SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2201.05587 · 2022

Transfer-Tuning: Reusing Auto-Schedules for Efficient Tensor Program Code Generation

arXiv · PDF · Open in the Atlas

Code that ran

9 of 12 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 5 have no contract verdict; 4 failures of our instrument; 3 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/2201.05587.json · badge: /api/ran/2201.05587.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
giclab/transfer-tuning canonical (classifier label)
classifier basis not recorded on the link
9 ran · 3 no recorded run
FunctionStatusWhere it lives
D_arch Ran
ran; our contract draft was wrong, not the code
giclab/transfer-tuning/src/models/gans/BigGAN/BigGAN.py
pointer only (licence: MIT); the stored copy get_code serves, from VICO-UoE/IT-GAN, is pointer only (licence: NONE) · get_code("85eb95cce203dfc7")
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 VICO-UoE/IT-GAN, 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.

D_arch Ran
ran; our contract draft was wrong, not the code
giclab/transfer-tuning/src/models/gans/BigGAN/BigGANdeep.py
code served (permissive licence) · get_code("a99a05ac52533be9")
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
import numpy as np

torch.manual_seed(0)

ARGS = {
    'ch': 64,
    'attention': '64',
    'ksize': '333333',
    'dilation': '111111'
}
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 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
G_arch Ran
ran; our contract draft was wrong, not the code
giclab/transfer-tuning/src/models/gans/BigGAN/BigGAN.py
pointer only (licence: MIT); the stored copy get_code serves, from VICO-UoE/IT-GAN, is pointer only (licence: NONE) · get_code("e8c308c9283407d7")
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 VICO-UoE/IT-GAN, 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.

G_arch Ran
ran; our contract draft was wrong, not the code
giclab/transfer-tuning/src/models/gans/BigGAN/BigGANdeep.py
code served (permissive licence) · get_code("44d6a6343f6eb182")
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
import numpy as np

torch.manual_seed(0)

ARGS = {
    'ch': 64,
    'attention': '64_128_256',
    'ksize': '333333',
    'dilation': '111111'
}
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 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
convert_to_hf_transformer_graphdef Ran
ran on a synthesized input
giclab/transfer-tuning/src/models/models.py
code served (permissive licence) · get_code("7fbfeaaf67384961")
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 GraphDef. 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 tensorflow as tf
import numpy as np

# Create a minimal TensorFlow model that satisfies the function's requirements
class MinimalTransformerModel(tf.Module):
    def __init__(self):
        super(MinimalTransformerModel, self).__init__()
        self.dense = tf.keras.layers.Dense(4)
        # Create a mock _saved_model_inputs_spec attribute
        self._saved_model_inputs_spec = {
            'input': tf.TensorSpec(shape=[None, None], dtype=tf.float32)
        }
    
    def __call__(self, x):
        return self.dense(x)

# Instantiate the model
model = MinimalTransformerModel()

# Define batch_size and seq_len
batch_size = 2
seq_len = 3

# Construct ARGS dictionary
ARGS = {
    "model": model,
    "batch_size": batch_size,
    "seq_len": seq_len
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type GraphDef — a GraphDef output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a GraphDef 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_ansor_times Ran
ran on a synthesized input
giclab/transfer-tuning/src/visualization/plot_fig5.py
code served (permissive licence) · get_code("cb17c38b6be78875")
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 with a tuning_info.json file
temp_dir = tempfile.mkdtemp()

tuning_data = {
    "kernel_1": {"tuned_time": 0.123},
    "kernel_2": {"tuned_time": 0.456},
    "kernel_3": {"tuned_time": 0.789}
}

tuning_info_path = os.path.join(temp_dir, 'tuning_info.json')
with open(tuning_info_path, 'w') as f:
    json.dump(tuning_data, f)

ARGS = {
    'path': temp_dir
}
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
get_main_results Ran
ran on a synthesized input
giclab/transfer-tuning/src/visualization/plot_fig5.py
code served (permissive licence) · get_code("8412fd6d4393d0ca")
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
import shutil

# Create a temporary directory with test JSON files
temp_dir = tempfile.mkdtemp()

# Create sample JSON files with the expected structure
test_data_1 = {
    "experiment_1": {
        "0": {
            "tt_time": 1.5,
            "search_time": 2.3
        }
    }
}

test_data_2 = {
    "experiment_2": {
        "0": {
            "tt_time": 0.8,
            "search_time": 1.2
        }
    }
}

test_data_3 = {
    "experiment_3": {
        "0": {
            "tt_time": 3.1,
            "search_time": 4.5
        }
    }
}

# Write test JSON files to the temporary directory
with open(os.path.join(temp_dir, "result_1.json"), "w") as f:
    json.dump(test_data_1, f)

with open(os.path.join(temp_dir, "result_2.json"), "w") as f:
    json.dump(test_data_2, f)

with open(os.path.join(temp_dir, "result_3.json"), "w") as f:
    json.dump(test_data_3, f)

# Also create a non-JSON file to ensure the function filters correctly
with open(os.path.join(temp_dir, "readme.txt"), "w") as f:
    f.write("This is not a JSON file")

ARGS = {
    "path": temp_dir
}
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
get_model_set_data Ran
ran on a synthesized input
giclab/transfer-tuning/src/models/model_sets.py
code served (permissive licence) · get_code("3e3421add486fc6a")
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 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 a minimal model_set dict with a few representative models
# The function only iterates over the dict and looks up names in the global dicts
class MinimalModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 2)
    
    def forward(self, x):
        return self.linear(x)

model_set = {
    'mobilenetv2': MinimalModel(),
    'resnet50': MinimalModel(),
    'bert-base-uncased-seq_class-128': MinimalModel(),
}

