SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1905.02249 · 2019

MixMatch: A Holistic Approach to Semi-Supervised Learning

Colin Raffel, David Berthelot, Nicholas Carlini, Nicolas Papernot, Avital Oliver, Ian Goodfellow

arXiv · PDF · Open in the Atlas

Code that ran

47 of 63 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; 8 honoured the contract we drafted, 0 violated it and 19 have no contract verdict; 20 failures of our instrument; 16 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/1905.02249.json · badge: /api/ran/1905.02249.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
ntozer/mixmatch-tensorflow2.0 community
Papers with Code archive
6 ran · 0 no recorded run
dhx000/DGM_project community
Papers with Code archive
5 ran · 5 no recorded run
viig99/mixmatch-freesound community (classifier label)
classifier basis not recorded on the link
5 ran · 3 no recorded run
FelixAbrahamsson/mixmatch-pytorch community
Papers with Code archive
5 ran · 0 no recorded run
gan3sh500/mixmatch-pytorch community
Papers with Code archive
3 ran · 2 no recorded run
atinghosh/mixmatch_pytorch community
Papers with Code archive
3 ran · 0 no recorded run
rit-git/Snippext_public community
Papers with Code archive
3 ran · 0 no recorded run
smkim7-kr/albu-MixMatch-pytorch community
Papers with Code archive
3 ran · 0 no recorded run
kevinghst/mixmatch community (classifier label)
classifier basis not recorded on the link
2 ran · 3 no recorded run
DamianValle/ScarceClassifier community
Papers with Code archive
2 ran · 1 no recorded run
yuxi120407/mixmatch_tensorflow community
Papers with Code archive
2 ran · 1 no recorded run
narendoraiswamy/MixMatch-pytorch-demo community (classifier label)
classifier basis not recorded on the link
2 ran · 1 no recorded run
dLutscher/MixMatch-TransferLearning community
Papers with Code archive
2 ran · 0 no recorded run
perrying/realistic-ssl-evaluation-pytorch community
Papers with Code archive
1 ran · 0 no recorded run
Jeffkang-94/Mixmatch-pytorch-SSL community
Papers with Code archive
1 ran · 0 no recorded run
YU1ut/MixMatch-pytorch community
Papers with Code archive
1 ran · 0 no recorded run
fbuchert/mixmatch-pytorch community
Papers with Code archive
1 ran · 0 no recorded run
FunctionStatusWhere it lives
Augment Ran
ran on a synthesized input
DamianValle/ScarceClassifier/src/mixmatch.py
pointer only (licence: NONE) · get_code("f64393cdb0ae8878")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, DamianValle/ScarceClassifier 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.

BasicBlock Ran
ran on a synthesized input · constructed an object; no result was computed
atinghosh/mixmatch_pytorch/network_mod.py
pointer only (licence: NONE) · get_code("e4dc4ca4b40de499")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, atinghosh/mixmatch_pytorch 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.

MixMatch Ran
ran on a synthesized input
perrying/realistic-ssl-evaluation-pytorch/lib/algs/mixmatch.py
code served (permissive licence) · get_code("1733a4add804ce42")
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 0, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

# Create a simple mock model that has update_batch_stats method
class MockModel(nn.Module):
    def __init__(self, input_dim=8, hidden_dim=6, num_classes=5):
        super().__init__()
        self.fc1 = nn.Linear(input_dim, hidden_dim)
        self.fc2 = nn.Linear(hidden_dim, num_classes)
        self.bn = nn.BatchNorm1d(hidden_dim)
        
    def forward(self, x):
        x = self.fc1(x)
        x = self.bn(x)
        x = torch.relu(x)
        x = self.fc2(x)
        return x
    
    def update_batch_stats(self, flag):
        # Toggle batch norm training mode
        self.bn.training = flag

# Initialize the mock model
model = MockModel(input_dim=8, hidden_dim=6, num_classes=5)

# INIT_ARGS for MixMatch
INIT_ARGS = {
    'temperature': 0.5,
    'n_augment': 2,
    'alpha': 0.75
}

# ARGS for forward method
batch_size = 7
input_dim = 8
num_classes = 5
n_unlabeled = 4

# Create input tensor (both labeled and unlabeled)
x = torch.randn(batch_size, input_dim)

# Create dummy labels (not used in forward but parameter exists)
y = torch.randn(batch_size, num_classes)

# Create mask: 1 for unlabeled, 0 for labeled
# We need at least n_unlabeled samples with mask==1
mask = torch.zeros(batch_size)
mask[:n_unlabeled] = 1

ARGS = {
    'x': x,
    'y': y,
    'model': model,
    'mask': mask
}
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 0, 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)
MixMatchLoss Ran
ran on a synthesized input · constructed an object; no result was computed
Jeffkang-94/Mixmatch-pytorch-SSL/SSL_loss/mixmatch.py
pointer only (licence: NONE) · get_code("4de37e26980b9f34")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, Jeffkang-94/Mixmatch-pytorch-SSL 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.

