SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2104.01231 · ICML · 2021

Diverse Gaussian Noise Consistency Regularization for Robustness and Uncertainty Calibration

Theodoros Tsiligkaridis, Athanasios Tsiligkaridis

arXiv · PDF · Open in the Atlas

Code that ran

45 of 52 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (8 of them constructed an object rather than computing a result; 3 honoured the contract we drafted, 1 violated it and 9 have no contract verdict; 32 failures of our instrument; 7 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/2104.01231.json · badge: /api/ran/2104.01231.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
theot1/dign canonical (classifier label)
classifier basis: classifier_2025
34 ran · 2 no recorded run
TheoT1/DiGN canonical (classifier label)
classifier basis: census
11 ran · 5 no recorded run
FunctionStatusWhere it lives
AugMaxModule Ran
ran on a synthesized input
theot1/dign/DiGN.py
code served (permissive licence) · get_code("1e6ccf5662228a75")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, class construct-then-call prompt, model not recorded) and returned Tensor, ndim 4, dtype family float; no output shape is stated. 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 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 torch
import torch.nn as nn

# INIT_ARGS for AugMaxModule.__init__
INIT_ARGS = {
    'device': 'cpu'
}

# ARGS for AugMaxModule.forward
# xs: tuple of Tensors. len(xs)=4 (x_ori, x_aug1, x_aug2, x_aug3). x_ori.size()=(N,C,W,H)
# m: Tensor. m.size=(N)
# q: Tensor. q.size()=(N,3). w = softmax(q)

N = 2
C = 3
W = 5
H = 7

x_ori = torch.randn(N, C, W, H)
x_aug1 = torch.randn(N, C, W, H)
x_aug2 = torch.randn(N, C, W, H)
x_aug3 = torch.randn(N, C, W, H)

xs = (x_ori, x_aug1, x_aug2, x_aug3)
m = torch.rand(N)
q = torch.randn(N, 3)

ARGS = {
    'xs': xs,
    'm': m,
    'q': q
}
Setup block drafted by
a language model, class construct-then-call prompt, model not recorded
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 — no output shape is stated: observe_signature.py observed this class with its constructor sizes canonicalised where it could, which the node does not record, so its shape is not stated as this setup block's
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: torch and numpy seeded to 0 by the harness immediately before the class was constructed from INIT_ARGS (fixture_check.py)
AugMix Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
theot1/dign/DiGN.py
code served (permissive licence) · get_code("9e4262567e4d68af")
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; no output shape is stated. 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 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
import torch
from PIL import Image, ImageOps, ImageEnhance

torch.manual_seed(0)
np.random.seed(0)

# Create a small batch of fake image data
# Shape: (batch_size, channels, height, width) = (2, 3, 4, 4)
x = torch.rand(2, 3, 4, 4)

ARGS = {
    'x': x,
    'width': 3,
    'depth': -1,
    'alpha': 1.0,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy, 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 — no output shape is stated: observe_signature.py could not observe it (call:AssertionError)
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
BasicBlock Ran
ran on a synthesized input · constructed an object; no result was computed
theot1/dign/DiGN.py
code served (permissive licence) · get_code("9fcfe39f01a6db07")
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 an instance of its own class, BasicBlock: the run constructed the object and did not call it, so no result was computed. 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 torch.nn as nn

torch.manual_seed(0)

# BasicBlock constructor takes: in_planes, planes, stride=1
# We need to provide valid integer values for these parameters

in_planes = 3  # number of input channels
planes = 4     # number of output channels
stride = 1     # stride for first conv

ARGS = {
    'in_planes': in_planes,
    'planes': planes,
    'stride': stride
}
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 BasicBlock — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a BasicBlock 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
Bottle2neck Ran
ran on a synthesized input · constructed an object; no result was computed
theot1/dign/DiGN.py
code served (permissive licence) · get_code("386e139e40ce0e23")
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 an instance of its own class, Bottle2neck: the run constructed the object and did not call it, so no result was computed. 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 torch.nn as nn
import random

torch.manual_seed(0)
random.seed(0)

# The function creates a Bottle2neck module and expects a forward pass
# We need to construct valid arguments for __init__

ARGS = {
    'inplanes': 3,
    'planes': 4,
    'stride': 1,
    'hidden_planes': 9,
    'scale': 4,
    'batch_size': 5
}
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 Bottle2neck — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a Bottle2neck 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
Bottleneck Ran
ran on a synthesized input · constructed an object; no result was computed
theot1/dign/DiGN.py
code served (permissive licence) · get_code("c5503c74c1ea7353")
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 an instance of its own class, Bottleneck: the run constructed the object and did not call it, so no result was computed. 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 torch.nn as nn

torch.manual_seed(0)

# Bottleneck.__init__ parameters:
# - self (implicit)
# - in_planes: int
# - planes: int
# - stride: int (default=1)

# Bottleneck.forward uses:
# - x: input tensor
# - fake_relu: bool (default=False)

# Create a small input tensor for testing
# Shape: (batch=2, channels=in_planes, height=4, width=4)
in_planes = 8
planes = 4
stride = 1

# Create input tensor with matching in_planes channels
input_tensor = torch.randn(2, in_planes, 4, 4)

ARGS = {
    'in_planes': in_planes,
    'planes': planes,
    'stride': stride,
}
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 Bottleneck — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a Bottleneck 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
DenseNet Ran
ran on a synthesized input · constructed an object; no result was computed
theot1/dign/DiGN.py
code served (permissive licence) · get_code("757ed71343ce4c3b")
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 an instance of its own class, DenseNet: the run constructed the object and did not call it, so no result was computed. 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 torch.nn as nn
import math

torch.manual_seed(0)

