SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1809.07454 · 2018

Conv-TasNet: Surpassing Ideal Time-Frequency Magnitude Masking for Speech Separation

arXiv · PDF · Open in the Atlas

Code that ran

28 of 36 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; 3 honoured the contract we drafted, 2 violated it and 17 have no contract verdict; 6 failures of our instrument; 8 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/1809.07454.json · badge: /api/ran/1809.07454.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
funcwj/conv-tasnet canonical (classifier label)
classifier basis: census
8 ran · 2 no recorded run
IliasDG/Demucs_Project community (classifier label)
classifier basis not recorded on the link
11 ran · 2 no recorded run
kaituoxu/Conv-TasNet reimplementation (classifier label)
classifier basis: classifier_2025
3 ran · 0 no recorded run
pfnet-research/meta-tasnet community (classifier label)
classifier basis not recorded on the link
2 ran · 3 no recorded run
xiaolianggeitazgh/Conv-Tasnet reimplementation (classifier label)
classifier basis: classifier_2025
2 ran · 0 no recorded run
JusperLee/Conv-TasNet reimplementation (classifier label)
classifier basis: classifier_2025
2 ran · 0 no recorded run
robieta/demucs community (classifier label)
classifier basis not recorded on the link
0 ran · 1 no recorded run
FunctionStatusWhere it lives
average_metric Ran
ran on a synthesized input
IliasDG/Demucs_Project/demucs/utils.py
pointer only (licence: MIT); the stored copy get_code serves, from zhaozhipeng1997/demucs_ascend910_pytorch, is pointer only (licence: MIT) · get_code("a68dbf50b2629a55")
How it ran

The body of this sample is withheld: MIT, as GitHub detects it on this paper's repository, IliasDG/Demucs_Project, is in the permissive set, but a licence decision keeps the body a pointer: the reason for that decision is not recorded with it. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

build_norm Ran
ran on a synthesized input
funcwj/conv-tasnet/nnet/conv_tas_net.py
code served (permissive licence) · get_code("cc51e06c5fd96123")
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 ChannelWiseLayerNorm. 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 = {
    'norm': 'cLN',
    'dim': 4
}
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 ChannelWiseLayerNorm — a ChannelWiseLayerNorm output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a ChannelWiseLayerNorm 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
cal_SISNR Ran
ran, violated the contract we drafted
kaituoxu/Conv-TasNet/src/evaluate.py
code served (permissive licence) · get_code("faac9c0b7e1a08c5")
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, 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

# Create simple numpy arrays of equal length for the SI-SNR calculation
np.random.seed(0)
T = 16000  # typical audio signal length

# Generate reference signal and output signal
ref_sig = np.random.randn(T).astype(np.float32)
out_sig = ref_sig + 0.1 * np.random.randn(T).astype(np.float32)  # slightly noisy version

ARGS = {
    'ref_sig': ref_sig,
    'out_sig': out_sig,
    'eps': 1e-08
}
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, 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
cal_SISNR Ran
ran, violated the contract we drafted
xiaolianggeitazgh/Conv-Tasnet/evaluate.py
code served (permissive licence) · get_code("891e417fc38edeb0")
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, 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

# Create two 1D numpy arrays of the same length
# These represent reference and output signals
np.random.seed(0)
ref_sig = np.random.randn(100)
out_sig = np.random.randn(100)

ARGS = {
    'ref_sig': ref_sig,
    'out_sig': out_sig,
    'eps': 1e-08
}
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, 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
cal_SISNRi Ran
ran, honoured the contract we drafted
kaituoxu/Conv-TasNet/src/evaluate.py
code served (permissive licence) · get_code("b19791aa06bb7b9b")
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, 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

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

# Create sample audio signals
# ref_sig and est_sig should be 1D arrays [T] where T is time samples
# src_ref should be [C, T] where C is number of sources (2 in this case)
# src_est should be [C, T] with same shape as src_ref
# mix should be [T]

T = 16000  # time samples
C = 2      # number of sources

# Generate synthetic audio signals
src_ref = np.random.randn(C, T).astype(np.float32)
src_est = src_ref + 0.1 * np.random.randn(C, T).astype(np.float32)  # estimated with some noise
mix = np.sum(src_ref, axis=0).astype(np.float32)  # mixture is sum of sources

