SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1904.08779 · 2019

SpecAugment: A Simple Data Augmentation Method for Automatic Speech Recognition

arXiv · PDF · Open in the Atlas

Code that ran

14 of 18 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 13 have no contract verdict; 1 failure of our instrument; 4 no recorded run), measured 2026-11-22; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/1904.08779.json · badge: /api/ran/1904.08779.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
cosmoquester/speech-recognition community (classifier label)
classifier basis not recorded on the link
5 ran · 0 no recorded run
DemisEom/SpecAugment community (classifier label)
classifier basis not recorded on the link
4 ran · 3 no recorded run
KimJeongSun/SpecAugment_numpy_scipy community (classifier label)
classifier basis not recorded on the link
2 ran · 0 no recorded run
kimjeongsun/specaugment community (classifier label)
classifier basis not recorded on the link
1 ran · 1 no recorded run
sh951011/Korean-Speech-Recognition community (classifier label)
classifier basis not recorded on the link
1 ran · 0 no recorded run
ZhengkunTian/OpenTransformer reimplementation (classifier label)
classifier basis: classifier_2025
1 ran · 0 no recorded run
FunctionStatusWhere it lives
check_envirionment Ran
ran on a synthesized input
sh951011/Korean-Speech-Recognition/kospeech/utils.py
code served (permissive licence) · get_code("8365bd2bebaeee71")
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 device. 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 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 logging

logging.basicConfig(level=logging.INFO)

torch.manual_seed(0)

ARGS = {
    "use_cuda": False
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type device — a device output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a device 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
freq_mask Ran
ran on a synthesized input
DemisEom/SpecAugment/SpecAugment/sparse_image_warp_pytorch.py
code served (permissive licence) · get_code("5c640f85e661f5c9")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 3, dtype family float, shape (1, 10, 20) 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 random

torch.manual_seed(0)

# Create a small spectrogram tensor
# Shape: (batch=1, mel_channels=10, time_steps=20)
spec = torch.randn(1, 10, 20)

ARGS = {
    "spec": spec,
    "F": 5,
    "num_masks": 2,
    "replace_with_zero": False
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 3, dtype family float, shape (1, 10, 20) — shape (1, 10, 20) 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
frequency_masking Ran
ran on a synthesized input
DemisEom/SpecAugment/SpecAugment/spec_augment_tensorflow.py
code served (permissive licence) · get_code("b8b1c17d15f3c463")
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 (1, 4, 8, 1) 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 mel spectrogram tensor
# Shape: (batch_size=1, frequency_bins=4, time_steps=8, channels=1)
mel_spectrogram = tf.constant(
    np.random.rand(1, 4, 8, 1).astype(np.float32)
)

# The 'v' parameter - representing the frequency dimension size
# Based on the function body, v is extracted from fbank_size[2]
# which is the 3rd dimension of mel_spectrogram (time_steps dimension)
v = tf.constant(8, dtype=tf.int32)

ARGS = {
    "mel_spectrogram": mel_spectrogram,
    "v": v,
    "frequency_masking_para": 3,
    "frequency_mask_num": 2
}
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 (1, 4, 8, 1) — shape (1, 4, 8, 1) 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
get_device_strategy Ran
ran on a synthesized input
cosmoquester/speech-recognition/speech_recognition/utils.py
code served (permissive licence) · get_code("3046ea943ba309f2")
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 OneDeviceStrategy. 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 os
import tensorflow as tf

# Set up a minimal environment for the function
# The function takes a 'device' parameter which should be a string

ARGS = {
    'device': 'CPU'
}
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 OneDeviceStrategy — a OneDeviceStrategy output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a OneDeviceStrategy 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_logger Ran
ran on a synthesized input
cosmoquester/speech-recognition/speech_recognition/utils.py
code served (permissive licence) · get_code("0c8faeedf5a5c01a")
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 Logger. 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 logging
import sys

ARGS = {
    "name": "test_logger"
}
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 Logger — a Logger output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Logger 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_rnn_cls Ran
ran on a synthesized input
cosmoquester/speech-recognition/speech_recognition/models/las.py
code served (permissive licence) · get_code("aa9757e8039b3c4a")
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 type. 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 tensorflow.keras.layers import GRU, LSTM, SimpleRNN

torch.manual_seed(0)

# The function get_rnn_cls takes a single parameter: rnn_type (str)
# It returns one of three RNN layer classes based on the input string
# Valid values are 'rnn', 'lstm', or 'gru'

ARGS = {
    'rnn_type': 'lstm'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type type — a type output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a type 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_tfrecord_dataset Ran
ran on a synthesized input
cosmoquester/speech-recognition/speech_recognition/data.py
code served (permissive licence) · get_code("ba904be9deeb14d4")
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 _MapDataset. 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 tempfile
import os
import tensorflow as tf

