SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2203.08243 · ICLR · 2022

Unified Visual Transformer Compression

Tianlong Chen, Zhangyang Wang, Ji Liu, Jiayi Shen, Shixing Yu, Huan Yuan, Jianchao Tan, Sen Yang

arXiv · PDF · Open in the Atlas

Code that ran

9 of 16 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (3 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 1 violated it and 5 have no contract verdict; 3 failures of our instrument; 7 no recorded run), measured 2026-11-22; 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/2203.08243.json · badge: /api/ran/2203.08243.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
vita-group/uvc canonical (classifier label)
classifier basis: classifier_2025
2 ran · 1 no recorded run
VITA-Group/UVC marked official in Papers with Code
Papers with Code archive
5 ran · 5 no recorded run
copy not recorded community
not in the paper's declared repositories
2 ran · 1 no recorded run
FunctionStatusWhere it lives
Attention Ran
ran on a synthesized input · constructed an object; no result was computed
VITA-Group/UVC/UVC/models/modeling.py
code served (permissive licence) · get_code("bfc8fe0ddf96d4c7")
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 an instance of its own class, Attention: the run constructed the object and did not call it, so no result was computed. 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 math
import torch
import torch.nn as nn
from torch.nn import CrossEntropyLoss, Dropout, Softmax, Linear, Conv2d, LayerNorm

torch.manual_seed(0)

# Create a minimal config class that provides the required attributes
class MinimalConfig:
    def __init__(self):
        self.hidden_size = 8
        self.transformer = {
            'num_heads': 2,
            'attention_dropout_rate': 0.1
        }

# Create config instance
config = MinimalConfig()

# vis parameter - boolean flag for visualization
vis = False

ARGS = {
    'config': config,
    'vis': vis
}
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 Attention — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a Attention 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
Mlp Ran
ran on a synthesized input · constructed an object; no result was computed
VITA-Group/UVC/UVC/models/modeling.py
code served (permissive licence) · get_code("3215e1b62b114566")
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 an instance of its own class, Mlp: the run constructed the object and did not call it, so no result was computed. 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
from torch.nn import Linear, Dropout

torch.manual_seed(0)

# Create a minimal config class with required attributes
class Config:
    def __init__(self):
        self.hidden_size = 4
        self.transformer = {
            'mlp_dim': 8,
            'dropout_rate': 0.1
        }

config = Config()

# Construct ARGS dictionary with the single parameter
ARGS = {
    'config': config
}
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 Mlp — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a Mlp 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
PreActBottleneck Ran
ran on a synthesized input
VITA-Group/UVC/UVC/models/modeling.py
code served (permissive licence) · get_code("df180144004badc3")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, class construct-then-call prompt, model not recorded) and returned Tensor, ndim 4, 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 numpy as np

# INIT_ARGS for PreActBottleneck.__init__(cin, cout=None, cmid=None, stride=1)
INIT_ARGS = {
    'cin': 64,
    'cout': 128,
    'cmid': 32,
    'stride': 2
}

# ARGS for PreActBottleneck.forward(x)
# Input tensor shape: (batch, cin, height, width)
ARGS = {
    'x': torch.randn(2, 64, 8, 8)
}
Setup block drafted by
a language model, class construct-then-call prompt, model not recorded
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 4, dtype family float — no output shape is stated: observe_signature.py observed this class with its constructor sizes canonicalised where it could, which the node does not record, so its shape is not stated as this setup block's
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: torch and numpy seeded to 0 by the harness immediately before the class was constructed from INIT_ARGS (fixture_check.py)
ResNetV2 Ran
ran on a synthesized input · constructed an object; no result was computed
VITA-Group/UVC/UVC/models/modeling.py
code served (permissive licence) · get_code("e91e85b73a67665a")
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 an instance of its own class, ResNetV2: the run constructed the object and did not call it, so no result was computed. 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
from collections import OrderedDict
import torch.nn.functional as F
import numpy as np

torch.manual_seed(0)

class StdConv2d(nn.Conv2d):
    def forward(self, x):
        w = self.weight
        v, m = torch.var_mean(w, dim=[1, 2, 3], keepdim=True, unbiased=False)
        w = (w - m) / torch.sqrt(v + 1e-05)
        return F.conv2d(x, w, self.bias, self.stride, self.padding, self.dilation, self.groups)

def conv1x1(cin, cout, stride=1, bias=False):
    return StdConv2d(cin, cout, kernel_size=1, stride=stride, padding=0, bias=bias)

def conv3x3(cin, cout, stride=1, groups=1, bias=False):
    return StdConv2d(cin, cout, kernel_size=3, stride=stride, padding=1, bias=bias, groups=groups)

def np2th(weights, conv=False):
    """Possibly convert HWIO to OIHW."""
    if conv:
        weights = weights.transpose([3, 2, 0, 1])
    return torch.from_numpy(weights)

class PreActBottleneck(nn.Module):
    """Pre-activation (v2) bottleneck block."""