ARGS = {
    "src_ref": src_ref,
    "src_est": src_est,
    "mix": mix,
}
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, 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
cal_SISNRi Ran
ran, honoured the contract we drafted
xiaolianggeitazgh/Conv-Tasnet/evaluate.py
code served (permissive licence) · get_code("ca6c3339d98f2a2c")
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, 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

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

# Create small synthetic audio signals
# src_ref: [C, T] - reference sources (2 channels, 100 time samples)
src_ref = np.random.randn(2, 100).astype(np.float32)

# src_est: [C, T] - estimated/separated sources (2 channels, 100 time samples)
src_est = src_ref + 0.1 * np.random.randn(2, 100).astype(np.float32)

# mix: [T] - mixture signal (100 time samples)
mix = src_ref[0] + src_ref[1]

ARGS = {
    "src_ref": src_ref,
    "src_est": src_est,
    "mix": mix,
}
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, 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
center_trim Ran
ran on a synthesized input
IliasDG/Demucs_Project/demucs/utils.py
pointer only (licence: MIT) · get_code("f89970356c8e9506")
How it ran

The body of this sample is withheld: MIT, as GitHub detects it on this paper's repository, IliasDG/Demucs_Project, is in the permissive set, but a licence decision keeps the body a pointer: the reason for that decision is not recorded with it. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

check_parameters Ran
ran, honoured the contract we drafted
JusperLee/Conv-TasNet/Conv_TasNet_Pytorch/Conv_TasNet.py
pointer only (licence: NONE) · get_code("bbd39ee3588090e3")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, JusperLee/Conv-TasNet 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.

chose_norm Ran
ran; our contract draft was wrong, not the code
kaituoxu/Conv-TasNet/src/conv_tasnet.py
code served (permissive licence) · get_code("22f921651bb6e769")
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 GlobalLayerNorm. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

torch.manual_seed(0)

# The function chose_norm takes two parameters:
# 1. norm_type: a string that can be 'gLN', 'cLN', or other (defaults to BatchNorm1d)
# 2. channel_size: an integer representing the channel dimension

ARGS = {
    'norm_type': 'gLN',
    'channel_size': 4
}
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 GlobalLayerNorm — a GlobalLayerNorm output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a GlobalLayerNorm 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
chose_norm Ran
ran; our contract draft was wrong, not the code
IliasDG/Demucs_Project/demucs/tasnet.py
pointer only (licence: MIT) · get_code("c20ada1396105c22")
How it ran

The body of this sample is withheld: MIT, as GitHub detects it on this paper's repository, IliasDG/Demucs_Project, is in the permissive set, but a licence decision keeps the body a pointer: the reason for that decision is not recorded with it. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

downsample Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
IliasDG/Demucs_Project/demucs/model.py
pointer only (licence: MIT) · get_code("07fa2ca4b8bf0830")
How it ran

The body of this sample is withheld: MIT, as GitHub detects it on this paper's repository, IliasDG/Demucs_Project, is in the permissive set, but a licence decision keeps the body a pointer: the reason for that decision is not recorded with it. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

fname Ran
ran on a synthesized input
IliasDG/Demucs_Project/dora.py
pointer only (licence: MIT) · get_code("9f54a0c5172ff77c")
How it ran

The body of this sample is withheld: MIT, as GitHub detects it on this paper's repository, IliasDG/Demucs_Project, is in the permissive set, but a licence decision keeps the body a pointer: the reason for that decision is not recorded with it. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

free_port Ran
ran on a synthesized input
IliasDG/Demucs_Project/demucs/utils.py
pointer only (licence: MIT); the stored copy get_code serves, from zhaozhipeng1997/demucs_ascend910_pytorch, is pointer only (licence: MIT) · get_code("f747e6cf6fc620a6")
How it ran

The body of this sample is withheld: MIT, as GitHub detects it on this paper's repository, IliasDG/Demucs_Project, is in the permissive set, but a licence decision keeps the body a pointer: the reason for that decision is not recorded with it. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

get_done Ran
ran on a synthesized input
IliasDG/Demucs_Project/dora.py
pointer only (licence: MIT) · get_code("7bd6c5376583922a")
How it ran

