Skip to main content

A tool to count the FLOPs of PyTorch model.

Project description


🚀 THOP: PyTorch-OpCounter

Welcome to the THOP repository, your comprehensive solution for profiling PyTorch models by computing the number of Multiply-Accumulate Operations (MACs) and parameters. This tool is essential for deep learning practitioners to evaluate model efficiency and performance.

GitHub Actions PyPI version Discord

📄 Description

THOP offers an intuitive API to profile PyTorch models by calculating the number of MACs and parameters. This functionality is crucial for assessing the computational efficiency and memory footprint of deep learning models.

📦 Installation

You can install THOP via pip:

pip install ultralytics-thop

Alternatively, install the latest version directly from GitHub:

pip install --upgrade git+https://github.com/ultralytics/thop.git

🛠 How to Use

Basic Usage

To profile a model, you can use the following example:

from torchvision.models import resnet50
from thop import profile
import torch

model = resnet50()
input = torch.randn(1, 3, 224, 224)
macs, params = profile(model, inputs=(input, ))

Define Custom Rules for Third-Party Modules

You can define custom rules for unsupported modules:

import torch.nn as nn

class YourModule(nn.Module):
    # your definition
    pass

def count_your_model(model, x, y):
    # your rule here
    pass

input = torch.randn(1, 3, 224, 224)
macs, params = profile(model, inputs=(input, ), custom_ops={YourModule: count_your_model})

Improve Output Readability

Use thop.clever_format for a more readable output:

from thop import clever_format
macs, params = clever_format([macs, params], "%.3f")

📊 Results of Recent Models

The following table presents the parameters and MACs for popular models. These results can be reproduced using the script benchmark/evaluate_famous_models.py.

Model Params(M) MACs(G)
alexnet 61.10 0.77
vgg11 132.86 7.74
vgg11_bn 132.87 7.77
vgg13 133.05 11.44
vgg13_bn 133.05 11.49
vgg16 138.36 15.61
vgg16_bn 138.37 15.66
vgg19 143.67 19.77
vgg19_bn 143.68 19.83
resnet18 11.69 1.82
resnet34 21.80 3.68
resnet50 25.56 4.14
resnet101 44.55 7.87
resnet152 60.19 11.61
wide_resnet101_2 126.89 22.84
wide_resnet50_2 68.88 11.46
Model Params(M) MACs(G)
resnext50_32x4d 25.03 4.29
resnext101_32x8d 88.79 16.54
densenet121 7.98 2.90
densenet161 28.68 7.85
densenet169 14.15 3.44
densenet201 20.01 4.39
squeezenet1_0 1.25 0.82
squeezenet1_1 1.24 0.35
mnasnet0_5 2.22 0.14
mnasnet0_75 3.17 0.24
mnasnet1_0 4.38 0.34
mnasnet1_3 6.28 0.53
mobilenet_v2 3.50 0.33
shufflenet_v2_x0_5 1.37 0.05
shufflenet_v2_x1_0 2.28 0.15
shufflenet_v2_x1_5 3.50 0.31
shufflenet_v2_x2_0 7.39 0.60
inception_v3 27.16 5.75

💡 Contribute

We welcome community contributions to enhance THOP. Please check our Contributing Guide for more details. Your feedback and suggestions are highly appreciated!

📄 License

THOP is licensed under the AGPL-3.0 License. For more information, see the LICENSE file.

📮 Contact

For bugs or feature requests, please open an issue on GitHub Issues. Join our community on Discord for discussions and support.


Ultralytics GitHub space Ultralytics LinkedIn space Ultralytics Twitter space Ultralytics YouTube space Ultralytics TikTok space Ultralytics Instagram space Ultralytics Discord

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

ultralytics_thop-0.0.3.tar.gz (54.8 kB view hashes)

Uploaded Source

Built Distribution

ultralytics_thop-0.0.3-py3-none-any.whl (42.0 kB view hashes)

Uploaded Python 3

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page