    def __init__(self, cin, cout=None, cmid=None, stride=1):
        super().__init__()
        cout = cout or cin
        cmid = cmid or cout // 4
        self.gn1 = nn.GroupNorm(32, cmid, eps=1e-06)
        self.conv1 = conv1x1(cin, cmid, bias=False)
        self.gn2 = nn.GroupNorm(32, cmid, eps=1e-06)
        self.conv2 = conv3x3(cmid, cmid, stride, bias=False)
        self.gn3 = nn.GroupNorm(32, cout, eps=1e-06)
        self.conv3 = conv1x1(cmid, cout, bias=False)
        self.relu = nn.ReLU(inplace=True)
        if stride != 1 or cin != cout:
            self.downsample = conv1x1(cin, cout, stride, bias=False)
            self.gn_proj = nn.GroupNorm(cout, cout)

    def forward(self, x):
        residual = x
        if hasattr(self, 'downsample'):
            residual = self.downsample(x)
            residual = self.gn_proj(residual)
        y = self.relu(self.gn1(self.conv1(x)))
        y = self.relu(self.gn2(self.conv2(y)))
        y = self.gn3(self.conv3(y))
        y = self.relu(residual + y)
        return y

    def load_from(self, weights, n_block, n_unit):
        pass

class ResNetV2(nn.Module):
    """Implementation of Pre-activation (v2) ResNet mode."""

    def __init__(self, block_units, width_factor):
        super().__init__()
        width = int(64 * width_factor)
        self.width = width
        self.root = nn.Sequential(OrderedDict([('conv', StdConv2d(3, width, kernel_size=7, stride=2, bias=False, padding=3)), ('gn', nn.GroupNorm(32, width, eps=1e-06)), ('relu', nn.ReLU(inplace=True)), ('pool', nn.MaxPool2d(kernel_size=3, stride=2, padding=0))]))
        self.body = nn.Sequential(OrderedDict([('block1', nn.Sequential(OrderedDict([('unit1', PreActBottleneck(cin=width, cout=width * 4, cmid=width))] + [(f'unit{i:d}', PreActBottleneck(cin=width * 4, cout=width * 4, cmid=width)) for i in range(2, block_units[0] + 1)]))), ('block2', nn.Sequential(OrderedDict([('unit1', PreActBottleneck(cin=width * 4, cout=width * 8, cmid=width * 2, stride=2))] + [(f'unit{i:d}', PreActBottleneck(cin=width * 8, cout=width * 8, cmid=width * 2)) for i in range(2, block_units[1] + 1)]))), ('block3', nn.Sequential(OrderedDict([('unit1', PreActBottleneck(cin=width * 8, cout=width * 16, cmid=width * 4, stride=2))] + [(f'unit{i:d}', PreActBottleneck(cin=width * 16, cout=width * 16, cmid=width * 4)) for i in range(2, block_units[2] + 1)])))]))

    def forward(self, x):
        x = self.root(x)
        x = self.body(x)
        return x

ARGS = {
    'block_units': [3, 4, 9],
    'width_factor': 1,
}
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 ResNetV2 — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a ResNetV2 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
StdConv2d Ran
ran on a synthesized input
VITA-Group/UVC/UVC/models/modeling.py
code served (permissive licence) · get_code("eadbbe03b5b32f01")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, class construct-then-call prompt, model not recorded) and returned Tensor, ndim 4, 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
import torch.nn.functional as F

class StdConv2d(nn.Conv2d):

    def forward(self, x):
        w = self.weight
        v, m = torch.var_mean(w, dim=[1, 2, 3], keepdim=True, unbiased=False)
        w = (w - m) / torch.sqrt(v + 1e-05)
        return F.conv2d(x, w, self.bias, self.stride, self.padding, self.dilation, self.groups)


INIT_ARGS = {
    'in_channels': 3,
    'out_channels': 4,
    'kernel_size': 3,
    'stride': 1,
    'padding': 1,
    'dilation': 1,
    'groups': 1,
    'bias': True,
    'padding_mode': 'zeros'
}