The body of this sample is withheld: MIT, as GitHub detects it on this paper's repository, IliasDG/Demucs_Project, is in the permissive set, but a licence decision keeps the body a pointer: the reason for that decision is not recorded with it. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

get_logger Ran
ran on a synthesized input
funcwj/conv-tasnet/nnet/libs/utils.py
code served (permissive licence) · get_code("d0a0123e9bc8a4c4")
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 tempfile
import os

# The function expects:
# - name: a string (logger name)
# - format_str: a string (format specification)
# - date_format: a string (date format specification)
# - file: a boolean (whether to use FileHandler or StreamHandler)

# Create a temporary file for the file-based logging test
temp_dir = tempfile.gettempdir()
temp_log_file = os.path.join(temp_dir, "test_logger.log")

ARGS = {
    "name": "test_logger",
    "format_str": "%(asctime)s [%(pathname)s:%(lineno)s - %(levelname)s ] %(message)s",
    "date_format": "%Y-%m-%d %H:%M:%S",
    "file": False
}
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_name Ran
ran on a synthesized input
IliasDG/Demucs_Project/demucs/parser.py
pointer only (licence: MIT); the stored copy get_code serves, from zhaozhipeng1997/demucs_ascend910_pytorch, is pointer only (licence: MIT) · get_code("dccae3e25e1c55a2")
How it ran

The body of this sample is withheld: MIT, as GitHub detects it on this paper's repository, IliasDG/Demucs_Project, is in the permissive set, but a licence decision keeps the body a pointer: the reason for that decision is not recorded with it. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

load_json Ran
ran on a synthesized input
funcwj/conv-tasnet/nnet/libs/utils.py
code served (permissive licence) · get_code("1a0aeff1eb7546d6")
How it ran

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

import os
import json
import tempfile

# Create a temporary directory and a valid JSON file
temp_dir = tempfile.mkdtemp()
test_json_path = os.path.join(temp_dir, "test.json")

# Write a valid JSON file
test_data = {
    "key1": "value1",
    "key2": [1, 2, 3],
    "key3": {"nested": "object"}
}

with open(test_json_path, 'w') as f:
    json.dump(test_data, f)

# Define arguments for the function
ARGS = {
    "fdir": temp_dir,
    "name": "test.json"
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
load_obj Ran
ran on a synthesized input
funcwj/conv-tasnet/nnet/libs/trainer.py
code served (permissive licence) · get_code("9b384ef1d2d3ea3f")
How it ran

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

import torch as th

th.manual_seed(0)

# Test case 1: nested dict and list structure with tensors
obj = {
    'tensor1': th.randn(2, 4),
    'tensor2': th.randn(3, 5),
    'nested_list': [
        th.randn(2, 4),
        {'inner_tensor': th.randn(1, 3)},
        'string_value'
    ],
    'nested_dict': {
        'deep_tensor': th.randn(2, 2),
        'scalar': 42
    },
    'plain_value': 3.14
}

device = 'cpu'

ARGS = {
    'obj': obj,
    'device': device
}
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 dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
overlap_and_add Ran
ran; our contract draft was wrong, not the code
IliasDG/Demucs_Project/demucs/tasnet.py
pointer only (licence: MIT) · get_code("6fc6cdb0729e7905")
How it ran

The body of this sample is withheld: MIT, as GitHub detects it on this paper's repository, IliasDG/Demucs_Project, is in the permissive set, but a licence decision keeps the body a pointer: the reason for that decision is not recorded with it. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

param Ran
ran on a synthesized input
funcwj/conv-tasnet/nnet/conv_tas_net.py
code served (permissive licence) · get_code("3a4ca7c24b1659fc")
How it ran

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

import torch
import torch.nn as nn

torch.manual_seed(0)

class SimpleNet(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear1 = nn.Linear(4, 3)
        self.linear2 = nn.Linear(3, 2)
    
    def forward(self, x):
        x = self.linear1(x)
        x = self.linear2(x)
        return x

nnet = SimpleNet()