NetworkBlock Ran
ran on a synthesized input · constructed an object; no result was computed
atinghosh/mixmatch_pytorch/network_mod.py
pointer only (licence: NONE) · get_code("2950d5425e9dd18b")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, atinghosh/mixmatch_pytorch 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.

RandomCrop Ran
ran on a synthesized input
DamianValle/ScarceClassifier/src/mixmatch.py
pointer only (licence: NONE) · get_code("3f2a5048d7bc9b95")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, DamianValle/ScarceClassifier 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.

Residual3x3Unit Ran
ran on a synthesized input · constructed an object; no result was computed
yuxi120407/mixmatch_tensorflow/model.py
pointer only (licence: NONE) · get_code("a50bcc606f57701c")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, yuxi120407/mixmatch_tensorflow 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.

ResidualBlock Ran
ran on a synthesized input · constructed an object; no result was computed
yuxi120407/mixmatch_tensorflow/model.py
pointer only (licence: NONE) · get_code("251561e345918b21")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, yuxi120407/mixmatch_tensorflow 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.

SemiLoss Ran
ran on a synthesized input · constructed an object; no result was computed
dLutscher/MixMatch-TransferLearning/utils/utils.py
pointer only (licence: NONE) · get_code("48a12c06cbdd6b70")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, dLutscher/MixMatch-TransferLearning 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.

WideResNet Ran
ran on a synthesized input · constructed an object; no result was computed
YU1ut/MixMatch-pytorch/models/wideresnet.py
code served (permissive licence) · get_code("c3d0f2f3671b4bbc")
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, WideResNet: 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)

ARGS = {
    "num_classes": 10,
    "depth": 28,
    "widen_factor": 2,
    "dropRate": 0.0,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type WideResNet — 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 WideResNet 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
Wide_ResNet Ran
ran on a synthesized input · constructed an object; no result was computed
atinghosh/mixmatch_pytorch/network_mod.py
pointer only (licence: NONE) · get_code("8608707dd0f937bd")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, atinghosh/mixmatch_pytorch 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.

augment Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
ntozer/mixmatch-tensorflow2.0/mixmatch.py
code served (permissive licence) · get_code("647302bd0cbc0cd4")
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 EagerTensor, ndim 4, dtype family float, shape (2, 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 tensorflow as tf
import numpy as np

# Create a small batch of image-like tensors
# The function expects input shape compatible with image operations
# Input: (batch_size, height, width, channels)
# After padding: (batch_size, height+8, width+8, channels)
# After crop: (batch_size, 32, 32, channels)
batch_size = 2
height = 28
width = 28
channels = 3

# Create a tensor with shape (2, 28, 28, 3) - small batch of 28x28 RGB images
x = tf.constant(np.random.randint(0, 256, size=(batch_size, height, width, channels), dtype=np.uint8), dtype=tf.uint8)

# Convert to float for augmentation operations if needed
x = tf.cast(x, tf.float32)

ARGS = {
    "x": x
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type EagerTensor, ndim 4, dtype family float, shape (2, 32, 32, 3) — shape (2, 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
augment Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
dhx000/DGM_project/mixmatch/cycle_mixmatch.py
pointer only (licence: NONE) · get_code("39827536f2c0ed9a")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, dhx000/DGM_project 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.

calculate_overall_lwlrap_sklearn Ran
ran on a synthesized input
viig99/mixmatch-freesound/analysis/compute_lwlrap.py
code served (permissive licence) · get_code("c9498d65f32e1148")
How it ran

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

import numpy as np
import sklearn.metrics

# Create small sample data for the function
# truth should be a binary matrix (num_samples, num_labels)
# scores should be a score matrix (num_samples, num_labels)

np.random.seed(0)

# Create truth matrix: 3 samples, 4 labels (binary)
truth = np.array([
    [1, 0, 1, 0],
    [0, 1, 1, 1],
    [1, 1, 0, 0]
], dtype=np.float32)

# Create scores matrix: 3 samples, 4 labels (continuous scores)
scores = np.array([
    [0.9, 0.2, 0.8, 0.1],
    [0.3, 0.7, 0.6, 0.5],
    [0.8, 0.6, 0.4, 0.2]
], dtype=np.float32)

ARGS = {
    "truth": truth,
    "scores": scores
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, sklearn — 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
categorize Ran
ran on a synthesized input
narendoraiswamy/MixMatch-pytorch-demo/dataset/_cifar10.py
code served (permissive licence) · get_code("d687dcb3c4ba5861")
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 1, dtype family int; 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 numpy as np

torch.manual_seed(0)