ARGS = {
    'x': torch.randn(2, 3, 5, 5)
}

if __name__ == '__main__':
    model = StdConv2d(**INIT_ARGS)
    output = model(**ARGS)
    print(f"Output shape: {output.shape}")
    print(f"Expected shape: torch.Size([2, 4, 5, 5])")
Setup block drafted by
a language model, class construct-then-call prompt, model not recorded
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 4, dtype family float — no output shape is stated: observe_signature.py observed this class with its constructor sizes canonicalised where it could, which the node does not record, so its shape is not stated as this setup block's
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: torch and numpy seeded to 0 by the harness immediately before the class was constructed from INIT_ARGS (fixture_check.py)
conv1x1 Ran
ran; our contract draft was wrong, not the code
this paper's copy was not recorded; identical code first harvested from maloadba/mgenseg_2d
pointer only; the stored copy get_code serves, from maloadba/mgenseg_2d, is pointer only (licence: NONE) · get_code("b0f916df39e78ae5")
How it ran

The body of this sample is withheld: this paper's copy of the code was not recorded (its link names no repository or source path), so no licence was read for 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.

conv3x3 Ran
ran; our contract draft was wrong, not the code
this paper's copy was not recorded; identical code first harvested from maloadba/mgenseg_2d
pointer only; the stored copy get_code serves, from maloadba/mgenseg_2d, is pointer only (licence: NONE) · get_code("4e5bf13dbdc4f008")
How it ran

The body of this sample is withheld: this paper's copy of the code was not recorded (its link names no repository or source path), so no licence was read for 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_tau Ran
ran, violated the contract we drafted
vita-group/uvc/UVC/joint_train.py
code served (permissive licence) · get_code("aa9253d77220768c")
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) 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 numpy as np

torch.manual_seed(0)

# Analyze the function:
# get_tau(max_tau, min_tau, ite, total)
# - max_tau: numeric value used in arithmetic
# - min_tau: numeric value used in arithmetic
# - ite: numeric value (iteration count) used in arithmetic
# - total: numeric value (total iterations) used in arithmetic
# All parameters are used in a simple linear interpolation formula

ARGS = {
    "max_tau": 1.0,
    "min_tau": 0.1,
    "ite": 5,
    "total": 10
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
numpy, 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
get_uvc_layers Ran
ran; our contract draft was wrong, not the code
vita-group/uvc/UVC/post_train.py
code served (permissive licence) · get_code("80da1ef156cc3a88")
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) and returned tuple, 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)

class SimpleTransformerBlock(nn.Module):
    def __init__(self):
        super().__init__()
        self.attn = nn.Module()
        self.attn.proj = nn.Linear(4, 3)
        self.mlp = nn.Module()
        self.mlp.fc1 = nn.Linear(4, 5)
        self.mlp.fc2 = nn.Linear(5, 3)
    
    def forward(self, x):
        return x

class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.block1 = SimpleTransformerBlock()
        self.block2 = SimpleTransformerBlock()
    
    def forward(self, x):
        return x

model = SimpleModel()

ARGS = {
    'model': model,
    'args': None
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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 1, 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
Block Not yet run
unverified (harvested, no recorded run)
VITA-Group/UVC/UVC/models/modeling.py
code served (permissive licence) · get_code("195dcd4e31d9f812")
How it ran

no invocation recorded.

Embeddings Not yet run
unverified (harvested, no recorded run)
VITA-Group/UVC/UVC/models/modeling.py
code served (permissive licence) · get_code("8020519bcd7e5931")
How it ran

no invocation recorded.

Encoder Not yet run
unverified (harvested, no recorded run)
VITA-Group/UVC/UVC/models/modeling.py
code served (permissive licence) · get_code("f5a8c54634cf9950")
How it ran

no invocation recorded.

Transformer Not yet run
unverified (harvested, no recorded run)
VITA-Group/UVC/UVC/models/modeling.py
code served (permissive licence) · get_code("a4485af62ba072e0")
How it ran

no invocation recorded.

VisionTransformer Not yet run
unverified (harvested, no recorded run)
VITA-Group/UVC/UVC/models/modeling.py
code served (permissive licence) · get_code("4900a340f86d38d4")
How it ran

no invocation recorded.

complex_accuracy Not yet run
unverified (harvested, no recorded run)
vita-group/uvc/UVC/joint_train.py
code served (permissive licence) · get_code("3e8918f575824271")
How it ran

no invocation recorded.

simple_accuracy Not yet run
unverified (harvested, no recorded run)
this paper's copy was not recorded; identical code first harvested from ink-usc/expl-refinement
pointer only · get_code("3c241ecfe3749a6d")
How it ran

The body of this sample is withheld: this paper's copy of the code was not recorded (its link names no repository or source path), so no licence was read for 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.