ARGS = {
    "nnet": nnet,
    "Mb": True
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type float, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
parse_scripts Ran
ran on a synthesized input
funcwj/conv-tasnet/nnet/libs/audio.py
code served (permissive licence) · get_code("77ec6163bd018420")
How it ran

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

import tempfile
import os

# Create a temporary .scp file with valid Kaldi script format
temp_dir = tempfile.mkdtemp()
scp_file = os.path.join(temp_dir, "test.scp")

# Write a valid .scp file with 2 tokens per line
with open(scp_file, 'w') as f:
    f.write("utt001 /path/to/audio1.wav\n")
    f.write("utt002 /path/to/audio2.wav\n")
    f.write("utt003 /path/to/audio3.wav\n")

# Define a simple value processor function
def simple_processor(x):
    """Simple processor that converts value to uppercase"""
    return x.upper() if isinstance(x, str) else x

ARGS = {
    "scp_path": scp_file,
    "value_processor": simple_processor,
    "num_tokens": 2
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
permute_si_snr Ran
ran on a synthesized input
funcwj/conv-tasnet/nnet/libs/metric.py
code served (permissive licence) · get_code("457f6fe40f154950")
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

np.random.seed(0)

# Create small reference signals (ground truth)
s1 = np.random.randn(100)
s2 = np.random.randn(100)
slist = [s1, s2]

# Create enhanced/separated signals (typically noisy or mixed versions)
x1 = s1 + 0.1 * np.random.randn(100)
x2 = s2 + 0.1 * np.random.randn(100)
xlist = [x1, x2]

ARGS = {
    "xlist": xlist,
    "slist": slist,
}
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
quantize Ran
ran on a synthesized input
IliasDG/Demucs_Project/quantize.py
pointer only (licence: MIT) · get_code("95678855edc371ba")
How it ran

The body of this sample is withheld: MIT, as GitHub detects it on this paper's repository, IliasDG/Demucs_Project, is in the permissive set, but a licence decision keeps the body a pointer: the reason for that decision is not recorded with it. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

sdr_objective Ran
ran on a synthesized input
pfnet-research/meta-tasnet/utility/loss.py
code served (permissive licence) · get_code("175694073fa2797e")
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 1, dtype family float, shape (4,) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# Create tiny tensors with shape (B, 4, 1, T)
# B=2 (batch), 4 (channels), 1 (spatial), T=3 (time)
B, C, spatial, T = 2, 4, 1, 3

estimation = torch.randn(B, C, spatial, T)
origin = torch.randn(B, C, spatial, T)

# Create boolean mask of shape (B, 4, 1)
# True where origin is 0.0, False otherwise
mask = torch.zeros(B, C, spatial, dtype=torch.bool)
mask[0, 0, 0] = True
mask[1, 2, 0] = True

ARGS = {
    "estimation": estimation,
    "origin": origin,
    "mask": mask,
}
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 1, dtype family float, shape (4,) — shape (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
select_norm Ran
ran; our contract draft was wrong, not the code
JusperLee/Conv-TasNet/Conv_TasNet_Pytorch/Conv_TasNet.py
pointer only (licence: NONE) · get_code("fd365b69d72fe0c9")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, JusperLee/Conv-TasNet 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.

si_snr Ran
ran on a synthesized input
funcwj/conv-tasnet/nnet/libs/metric.py
code served (permissive licence) · get_code("23b2bdd3e0fe81c3")
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 float32, 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 torch

torch.manual_seed(0)

# Create small numpy vectors for x and s
# x: enhanced/separated signal
x = np.array([1.2, 2.3, 0.5, -0.8, 1.1], dtype=np.float32)

# s: reference signal (ground truth)
s = np.array([1.0, 2.5, 0.3, -0.7, 1.3], dtype=np.float32)

# remove_dc: boolean flag
remove_dc = True

ARGS = {
    "x": x,
    "s": s,
    "remove_dc": remove_dc
}
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 float32, 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
similarity_loss Ran
ran on a synthesized input
pfnet-research/meta-tasnet/utility/loss.py
code served (permissive licence) · get_code("865adc65cb86cc62")
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
import torch.nn.functional as F

torch.manual_seed(0)

