{"author":null,"author_email":"Ligeng Zhu ","bugtrack_url":null,"classifiers":["Development Status :: 4 - Beta","Intended Audience :: Developers","Intended Audience :: Education","Intended Audience :: Science/Research","License :: OSI Approved :: GNU Affero General Public License v3 or later (AGPLv3+)","Operating System :: MacOS","Operating System :: Microsoft :: Windows","Operating System :: POSIX :: Linux","Programming Language :: Python :: 3","Programming Language :: Python :: 3.10","Programming Language :: Python :: 3.11","Programming Language :: Python :: 3.12","Programming Language :: Python :: 3.13","Programming Language :: Python :: 3.8","Programming Language :: Python :: 3.9","Topic :: Scientific/Engineering","Topic :: Scientific/Engineering :: Artificial Intelligence","Topic :: Scientific/Engineering :: Image Recognition","Topic :: Software Development"],"description":"
\n $\"Ultralytics$ \n\n# 🚀 THOP: PyTorch-OpCounter\n\nWelcome to the [THOP](https://github.com/ultralytics/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.\n\n[![GitHub Actions](https://github.com/ultralytics/thop/actions/workflows/format.yml/badge.svg)](https://github.com/ultralytics/thop/actions/workflows/main.yml) $\"Discord\"$ $\"Ultralytics$ $\"Ultralytics$ \n\n## 📄 Description\n\nTHOP 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.\n\n## 📦 Installation\n\nYou can install THOP via pip:\n\n[![PyPI - Version](https://img.shields.io/pypi/v/ultralytics-thop?logo=pypi&logoColor=white)](https://pypi.org/project/ultralytics-thop/) [![Downloads](https://static.pepy.tech/badge/ultralytics-thop)](https://www.pepy.tech/projects/ultralytics-thop) [![PyPI - Python Version](https://img.shields.io/pypi/pyversions/ultralytics-thop?logo=python&logoColor=gold)](https://pypi.org/project/ultralytics-thop/)\n\n```bash\npip install ultralytics-thop\n```\n\nAlternatively, install the latest version directly from GitHub:\n\n```bash\npip install --upgrade git+https://github.com/ultralytics/thop.git\n```\n\n## 🛠 How to Use\n\n### Basic Usage\n\nTo profile a model, you can use the following example:\n\n```python\nimport torch\nfrom torchvision.models import resnet50\n\nfrom thop import profile\n\nmodel = resnet50()\ninput = torch.randn(1, 3, 224, 224)\nmacs, params = profile(model, inputs=(input,))\n```\n\n### Define Custom Rules for Third-Party Modules\n\nYou can define custom rules for unsupported modules:\n\n```python\nimport torch.nn as nn\n\n\nclass YourModule(nn.Module):\n # your definition\n pass\n\n\ndef count_your_model(model, x, y):\n # your rule here\n pass\n\n\ninput = torch.randn(1, 3, 224, 224)\nmacs, params = profile(model, inputs=(input,), custom_ops={YourModule: count_your_model})\n```\n\n### Improve Output Readability\n\nUse `thop.clever_format` for a more readable output:\n\n```python\nfrom thop import clever_format\n\nmacs, params = clever_format([macs, params], \"%.3f\")\n```\n\n## 📊 Results of Recent Models\n\nThe following table presents the parameters and MACs for popular models. These results can be reproduced using the script `benchmark/evaluate_famous_models.py`.\n\n\n\n\n\n\n

\n\n| Model | Params(M) | MACs(G) |\n| ---------------- | --------- | ------- |\n| alexnet | 61.10 | 0.77 |\n| vgg11 | 132.86 | 7.74 |\n| vgg11_bn | 132.87 | 7.77 |\n| vgg13 | 133.05 | 11.44 |\n| vgg13_bn | 133.05 | 11.49 |\n| vgg16 | 138.36 | 15.61 |\n| vgg16_bn | 138.37 | 15.66 |\n| vgg19 | 143.67 | 19.77 |\n| vgg19_bn | 143.68 | 19.83 |\n| resnet18 | 11.69 | 1.82 |\n| resnet34 | 21.80 | 3.68 |\n| resnet50 | 25.56 | 4.14 |\n| resnet101 | 44.55 | 7.87 |\n| resnet152 | 60.19 | 11.61 |\n| wide_resnet101_2 | 126.89 | 22.84 |\n| wide_resnet50_2 | 68.88 | 11.46 |\n\n

