SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2106.11257 · ICML · 2022

Secure Distributed Training at Scale

Eduard Gorbunov, Max Ryabinin, Michael Diskin, Alexander Borzunov

arXiv · PDF · Open in the Atlas

Code that ran

6 of 9 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 6 have no contract verdict; 0 failures of our instrument; 3 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/2106.11257.json · badge: /api/ran/2106.11257.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
yandex-research/btard canonical (classifier label)
classifier basis: census
6 ran · 3 no recorded run
FunctionStatusWhere it lives
centered_clip Ran
ran on a synthesized input
yandex-research/btard/resnet/centered_clip.py
code served (permissive licence) · get_code("92df667d07531d60")
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 CenteredClipResult. 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 dataclasses

torch.manual_seed(0)

# Create small tensors
# input_tensors: shape (batch_size, *result_shape)
# Let's use batch_size=2, and result_shape=(4,) for a 1D result
batch_size = 2
result_dim = 4

input_tensors = torch.randn(batch_size, result_dim)

# weights: should match batch dimension
weights = torch.ones(batch_size)

# tau: a float threshold parameter
tau = 1.5

# n_iters: number of iterations (optional, has default)
n_iters = 20

# eps: convergence threshold (optional, has default)
eps = 1e-6

ARGS = {
    "input_tensors": input_tensors,
    "weights": weights,
    "tau": tau,
    "n_iters": n_iters,
    "eps": eps,
}
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 CenteredClipResult — a CenteredClipResult output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a CenteredClipResult 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
evaluate_accuracy Ran
ran on a synthesized input
yandex-research/btard/resnet/training_utils.py
code served (permissive licence) · get_code("a904de8f2aa42d6f")
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 as nn
import torch.utils.data

torch.manual_seed(0)

# Define a simple model
class SimpleClassifier(nn.Module):
    def __init__(self, input_dim=4, num_classes=3):
        super().__init__()
        self.linear = nn.Linear(input_dim, num_classes)
    
    def forward(self, x):
        return self.linear(x)

# Create model on CPU
model = SimpleClassifier(input_dim=4, num_classes=3)
model.eval()

# Create a simple dataset
class SimpleDataset(torch.utils.data.Dataset):
    def __init__(self, num_samples=8, input_dim=4, num_classes=3):
        self.num_samples = num_samples
        self.input_dim = input_dim
        self.num_classes = num_classes
        self.inputs = torch.randn(num_samples, input_dim)
        self.targets = torch.randint(0, num_classes, (num_samples,))
    
    def __len__(self):
        return self.num_samples
    
    def __getitem__(self, idx):
        return self.inputs[idx], self.targets[idx]

dataset = SimpleDataset(num_samples=8, input_dim=4, num_classes=3)

# Define ARGS dictionary
ARGS = {
    "model": model,
    "dataset": dataset,
    "batch_size": 2,
    "num_workers": 0
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 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
restore_from_parts Ran
ran on a synthesized input
yandex-research/btard/resnet/centered_clip.py
code served (permissive licence) · get_code("96f98edcffa1a140")
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 torch

torch.manual_seed(0)

# Create sample chunks and shapes
# The function expects chunks (flattened tensors) and their corresponding shapes
chunk1 = torch.randn(6)  # 2x3 when reshaped
chunk2 = torch.randn(8)  # 2x4 when reshaped
chunk3 = torch.randn(4)  # 2x2 when reshaped

chunks = [chunk1, chunk2, chunk3]
shapes = [torch.Size([2, 3]), torch.Size([2, 4]), torch.Size([2, 2])]

ARGS = {
    "chunks": chunks,
    "shapes": shapes,
}
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 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
sample_tensors Ran
ran on a synthesized input
yandex-research/btard/albert/hivemind/benchmarks/benchmark_averaging.py
code served (permissive licence) · get_code("df3533d0b1a9d882")
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 torch

torch.manual_seed(0)

