SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2005.10481 · 2020

AOWS: Adaptive and optimal network width search with latency constraints

arXiv · PDF · Open in the Atlas

Code that ran

7 of 7 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 7 have no contract verdict; 0 failures of our instrument; 0 no recorded run), measured 2026-11-21; 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/2005.10481.json · badge: /api/ran/2005.10481.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
bermanmaxim/AOWS canonical (classifier label)
classifier basis: census
7 ran · 0 no recorded run
FunctionStatusWhere it lives
accuracy Ran
ran on a synthesized input
bermanmaxim/AOWS/misc.py
code served (permissive licence) · get_code("05edb495f972da82")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned list, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

torch.manual_seed(0)

# Create mock tensors
batch_size = 2
num_classes = 4

# output: model predictions (logits), shape [batch_size, num_classes]
output = torch.randn(batch_size, num_classes)

# target: ground truth labels, shape [batch_size]
target = torch.randint(0, num_classes, (batch_size,))

# topk: tuple of k values to compute accuracy for
topk = (1, 2)

# return_correct_k: whether to return the correct_k tensors
return_correct_k = False

ARGS = {
    "output": output,
    "target": target,
    "topk": topk,
    "return_correct_k": return_correct_k,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family float — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
aows_temp Ran
ran on a synthesized input
bermanmaxim/AOWS/train_nas.py
code served (permissive licence) · get_code("580670be3c585327")
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

class Args:
    def __init__(self):
        self.AOWS_warmup = 2
        self.AOWS_min_temp = 0.0001
        self.epochs = 20

ARGS = {
    'epoch': 5,
    'epoch_len': 100,
    'iteration': 50,
    'args': Args()
}
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
complete Ran
ran on a synthesized input
bermanmaxim/AOWS/viterbi.py
code served (permissive licence) · get_code("eefdfc0fa86b3dd0")
How it ran

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

import numpy as np

# Build small test data for the complete() function
# unary: list of N arrays, each with varying number of states
# pairwise: list of N-1 matrices, each pairwise[i] connects states of unary[i] and unary[i+1]

unary = [
    np.array([1.0, 2.0, 3.0]),           # 3 states for node 0
    np.array([4.0, 5.0]),                 # 2 states for node 1
    np.array([6.0, 7.0, 8.0, 9.0])       # 4 states for node 2
]

pairwise = [
    np.array([[0.1, 0.2],                 # 3x2 pairwise between node 0 and node 1
              [0.3, 0.4],
              [0.5, 0.6]]),
    np.array([[1.0, 2.0, 3.0, 4.0],       # 2x4 pairwise between node 1 and node 2
              [5.0, 6.0, 7.0, 8.0]])
]

ARGS = {
    "unary": unary,
    "pairwise": pairwise,
    "fill": -np.inf
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
maxsum Ran
ran on a synthesized input
bermanmaxim/AOWS/viterbi.py
code served (permissive licence) · get_code("ee2f4aa1e69125ab")
How it ran

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

import numpy as np

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

# Define parameters for maxsum function
# unary: shape (N, S) - N is number of time steps, S is number of states
# pairwise: shape (N-1, S, S) - transition scores between states
# states: array of valid state counts at each time step

N = 3  # number of time steps
S = 4  # maximum number of states

# Create unary potentials: (N, S) array
unary = np.random.randn(N, S).astype(np.float64)

# Create pairwise potentials: (N-1, S, S) array
pairwise = np.random.randn(N - 1, S, S).astype(np.float64)

# Create states array: valid number of states at each time step
# states should have length N, with values indicating how many states are valid at each step
states = np.array([4, 3, 2], dtype=np.int64)

ARGS = {
    "unary": unary,
    "pairwise": pairwise,
    "states": states,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numba, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
score Ran
ran on a synthesized input
bermanmaxim/AOWS/viterbi.py
code served (permissive licence) · get_code("dfc753382f18458e")
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 numpy as np

torch.manual_seed(0)