# Construct latents tensor of shape (B, 1, T', N)
# B=2 (batch), 1 (fixed), T'=3 (time), N=4 (latent dim)
latents = torch.randn(2, 1, 3, 4)

# Construct mask tensor of shape (B, 4)
# B=2 (batch), 4 (instruments)
# Boolean mask: True when signal is 0.0
mask = torch.tensor([[True, False, True, False],
                     [False, True, False, True]], dtype=torch.bool)

ARGS = {
    "latents": latents,
    "mask": mask
}
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
upsample Ran
ran; our contract draft was wrong, not the code
IliasDG/Demucs_Project/demucs/model.py
pointer only (licence: MIT) · get_code("5cf51541da88d14a")
How it ran

The body of this sample is withheld: MIT, as GitHub detects it on this paper's repository, IliasDG/Demucs_Project, is in the permissive set, but a licence decision keeps the body a pointer: the reason for that decision is not recorded with it. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

Conv1dWrapper Not yet run
unverified (harvested, no recorded run)
pfnet-research/meta-tasnet/model/conv1d.py
code served (permissive licence) · get_code("b80414a2c961fe63")
How it ran

no invocation recorded.

GroupNormWrapper Not yet run
unverified (harvested, no recorded run)
pfnet-research/meta-tasnet/model/group_norm.py
code served (permissive licence) · get_code("60382cdca01c9e16")
How it ran

no invocation recorded.

build_metadata Not yet run
unverified (harvested, no recorded run)
IliasDG/Demucs_Project/demucs/compressed.py
pointer only (licence: MIT); the stored copy get_code serves, from zhaozhipeng1997/demucs_ascend910_pytorch, is pointer only (licence: MIT) · get_code("ffd24b3bb5a4d744")
How it ran

The body of this sample is withheld: MIT, as GitHub detects it on this paper's repository, IliasDG/Demucs_Project, is in the permissive set, but a licence decision keeps the body a pointer: the reason for that decision is not recorded with it. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

build_metadata Not yet run
unverified (harvested, no recorded run)
robieta/demucs/demucs/compressed.py
pointer only (licence: MIT) · get_code("b482cd70c4311bcc")
How it ran

The body of this sample is withheld: MIT, as GitHub detects it on this paper's repository, robieta/demucs, is in the permissive set, but a licence decision keeps the body a pointer: the reason for that decision is not recorded with it. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

dissimilarity_loss Not yet run
unverified (harvested, no recorded run)
pfnet-research/meta-tasnet/utility/loss.py
code served (permissive licence) · get_code("6d72bd968efb3b87")
How it ran

no invocation recorded.

get_sid Not yet run
unverified (harvested, no recorded run)
IliasDG/Demucs_Project/dora.py
pointer only (licence: MIT) · get_code("4678393521f60867")
How it ran

The body of this sample is withheld: MIT, as GitHub detects it on this paper's repository, IliasDG/Demucs_Project, is in the permissive set, but a licence decision keeps the body a pointer: the reason for that decision is not recorded with it. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

make_dataloader Not yet run
unverified (harvested, no recorded run)
funcwj/conv-tasnet/nnet/libs/dataset.py
code served (permissive licence) · get_code("4e23d73a9b135dab")
How it ran

no invocation recorded.

read_wav Not yet run
unverified (harvested, no recorded run)
funcwj/conv-tasnet/nnet/libs/audio.py
code served (permissive licence) · get_code("455d88ef82f00d25")
How it ran

no invocation recorded.