# The function iterates over cifar_targets and looks up each element in id_to_trainid dict
# cifar_targets should be an iterable of integers in range [0, 9]
cifar_targets = [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]

ARGS = {
    "cifar_targets": cifar_targets
}
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 list, ndim 1, dtype family int — 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
classifier_criterion Ran
ran; our contract draft was wrong, not the code
rit-git/Snippext_public/snippext/mixmatchnl.py
code served (permissive licence) · get_code("3cac74cd5cc4e2f1")
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 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# Create small tensors for the classification task
# pred_labeled: predicted float tensor (batch_size=2, num_classes=4)
# y_labeled: ground truth float tensor (batch_size=2, num_classes=4)
# Both should be probability-like tensors (non-negative, sum to 1 for proper cross entropy)

batch_size = 2
num_classes = 4

# Create ground truth labels as one-hot encoded vectors
y_labeled = torch.zeros(batch_size, num_classes)
y_labeled[0, 0] = 1.0  # First sample is class 0
y_labeled[1, 2] = 1.0  # Second sample is class 2

# Create predicted logits and convert to log probabilities
pred_logits = torch.randn(batch_size, num_classes)
pred_labeled = torch.log_softmax(pred_logits, dim=-1)

ARGS = {
    "pred_labeled": pred_labeled,
    "y_labeled": y_labeled,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 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
collate_fn Ran
ran on a synthesized input
viig99/mixmatch-freesound/dataset/freesound.py
code served (permissive licence) · get_code("c9ad1e68bbc3805b")
How it ran

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

import numpy as np
import torch

torch.manual_seed(0)

# The function expects a batch parameter which is an iterable of (spec, label) tuples
# Each spec is array-like that supports indexing with [..., ::2]
# Each label is array-like that can be stacked

# Create sample specs (small arrays with distinct dimensions)
spec1 = np.random.randn(5, 8)  # 5x8 spec
spec2 = np.random.randn(5, 8)  # 5x8 spec

# Create sample labels
label1 = np.array([0])
label2 = np.array([1])

# Create batch as list of tuples
batch = [
    (spec1, label1),
    (spec2, label2),
]

ARGS = {
    "batch": batch,
}
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 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
collate_fn_unlabbelled Ran
ran on a synthesized input
viig99/mixmatch-freesound/dataset/freesound.py
code served (permissive licence) · get_code("45f16173aaa12967")
How it ran

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

import numpy as np
import torch

torch.manual_seed(0)

# Create sample batch items
# Each item is (spec, label) where spec is a tuple of two spectrograms
batch_items = []

for i in range(2):
    # spec is a tuple of (spec1, spec2)
    spec1 = np.random.randn(4, 8).astype(np.float32)  # shape (4, 8)
    spec2 = np.random.randn(4, 8).astype(np.float32)  # shape (4, 8)
    spec = (spec1, spec2)
    
    # label is a single value (e.g., class index)
    label = np.array(i, dtype=np.int64)
    
    batch_items.append((spec, label))

batch = batch_items

ARGS = {
    "batch": batch
}
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 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
get_weak_augmentation Ran
ran; our contract draft was wrong, not the code
fbuchert/mixmatch-pytorch/augmentation/augmentations.py
code served (permissive licence) · get_code("fbc0e34ee7b8cdb5")
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 Compose. 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 torchvision import transforms

torch.manual_seed(0)

ARGS = {
    'img_size': 32,
    'padding': 4,
    'padding_mode': 'reflect'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
torch, torchvision — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Compose — a Compose output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Compose 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
guess_labels Ran
ran; our contract draft was wrong, not the code
ntozer/mixmatch-tensorflow2.0/mixmatch.py
code served (permissive licence) · get_code("e364b83801e8d8c9")
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 EagerTensor, ndim 2, dtype family float, shape (2, 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 tensorflow as tf
import numpy as np

# Create a simple model that takes input and outputs logits
class SimpleClassifier(tf.keras.Model):
    def __init__(self, num_classes=3):
        super(SimpleClassifier, self).__init__()
        self.dense = tf.keras.layers.Dense(num_classes)
    
    def call(self, x):
        return self.dense(x)

# Initialize the model
model = SimpleClassifier(num_classes=3)

# Create K augmented versions of input data
# u_aug is a list of K tensors, each with shape (batch_size, feature_dim)
K = 3
batch_size = 2
feature_dim = 4

u_aug = [
    tf.constant(np.random.randn(batch_size, feature_dim), dtype=tf.float32),
    tf.constant(np.random.randn(batch_size, feature_dim), dtype=tf.float32),
    tf.constant(np.random.randn(batch_size, feature_dim), dtype=tf.float32),
]

