SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1803.09877 · 2018

DRACO: Byzantine-resilient Distributed Training via Redundant Gradients

arXiv · PDF · Open in the Atlas

Code that ran

4 of 10 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 3 have no contract verdict; 1 failure of our instrument; 6 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/1803.09877.json · badge: /api/ran/1803.09877.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
hwang595/Draco canonical (classifier label)
classifier basis: census
4 ran · 6 no recorded run
FunctionStatusWhere it lives
add_fit_args Ran
ran; our contract draft was wrong, not the code
hwang595/Draco/src/distributed_evaluator.py
pointer only (licence: MIT); the stored copy get_code serves, from hwang595/DETOX, is pointer only (licence: NONE) · get_code("3b55ad00f11e2fbf")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from hwang595/DETOX, and GitHub detected no licence in that repository when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

add_fit_args Ran
ran on a synthesized input
hwang595/Draco/src/distributed_nn.py
code served (permissive licence) · get_code("b8d6643e20276b6a")
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 Namespace. 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 argparse
import torch
import torch.nn as nn

# Create a minimal parser
parser = argparse.ArgumentParser(description='Test parser for add_fit_args')

# Define the function from the problem
def add_fit_args(parser):
    """
    parser : argparse.ArgumentParser
    return a parser added with args required by fit
    """
    parser.add_argument('--batch-size', type=int, default=128, metavar='N', help='input batch size for training (default: 64)')
    parser.add_argument('--test-batch-size', type=int, default=100, metavar='N', help='input batch size for testing (default: 1000)')
    parser.add_argument('--max-steps', type=int, default=10000, metavar='N', help='the maximum number of iterations')
    parser.add_argument('--epochs', type=int, default=100, metavar='N', help='number of epochs to train (default: 10)')
    parser.add_argument('--lr', type=float, default=0.01, metavar='LR', help='learning rate (default: 0.01)')
    parser.add_argument('--momentum', type=float, default=0.5, metavar='M', help='SGD momentum (default: 0.5)')
    parser.add_argument('--no-cuda', action='store_true', default=False, help='disables CUDA training')
    parser.add_argument('--seed', type=int, default=1, metavar='S', help='random seed (default: 1)')
    parser.add_argument('--log-interval', type=int, default=10, metavar='N', help='how many batches to wait before logging training status')
    parser.add_argument('--network', type=str, default='LeNet', metavar='N', help='which kind of network we are going to use, support LeNet and ResNet currently')
    parser.add_argument('--mode', type=str, default='normal', metavar='N', help='determine if we use normal averaged gradients or geometric median (in normal mode)                         or whether we use normal/majority vote in coded mode to udpate the model')
    parser.add_argument('--dataset', type=str, default='MNIST', metavar='N', help='which dataset used in training, MNIST and Cifar10 supported currently')
    parser.add_argument('--comm-type', type=str, default='Bcast', metavar='N', help='which kind of method we use during the mode fetching stage')
    parser.add_argument('--err-mode', type=str, default='rev_grad', metavar='N', help='which type of byzantine err we are going to simulate rev_grad/constant/random are supported')
    parser.add_argument('--approach', type=str, default='maj_vote', metavar='N', help='method used to achieve byzantine tolerence, currently majority vote is supported set to normal will return to normal mode')
    parser.add_argument('--num-aggregate', type=int, default=5, metavar='N', help='how many number of gradients we wish to gather at each iteration')
    parser.add_argument('--eval-freq', type=int, default=50, metavar='N', help='it determines per how many step the model should be evaluated')
    parser.add_argument('--train-dir', type=str, default='output/models/', metavar='N', help='directory to save the temp model during the training process for evaluation')
    parser.add_argument('--adversarial', type=int, default=1, metavar='N', help='how much adversary we want to add to a certain worker')
    parser.add_argument('--worker-fail', type=int, default=2, metavar='N', help='how many number of worker nodes we want to simulate byzantine error on')
    parser.add_argument('--group-size', type=int, default=5, metavar='N', help='in majority vote how many worker nodes are in a certain group')
    parser.add_argument('--compress-grad', type=str, default='compress', metavar='N', help='compress/none indicate if we compress the gradient matrix before communication')
    parser.add_argument('--checkpoint-step', type=int, default=0, metavar='N', help='which step to proceed the training process')
    args = parser.parse_args()
    return args

# Build ARGS dict with values for each parameter
ARGS = {
    'parser': parser
}
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 Namespace — a Namespace output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Namespace 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
add_fit_args Ran
ran on a synthesized input
hwang595/Draco/src/single_machine.py
code served (permissive licence) · get_code("98bdc33dfac0b80d")
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 Namespace. 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 argparse
import torch
import numpy as np

torch.manual_seed(0)

# Create a minimal argument parser
parser = argparse.ArgumentParser(description='Test parser for add_fit_args')

# Define the ARGS dict with all parameters the function expects
ARGS = {
    'parser': parser
}
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 Namespace — a Namespace output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Namespace 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
default_collate Ran
ran on a synthesized input
hwang595/Draco/src/data_loader_ops/my_data_loader.py
code served (permissive licence) · get_code("eb5ca9515b0fe778")
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 (2, 3, 4) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import collections
import numpy as np

torch.manual_seed(0)