\n\n| Model | Params(M) | MACs(G) |\n| ------------------ | --------- | ------- |\n| resnext50_32x4d | 25.03 | 4.29 |\n| resnext101_32x8d | 88.79 | 16.54 |\n| densenet121 | 7.98 | 2.90 |\n| densenet161 | 28.68 | 7.85 |\n| densenet169 | 14.15 | 3.44 |\n| densenet201 | 20.01 | 4.39 |\n| squeezenet1_0 | 1.25 | 0.82 |\n| squeezenet1_1 | 1.24 | 0.35 |\n| mnasnet0_5 | 2.22 | 0.14 |\n| mnasnet0_75 | 3.17 | 0.24 |\n| mnasnet1_0 | 4.38 | 0.34 |\n| mnasnet1_3 | 6.28 | 0.53 |\n| mobilenet_v2 | 3.50 | 0.33 |\n| shufflenet_v2_x0_5 | 1.37 | 0.05 |\n| shufflenet_v2_x1_0 | 2.28 | 0.15 |\n| shufflenet_v2_x1_5 | 3.50 | 0.31 |\n| shufflenet_v2_x2_0 | 7.39 | 0.60 |\n| inception_v3 | 27.16 | 5.75 |\n\n

\n\n## 💡 Contribute\n\nWe welcome community contributions to enhance THOP. Please check our [Contributing Guide](https://docs.ultralytics.com/help/contributing/) for more details. Your feedback and suggestions are highly appreciated!\n\n## 📄 License\n\nTHOP is licensed under the AGPL-3.0 License. For more information, see the [LICENSE](https://github.com/ultralytics/thop/blob/main/LICENSE) file.\n\n## 📮 Contact\n\nFor bugs or feature requests, please open an issue on [GitHub Issues](https://github.com/ultralytics/thop/pulls). Join our community on [Discord](https://discord.com/invite/ultralytics) for discussions and support.\n\n
\n

\n $\"Ultralytics$ \n $\"space\"$ \n $\"Ultralytics$ \n $\"space\"$ \n $\"Ultralytics$ \n $\"space\"$ \n $\"Ultralytics$ \n $\"space\"$ \n $\"Ultralytics$ \n $\"space\"$ \n $\"Ultralytics$ \n $\"space\"$ \n $\"Ultralytics$ \n

\n","description_content_type":"text/markdown","docs_url":null,"download_url":null,"downloads":{"last_day":-1,"last_month":-1,"last_week":-1},"dynamic":null,"home_page":null,"keywords":"FLOPs, PyTorch, Model Analysis","license":"AGPL-3.0","license_expression":null,"license_files":null,"maintainer":null,"maintainer_email":"Ultralytics ","name":"ultralytics-thop","package_url":"https://pypi.org/project/ultralytics-thop/","platform":null,"project_url":"https://pypi.org/project/ultralytics-thop/","project_urls":{"Bug Reports":"https://github.com/ultralytics/thop/issues","Changelog":"https://github.com/ultralytics/thop/releases","Documentation":"https://docs.ultralytics.com","Homepage":"https://ultralytics.com","Source":"https://github.com/ultralytics/thop"},"provides_extra":null,"release_url":"https://pypi.org/project/ultralytics-thop/2.0.14/","requires_dist":["numpy","torch"],"requires_python":">=3.8","summary":"Ultralytics THOP package for fast computation of PyTorch model FLOPs and parameters.","version":"2.0.14","yanked":false,"yanked_reason":null}

Dependencies of ultralytics-thop (2).
Dependency	Optional
numpy	false
torch	false

Package search

ultralytics-thop

project_urls

Dependencies

All package metadata

Releases