# Define a minimal DenseBlock that satisfies the interface
class DenseBlock(nn.Module):
    def __init__(self, in_planes, growth_rate):
        super(DenseBlock, self).__init__()
        self.bn = nn.BatchNorm2d(in_planes)
        self.conv = nn.Conv2d(in_planes, growth_rate, kernel_size=3, padding=1, bias=False)
    
    def forward(self, x):
        out = self.conv(torch.relu(self.bn(x)))
        return torch.cat([x, out], 1)

# Create arguments
block = DenseBlock
nblocks = [2, 2, 2, 2]
growth_rate = 12
reduction = 0.5
num_classes = 10

ARGS = {
    'block': block,
    'nblocks': nblocks,
    'growth_rate': growth_rate,
    'reduction': reduction,
    'num_classes': num_classes
}
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 DenseNet — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a DenseNet 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
DenseNet121 Ran
ran; our contract draft was wrong, not the code
theot1/dign/DiGN.py
code served (permissive licence) · get_code("e18432664d73b55f")
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 DenseNet. 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 torch.nn as nn

torch.manual_seed(0)

ARGS = {
    'num_classes': 10
}
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 DenseNet — a DenseNet output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a DenseNet 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
DenseNet121 Ran
ran; our contract draft was wrong, not the code
TheoT1/DiGN/densenet.py
code served (permissive licence) · get_code("9ccdb4ad5a50fc5c")
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 DenseNet. 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 torch.nn as nn

torch.manual_seed(0)

# Create a small input tensor for DenseNet121
# DenseNet121 expects 3-channel input images
# Batch size 2, 3 channels, 32x32 spatial dimensions
input_tensor = torch.randn(2, 3, 32, 32)

# num_classes parameter
num_classes = 10

# Additional keyword arguments that can be passed to DenseNet121
# Based on the DenseNet.__init__ signature, these are supported:
# growth_rate, reduction, num_classes
kwargs = {
    'growth_rate': 12,
    'reduction': 0.5,
}

ARGS = {
    'num_classes': num_classes,
}
ARGS.update(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 DenseNet — a DenseNet output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a DenseNet 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
FakeReLU Ran
ran on a synthesized input · constructed an object; no result was computed
theot1/dign/DiGN.py
code served (permissive licence) · get_code("06935f102e16e291")
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 an instance of its own class, FakeReLU: the run constructed the object and did not call it, so no result was computed. 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)

# FakeReLU.forward takes ctx and input
# FakeReLU.backward takes ctx and grad_output
# Both are static methods, so we need to construct the input tensor

input_tensor = torch.randn(2, 4)

ARGS = {
    "input": input_tensor
}
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 FakeReLU — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a FakeReLU 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
Res2Net Ran
ran on a synthesized input · constructed an object; no result was computed
theot1/dign/DiGN.py
code served (permissive licence) · get_code("2f365ca72da7a611")
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 an instance of its own class, Res2Net: the run constructed the object and did not call it, so no result was computed. 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 torch.nn as nn

torch.manual_seed(0)

# The Res2Net forward method expects:
# - x: a tensor that can be passed through Bottle2neck blocks
# - The blocks expect input shape (1, inplanes*batch_size, H, W)
# Looking at Bottle2neck, it uses inplanes=3, planes=3, batch_size=5
# So input should be (1, 15, H, W) where H, W are spatial dimensions

# Create a small input tensor
# batch_size in Res2Net is implicitly 5 (from Bottle2neck initialization)
# inplanes and planes are 3
# So we need (1, 3*5, H, W) = (1, 15, H, W)
x = torch.randn(1, 15, 4, 4)

ARGS = {
    "epsilon": 0.2,
    "hidden_planes": 16,
    "batch_size": 5,
}
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 Res2Net — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a Res2Net 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
ResNet Ran
ran on a synthesized input · constructed an object; no result was computed
theot1/dign/DiGN.py
code served (permissive licence) · get_code("b3e481f3ef6524ff")
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 an instance of its own class, ResNet: the run constructed the object and did not call it, so no result was computed. 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 torch.nn as nn

torch.manual_seed(0)

class BasicBlock(nn.Module):
    expansion = 1
    
    def __init__(self, in_planes, planes, stride=1):
        super(BasicBlock, self).__init__()
        self.conv1 = nn.Conv2d(in_planes, planes, kernel_size=3, stride=stride, padding=1, bias=False)
        self.bn1 = nn.BatchNorm2d(planes)
        self.conv2 = nn.Conv2d(planes, planes, kernel_size=3, stride=1, padding=1, bias=False)
        self.bn2 = nn.BatchNorm2d(planes)
        
        self.shortcut = nn.Sequential()
        if stride != 1 or in_planes != planes * self.expansion:
            self.shortcut = nn.Sequential(
                nn.Conv2d(in_planes, planes * self.expansion, kernel_size=1, stride=stride, bias=False),
                nn.BatchNorm2d(planes * self.expansion)
            )
    
    def forward(self, x, fake_relu=False, **kwargs):
        out = torch.relu(self.bn1(self.conv1(x)))
        out = self.bn2(self.conv2(out))
        out += self.shortcut(x)
        if not fake_relu:
            out = torch.relu(out)
        return out

ARGS = {
    'block': BasicBlock,
    'num_blocks': [2, 2, 2, 2],
    'num_classes': 10,
    'feat_scale': 1,
    'wm': 1,
    'd_input': 32
}
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 ResNet — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a ResNet 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
ResNet18 Ran
ran; our contract draft was wrong, not the code
theot1/dign/DiGN.py
code served (permissive licence) · get_code("6dfc13a796fd363a")
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 ResNet. 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 torch.nn as nn

torch.manual_seed(0)

ARGS = {
    'num_classes': 10,
}
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 ResNet — a ResNet output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a ResNet 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
ResNet18 Ran
ran; our contract draft was wrong, not the code
TheoT1/DiGN/resnet.py
code served (permissive licence) · get_code("1fb78623e6d1393a")
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 ResNet. 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 torch.nn as nn

torch.manual_seed(0)