# path: a list of indices representing a sequence of states
path = [0, 1, 2]

# unary: indexed by timestep, then by state index
# unary[t][state] gives the unary score for that state at time t
unary = [
    torch.tensor([1.0, 2.0, 3.0, 4.0]),  # t=0: scores for states 0,1,2,3
    torch.tensor([0.5, 1.5, 2.5, 3.5]),  # t=1
    torch.tensor([0.2, 1.2, 2.2, 3.2]),  # t=2
]

# pairwise: indexed by timestep (transition step), then by [prev_state, curr_state]
# pairwise[i][prev, curr] gives the pairwise score for transitioning from prev to curr
pairwise = [
    torch.tensor([[0.1, 0.2, 0.3, 0.4],
                  [0.15, 0.25, 0.35, 0.45],
                  [0.12, 0.22, 0.32, 0.42],
                  [0.11, 0.21, 0.31, 0.41]]),  # transition from t=0 to t=1
    torch.tensor([[0.05, 0.15, 0.25, 0.35],
                  [0.06, 0.16, 0.26, 0.36],
                  [0.07, 0.17, 0.27, 0.37],
                  [0.08, 0.18, 0.28, 0.38]]),  # transition from t=1 to t=2
]

detail = False

ARGS = {
    "path": path,
    "unary": unary,
    "pairwise": pairwise,
    "detail": detail,
}
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 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
soft_cross_entropy Ran
ran on a synthesized input
bermanmaxim/AOWS/misc.py
code served (permissive licence) · get_code("49afece34f1fc351")
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 (2,) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# Create small tensors for the function
# output: logits from model, shape (batch_size, num_classes)
# target: soft target probabilities, shape (batch_size, num_classes)

batch_size = 2
num_classes = 4

output = torch.randn(batch_size, num_classes)
target = torch.randn(batch_size, num_classes)
target = torch.softmax(target, dim=1)  # normalize to valid probabilities

ARGS = {
    "output": output,
    "target": target,
}
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 (2,) — shape (2,) 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
tuplify Ran
ran on a synthesized input
bermanmaxim/AOWS/misc.py
code served (permissive licence) · get_code("e5242ed4761b8e1b")
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)

# The function tuplify recursively converts lists to tuples and recursively
# processes dict values. It handles lists, dicts, and returns other types as-is.

# Test with a nested structure containing lists, dicts, and primitive values
listything = [
    1,
    [2, 3],
    {"a": [4, 5], "b": 6},
    torch.tensor([1.0, 2.0, 3.0, 4.0]),
    {"nested": {"list": [7, 8, 9]}},
]

ARGS = {
    "listything": listything,
}
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

Also present in other papers' code

Of 7 samples lifted from this paper's linked repositories, 7 carry a function identity and 0 do not. Of the 7 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 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.

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 2026-11-21.

Repositories linked to this paper

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

Abstract

Neural architecture search (NAS) approaches aim at automatically finding novel CNN architectures that fit computational constraints while maintaining a good performance on the target platform. We introduce a novel efficient one-shot NAS approach to optimally search for channel numbers, given latency constraints on a specific hardware. We first show that we can use a black-box approach to estimate a realistic latency model for a specific inference platform, without the need for low-level access to the inference computation. Then, we design a pairwise MRF to score any channel configuration and use dynamic programming to efficiently decode the best performing configuration, yielding an optimal solution for the network width search. Finally, we propose an adaptive channel configuration sampling scheme to gradually specialize the training phase to the target computational constraints. Experiments on ImageNet classification show that our approach can find networks fitting the resource constraints on different target platforms while improving accuracy over the state-of-the-art efficient networks.

For agents

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

get_harvested_code_for_paper("2005.10481")
get_code_for_paper("2005.10481")
have("2005.10481")

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

curl https://syntology.ai/api/ran/2005.10481.json

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

[![Syntology run record](https://syntology.ai/api/ran/2005.10481.svg)](https://syntology.ai/paper/2005.10481)

Connect an agent — have() is free.