ARGS = {
    "u_aug": u_aug,
    "model": model,
    "K": K,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type EagerTensor, ndim 2, dtype family float, shape (2, 3) — shape (2, 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
guess_targets Ran
ran; our contract draft was wrong, not the code
FelixAbrahamsson/mixmatch-pytorch/mixmatch_pytorch/mixmatch_batch.py
pointer only (licence: NONE) · get_code("568ec1709ff7ca07")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, FelixAbrahamsson/mixmatch-pytorch 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.

interleave Ran
ran; our contract draft was wrong, not the code
smkim7-kr/albu-MixMatch-pytorch/src/utils.py
code served (permissive licence) · get_code("87035ef75e16edc6")
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 3, 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)

# xy is a list of tensors
# The function indexes into each tensor using offsets, so they need to support indexing along dim 0
# Each tensor in xy should have batch dimension as first dimension
batch = 8
nu = 2  # len(xy) - 1 = 3, so nu = 2

# Create 3 tensors (nu + 1 = 3)
# Each tensor has shape (batch, feature_dim)
xy = [
    torch.randn(batch, 4),
    torch.randn(batch, 4),
    torch.randn(batch, 4),
]

ARGS = {
    "xy": xy,
    "batch": batch,
}
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 3, 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
interleave Ran
ran; our contract draft was wrong, not the code
ntozer/mixmatch-tensorflow2.0/mixmatch.py
code served (permissive licence) · get_code("dc463911b8db674e")
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 3, 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 tensorflow as tf

# Create small test tensors - the function expects a list of tensors
# where each tensor has batch as the first dimension
batch_size = 8
feature_dim = 4

# Create 3 tensors (so nu = len(xy) - 1 = 2)
xy = [
    tf.random.normal((batch_size, feature_dim)),
    tf.random.normal((batch_size, feature_dim)),
    tf.random.normal((batch_size, feature_dim))
]

ARGS = {
    "xy": xy,
    "batch": batch_size
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 3, 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
interleave_offsets Ran
ran, honoured the contract we drafted
smkim7-kr/albu-MixMatch-pytorch/src/utils.py
code served (permissive licence) · get_code("73e81fa7395d8998")
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 1, dtype family int; 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)

ARGS = {
    "batch": 10,
    "nu": 2
}
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 1, dtype family int — 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
label_binarizer Ran
ran on a synthesized input
viig99/mixmatch-freesound/dataset/freesound.py
code served (permissive licence) · get_code("ad7aa34a23b546b2")
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 MultiLabelBinarizer. 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 sklearn import preprocessing

# MultiLabelBinarizer.fit() expects an iterable of iterables
# where each element represents a sample with multiple labels
# We'll create a small dataset with multi-label data

data = [
    ['label_0', 'label_1'],
    ['label_1', 'label_2'],
    ['label_0', 'label_2'],
    ['label_1']
]

ARGS = {
    'data': data
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, sklearn, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type MultiLabelBinarizer — a MultiLabelBinarizer output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a MultiLabelBinarizer 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
linear_rampup Ran
ran, honoured the contract we drafted
dLutscher/MixMatch-TransferLearning/utils/utils.py
pointer only (licence: NONE) · get_code("0aaac678344cd89f")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, dLutscher/MixMatch-TransferLearning 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.

mixmatch Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
ntozer/mixmatch-tensorflow2.0/mixmatch.py
code served (permissive licence) · get_code("91c88dac1ac09bee")
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 tensorflow as tf
import numpy as np

# Build a minimal model that takes input and returns logits
class SimpleModel(tf.Module):
    def __init__(self):
        super(SimpleModel, self).__init__()
        self.dense = tf.keras.layers.Dense(5)  # 5 classes for output logits
    
    def __call__(self, x):
        # x shape: (batch, 32, 32, 3)
        # Flatten and pass through dense layer
        batch_size = tf.shape(x)[0]
        x_flat = tf.reshape(x, [batch_size, -1])
        logits = self.dense(x_flat)
        return logits

# Create model instance
model = SimpleModel()

# Create small tensors with distinct dimensions
batch_size = 2
image_height = 32
image_width = 32
channels = 3
num_classes = 5

# x: labeled data (batch_size, 32, 32, 3)
x = tf.constant(np.random.randn(batch_size, image_height, image_width, channels).astype(np.float32))

# y: labeled targets (batch_size, num_classes) - one-hot encoded
y = tf.constant(np.eye(num_classes)[np.random.randint(0, num_classes, batch_size)].astype(np.float32))

# u: unlabeled data (batch_size, 32, 32, 3)
u = tf.constant(np.random.randn(batch_size, image_height, image_width, channels).astype(np.float32))

