SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2002.07136 · ICLR · 2020

Precision Gating: Improving Neural Network Efficiency with Dynamic Dual-Precision Activations

Yichi Zhang, Ritchie Zhao, Weizhe Hua, Nayun Xu, G Suh, Zhiru Zhang

arXiv · PDF · Open in the Atlas

Code that ran

9 of 11 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (8 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 9 have no contract verdict; 0 failures of our instrument; 2 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/2002.07136.json · badge: /api/ran/2002.07136.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
cornell-zhang/dnn-gating canonical (classifier label)
classifier basis: classifier_2025
9 ran · 2 no recorded run
FunctionStatusWhere it lives
Clamp Ran
ran on a synthesized input · constructed an object; no result was computed
cornell-zhang/dnn-gating/utils/pg_utils.py
code served (permissive licence) · get_code("3a4dabc50bed5ec4")
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, Clamp: 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

torch.manual_seed(0)

ARGS = {
    "input": torch.randn(2, 4),
    "min": -1.0,
    "max": 1.0
}
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 Clamp — 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 Clamp 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
Floor Ran
ran on a synthesized input · constructed an object; no result was computed
cornell-zhang/dnn-gating/utils/pg_utils.py
code served (permissive licence) · get_code("79d905dab5127a06")
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, Floor: 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

torch.manual_seed(0)

# The Floor.forward method takes an input tensor and applies torch.floor to it
# The Floor.backward method receives grad_output and returns a clone of it
# So we need to construct a small input tensor for testing

input_tensor = torch.randn(2, 4, requires_grad=True)

ARGS = {
    "input": input_tensor
}
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 Floor — 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 Floor 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
GreaterThan Ran
ran on a synthesized input · constructed an object; no result was computed
cornell-zhang/dnn-gating/utils/pg_utils.py
code served (permissive licence) · get_code("c3da4fa92190d610")
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, GreaterThan: 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

torch.manual_seed(0)

# The GreaterThan.forward method takes:
# - ctx: context object (provided by autograd)
# - input: a Tensor to compare
# - threshold: a scalar value to compare against

# Create a small input tensor
input_tensor = torch.randn(2, 4)

# Create a threshold scalar
threshold_value = 0.5

ARGS = {
    "input": input_tensor,
    "threshold": threshold_value,
}
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 GreaterThan — 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 GreaterThan 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
PGConv2d Ran
ran on a synthesized input · constructed an object; no result was computed
cornell-zhang/dnn-gating/utils/pg_utils.py
code served (permissive licence) · get_code("485e5f334e040599")
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, PGConv2d: 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

torch.manual_seed(0)

# Since the function definitions are incomplete, I'll create a minimal PGConv2d
# class based on typical conv2d patterns and custom autograd functions

class PGConv2d(nn.Module):
    def __init__(self, in_channels, out_channels, kernel_size, stride=1, padding=0):
        super(PGConv2d, self).__init__()
        self.in_channels = in_channels
        self.out_channels = out_channels
        self.kernel_size = kernel_size
        self.stride = stride
        self.padding = padding
        self.weight = nn.Parameter(torch.randn(out_channels, in_channels, kernel_size, kernel_size))
        self.bias = nn.Parameter(torch.randn(out_channels))
    
    def forward(self, x):
        return nn.functional.conv2d(x, self.weight, self.bias, stride=self.stride, padding=self.padding)

# Create minimal valid arguments
ARGS = {
    "in_channels": 3,
    "out_channels": 4,
    "kernel_size": 3,
    "stride": 1,
    "padding": 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 PGConv2d — 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 PGConv2d 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
Round Ran
ran on a synthesized input · constructed an object; no result was computed
cornell-zhang/dnn-gating/utils/pg_utils.py
code served (permissive licence) · get_code("6d47a5d78601c488")
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, Round: 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

torch.manual_seed(0)

# The Round class is a custom autograd Function
# The forward method takes ctx and input (a Tensor)
# The backward method takes ctx and grad_output (a Tensor)
# We need to construct arguments to instantiate and use this function

# Create a small input tensor with requires_grad=True so we can compute gradients
input_tensor = torch.randn(2, 4, requires_grad=True)