# ResNet18 function signature: ResNet18(num_classes=10, **kwargs)
# The **kwargs are passed to ResNet.__init__, which accepts:
# block, num_blocks, num_classes=10, feat_scale=1, wm=1, d_input=32

# Build a small input tensor for testing (batch_size=2, channels=3, height=32, width=32)
x = torch.randn(2, 3, 32, 32)

ARGS = {
    'num_classes': 10,
    # kwargs for ResNet:
    'feat_scale': 1,
    'wm': 1,
    'd_input': 32,
}
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 ResNet — a ResNet output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a ResNet 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
ResNet18Wide Ran
ran; our contract draft was wrong, not the code
theot1/dign/DiGN.py
code served (permissive licence) · get_code("09e581aa9c0b8963")
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 ResNet. 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 torch.nn as nn

torch.manual_seed(0)

ARGS = {
    'num_classes': 10,
}
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 ResNet — a ResNet output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a ResNet 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
ResNet18Wide Ran
ran; our contract draft was wrong, not the code
TheoT1/DiGN/resnet.py
code served (permissive licence) · get_code("46fb144394e21588")
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 ResNet. 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 torch.nn as nn

torch.manual_seed(0)

# ResNet18Wide takes num_classes as the main parameter
# It calls ResNet which takes block, num_blocks, num_classes, feat_scale, wm, d_input
# ResNet18Wide passes BasicBlock and [2, 2, 2, 2] as num_blocks, wm=5
# and accepts **kwargs for num_classes, feat_scale, d_input

# Looking at ResNet.forward, it expects:
# - x: input tensor (batch, 3, height, width)
# - with_latent: bool (optional)
# - fake_relu: bool (optional)
# - no_relu: bool (optional)

# Create a small input tensor: batch_size=2, channels=3, height=4, width=4
x = torch.randn(2, 3, 4, 4)

ARGS = {
    "num_classes": 10,
}
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 ResNet — a ResNet output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a ResNet 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
ResNet18Wide_64 Ran
ran; our contract draft was wrong, not the code
theot1/dign/DiGN.py
code served (permissive licence) · get_code("369c908be6ae465b")
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 ResNet. 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 torch.nn as nn

torch.manual_seed(0)

# ResNet18Wide_64 has parameters: num_classes and **kwargs
# Looking at the function signature and how it's used:
# - num_classes: int, passed to ResNet
# - **kwargs: passed to ResNet, which accepts feat_scale and wm (but wm=5 is hardcoded in ResNet18Wide_64)
#   From ResNet.__init__, it accepts: block, num_blocks, num_classes, feat_scale, wm, d_input
#   The function hardcodes: block=BasicBlock, num_blocks=[2,2,2,2], wm=5, d_input=64
#   So kwargs could contain feat_scale or other params that ResNet accepts

# Create minimal valid arguments
ARGS = {
    'num_classes': 200,
}
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 ResNet — a ResNet output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a ResNet 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
ResNet18_64 Ran
ran; our contract draft was wrong, not the code
theot1/dign/DiGN.py
code served (permissive licence) · get_code("970d810339028979")
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 ResNet. 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 torch.nn as nn

torch.manual_seed(0)

# ResNet18_64 expects:
# - num_classes: int
# - **kwargs: which may include feat_scale, wm, d_input (though d_input is hardcoded to 64)

ARGS = {
    "num_classes": 10,
}
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 ResNet — a ResNet output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a ResNet 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
ResNet18_64 Ran
ran; our contract draft was wrong, not the code
TheoT1/DiGN/resnet.py
code served (permissive licence) · get_code("7bc30cc42373d340")
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 ResNet. 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 torch.nn as nn

torch.manual_seed(0)

ARGS = {
    'num_classes': 10,
}
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 ResNet — a ResNet output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a ResNet 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
ResNet50 Ran
ran; our contract draft was wrong, not the code
theot1/dign/DiGN.py
code served (permissive licence) · get_code("a19e7b98eeb4b6cd")
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 ResNet. 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 torch.nn as nn

torch.manual_seed(0)

ARGS = {
    'num_classes': 10,
}
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 ResNet — a ResNet output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a ResNet 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
Transition Ran
ran on a synthesized input · constructed an object; no result was computed
theot1/dign/DiGN.py
code served (permissive licence) · get_code("da9155d90cea93d6")
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 an instance of its own class, Transition: the run constructed the object and did not call it, so no result was computed. 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 torch.nn as nn

torch.manual_seed(0)

# The Transition module takes in_planes and out_planes as constructor arguments
# in_planes: number of input channels
# out_planes: number of output channels
# These are used to construct BatchNorm2d and Conv2d layers

# For testing, we'll use small distinct values
in_planes = 4
out_planes = 3

ARGS = {
    "in_planes": in_planes,
    "out_planes": out_planes,
}
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 Transition — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a Transition 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
apply_op Ran
ran; our contract draft was wrong, not the code
theot1/dign/DiGN.py
code served (permissive licence) · get_code("cc5cf9b07ce0d984")
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 2, dtype family float, shape (4, 3) 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
from PIL import Image

# Create a simple operation function that takes a PIL image, severity, and d_input
def dummy_op(pil_img, severity, d_input):
    """Dummy operation that returns the image unchanged"""
    return pil_img

# Create a small test image as numpy array (normalized 0-1)
# Using shape (4, 3) for height x width to be distinct
test_image = np.random.rand(4, 3).astype(np.float32)

ARGS = {
    "image": test_image,
    "op": dummy_op,
    "severity": 0.5,
    "d_input": {"key": "value"}
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 2, dtype family float, shape (4, 3) — shape (4, 3) 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
augment Ran
ran; our contract draft was wrong, not the code
theot1/dign/DiGN.py
code served (permissive licence) · get_code("269d965e70b2c96f")
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, ndim 4, 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
import numpy.random as npr