# T: temperature for sharpening (scalar)
T = 0.5

# K: number of augmentations
K = 2

# beta: mixing coefficient (scalar)
beta = tf.constant(0.75, dtype=tf.float32)

ARGS = {
    'model': model,
    'x': x,
    'y': y,
    'u': u,
    'T': T,
    'K': K,
    'beta': beta
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
numpy, tensorflow — 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
mixmatch Ran
ran, honoured the contract we drafted
rit-git/Snippext_public/snippext/mixmatchnl.py
code served (permissive licence) · get_code("9f0765feb50790dd")
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 Tensor, 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 torch.nn as nn
import torch.nn.functional as F
import numpy as np

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

class MultiTaskNet(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 3)
    
    def forward(self, x, y, task=None, get_enc=False, augment_batch=None, aug_enc=None, x_enc=None, second_batch=None):
        """
        Args:
            x: input tensor
            y: label tensor
            task: task name string
            get_enc: whether to return encoding
            augment_batch: tuple of (aug_x, lam) for augmentation
            aug_enc: augmentation encoding
            x_enc: pre-computed encoding
            second_batch: tuple of (index, lam) for mixup
        
        Returns:
            logits, y, something, encoding (if get_enc=True)
            or logits, y, something (if get_enc=False)
        """
        batch_size = x.size(0)
        logits = self.linear(x)
        
        if get_enc:
            enc = torch.randn(batch_size, 4)
            return logits, y, None, enc
        else:
            return logits, y, None

# Create model
model = MultiTaskNet()

# Create batch components
batch_size = 2
num_aug = 2
vocab_size = 3
feature_dim = 4

# Construct x: [labeled, unlabeled_0, aug_x, u_aug_1, u_aug_2]
# Total size: batch_size * (num_aug + 3) = 2 * 5 = 10
x = torch.randn(batch_size * (num_aug + 3), feature_dim)

# y: labeled targets (batch_size,)
y = torch.randint(0, vocab_size, (batch_size,))

# mask: dummy mask
mask = torch.ones(batch_size)

# Create batch tuple: (_, x, _, _, mask, y, _, taskname)
batch = (
    None,  # index 0
    x,     # index 1: x
    None,  # index 2
    None,  # index 3
    mask,  # index 4: mask
    y,     # index 5: y
    None,  # index 6
    ["classification"]  # index 7: taskname (list so we can index [0])
)

ARGS = {
    "model": model,
    "batch": batch,
    "num_aug": 2,
    "alpha": 0.4,
    "alpha_aug": 0.4,
    "u_lambda": 0.5
}
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 Tensor, 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
mixmatch Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
dhx000/DGM_project/mixmatch/cycle_mixmatch.py
pointer only (licence: NONE) · get_code("4350f70a89ff949b")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, dhx000/DGM_project 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.

mixmatch_batch Ran
ran; our contract draft was wrong, not the code
FelixAbrahamsson/mixmatch-pytorch/mixmatch_pytorch/mixmatch_batch.py
pointer only (licence: NONE) · get_code("3164897bcf3781a2")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, FelixAbrahamsson/mixmatch-pytorch 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.

mixup Ran
ran; our contract draft was wrong, not the code
smkim7-kr/albu-MixMatch-pytorch/src/utils.py
code served (permissive licence) · get_code("a083b3ef65c4a3c1")
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 torch
import numpy as np

torch.manual_seed(0)

class Args:
    def __init__(self):
        self.alpha = 1.0
        self.batch_size = 2

args = Args()

# Create small tensors for testing
# all_imgs: batch_size=2, channels=3, height=4, width=4
all_imgs = torch.randn(2, 3, 4, 4)

# all_targets: batch_size=2, num_classes=5 (soft targets for mixup)
all_targets = torch.randn(2, 5)

ARGS = {
    "args": args,
    "all_imgs": all_imgs,
    "all_targets": all_targets
}
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 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
mixup Ran
ran, honoured the contract we drafted
ntozer/mixmatch-tensorflow2.0/mixmatch.py
code served (permissive licence) · get_code("de552909167b4340")
How it ran

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

import tensorflow as tf
import numpy as np

# Create small sample tensors for mixup function
batch_size = 2
feature_dim = 4
num_classes = 3

# Create input data
x1 = tf.constant(np.random.randn(batch_size, feature_dim), dtype=tf.float32)
x2 = tf.constant(np.random.randn(batch_size, feature_dim), dtype=tf.float32)

# Create labels (one-hot encoded)
y1 = tf.constant(np.random.rand(batch_size, num_classes), dtype=tf.float32)
y2 = tf.constant(np.random.rand(batch_size, num_classes), dtype=tf.float32)

# Create beta value between 0 and 1
beta = tf.constant(0.7, dtype=tf.float32)

ARGS = {
    "x1": x1,
    "x2": x2,
    "y1": y1,
    "y2": y2,
    "beta": beta,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — 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
mixup Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
dhx000/DGM_project/mixmatch/cycle_mixmatch.py
pointer only (licence: NONE) · get_code("65eae886ef202b56")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, dhx000/DGM_project 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.