# The function expects a 'batch' parameter which is a list of items
# Let's create different test cases for the various branches

# Test case 1: batch of tensors
batch_of_tensors = [torch.randn(3, 4) for _ in range(2)]

# Test case 2: batch of numpy arrays
batch_of_numpy = [np.random.randn(3, 4).astype(np.float32) for _ in range(2)]

# Test case 3: batch of integers
batch_of_ints = [1, 2]

# Test case 4: batch of floats
batch_of_floats = [1.5, 2.5]

# Test case 5: batch of dicts
batch_of_dicts = [
    {'a': torch.randn(3, 4), 'b': torch.randn(2, 5)},
    {'a': torch.randn(3, 4), 'b': torch.randn(2, 5)}
]

# Test case 6: batch of sequences/tuples
batch_of_sequences = [
    (torch.randn(3, 4), torch.randn(2, 5)),
    (torch.randn(3, 4), torch.randn(2, 5))
]

# Pick one for the main test - using tensors as primary case
ARGS = {
    'batch': batch_of_tensors
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 3, dtype family float, shape (2, 3, 4) — shape (2, 3, 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
accuracy Not yet run
unverified (harvested, no recorded run)
hwang595/Draco/src/distributed_evaluator.py
pointer only (licence: MIT); the stored copy get_code serves, from kkonstantinidis/ByzShield, is pointer only (licence: NONE) · get_code("4d3faaaa1c706b0b")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from kkonstantinidis/ByzShield, and GitHub detected no licence in that repository when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

get_batch Not yet run
unverified (harvested, no recorded run)
hwang595/Draco/src/datasets/utils.py
code served (permissive licence) · get_code("d5ef5fcb7374c455")
How it ran

no invocation recorded.

group_assign Not yet run
unverified (harvested, no recorded run)
hwang595/Draco/src/util.py
code served (permissive licence) · get_code("edc4f2fa379308d5")
How it ran

no invocation recorded.

load_data Not yet run
unverified (harvested, no recorded run)
hwang595/Draco/src/util.py
code served (permissive licence) · get_code("84d7005b5d219644")
How it ran

no invocation recorded.

pin_memory_batch Not yet run
unverified (harvested, no recorded run)
hwang595/Draco/src/data_loader_ops/my_data_loader.py
code served (permissive licence) · get_code("6d8c67dc116f1f8a")
How it ran

no invocation recorded.

search_w Not yet run
unverified (harvested, no recorded run)
hwang595/Draco/src/coding.py
code served (permissive licence) · get_code("c8d56eb161c04336")
How it ran

no invocation recorded.

Also present in other papers' code

Of 10 samples lifted from this paper's linked repositories, 10 carry a function identity and 0 do not. Of the 10 distinct functions with an identity, 4 are also present in other papers whose code we hold (3 with at least one paper dated earlier, 0 the same month, 4 later) and 6 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
accuracy 23 other papers
3 dated earlier, 20 later
1603.07285 (2016-03)
A guide to convolution arithmetic for deep learning
1608.07017 (2016-08)
Ambient Sound Provides Supervision for Visual Learning
1608.08710 (2016-08)
Pruning Filters for Efficient ConvNets
and 20 more not named here
default_collate 4 other papers
1 dated earlier, 3 later
1703.05593 (2017-03)
Convolutional neural network architecture for geometric matching
1810.10510 (2018-10)
Neighbourhood Consensus Networks
2003.12059 (2020-03)
Correspondence Networks with Adaptive Neighbourhood Consensus
and 1 more not named here
pin_memory_batch 4 other papers
1 dated earlier, 3 later
1703.05593 (2017-03)
Convolutional neural network architecture for geometric matching
1810.10510 (2018-10)
Neighbourhood Consensus Networks
2003.12059 (2020-03)
Correspondence Networks with Adaptive Neighbourhood Consensus
and 1 more not named here
add_fit_args 1 other paper
1 later
1907.12205 (2019-07)
DETOX: A Redundancy-based Framework for Faster and More Robust Gradient Aggregation

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

Distributed model training is vulnerable to byzantine system failures and adversarial compute nodes, i.e., nodes that use malicious updates to corrupt the global model stored at a parameter server (PS). To guarantee some form of robustness, recent work suggests using variants of the geometric median as an aggregation rule, in place of gradient averaging. Unfortunately, median-based rules can incur a prohibitive computational overhead in large-scale settings, and their convergence guarantees often require strong assumptions. In this work, we present DRACO, a scalable framework for robust distributed training that uses ideas from coding theory. In DRACO, each compute node evaluates redundant gradients that are used by the parameter server to eliminate the effects of adversarial updates. DRACO comes with problem-independent robustness guarantees, and the model that it trains is identical to the one trained in the adversary-free setup. We provide extensive experiments on real datasets and distributed setups across a variety of large-scale models, where we show that DRACO is several times, to orders of magnitude faster than median-based approaches.

For agents

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

get_harvested_code_for_paper("1803.09877")
get_code_for_paper("1803.09877")
have("1803.09877")

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

curl https://syntology.ai/api/ran/1803.09877.json

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

[![Syntology run record](https://syntology.ai/api/ran/1803.09877.svg)](https://syntology.ai/paper/1803.09877)

Connect an agent — have() is free.