# Set seed for reproducibility
npr.seed(0)

# Create a small sample image as float32 numpy array
# Shape: (height, width, channels)
# Using normalized values between 0 and 1
sample_image = np.random.uniform(0.1, 0.9, size=(32, 32, 3)).astype(np.float32)

ARGS = {
    'image': sample_image,
    'severity': 3,
    'width': 2,
    'depth': 2
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 4, dtype family float — a list output has no array shape of its own; none is stated
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
augment_and_mix Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
theot1/dign/DiGN.py
code served (permissive licence) · get_code("ec1ce1fd53588761")
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 float, shape (32, 32, 3) 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
from PIL import Image, ImageOps, ImageEnhance

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

# Create a small test image as float32 numpy array with shape (h, w, c)
# Using shape (32, 32, 3) for a small RGB image normalized to [0, 1]
test_image = np.random.rand(32, 32, 3).astype(np.float32)

# Define ARGS dict with all parameters for augment_and_mix
ARGS = {
    "image": test_image,
    "severity": 3,
    "width": 3,
    "depth": -1,
    "alpha": 1.0
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, 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 float, shape (32, 32, 3) — shape (32, 32, 3) 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
autocontrast Ran
ran; our contract draft was wrong, not the code
theot1/dign/DiGN.py
code served (permissive licence) · get_code("f9666ba13dd406de")
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 Image. 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 PIL import Image
import numpy as np

# Create a small PIL Image for testing
# Create a simple RGB image with distinct pixel values
img_array = np.array([
    [[255, 0, 0], [0, 255, 0]],
    [[0, 0, 255], [128, 128, 128]]
], dtype=np.uint8)
pil_img = Image.fromarray(img_array, mode='RGB')

# Second parameter is unused (underscore), so we pass None or any value
unused_param = None

# IMAGE_SIZE parameter - this is not used in the function body
IMAGE_SIZE = 224

ARGS = {
    'pil_img': pil_img,
    '_': unused_param,
    'IMAGE_SIZE': IMAGE_SIZE
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Image — a Image output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Image output the recorded ndim 3 and dtype family int 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
calibr_metrics Ran
ran; our contract draft was wrong, not the code
theot1/dign/test-cal-c.py
code served (permissive licence) · get_code("09d8f6afc431411e")
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) 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

# Create small test data
np.random.seed(0)

# p_pred: shape (n_samples, n_classes) - predicted probabilities
# Should be valid probabilities (sum to 1 along axis 1)
n_samples = 2
n_classes = 4
p_pred = np.random.dirichlet(np.ones(n_classes), size=n_samples)

# y_true: shape (n_samples,) - true class labels
y_true = np.array([0, 2])

ARGS = {
    'p_pred': p_pred,
    'y_true': y_true
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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
conv1x1 Ran
ran; our contract draft was wrong, not the code
TheoT1/DiGN/imagenet_models/leaky_resnet.py
code served (permissive licence) · get_code("d9def42110729a85")
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) and returned Conv2d. 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 torch.nn as nn

torch.manual_seed(0)

# The function conv1x1 takes in_planes, out_planes, and stride as parameters
# It returns an nn.Conv2d layer configured for 1x1 convolution
# We need to provide valid integer arguments for these parameters

ARGS = {
    "in_planes": 3,
    "out_planes": 64,
    "stride": 1
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Conv2d — a Conv2d output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Conv2d 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
conv3x3 Ran
ran; our contract draft was wrong, not the code
TheoT1/DiGN/imagenet_models/resnet.py
code served (permissive licence) · get_code("160bb14bd76201b4")
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) and returned Conv2d. 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 torch.nn as nn

torch.manual_seed(0)

ARGS = {
    "in_planes": 3,
    "out_planes": 64,
    "stride": 1,
    "groups": 1,
    "dilation": 1,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Conv2d — a Conv2d output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Conv2d 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
conv3x3 Ran
ran; our contract draft was wrong, not the code
TheoT1/DiGN/imagenet_models/leaky_resnet.py
pointer only (licence: MIT); the stored copy get_code serves, from lixiaotong97/DSU, is pointer only (licence: NONE) · get_code("fac5364e2f53c6db")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from lixiaotong97/DSU, 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.

equalize Ran
ran; our contract draft was wrong, not the code
theot1/dign/DiGN.py
code served (permissive licence) · get_code("d1d727ae7d403f9d")
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 Image. 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
from PIL import Image
import io

torch.manual_seed(0)

# Create a small PIL image
pil_img = Image.new('RGB', (4, 3), color=(100, 150, 200))

# The function signature is: equalize(pil_img, _, IMAGE_SIZE)
# - pil_img: PIL Image object (used with ImageOps.equalize)
# - _: second parameter (unused in function body)
# - IMAGE_SIZE: third parameter (unused in function body)

ARGS = {
    'pil_img': pil_img,
    '_': None,
    'IMAGE_SIZE': 224
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Image — a Image output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Image output the recorded ndim 3 and dtype family int 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_parameter Ran
ran, honoured the contract we drafted
theot1/dign/DiGN.py
code served (permissive licence) · get_code("340cb3797f4ed00a")
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) 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
import numpy as np

torch.manual_seed(0)

# Analyzing the function:
# - level: a numeric value between 0 and PARAMETER_MAX (appears to be 10 based on division by 10.0)
# - maxval: a numeric value representing the maximum value to scale
# The function returns: float(level) * maxval / 10.0