mixup_mod Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
gan3sh500/mixmatch-pytorch/mixmatch_utils.py
pointer only (licence: NONE) · get_code("77308af8bcbb2b2b")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, gan3sh500/mixmatch-pytorch 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.

mixup_samples Ran
ran; our contract draft was wrong, not the code
FelixAbrahamsson/mixmatch-pytorch/mixmatch_pytorch/mixmatch_batch.py
pointer only (licence: NONE) · get_code("9d112ebaba96ddb1")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, FelixAbrahamsson/mixmatch-pytorch 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.

plot_overlap Ran
ran; our contract draft was wrong, not the code
kevinghst/mixmatch/utils/logger.py
code served (permissive licence) · get_code("50afa2863ffd9fde")
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 list, 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 matplotlib.pyplot as plt
import numpy as np

class MockLogger:
    def __init__(self):
        self.title = "TestPlot"
        self.names = ["loss", "accuracy", "f1_score"]
        self.numbers = {
            "loss": [0.5, 0.4, 0.3, 0.2],
            "accuracy": [0.7, 0.75, 0.8, 0.85],
            "f1_score": [0.65, 0.72, 0.78, 0.82]
        }

logger = MockLogger()

ARGS = {
    "logger": logger,
    "names": None
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
matplotlib, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, 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
sharpen Ran
ran, honoured the contract we drafted
ntozer/mixmatch-tensorflow2.0/mixmatch.py
code served (permissive licence) · get_code("2777c70138515c4d")
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 EagerTensor, 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 tensorflow as tf
import numpy as np

# Create tiny probability tensor with batch size 2 and 4 dimensions
# Values should sum to 1 along axis 1 for valid probabilities
p_values = np.array([
    [0.1, 0.2, 0.3, 0.4],
    [0.25, 0.25, 0.25, 0.25]
], dtype=np.float32)

p = tf.constant(p_values)

# Temperature parameter - scalar value that controls sharpening
T = tf.constant(0.5, dtype=tf.float32)

ARGS = {
    "p": p,
    "T": T
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type EagerTensor, 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
sharpen Ran
ran, honoured the contract we drafted
FelixAbrahamsson/mixmatch-pytorch/mixmatch_pytorch/mixmatch_batch.py
pointer only (licence: NONE) · get_code("cbc1955a34800714")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, FelixAbrahamsson/mixmatch-pytorch 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.

sharpen Ran
ran, honoured the contract we drafted
gan3sh500/mixmatch-pytorch/mixmatch_utils.py
pointer only (licence: NONE) · get_code("6d22d34f03eff901")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, gan3sh500/mixmatch-pytorch 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.

sharpen Ran
ran, honoured the contract we drafted
dhx000/DGM_project/mixmatch/cycle_mixmatch.py
pointer only (licence: NONE) · get_code("63688837810c6e5a")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, dhx000/DGM_project 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.

should_apply_transform Ran
ran on a synthesized input
viig99/mixmatch-freesound/kaggle_submission.py
code served (permissive licence) · get_code("06ac2737867f1b5b")
How it ran

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

import random

random.seed(42)

ARGS = {
    "prob": 0.3
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type bool, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
split_U Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
dhx000/DGM_project/mixmatch/cycle_mixmatch.py
pointer only (licence: NONE) · get_code("3c7a33c95b913c66")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, dhx000/DGM_project 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.

tagging_criterion Ran
ran; our contract draft was wrong, not the code
rit-git/Snippext_public/snippext/mixmatchnl.py
code served (permissive licence) · get_code("670937fdbe1b61da")
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 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# Create tiny tensors for the tagging criterion function
# The function expects:
# - pred_labeled: predicted float tensor, shape (batch_size, seq_len)
# - y_labeled: groundtruth float tensor, same shape as pred_labeled
# The function computes cross entropy loss by:
#   1. Ignoring the 0'th dimension (column 0) with [:, 1:]
#   2. Computing -y_labeled[:, 1:] * log(pred_labeled[:, 1:])
#   3. Summing over last dimension and taking mean

batch_size = 2
seq_len = 4

# pred_labeled should be log probabilities (or at least floats where log can be taken)
# We'll use softmax output to ensure all values are positive and in (0, 1)
pred_labeled = torch.softmax(torch.randn(batch_size, seq_len), dim=-1)

# y_labeled should be probability-like values (non-negative, typically in [0, 1])
y_labeled = torch.softmax(torch.randn(batch_size, seq_len), dim=-1)