# Create temporary directory and a minimal TFRecord file
temp_dir = tempfile.mkdtemp()

# Create a simple TFRecord file with gzip compression
tfrecord_path = os.path.join(temp_dir, "test.tfrecord")

# Write a minimal TFRecord with two examples
writer = tf.io.TFRecordWriter(tfrecord_path, options=tf.io.TFRecordOptions(compression_type='GZIP'))

# Create sample data: (float32 tensor, int32 tensor)
# First tensor: shape (2, 4)
# Second tensor: shape (3,)
float_tensor = tf.constant([[1.0, 2.0, 3.0, 4.0], [5.0, 6.0, 7.0, 8.0]], dtype=tf.float32)
int_tensor = tf.constant([1, 2, 3], dtype=tf.int32)

# Serialize as strings within a list
serialized_float = tf.io.serialize_tensor(float_tensor).numpy()
serialized_int = tf.io.serialize_tensor(int_tensor).numpy()

# Create feature dict
feature = {
    'float_data': tf.train.Feature(bytes_list=tf.train.BytesList(value=[serialized_float])),
    'int_data': tf.train.Feature(bytes_list=tf.train.BytesList(value=[serialized_int]))
}

# Write two examples
for _ in range(2):
    example = tf.train.Example(features=tf.train.Features(feature=feature))
    writer.write(example.SerializeToString())

writer.close()

# Create the dataset_paths pattern that glob will match
dataset_paths = os.path.join(temp_dir, "*.tfrecord")

ARGS = {
    "dataset_paths": dataset_paths
}
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 _MapDataset — a _MapDataset output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a _MapDataset 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
levenshtein_distance Ran
ran on a synthesized input
cosmoquester/speech-recognition/speech_recognition/utils.py
code served (permissive licence) · get_code("2046b512475f2561")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned float64, ndim 0, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# Test with strings
truth_str = "kitten"
hypothesis_str = "sitting"

# Test with iterables (lists)
truth_list = [1, 2, 3, 4]
hypothesis_list = [1, 2, 5, 6]

# Create ARGS dict with string inputs and normalize=True
ARGS = {
    "truth": truth_str,
    "hypothesis": hypothesis_str,
    "normalize": True
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type float64, ndim 0, dtype family float — no output shape is stated: observe_signature.py recorded float ndim 0, which disagrees with the fixture run
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
liftPts Ran
ran on a synthesized input
KimJeongSun/SpecAugment_numpy_scipy/specaugment.py
code served (permissive licence) · get_code("be893d0ee5559ea0")
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 (2, 6) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

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

# Create small sample data
# p: N points in 2D space (N=2, dimension=2)
p = np.array([
    [0.5, 1.2],
    [2.3, 0.8]
], dtype=np.float64)

# cp: K control points in 2D space (K=3, dimension=2)
cp = np.array([
    [0.0, 0.0],
    [1.0, 1.0],
    [2.0, 0.5]
], dtype=np.float64)

ARGS = {
    'p': p,
    'cp': cp
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy — 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 (2, 6) — shape (2, 6) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
makeT Ran
ran on a synthesized input
KimJeongSun/SpecAugment_numpy_scipy/specaugment.py
code served (permissive licence) · get_code("7fd389e1e9203afa")
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 (7, 7) 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 scipy.spatial.distance import pdist, cdist, squareform

# Create a small control points array
# The function expects cp to have shape (K, 2) based on how it's used
# K is the number of control points, and they should be 2D points
cp = np.array([
    [0.0, 0.0],
    [1.0, 1.0],
    [2.0, 0.5],
    [1.5, 2.0]
], dtype=np.float64)

ARGS = {
    'cp': cp
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy — 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 (7, 7) — shape (7, 7) 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
spec_augment Ran
ran; our contract draft was wrong, not the code
ZhengkunTian/OpenTransformer/otrans/data/augment.py
code served (permissive licence) · get_code("aabacfe520817be6")
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 (128, 32) 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
import numpy as np

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

# Create a small mel spectrogram (time dimension x frequency dimension)
# tau = 128 (time), v = 32 (frequency bins)
mel_spectrogram = np.random.randn(128, 32).astype(np.float32)

ARGS = {
    "mel_spectrogram": mel_spectrogram,
    "freq_mask_num": 2,
    "time_mask_num": 2,
    "freq_mask_rate": 0.3,
    "time_mask_rate": 0.05,
    "max_mask_time_len": 100,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 2, dtype family float, shape (128, 32) — shape (128, 32) 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
spec_augment Ran
ran on a synthesized input
kimjeongsun/specaugment/SpecAugment/spec_augment_pytorch.py
code served (permissive licence) · get_code("ed8e1a9c22ce7ac6")
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 (64, 128) 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 random

np.random.seed(0)
random.seed(0)

# Create a small mel spectrogram: shape (v, tau) = (64, 128)
# v = number of frequency bins, tau = number of time steps
mel_spectrogram = np.random.randn(64, 128).astype(np.float32)