ARGS = {
    'model_set': model_set,
}
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 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
get_untuned_times Ran
ran on a synthesized input
giclab/transfer-tuning/src/visualization/plot_fig5.py
code served (permissive licence) · get_code("4e12644423bc0ca1")
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 with a tuning_info.json file
temp_dir = tempfile.mkdtemp()

# Create sample tuning_info.json data
tuning_data = {
    "kernel_1": {
        "untuned_time": 0.125,
        "other_field": "ignored"
    },
    "kernel_2": {
        "untuned_time": 0.234,
        "other_field": "ignored"
    },
    "kernel_3": {
        "untuned_time": 0.456,
        "other_field": "ignored"
    }
}

# Write the JSON file
json_path = os.path.join(temp_dir, 'tuning_info.json')
with open(json_path, 'w') as f:
    json.dump(tuning_data, f)

ARGS = {
    'path': temp_dir
}
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
get_hf_tf_model Not yet run
unverified (harvested, no recorded run)
giclab/transfer-tuning/src/models/models.py
code served (permissive licence) · get_code("efe6d2a9c6142882")
How it ran

no invocation recorded.

load_keras_transformer_model Not yet run
unverified (harvested, no recorded run)
giclab/transfer-tuning/src/models/models.py
code served (permissive licence) · get_code("f2c49fb545926598")
How it ran

no invocation recorded.

pytorch_onnx_exporter Not yet run
unverified (harvested, no recorded run)
giclab/transfer-tuning/src/models/export_model.py
code served (permissive licence) · get_code("1b2a9e6ff029028e")
How it ran

no invocation recorded.

Also present in other papers' code

Of 12 samples lifted from this paper's linked repositories, 12 carry a function identity and 0 do not. Of the 12 distinct functions with an identity, 4 are also present in other papers whose code we hold (4 with at least one paper dated earlier, 4 the same month, 4 later) and 8 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
G_arch 18 other papers
9 dated earlier, 1 the same month, 7 later, 1 same year, order not determinable · 3 only as an assembly brick
1809.11096 (2018-09)
Large Scale GAN Training for High Fidelity Natural Image Synthesis
1811.10597 (2018-11)
GAN Dissection: Visualizing and Understanding Generative Adversarial Networks
1912.05270 (2019-12)
MineGAN: effective knowledge transfer from GANs to target domains with few images
and 15 more not named here
D_arch 15 other papers
7 dated earlier, 1 the same month, 6 later, 1 same year, order not determinable
1809.11096 (2018-09)
Large Scale GAN Training for High Fidelity Natural Image Synthesis
1811.10597 (2018-11)
GAN Dissection: Visualizing and Understanding Generative Adversarial Networks
1912.05270 (2019-12)
MineGAN: effective knowledge transfer from GANs to target domains with few images
and 12 more not named here
G_arch 9 other papers
3 dated earlier, 1 the same month, 4 later, 1 same year, order not determinable
1912.05270 (2019-12)
MineGAN: effective knowledge transfer from GANs to target domains with few images
2008.10599 (2020-08)
The Hessian Penalty: A Weak Prior for Unsupervised Disentanglement
2108.07668 (2021-08)
Orthogonal Jacobian Regularization for Unsupervised Disentanglement in Image Generation
and 6 more not named here
D_arch 8 other papers
3 dated earlier, 1 the same month, 3 later, 1 same year, order not determinable
1912.05270 (2019-12)
MineGAN: effective knowledge transfer from GANs to target domains with few images
2008.10599 (2020-08)
The Hessian Penalty: A Weak Prior for Unsupervised Disentanglement
2108.07668 (2021-08)
Orthogonal Jacobian Regularization for Unsupervised Disentanglement in Image Generation
and 5 more not named here

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

Auto-scheduling for tensor programs is a process where a search algorithm automatically explores candidate schedules (program transformations) for a given program on a target hardware platform to improve its performance. However this can be a very time consuming process depending on the complexity of the tensor program and the capacity of the target device, with often many thousands of program variants being explored. To address this, in this paper we introduce the idea of transfer-tuning, a novel approach to identify and reuse auto-schedules between tensor programs. We demonstrate this concept using Deep Neural Networks (DNNs), taking sets of auto-schedules from pre-tuned DNNs and using them to reduce the inference time of a new DNN. We compare transfer-tuning against the state-of-the-art Ansor auto-scheduler, defining the maximum possible speedup for a given DNN model as what Ansor achieves using its recommended full tuning time. On a server-class CPU and across 11 widely used DNN models, we observe that transfer-tuning achieves up to $88.41\%$ ($49.13\%$ on average) of this maximum speedup, while Ansor requires $6.5\times$ more search time on average to match it. We also evaluate transfer-tuning on a constrained edge CPU and observe that the differences in search time are exacerbated, with Ansor requiring $10.8\times$ more time on average to match transfer-tuning's speedup, which further demonstrates its value. Our code is available at https://www.github.com/gicLAB/transfer-tuning

For agents

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

get_harvested_code_for_paper("2201.05587")
get_code_for_paper("2201.05587")
have("2201.05587")

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

curl https://syntology.ai/api/ran/2201.05587.json

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

[![Syntology run record](https://syntology.ai/api/ran/2201.05587.svg)](https://syntology.ai/paper/2201.05587)

Connect an agent — have() is free.