ARGS = {
    "pred_labeled": pred_labeled,
    "y_labeled": y_labeled,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 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
tile_adjacent Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
FelixAbrahamsson/mixmatch-pytorch/mixmatch_pytorch/mixmatch_batch.py
pointer only (licence: NONE) · get_code("caacacf94d563a78")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, FelixAbrahamsson/mixmatch-pytorch 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.

to_categorical Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
gan3sh500/mixmatch-pytorch/mixmatch_utils.py
pointer only (licence: NONE) · get_code("7ccbcdef1f97bdbf")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, gan3sh500/mixmatch-pytorch 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.

train_val_split Ran
ran on a synthesized input
kevinghst/mixmatch/dataset/cifar10.py
code served (permissive licence) · get_code("035b6b2199fe55d1")
How it ran

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

import numpy as np

np.random.seed(0)

# The function expects:
# 1. labels: array-like of class labels (0-9), needs at least 500 samples per class
# 2. n_labeled_per_class: int for how many labeled samples per class

# Create labels for 10 classes
# Each class needs: n_labeled_per_class + some unlabeled + 500 validation
# Let's use n_labeled_per_class = 10, so each class needs at least 510 + 10 = 520 samples
# Let's create 1000 samples per class for safety
labels_list = []
for class_id in range(10):
    labels_list.extend([class_id] * 1000)

labels = labels_list

n_labeled_per_class = 10

ARGS = {
    "labels": labels,
    "n_labeled_per_class": n_labeled_per_class,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
train_val_split Ran
ran on a synthesized input
narendoraiswamy/MixMatch-pytorch-demo/dataset/_cifar10.py
code served (permissive licence) · get_code("41670d21181a8756")
How it ran

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

import numpy as np

np.random.seed(0)

# The function expects:
# 1. labels: array-like of class labels (function converts to np.array)
# 2. n_labeled_per_class: int specifying how many labeled examples per class

# Create a simple labels array with 4 classes
# Need enough samples per class to satisfy the splits:
# - n_labeled_per_class samples for training labeled
# - remaining - 500 samples for training unlabeled
# - 500 samples for validation
# So each class needs at least (n_labeled_per_class + 500) samples

n_labeled_per_class = 10
samples_per_class = n_labeled_per_class + 600  # 10 labeled + 90 unlabeled + 500 validation

# Create labels: 4 classes with enough samples each
labels = np.repeat(np.arange(4), samples_per_class)

ARGS = {
    "labels": labels,
    "n_labeled_per_class": n_labeled_per_class,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
MixMatch Not yet run
unverified (harvested, no recorded run)
DamianValle/ScarceClassifier/src/mixmatch.py
pointer only (licence: NONE) · get_code("7397979be7d7eeb1")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, DamianValle/ScarceClassifier 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. No invocation recorded.

WideResNet Not yet run
unverified (harvested, no recorded run)
yuxi120407/mixmatch_tensorflow/model.py
pointer only (licence: NONE) · get_code("6e4bb6cf26d5d34d")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, yuxi120407/mixmatch_tensorflow 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. No invocation recorded.

augment_labelData Not yet run
unverified (harvested, no recorded run)
dhx000/DGM_project/mixmatch/cycle_mixmatch.py
pointer only (licence: NONE) · get_code("cd590be04d9864e1")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, dhx000/DGM_project 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. No invocation recorded.

augment_unlabelData Not yet run
unverified (harvested, no recorded run)
dhx000/DGM_project/mixmatch/cycle_mixmatch.py
pointer only (licence: NONE) · get_code("21f82c5babfc89be")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, dhx000/DGM_project 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. No invocation recorded.

calculate_per_class_lwlrap Not yet run
unverified (harvested, no recorded run)
viig99/mixmatch-freesound/analysis/compute_lwlrap.py
code served (permissive licence) · get_code("033283e1fa3d96fe")
How it ran

no invocation recorded.

create_model Not yet run
unverified (harvested, no recorded run)
viig99/mixmatch-freesound/kaggle_submission_checker.py
code served (permissive licence) · get_code("f8b6744d7114ac81")
How it ran

no invocation recorded.

expand_ulabel Not yet run
unverified (harvested, no recorded run)
dhx000/DGM_project/mixmatch/cycle_mixmatch.py
pointer only (licence: NONE) · get_code("44bd7219e3206702")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, dhx000/DGM_project 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. No invocation recorded.

get_cifar10 Not yet run
unverified (harvested, no recorded run)
kevinghst/mixmatch/dataset/cifar10.py
code served (permissive licence) · get_code("a0e25dd8565d9d94")
How it ran

no invocation recorded.

get_cifar10 Not yet run
unverified (harvested, no recorded run)
narendoraiswamy/MixMatch-pytorch-demo/dataset/_cifar10.py
code served (permissive licence) · get_code("0f1e6cdf5451e58d")
How it ran

no invocation recorded.

