317f7")
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 Tensor, ndim 4, dtype family float, shape (2, 3, 5, 4) as a later run recorded it. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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)

# z should be a 3D tensor with shape (v, s, dim)
# where v is number of vectors, s is number of steps, dim is latent dimension
z = torch.randn(2, 3, 4)

# n_steps is the number of interpolation steps
n_steps = 5

ARGS = {
    "z": z,
    "n_steps": n_steps,
}
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 Tensor, ndim 4, dtype family float, shape (2, 3, 5, 4) — shape (2, 3, 5, 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
time_string Ran
ran; our contract draft was wrong, not the code
wpeebles/g.pt/Gpt/meters.py
code served (permissive licence) · get_code("803bcd86e3301891")
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

torch.manual_seed(0)

# Looking at the time_string function:
# - It takes one parameter: seconds
# - It performs divmod operations on an integer version of seconds
# - It formats the result as a string with days, hours, minutes, seconds

ARGS = {
    "seconds": 123456.789
}
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 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
zero_pad Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
wpeebles/g.pt/data_gen/train_cifar10.py
pointer only (licence: BSD-2-Clause); the stored copy get_code serves, from blackfeather-wang/AdaFocus, is pointer only (licence: NONE) · get_code("8e712c383c677282")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (BSD-2-Clause), but get_code serves the stored copy, first harvested from blackfeather-wang/AdaFocus, and GitHub detected no licence in that repository when it was checked. 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.

construct_loader Not yet run
unverified (harvested, no recorded run)
wpeebles/g.pt/Gpt/utils.py
code served (permissive licence) · get_code("7bef924189a2a628")
How it ran

no invocation recorded.

setup_env Not yet run
unverified (harvested, no recorded run)
wpeebles/g.pt/Gpt/utils.py
code served (permissive licence) · get_code("6f53e4795c83768e")
How it ran

no invocation recorded.

synth Not yet run
unverified (harvested, no recorded run)
wpeebles/g.pt/Gpt/vis.py
code served (permissive licence) · get_code("8c81a09dfbb6d408")
How it ran

no invocation recorded.

top1_error Not yet run
unverified (harvested, no recorded run)
wpeebles/g.pt/data_gen/train_mnist.py
code served (permissive licence) · get_code("093a895f2d9c1921")
How it ran

no invocation recorded.

Also present in other papers' code

Of 20 samples lifted from this paper's linked repositories, 20 carry a function identity and 0 do not. Of the 20 distinct functions with an identity, 3 are also present in other papers whose code we hold (2 with at least one paper dated earlier, 0 the same month, 3 later) and 17 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
time_string 3 other papers
2 dated earlier, 1 later
1905.13214 (2019-05)
On Network Design Spaces for Visual Recognition
2108.02927 (2021-08)
DOLG: Single-Stage Image Retrieval with Deep Orthogonal Fusion of Local and Global Features
2308.04008 (2023-08)
Coarse-to-Fine: Learning Compact Discriminative Representation for Single-Stage Image Retrieval
zero_pad 3 other papers
1 dated earlier, 2 later
2105.03245 (2021-05)
Adaptive Focus for Efficient Video Recognition
2308.04008 (2023-08)
Coarse-to-Fine: Learning Compact Discriminative Representation for Single-Stage Image Retrieval
2408.03282 (2024-08)
AMES: Asymmetric and Memory-Efficient Similarity Estimation for Instance-level Retrieval
color_norm 1 other paper
1 later
2408.03282 (2024-08)
AMES: Asymmetric and Memory-Efficient Similarity Estimation for Instance-level Retrieval

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 explore a data-driven approach for learning to optimize neural networks. We construct a dataset of neural network checkpoints and train a generative model on the parameters. In particular, our model is a conditional diffusion transformer that, given an initial input parameter vector and a prompted loss, error, or return, predicts the distribution over parameter updates that achieve the desired metric. At test time, it can optimize neural networks with unseen parameters for downstream tasks in just one update. We find that our approach successfully generates parameters for a wide range of loss prompts. Moreover, it can sample multimodal parameter solutions and has favorable scaling properties. We apply our method to different neural network architectures and tasks in supervised and reinforcement learning.

For agents

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

get_harvested_code_for_paper("2209.12892")
get_code_for_paper("2209.12892")
have("2209.12892")

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

curl https://syntology.ai/api/ran/2209.12892.json

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

[![Syntology run record](https://syntology.ai/api/ran/2209.12892.svg)](https://syntology.ai/paper/2209.12892)

Connect an agent — have() is free.