# Simple test values
level = 5
maxval = 20.0

ARGS = {
    "level": level,
    "maxval": maxval,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
numpy, 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
genAugs Ran
ran; our contract draft was wrong, not the code
theot1/dign/DiGN.py
code served (permissive licence) · get_code("6628458b28e57fbb")
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, ndim 5, dtype family float; no output shape is stated. 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 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
import torch
from PIL import Image
from PIL import Image, ImageOps, ImageEnhance
import types as _types

torch.manual_seed(0)

def autocontrast(pil_img, _, IMAGE_SIZE):
    return ImageOps.autocontrast(pil_img)

def equalize(pil_img, _, IMAGE_SIZE):
    return ImageOps.equalize(pil_img)

def float_parameter(level, maxval):
    return float(level) * maxval / 10.0

def int_parameter(level, maxval):
    return int(level * maxval / 10)

def posterize(pil_img, level, IMAGE_SIZE):
    level = int_parameter(sample_level(level), 4)
    return ImageOps.posterize(pil_img, 4 - level)

def rotate(pil_img, level, IMAGE_SIZE):
    degrees = int_parameter(sample_level(level), 30)
    if np.random.uniform() > 0.5:
        degrees = -degrees
    return pil_img.rotate(degrees, resample=Image.BILINEAR)

def sample_level(n):
    return np.random.uniform(low=0.1, high=n)

def shear_x(pil_img, level, IMAGE_SIZE):
    level = float_parameter(sample_level(level), 0.3)
    if np.random.uniform() > 0.5:
        level = -level
    return pil_img.transform((IMAGE_SIZE, IMAGE_SIZE), Image.AFFINE, (1, level, 0, 0, 1, 0), resample=Image.BILINEAR)

def shear_y(pil_img, level, IMAGE_SIZE):
    level = float_parameter(sample_level(level), 0.3)
    if np.random.uniform() > 0.5:
        level = -level
    return pil_img.transform((IMAGE_SIZE, IMAGE_SIZE), Image.AFFINE, (1, 0, 0, level, 1, 0), resample=Image.BILINEAR)

def solarize(pil_img, level, IMAGE_SIZE):
    level = int_parameter(sample_level(level), 256)
    return ImageOps.solarize(pil_img, 256 - level)

def translate_x(pil_img, level, IMAGE_SIZE):
    level = int_parameter(sample_level(level), IMAGE_SIZE / 3)
    if np.random.random() > 0.5:
        level = -level
    return pil_img.transform((IMAGE_SIZE, IMAGE_SIZE), Image.AFFINE, (1, 0, level, 0, 1, 0), resample=Image.BILINEAR)

def translate_y(pil_img, level, IMAGE_SIZE):
    level = int_parameter(sample_level(level), IMAGE_SIZE / 3)
    if np.random.random() > 0.5:
        level = -level
    return pil_img.transform((IMAGE_SIZE, IMAGE_SIZE), Image.AFFINE, (1, 0, 0, 0, 1, level), resample=Image.BILINEAR)

augmentations_list = [autocontrast, equalize, posterize, rotate, solarize, shear_x, shear_y, translate_x, translate_y]

def apply_op(image, op, severity, d_input):
    image = np.clip(image * 255.0, 0, 255).astype(np.uint8)
    pil_img = Image.fromarray(image)
    pil_img = op(pil_img, severity, d_input)
    return np.asarray(pil_img) / 255.0

def augment(image, severity=3, width=3, depth=-1):
    d_input = image.shape[0]
    augs_list = []
    for i in range(width):
        image_aug = image.copy()
        d = depth if depth > 0 else np.random.randint(1, 4)
        for _ in range(d):
            op = np.random.choice(augmentations_list)
            image_aug = apply_op(image_aug, op, severity, d_input)
        augs_list.append(image_aug)
    return augs_list

def genAugs(x, width=3, depth=-1):
    n = x.shape[0]
    x_augs = []
    for j in range(width):
        x_augs.append(torch.zeros_like(x).cpu())
    for i in range(n):
        x_np = x[i].cpu().numpy()
        severity = np.random.randint(1, 11)
        x_aug_list = augment(x_np.transpose(1, 2, 0), severity=severity, width=width)
        for j in range(width):
            x_augs[j][i] = torch.from_numpy(x_aug_list[j].transpose(2, 0, 1))
    return x_augs

x = torch.rand(2, 3, 4, 4)

ARGS = {
    "x": x,
    "width": 3,
    "depth": -1,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 5, dtype family float — a list output has no array shape of its own; none is stated
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
get_loader_from_path Ran
ran; our contract draft was wrong, not the code
TheoT1/DiGN/DataLoad.py
code served (permissive licence) · get_code("66d02694965ae38a")
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
import tempfile
import os
from PIL import Image
import numpy as np

torch.manual_seed(0)

# Create a temporary directory with a simple image dataset structure
temp_dir = tempfile.mkdtemp()
class_dir = os.path.join(temp_dir, "class_0")
os.makedirs(class_dir, exist_ok=True)

# Create a small dummy image file
img_array = np.random.randint(0, 256, (32, 32, 3), dtype=np.uint8)
img = Image.fromarray(img_array)
img.save(os.path.join(class_dir, "dummy_0.jpg"))

# Create another dummy image
img_array2 = np.random.randint(0, 256, (32, 32, 3), dtype=np.uint8)
img2 = Image.fromarray(img_array2)
img2.save(os.path.join(class_dir, "dummy_1.jpg"))

ARGS = {
    "path": temp_dir,
    "batch_size": 2,
    "num_workers": 0,
    "d_input": 64,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy, torch, torchvision — 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
get_loaders_imagenette Ran
ran; our contract draft was wrong, not the code
TheoT1/DiGN/DataLoad.py
code served (permissive licence) · get_code("27fddf2dbd26fc4b")
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, ndim 1, 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 torch
import tempfile
import os
from PIL import Image
import numpy as np

torch.manual_seed(0)