ARGS = {
    "input": input_tensor
}
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 Round — 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 Round 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
SparseGreaterThan Ran
ran on a synthesized input · constructed an object; no result was computed
cornell-zhang/dnn-gating/utils/pg_utils.py
code served (permissive licence) · get_code("e31dcb7b107eb117")
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, SparseGreaterThan: 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

torch.manual_seed(0)

# Create a small input tensor (batch_size=2, features=4)
input_tensor = torch.randn(2, 4, requires_grad=True)

# Define a threshold value
threshold_value = 0.5

ARGS = {
    'input': input_tensor,
    'threshold': threshold_value
}
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 SparseGreaterThan — 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 SparseGreaterThan 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
TorchBinarize Ran
ran on a synthesized input
cornell-zhang/dnn-gating/utils/pg_utils.py
code served (permissive licence) · get_code("3aaece55f3d40d52")
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 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
import torch.nn as nn

# INIT_ARGS for TorchBinarize.__init__
# The __init__ method takes no parameters beyond self
INIT_ARGS = {}

# ARGS for TorchBinarize.forward
# The forward method takes 'input' as parameter
# Create a tensor with values in a range that will test the binarization
# Using values both inside and outside [-1, 1] to test clamping
ARGS = {
    'input': torch.tensor([[-1.5, -0.7, 0.3, 1.2],
                           [0.0, -0.1, 0.9, -1.0]], dtype=torch.float32)
}
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 2, 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)
TorchQuantize Ran
ran on a synthesized input · constructed an object; no result was computed
cornell-zhang/dnn-gating/utils/pg_utils.py
code served (permissive licence) · get_code("def5e6826bd88736")
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, TorchQuantize: 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

torch.manual_seed(0)

ARGS = {
    "bits": 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 TorchQuantize — 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 TorchQuantize 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
TorchRoundToBits Ran
ran on a synthesized input · constructed an object; no result was computed
cornell-zhang/dnn-gating/utils/pg_utils.py
code served (permissive licence) · get_code("a25443ef8bc781e7")
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, TorchRoundToBits: 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

torch.manual_seed(0)

# Create a small input tensor for testing
# Shape: (2, 4) - batch size 2, feature dimension 4
input_tensor = torch.randn(2, 4)

# The TorchRoundToBits module takes one parameter: bits
# We'll test with bits=2 (the default and minimum valid value > 1)
model = nn.Module()

ARGS = {
    'bits': 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 TorchRoundToBits — 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 TorchRoundToBits 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
TorchTruncate Not yet run
unverified (harvested, no recorded run)
cornell-zhang/dnn-gating/utils/pg_utils.py
code served (permissive licence) · get_code("13265ea7eba5f4a4")
How it ran

no invocation recorded.

generate_model Not yet run
unverified (harvested, no recorded run)
cornell-zhang/dnn-gating/pg-cifar10.py
code served (permissive licence) · get_code("ceb79d9355fccd93")
How it ran

no invocation recorded.

Also present in other papers' code

Of 11 samples lifted from this paper's linked repositories, 11 carry a function identity and 0 do not. Of the 11 distinct functions with an identity, 1 are also present in other papers whose code we hold (1 with at least one paper dated earlier, 0 the same month, 0 later) and 10 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
generate_model 1 other paper
1 dated earlier
1805.12549 (2018-05)
Channel Gating Neural Networks

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

We propose precision gating (PG), an end-to-end trainable dynamic dual-precision quantization technique for deep neural networks. PG computes most features in a low precision and only a small proportion of important features in a higher precision to preserve accuracy. The proposed approach is applicable to a variety of DNN architectures and significantly reduces the computational cost of DNN execution with almost no accuracy loss. Our experiments indicate that PG achieves excellent results on CNNs, including statically compressed mobile-friendly networks such as ShuffleNet. Compared to the state-of-the-art prediction-based quantization schemes, PG achieves the same or higher accuracy with 2.4× less compute on ImageNet. PG furthermore applies to RNNs. Compared to 8-bit uniform quantization, PG obtains a 1.2% improvement in perplexity per word with 2.7× computational cost reduction on LSTM on the Penn Tree Bank dataset. 1

For agents

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

get_harvested_code_for_paper("2002.07136")
get_code_for_paper("2002.07136")
have("2002.07136")

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

curl https://syntology.ai/api/ran/2002.07136.json

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

[![Syntology run record](https://syntology.ai/api/ran/2002.07136.svg)](https://syntology.ai/paper/2002.07136)

Connect an agent — have() is free.