Also present in other papers' code

Of 16 samples lifted from this paper's linked repositories, 16 carry a function identity and 0 do not. Of the 16 distinct functions with an identity, 10 are also present in other papers whose code we hold (9 with at least one paper dated earlier, 2 the same month, 7 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
simple_accuracy 59 other papers
41 dated earlier, 18 later
1506.05865 (2015-06)
LCSTS: A Large Scale Chinese Short Text Summarization Dataset
1509.01626 (2015-09)
Character-level Convolutional Networks for Text Classification *
1810.04805 (2018-10)
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
and 56 more not named here
conv1x1
brick of this paper's assembly: helper
33 other papers
11 dated earlier, 2 the same month, 20 later · 3 only as an assembly brick
1912.11370 (2019-12)
Big Transfer (BiT): General Visual Representation Learning
2010.11929 (2020-10)
An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale
2102.04306 (2021-02)
TransUNet: Transformers Make Strong Encoders for Medical Image Segmentation
and 30 more not named here
conv3x3
brick of this paper's assembly: helper
33 other papers
11 dated earlier, 2 the same month, 20 later · 3 only as an assembly brick
1912.11370 (2019-12)
Big Transfer (BiT): General Visual Representation Learning
2010.11929 (2020-10)
An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale
2102.04306 (2021-02)
TransUNet: Transformers Make Strong Encoders for Medical Image Segmentation
and 30 more not named here
StdConv2d
brick of this paper's assembly: helper
5 other papers
2 dated earlier, 3 later · 4 only as an assembly brick
2010.11929 (2020-10)
An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale
2103.12091 (2021-03)
Transformer-Based Attention Networks for Continuous Pixel-Wise Prediction
2403.02746 (2024-03)
Learning without Exact Guidance: Updating Large-scale High-resolution Land Cover Maps from Low-resolution Historical Labels
and 2 more not named here
PreActBottleneck
brick of this paper's assembly: helper
4 other papers
2 dated earlier, 2 later · 2 only as an assembly brick
2010.11929 (2020-10)
An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale
2103.12091 (2021-03)
Transformer-Based Attention Networks for Continuous Pixel-Wise Prediction
2501.18936 (2025-01)
Revisit Visual Prompt Tuning: The Expressiveness of Prompt Experts
and 1 more not named here
Encoder
brick of this paper's assembly: model
3 other papers
3 dated earlier
2010.11929 (2020-10)
An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale
2103.12091 (2021-03)
Transformer-Based Attention Networks for Continuous Pixel-Wise Prediction
2103.15691 (2021-03)
ViViT: A Video Vision Transformer
ResNetV2
brick of this paper's assembly: helper
3 other papers
1 dated earlier, 2 later
2010.11929 (2020-10)
An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale
2501.18936 (2025-01)
Revisit Visual Prompt Tuning: The Expressiveness of Prompt Experts
2510.16446 (2025-10)
VIPAMIN: Visual Prompt Initialization via Embedding Selection and Subspace Expansion
Transformer
brick of this paper's assembly: model
1 other paper
1 dated earlier
2010.11929 (2020-10)
An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale
VisionTransformer
brick of this paper's assembly: entry
1 other paper
1 dated earlier
2010.11929 (2020-10)
An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale
get_tau 1 other paper
1 later
2407.02068 (2024-07)
LPViT: Low-Power Semi-structured Pruning for Vision Transformers

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

Vision transformers (ViTs) have gained popularity recently. Even without customized image operators such as convolutions, ViTs can yield competitive performance when properly trained on massive data. However, the computational overhead of ViTs remains prohibitive, due to stacking multi-head self-attention modules and else. Compared to the vast literature and prevailing success in compressing convolutional neural networks, the study of Vision Transformer compression has also just emerged, and existing works focused on one or two aspects of compression. This paper proposes a unified ViT compression framework that seamlessly assembles three effective techniques: pruning, layer skipping, and knowledge distillation. We formulate a budget-constrained, end-to-end optimization framework, targeting jointly learning model weights, layer-wise pruning ratios/masks, and skip configurations, under a distillation loss. The optimization problem is then solved using the primal-dual algorithm. Experiments are conducted with several ViT variants, e.g. DeiT and T2T-ViT backbones on the ImageNet dataset, and our approach consistently outperforms recent competitors. For example, DeiT-Tiny can be trimmed down to 50% of the original FLOPs almost without losing accuracy. Codes are available online: https://github.com/VITA-Group/UVC.

For agents

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

get_harvested_code_for_paper("2203.08243")
get_code_for_paper("2203.08243")
have("2203.08243")

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

curl https://syntology.ai/api/ran/2203.08243.json

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

[![Syntology run record](https://syntology.ai/api/ran/2203.08243.svg)](https://syntology.ai/paper/2203.08243)

Connect an agent — have() is free.