# Create a temporary directory structure mimicking ImageNette
temp_dir = tempfile.mkdtemp()
train_dir = os.path.join(temp_dir, 'train')
val_dir = os.path.join(temp_dir, 'val')

os.makedirs(train_dir, exist_ok=True)
os.makedirs(val_dir, exist_ok=True)

# Create class subdirectories and dummy images
for split_dir in [train_dir, val_dir]:
    for class_idx in range(2):
        class_dir = os.path.join(split_dir, str(class_idx))
        os.makedirs(class_dir, exist_ok=True)
        
        # Create 2 dummy images per class
        for img_idx in range(2):
            img_array = np.random.randint(0, 256, (32, 32, 3), dtype=np.uint8)
            img = Image.fromarray(img_array)
            img_path = os.path.join(class_dir, f'image_{img_idx}.jpg')
            img.save(img_path)

ARGS = {
    'data_path': temp_dir,
    'batch_size_train': 2,
    'batch_size_val': 2,
    'num_workers': 0,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy, torch, torchvision — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family float — a tuple output has no array shape of its own; none is stated
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
get_loaders_tinyimagenet Ran
ran; our contract draft was wrong, not the code
TheoT1/DiGN/DataLoad.py
code served (permissive licence) · get_code("32bd96a059966be0")
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, ndim 1, 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 torch
import tempfile
import os
from PIL import Image
import shutil

torch.manual_seed(0)

# Create a temporary directory structure mimicking Tiny-ImageNet
temp_dir = tempfile.mkdtemp()
train_dir = os.path.join(temp_dir, 'train')
val_dir = os.path.join(temp_dir, 'val')

os.makedirs(train_dir)
os.makedirs(val_dir)

# Create dummy class directories and images for train set
for class_id in range(2):
    class_dir = os.path.join(train_dir, f'class_{class_id}')
    os.makedirs(class_dir)
    for img_id in range(2):
        img = Image.new('RGB', (64, 64), color=(73, 109, 137))
        img.save(os.path.join(class_dir, f'img_{img_id}.jpg'))

# Create dummy class directories and images for val set
for class_id in range(2):
    class_dir = os.path.join(val_dir, f'class_{class_id}')
    os.makedirs(class_dir)
    for img_id in range(2):
        img = Image.new('RGB', (64, 64), color=(73, 109, 137))
        img.save(os.path.join(class_dir, f'img_{img_id}.jpg'))

ARGS = {
    'data_path': temp_dir,
    'batch_size_train': 2,
    'batch_size_val': 2,
    'num_workers': 0,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, torch, torchvision — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family float — a tuple output has no array shape of its own; none is stated
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
int_parameter Ran
ran, honoured the contract we drafted
theot1/dign/DiGN.py
code served (permissive licence) · get_code("8b20dcbdf9e59b56")
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) 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 torch

torch.manual_seed(0)

ARGS = {
    "level": 5,
    "maxval": 100
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
torch — 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
inv_normalize Ran
ran; our contract draft was wrong, not the code
theot1/dign/DiGN.py
code served (permissive licence) · get_code("63aa40a3b7aafb93")
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, ndim 2, 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 torch

torch.manual_seed(0)

# The function inv_normalize iterates over tensor, mean, and std with zip,
# and calls mul_ and add_ on each element of tensor.
# So tensor should be iterable of tensors, and mean/std should be iterables of scalars.

# Create a batch of 2 tensors, each of dimension 4
tensor = [torch.randn(4), torch.randn(4)]

# Create mean and std as lists of scalars (one per tensor in the batch)
mean = [0.5, 0.3]
std = [0.2, 0.1]

ARGS = {
    "tensor": tensor,
    "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 list, ndim 2, dtype family float — a list output has no array shape of its own; none is stated
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
normalize Ran
ran, honoured the contract we drafted
theot1/dign/DiGN.py
code served (permissive licence) · get_code("12e0fe0c3812d359")
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
import numpy as np

torch.manual_seed(0)

# The function normalizes a tensor by subtracting mean and dividing by std
# It iterates over tensor, mean, std using zip
# Each element is modified in-place with sub_ and div_ operations
# This suggests tensor, mean, std are all iterables of tensors/numbers

# Create a batch of 2 samples, each with 4 features
tensor = torch.tensor([
    [1.0, 2.0, 3.0, 4.0],
    [5.0, 6.0, 7.0, 8.0]
], dtype=torch.float32)

# Mean values for each of the 4 features
mean = torch.tensor([1.0, 2.0, 3.0, 4.0], dtype=torch.float32)

# Standard deviation values for each of the 4 features
std = torch.tensor([1.0, 1.0, 1.0, 1.0], dtype=torch.float32)

ARGS = {
    "tensor": tensor,
    "mean": mean,
    "std": std
}
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 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
posterize Ran
ran; our contract draft was wrong, not the code
theot1/dign/DiGN.py
code served (permissive licence) · get_code("7762e73a88022552")
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 Image. 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
from PIL import Image, ImageOps, ImageEnhance

# Create a small PIL image for testing
pil_img = Image.new('RGB', (64, 64), color='red')

# Define arguments for the posterize function
ARGS = {
    'pil_img': pil_img,
    'level': 5,
    'IMAGE_SIZE': 64
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Image — a Image output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Image output the recorded ndim 3 and dtype family int 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
rotate Ran
ran; our contract draft was wrong, not the code
theot1/dign/DiGN.py
code served (permissive licence) · get_code("427c26c8b0d1e858")
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 Image. 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
from PIL import Image

# Create a small PIL image for testing
# Using RGB mode with a 32x32 image
pil_img = Image.new('RGB', (32, 32), color=(100, 150, 200))

# Define parameters for the rotate function
level = 5  # This will be used to sample and scale the rotation degrees
IMAGE_SIZE = 32  # Size of the image