ARGS = {
    "hid_size": 4,
    "num_layers": 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 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
split_into_parts Ran
ran on a synthesized input
yandex-research/btard/resnet/centered_clip.py
code served (permissive licence) · get_code("52afa7d9072a594d")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple, ndim 2, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# Create a sequence of small tensors with distinct shapes
tensor1 = torch.randn(2, 3)  # 6 elements
tensor2 = torch.randn(4, 2)  # 8 elements
tensor3 = torch.randn(1, 5)  # 5 elements
# Total: 19 elements

tensors = [tensor1, tensor2, tensor3]
num_parts = 4

ARGS = {
    "tensors": tensors,
    "num_parts": num_parts,
}
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 tuple, ndim 2, dtype family float — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
verify_equal_parameters Ran
ran on a synthesized input
yandex-research/btard/resnet/training_utils.py
code served (permissive licence) · get_code("74f2dafa5b436318")
How it ran

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

import torch
import torch.nn as nn
import torch.distributed as dist
from unittest.mock import MagicMock, patch

torch.manual_seed(0)

# Create a minimal model with parameters
class SimpleModel(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

model = SimpleModel()

# Mock the distributed functions since we're not in a real distributed setting
original_get_world_size = dist.get_world_size
original_all_gather = dist.all_gather

def mock_get_world_size():
    return 2

def mock_all_gather(tensor_list, tensor):
    # Simulate gathering checksums from 2 workers
    # Make them equal so the function passes
    checksum_val = tensor.clone()
    tensor_list[0] = checksum_val
    tensor_list[1] = checksum_val

# Patch the distributed functions
dist.get_world_size = mock_get_world_size
dist.all_gather = mock_all_gather

ARGS = {
    "model": model
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type bool, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
create_instances_from_document Not yet run
unverified (harvested, no recorded run)
yandex-research/btard/albert/experiments/tokenize_wikitext103.py
code served (permissive licence) · get_code("b6beffa06be72594")
How it ran

no invocation recorded.

get_model Not yet run
unverified (harvested, no recorded run)
yandex-research/btard/albert/experiments/run_trainer.py
code served (permissive licence) · get_code("1960851ee3b8f1ba")
How it ran

no invocation recorded.

tokenize_function Not yet run
unverified (harvested, no recorded run)
yandex-research/btard/albert/experiments/tokenize_wikitext103.py
code served (permissive licence) · get_code("d2b88e48b87e9450")
How it ran

no invocation recorded.

Also present in other papers' code

Of 9 samples lifted from this paper's linked repositories, 9 carry a function identity and 0 do not. Of the 9 distinct functions with an identity, 2 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 2 the same month, 0 later) and 7 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
create_instances_from_document 1 other paper
1 the same month
2106.10207 (2021-06)
Distributed Deep Learning In Open Collaborations
tokenize_function 1 other paper
1 the same month
2106.10207 (2021-06)
Distributed Deep Learning In Open Collaborations

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

Many areas of deep learning benefit from using increasingly larger neural networks trained on public data, as is the case for pre-trained models for NLP and computer vision. Training such models requires a lot of computational resources (e.g., HPC clusters) that are not available to small research groups and independent researchers. One way to address it is for several smaller groups to pool their computational resources together and train a model that benefits all participants. Unfortunately, in this case, any participant can jeopardize the entire training run by sending incorrect updates, deliberately or by mistake. Training in presence of such peers requires specialized distributed training algorithms with Byzantine tolerance. These algorithms often sacrifice efficiency by introducing redundant communication or passing all updates through a trusted server, making it infeasible to apply them to large-scale deep learning, where models can have billions of parameters. In this work, we propose a novel protocol for secure (Byzantine-tolerant) decentralized training that emphasizes communication efficiency.

For agents

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

get_harvested_code_for_paper("2106.11257")
get_code_for_paper("2106.11257")
have("2106.11257")

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

curl https://syntology.ai/api/ran/2106.11257.json

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

[![Syntology run record](https://syntology.ai/api/ran/2106.11257.svg)](https://syntology.ai/paper/2106.11257)

Connect an agent — have() is free.