get_mean_and_std Not yet run
unverified (harvested, no recorded run)
kevinghst/mixmatch/utils/misc.py
pointer only (licence: MIT); the stored copy get_code serves, from LeeHyuck/CDMAD, is pointer only (licence: NONE) · get_code("1d6e2edb8832d7b0")
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 LeeHyuck/CDMAD, 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. No invocation recorded.

guess_label_with_logit Not yet run
unverified (harvested, no recorded run)
dhx000/DGM_project/mixmatch/cycle_mixmatch.py
pointer only (licence: NONE) · get_code("71516924003b7081")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, dhx000/DGM_project 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. No invocation recorded.

label_guessing Not yet run
unverified (harvested, no recorded run)
gan3sh500/mixmatch-pytorch/mixmatch_utils.py
pointer only (licence: NONE) · get_code("0ad756f5a68bcd04")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, gan3sh500/mixmatch-pytorch 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. No invocation recorded.

load_state_dict Not yet run
unverified (harvested, no recorded run)
viig99/mixmatch-freesound/kaggle_submission.py
code served (permissive licence) · get_code("197f711034e07fc3")
How it ran

no invocation recorded.

Showing the first 60. An agent gets every one from get_harvested_code_for_paper("1905.02249").

Also present in other papers' code

Of 63 samples lifted from this paper's linked repositories, 63 carry a function identity and 0 do not. Of the 63 distinct functions with an identity, 7 are also present in other papers whose code we hold (2 with at least one paper dated earlier, 2 the same month, 7 later) and 56 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
plot_overlap 31 other papers
6 dated earlier, 1 the same month, 24 later
1612.06321 (2016-12)
Large-Scale Image Retrieval with Attentive Deep Local Features
1805.07925 (2018-05)
Batch-Instance Normalization for Adaptively Style-Invariant Neural Networks
1806.04606 (2018-06)
Knowledge Distillation by On-the-Fly Native Ensemble
and 28 more not named here
get_mean_and_std 21 other papers
5 dated earlier, 1 the same month, 15 later
1805.07925 (2018-05)
Batch-Instance Normalization for Adaptively Style-Invariant Neural Networks
1806.04606 (2018-06)
Knowledge Distillation by On-the-Fly Native Ensemble
1811.05850 (2018-11)
Drop-Activation: Implicit Parameter Reduction and Harmonic Regularization
and 18 more not named here
normalise 8 other papers
8 later
2003.14297 (2020-03)
Generative Latent Implicit Conditional Optimization when Learning from Small Sample
2006.12135 (2020-06)
Learning to Generate Noise for Multi-Attack Robustness
2103.14222 (2021-03)
Adversarial Attacks are Reversible with Natural Supervision
and 5 more not named here
get_weak_augmentation
brick of this paper's assembly: entry
1 other paper
1 later
2001.07685 (2020-01)
FixMatch: Simplifying Semi-Supervised Learning with Consistency and Confidence
interleave
brick of this paper's assembly: helper
1 other paper
1 later
2510.03993 (2025-10)
Keep It on a Leash: Controllable Pseudo-label Generation Towards Realistic Long-Tailed Semi-Supervised Learning
interleave_offsets
brick of this paper's assembly: helper
1 other paper
1 later
2510.03993 (2025-10)
Keep It on a Leash: Controllable Pseudo-label Generation Towards Realistic Long-Tailed Semi-Supervised Learning
to_categorical
brick of this paper's assembly: helper
1 other paper
1 later
2203.12297 (2022-03)
Increasing the accuracy and resolution of precipitation forecasts using deep generative models

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

Semi-supervised learning has proven to be a powerful paradigm for leveraging unlabeled data to mitigate the reliance on large labeled datasets. In this work, we unify the current dominant approaches for semi-supervised learning to produce a new algorithm, MixMatch, that guesses low-entropy labels for data-augmented unlabeled examples and mixes labeled and unlabeled data using MixUp. MixMatch obtains state-of-the-art results by a large margin across many datasets and labeled data amounts. For example, on CIFAR-10 with 250 labels, we reduce error rate by a factor of 4 (from 38% to 11%) and by a factor of 2 on STL-10. We also demonstrate how MixMatch can help achieve a dramatically better accuracy-privacy trade-off for differential privacy. Finally, we perform an ablation study to tease apart which components of MixMatch are most important for its success. We release all code used in our experiments. 1

For agents

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

get_harvested_code_for_paper("1905.02249")
get_code_for_paper("1905.02249")
have("1905.02249")

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

curl https://syntology.ai/api/ran/1905.02249.json

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

[![Syntology run record](https://syntology.ai/api/ran/1905.02249.svg)](https://syntology.ai/paper/1905.02249)

Connect an agent — have() is free.