ARGS = {
    'pil_img': pil_img,
    'level': level,
    'IMAGE_SIZE': IMAGE_SIZE
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Image — a Image output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Image output the recorded ndim 3 and dtype family int 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
sample_level Ran
ran, violated the contract we drafted
TheoT1/DiGN/augmentations.py
code served (permissive licence) · get_code("5cc8d4764ac07a35")
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) 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 numpy as np

def sample_level(n):
    return np.random.uniform(low=0.1, high=n)

ARGS = {
    "n": 5
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
numpy — 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
shear_x Ran
ran; our contract draft was wrong, not the code
theot1/dign/DiGN.py
code served (permissive licence) · get_code("8baad92985baea5d")
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 Image. 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
from PIL import Image

# Create a small PIL image for testing
# The function expects a PIL Image object
test_image = Image.new('RGB', (64, 64), color='red')

# Set random seed for reproducibility
np.random.seed(42)

# Define the arguments dictionary
ARGS = {
    'pil_img': test_image,
    'level': 5,
    'IMAGE_SIZE': 64
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Image — a Image output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Image output the recorded ndim 3 and dtype family int 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
shear_y Ran
ran; our contract draft was wrong, not the code
theot1/dign/DiGN.py
code served (permissive licence) · get_code("0af81fabe08c27e1")
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 Image. 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
from PIL import Image

# Create a small PIL image for testing
# Using RGB mode with a small 32x32 image
pil_img = Image.new('RGB', (32, 32), color=(100, 150, 200))

# Define parameters
level = 5  # Level between 0 and 10 (used in sample_level which samples uniform(0.1, level))
IMAGE_SIZE = 32  # Must match the image dimensions for the transform

ARGS = {
    'pil_img': pil_img,
    'level': level,
    'IMAGE_SIZE': IMAGE_SIZE
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Image — a Image output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Image output the recorded ndim 3 and dtype family int 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
solarize Ran
ran; our contract draft was wrong, not the code
theot1/dign/DiGN.py
code served (permissive licence) · get_code("58bdedec5152aaaa")
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 Image. 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
from PIL import Image

# Create a small test image
test_image = Image.new('RGB', (64, 64), color=(100, 150, 200))

ARGS = {
    'pil_img': test_image,
    'level': 5,
    'IMAGE_SIZE': 64
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Image — a Image output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Image output the recorded ndim 3 and dtype family int 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
translate_x Ran
ran; our contract draft was wrong, not the code
theot1/dign/DiGN.py
code served (permissive licence) · get_code("bb8ddf80c6f21c90")
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 Image. 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
from PIL import Image

# Create a small PIL image for testing
IMAGE_SIZE = 32
pil_img = Image.new('RGB', (IMAGE_SIZE, IMAGE_SIZE), color='red')

# Define ARGS dict with all parameters for translate_x
ARGS = {
    'pil_img': pil_img,
    'level': 5,
    'IMAGE_SIZE': IMAGE_SIZE
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Image — a Image output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Image output the recorded ndim 3 and dtype family int 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
translate_y Ran
ran; our contract draft was wrong, not the code
theot1/dign/DiGN.py
code served (permissive licence) · get_code("6785662f742188b4")
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 Image. 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
from PIL import Image

# Create a small PIL image for testing
# IMAGE_SIZE will be used in the function, so we create a square image
IMAGE_SIZE = 32
pil_img = Image.new('RGB', (IMAGE_SIZE, IMAGE_SIZE), color='red')

# Define the arguments dictionary
ARGS = {
    'pil_img': pil_img,
    'level': 5,
    'IMAGE_SIZE': IMAGE_SIZE
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Image — a Image output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Image output the recorded ndim 3 and dtype family int 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
DiGN Not yet run
unverified (harvested, no recorded run)
theot1/dign/DiGN.py
code served (permissive licence) · get_code("7cdbd71b633d65b8")
How it ran

no invocation recorded.

SequentialWithArgs Not yet run
unverified (harvested, no recorded run)
theot1/dign/DiGN.py
code served (permissive licence) · get_code("528114b7552b6911")
How it ran

no invocation recorded.

densenet121 Not yet run
unverified (harvested, no recorded run)
TheoT1/DiGN/imagenet_models/densenet.py
code served (permissive licence) · get_code("85c4757ab3246b61")
How it ran

no invocation recorded.

densenet161 Not yet run
unverified (harvested, no recorded run)
TheoT1/DiGN/imagenet_models/densenet.py
code served (permissive licence) · get_code("6b70762b5419b176")
How it ran

no invocation recorded.

densenet169 Not yet run
unverified (harvested, no recorded run)
TheoT1/DiGN/imagenet_models/densenet.py
code served (permissive licence) · get_code("019686268f24b3b5")
How it ran

no invocation recorded.

leaky_resnet18 Not yet run
unverified (harvested, no recorded run)
TheoT1/DiGN/imagenet_models/leaky_resnet.py
code served (permissive licence) · get_code("d5b851531e5607ed")
How it ran

no invocation recorded.

resnet18 Not yet run
unverified (harvested, no recorded run)
TheoT1/DiGN/imagenet_models/resnet.py
code served (permissive licence) · get_code("ba4c156b3db83275")
How it ran

no invocation recorded.