ARGS = {
    "mel_spectrogram": mel_spectrogram,
    "time_warping_para": 80,
    "frequency_masking_para": 27,
    "time_masking_para": 100,
    "frequency_mask_num": 1,
    "time_mask_num": 1,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 2, dtype family float, shape (64, 128) — shape (64, 128) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
time_mask Ran
ran on a synthesized input
DemisEom/SpecAugment/SpecAugment/sparse_image_warp_pytorch.py
code served (permissive licence) · get_code("649daef969b655ed")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 3, dtype family float, shape (1, 4, 20) 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 random

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

# Create a small spectrogram tensor
# Shape: (batch=1, freq_bins=4, time_steps=20)
# The function accesses spec.shape[2] for time dimension
# and uses spec.clone(), spec.mean(), and indexing
spec = torch.randn(1, 4, 20)

ARGS = {
    "spec": spec,
    "T": 15,
    "num_masks": 1,
    "replace_with_zero": False
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 3, dtype family float, shape (1, 4, 20) — shape (1, 4, 20) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
time_warp Ran
ran on a synthesized input
DemisEom/SpecAugment/SpecAugment/sparse_image_warp_pytorch.py
code served (permissive licence) · get_code("d94ba36bc01eda60")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 3, dtype family float; 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 random
from unittest.mock import MagicMock

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

# Create a mock SparseImageWarp class with sparse_image_warp method
class SparseImageWarp:
    @staticmethod
    def sparse_image_warp(spec, src_pts, dest_pts):
        # Return a warped spectro with shape (batch, height, width, 1)
        # and dummy dense_flows
        warped = torch.randn(spec.shape[0], spec.shape[1], spec.shape[2], 1)
        dense_flows = torch.randn(spec.shape[0], spec.shape[1], spec.shape[2], 2)
        return warped, dense_flows

# Inject the mock into the function's namespace
import sys
current_module = sys.modules[__name__]
setattr(current_module, 'SparseImageWarp', SparseImageWarp)

# Create a small spectrogram tensor
# Shape: (num_rows, spec_len) where num_rows is even (for y = num_rows // 2)
# After view(1, spec.shape[0], spec.shape[1]), it becomes (1, num_rows, spec_len)
spec = torch.randn(4, 20)  # 4 rows, 20 frequency bins

ARGS = {
    'spec': spec,
    'W': 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 Tensor, ndim 3, dtype family float — no output shape is stated: observe_signature.py could not observe it (fixture:KeyError)
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
dense_image_warp Not yet run
unverified (harvested, no recorded run)
DemisEom/SpecAugment/SpecAugment/sparse_image_warp_np.py
code served (permissive licence) · get_code("a67846a111580cfb")
How it ran

no invocation recorded.

sparse_image_warp Not yet run
unverified (harvested, no recorded run)
kimjeongsun/specaugment/SpecAugment/sparse_image_warp_pytorch.py
code served (permissive licence) · get_code("aa191294fe80e2d4")
How it ran

no invocation recorded.

sparse_image_warp_np Not yet run
unverified (harvested, no recorded run)
DemisEom/SpecAugment/SpecAugment/sparse_image_warp_np.py
code served (permissive licence) · get_code("ff652f83fe897b32")
How it ran

no invocation recorded.

time_masking Not yet run
unverified (harvested, no recorded run)
DemisEom/SpecAugment/SpecAugment/spec_augment_tensorflow.py
code served (permissive licence) · get_code("fdad4c356de2036b")
How it ran

no invocation recorded.

Also present in other papers' code

Of 18 samples lifted from this paper's linked repositories, 18 carry a function identity and 0 do not. Of the 18 distinct functions with an identity, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 18 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.

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

Repositories linked to this paper

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

Abstract

We present SpecAugment, a simple data augmentation method for speech recognition. SpecAugment is applied directly to the feature inputs of a neural network (i.e., filter bank coefficients). The augmentation policy consists of warping the features, masking blocks of frequency channels, and masking blocks of time steps. We apply SpecAugment on Listen, Attend and Spell networks for end-to-end speech recognition tasks. We achieve state-of-the-art performance on the LibriSpeech 960h and Swichboard 300h tasks, outperforming all prior work. On LibriSpeech, we achieve 6.8% WER on test-other without the use of a language model, and 5.8% WER with shallow fusion with a language model. This compares to the previous state-of-the-art hybrid system of 7.5% WER. For Switchboard, we achieve 7.2%/14.6% on the Switchboard/CallHome portion of the Hub5'00 test set without the use of a language model, and 6.8%/14.1% with shallow fusion, which compares to the previous state-of-the-art hybrid system at 8.3%/17.3% WER.

For agents

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

get_harvested_code_for_paper("1904.08779")
get_code_for_paper("1904.08779")
have("1904.08779")

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

curl https://syntology.ai/api/ran/1904.08779.json

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

[![Syntology run record](https://syntology.ai/api/ran/1904.08779.svg)](https://syntology.ai/paper/1904.08779)

Connect an agent — have() is free.