Also present in other papers' code

Of 36 samples lifted from this paper's linked repositories, 36 carry a function identity and 0 do not. Of the 34 distinct functions with an identity, 13 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 13 later) and 21 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
cal_SISNRi 4 other papers
4 later
1910.06379 (2019-10)
Dual-path RNN: efficient long sequence modeling for time-domain single-channel speech separation
1910.14104 (2019-10)
End-to-end Microphone Permutation and Number Invariant Multi-channel Speech Separation
2003.01531 (2020-03)
Voice Separation with an Unknown Number of Multiple Speakers
and 1 more not named here
cal_SISNR 3 other papers
3 later
1910.06379 (2019-10)
Dual-path RNN: efficient long sequence modeling for time-domain single-channel speech separation
2003.01531 (2020-03)
Voice Separation with an Unknown Number of Multiple Speakers
2007.13975 (2020-07)
Dual-Path Transformer Network: Direct Context-Aware Modeling for End-to-End Monaural Speech Separation
center_trim 2 other papers
2 later
2007.06833 (2020-07)
Sudo rm -rf: Efficient Networks for Universal Audio Source Separation
2211.08553 (2022-11)
Hybrid Transformers for Music Source Separation
check_parameters 2 other papers
2 later
2002.08688 (2020-02)
An empirical study of Conv-TasNet
2404.02063 (2024-04)
SPMamba: State-space model is all you need in speech separation
chose_norm 2 other papers
2 later
2211.08553 (2022-11)
Hybrid Transformers for Music Source Separation
2603.14889 (2026-03)
SDiaReward: Modeling and Benchmarking Spoken Dialogue Rewards with Modality and Colloquialness
downsample 2 other papers
2 later
2211.08553 (2022-11)
Hybrid Transformers for Music Source Separation
2603.14889 (2026-03)
SDiaReward: Modeling and Benchmarking Spoken Dialogue Rewards with Modality and Colloquialness
overlap_and_add 2 other papers
2 later
2211.08553 (2022-11)
Hybrid Transformers for Music Source Separation
2603.14889 (2026-03)
SDiaReward: Modeling and Benchmarking Spoken Dialogue Rewards with Modality and Colloquialness
upsample 2 other papers
2 later
2211.08553 (2022-11)
Hybrid Transformers for Music Source Separation
2603.14889 (2026-03)
SDiaReward: Modeling and Benchmarking Spoken Dialogue Rewards with Modality and Colloquialness
average_metric 1 other paper
1 later
2211.08553 (2022-11)
Hybrid Transformers for Music Source Separation
build_metadata 1 other paper
1 later
2211.08553 (2022-11)
Hybrid Transformers for Music Source Separation
free_port 1 other paper
1 later
2211.08553 (2022-11)
Hybrid Transformers for Music Source Separation
get_name 1 other paper
1 later
2211.08553 (2022-11)
Hybrid Transformers for Music Source Separation
select_norm 1 other paper
1 later
2002.08688 (2020-02)
An empirical study of Conv-TasNet

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

Single-channel, speaker-independent speech separation methods have recently seen great progress. However, the accuracy, latency, and computational cost of such methods remain insufficient. The majority of the previous methods have formulated the separation problem through the time-frequency representation of the mixed signal, which has several drawbacks, including the decoupling of the phase and magnitude of the signal, the suboptimality of time-frequency representation for speech separation, and the long latency in calculating the spectrograms. To address these shortcomings, we propose a fully-convolutional time-domain audio separation network (Conv-TasNet), a deep learning framework for end-to-end time-domain speech separation. Conv-TasNet uses a linear encoder to generate a representation of the speech waveform optimized for separating individual speakers. Speaker separation is achieved by applying a set of weighting functions (masks) to the encoder output. The modified encoder representations are then inverted back to the waveforms using a linear decoder. The masks are found using a temporal convolutional network (TCN) consisting of stacked 1-D dilated convolutional blocks, which allows the network to model the long-term dependencies of the speech signal while maintaining a small model size. The proposed Conv-TasNet system significantly outperforms previous time-frequency masking methods in separating two- and three-speaker mixtures. Additionally, Conv-TasNet surpasses several ideal time-frequency magnitude masks in two-speaker speech separation as evaluated by both objective distortion measures and subjective quality assessment by human listeners. Finally, Conv-TasNet has a significantly smaller model size and a shorter minimum latency, making it a suitable solution for both offline and real-time speech separation applications.

For agents

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

get_harvested_code_for_paper("1809.07454")
get_code_for_paper("1809.07454")
have("1809.07454")

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

curl https://syntology.ai/api/ran/1809.07454.json

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

[![Syntology run record](https://syntology.ai/api/ran/1809.07454.svg)](https://syntology.ai/paper/1809.07454)

Connect an agent — have() is free.