Also present in other papers' code

Of 52 samples lifted from this paper's linked repositories, 52 carry a function identity and 0 do not. Of the 48 distinct functions with an identity, 12 are also present in other papers whose code we hold (11 with at least one paper dated earlier, 4 the same month, 11 later) and 36 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
conv3x3 686 other papers
285 dated earlier, 19 the same month, 376 later, 6 same year, order not determinable · 31 only as an assembly brick
1406.2661 (2014-06)
Generative Adversarial Networks
1409.0575 (2014-09)
ImageNet Large Scale Visual Recognition Challenge
1412.6572 (2014-12)
Explaining and Harnessing Adversarial Examples
and 683 more not named here
conv1x1 431 other papers
114 dated earlier, 13 the same month, 298 later, 5 same year, order not determinable, 1 undated · 38 only as an assembly brick
1409.1556 (2014-09)
Very Deep Convolutional Networks for Large-Scale Image Recognition
1506.01186 (2015-06)
Cyclical Learning Rates for Training Neural Networks
1507.05717 (2015-07)
An End-to-End Trainable Neural Network for Image-based Sequence Recognition and Its Application to Scene Text Recognition
and 428 more not named here
conv3x3 312 other papers
66 dated earlier, 10 the same month, 232 later, 3 same year, order not determinable, 1 undated · 31 only as an assembly brick
1409.1556 (2014-09)
Very Deep Convolutional Networks for Large-Scale Image Recognition
1507.05717 (2015-07)
An End-to-End Trainable Neural Network for Image-based Sequence Recognition and Its Application to Scene Text Recognition
1512.03385 (2015-12)
Deep Residual Learning for Image Recognition
and 309 more not named here
resnet18 95 other papers
23 dated earlier, 3 the same month, 66 later, 2 same year, order not determinable, 1 undated
1805.00907 (2018-05)
Glow: Graph Lowering Compiler Techniques for Neural Networks
1902.08153 (2019-02)
Learned Step Size Quantization
1904.01355 (2019-04)
FCOS: Fully Convolutional One-Stage Object Detection
and 92 more not named here
float_parameter
brick of this paper's assembly: helper
19 other papers
3 dated earlier, 15 later, 1 same year, order not determinable · 2 only as an assembly brick
1912.02781 (2019-12)
AugMix: A Simple Data Processing Method to Improve Robustness and Uncertainty
2006.16241 (2020-06)
The Many Faces of Robustness: A Critical Analysis of Out-of-Distribution Generalization
2009.09238 (2020-09)
EfficientDeRain: Learning Pixel-wise Dilation Filtering for High-Efficiency Single-Image Deraining
and 16 more not named here
int_parameter
brick of this paper's assembly: helper
19 other papers
3 dated earlier, 15 later, 1 same year, order not determinable · 2 only as an assembly brick
1912.02781 (2019-12)
AugMix: A Simple Data Processing Method to Improve Robustness and Uncertainty
2006.16241 (2020-06)
The Many Faces of Robustness: A Critical Analysis of Out-of-Distribution Generalization
2009.09238 (2020-09)
EfficientDeRain: Learning Pixel-wise Dilation Filtering for High-Efficiency Single-Image Deraining
and 16 more not named here
sample_level
brick of this paper's assembly: helper
17 other papers
3 dated earlier, 13 later, 1 same year, order not determinable · 3 only as an assembly brick
1912.02781 (2019-12)
AugMix: A Simple Data Processing Method to Improve Robustness and Uncertainty
2006.16241 (2020-06)
The Many Faces of Robustness: A Critical Analysis of Out-of-Distribution Generalization
2009.09238 (2020-09)
EfficientDeRain: Learning Pixel-wise Dilation Filtering for High-Efficiency Single-Image Deraining
and 14 more not named here
densenet121 8 other papers
3 dated earlier, 5 later
1905.00641 (2019-05)
RetinaFace: Single-stage Dense Face Localisation in the Wild
1912.00385 (2019-12)
The Group Loss for Deep Metric Learning
2010.11742 (2020-10)
Learning Black-Box Attackers with Transferable Priors and Query Feedback
and 5 more not named here
densenet161 8 other papers
4 dated earlier, 4 later
1905.00641 (2019-05)
RetinaFace: Single-stage Dense Face Localisation in the Wild
1905.11926 (2019-05)
Network Deconvolution
1912.00385 (2019-12)
The Group Loss for Deep Metric Learning
and 5 more not named here
densenet169 8 other papers
4 dated earlier, 4 later
1905.00641 (2019-05)
RetinaFace: Single-stage Dense Face Localisation in the Wild
1905.11926 (2019-05)
Network Deconvolution
1912.00385 (2019-12)
The Group Loss for Deep Metric Learning
and 5 more not named here
SequentialWithArgs
brick of this paper's assembly: helper
1 other paper
1 dated earlier · 1 only as an assembly brick
1512.03385 (2015-12)
Deep Residual Learning for Image Recognition
leaky_resnet18 1 other paper
1 later
2211.08583 (2022-11)
Empirical Study on Optimizer Selection for Out-of-Distribution Generalization

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

Deep neural networks achieve high prediction accuracy when the train and test distributions coincide. In practice though, various types of corruptions occur which deviate from this setup and cause severe performance degradations. Few methods have been proposed to address generalization in the presence of unforeseen domain shifts. In particular, digital noise corruptions arise commonly in practice during the image acquisition stage and present a significant challenge for current methods. In this paper, we propose a diverse Gaussian noise consistency regularization method for improving robustness of image classifiers under a variety of corruptions while still maintaining high clean accuracy. We derive bounds to motivate and understand the behavior of our Gaussian noise consistency regularization using a local loss landscape analysis. Our approach improves robustness against unforeseen noise corruptions by 4.2-18.4% over adversarial training and other strong diverse data augmentation baselines across several benchmarks. Furthermore, it improves robustness and uncertainty calibration by 3.7% and 5.5%, respectively, against all common corruptions (weather, digital, blur, noise) when combined with state-of-the-art diverse data augmentations. Code is available at https://github.com/TheoT1/DiGN.

For agents

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

get_harvested_code_for_paper("2104.01231")
get_code_for_paper("2104.01231")
have("2104.01231")

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

curl https://syntology.ai/api/ran/2104.01231.json

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

[![Syntology run record](https://syntology.ai/api/ran/2104.01231.svg)](https://syntology.ai/paper/2104.01231)

Connect an agent — have() is free.