# THOP: PyTorch-OpCounter
## How to install
`pip install thop` (now continously intergrated on [Github actions](https://github.com/features/actions))
OR
`pip install --upgrade git+https://github.com/Lyken17/pytorch-OpCounter.git`
## How to use
* Basic usage
```python
from torchvision.models import resnet50
from thop import profile
model = resnet50()
input = torch.randn(1, 3, 224, 224)
macs, params = profile(model, inputs=(input, ))
```
* Define the rule for 3rd party module.
```python
class YourModule(nn.Module):
# your definition
def count_your_model(model, x, y):
# your rule here
input = torch.randn(1, 3, 224, 224)
macs, params = profile(model, inputs=(input, ),
custom_ops={YourModule: count_your_model})
```
* Improve the output readability
Call `thop.clever_format` to give a better format of the output.
```python
from thop import clever_format
macs, params = clever_format([macs, params], "%.3f")
```
## Results of Recent Models
The implementation are adapted from `torchvision`. Following results can be obtained using [benchmark/evaluate_famous_models.py](benchmark/evaluate_famous_models.py).
<p align="center">
<table>
<tr>
<td>
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
</td>
<td>
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
</td>
</tr>
</p>
Raw data
{
"_id": null,
"home_page": "https://github.com/Lyken17/pytorch-OpCounter/",
"name": "thop",
"maintainer": "",
"docs_url": null,
"requires_python": "",
"maintainer_email": "",
"keywords": "",
"author": "Ligeng Zhu",
"author_email": "ligeng.zhu+github@gmail.com",
"download_url": "",
"platform": null,
"description": "# THOP: PyTorch-OpCounter\n\n## How to install \n\n`pip install thop` (now continously intergrated on [Github actions](https://github.com/features/actions))\n\nOR\n\n`pip install --upgrade git+https://github.com/Lyken17/pytorch-OpCounter.git`\n\n## How to use \n* Basic usage \n ```python\n from torchvision.models import resnet50\n from thop import profile\n model = resnet50()\n input = torch.randn(1, 3, 224, 224)\n macs, params = profile(model, inputs=(input, ))\n ``` \n\n* Define the rule for 3rd party module.\n ```python\n class YourModule(nn.Module):\n # your definition\n def count_your_model(model, x, y):\n # your rule here\n\n input = torch.randn(1, 3, 224, 224)\n macs, params = profile(model, inputs=(input, ), \n custom_ops={YourModule: count_your_model})\n ```\n\n* Improve the output readability\n\n Call `thop.clever_format` to give a better format of the output.\n ```python\n from thop import clever_format\n macs, params = clever_format([macs, params], \"%.3f\")\n ``` \n\n## Results of Recent Models\n\nThe implementation are adapted from `torchvision`. Following results can be obtained using [benchmark/evaluate_famous_models.py](benchmark/evaluate_famous_models.py).\n\n<p align=\"center\">\n<table>\n<tr>\n<td>\n\nModel | Params(M) | MACs(G)\n---|---|---\nalexnet | 61.10 | 0.77\nvgg11 | 132.86 | 7.74\nvgg11_bn | 132.87 | 7.77\nvgg13 | 133.05 | 11.44\nvgg13_bn | 133.05 | 11.49\nvgg16 | 138.36 | 15.61\nvgg16_bn | 138.37 | 15.66\nvgg19 | 143.67 | 19.77\nvgg19_bn | 143.68 | 19.83\nresnet18 | 11.69 | 1.82\nresnet34 | 21.80 | 3.68\nresnet50 | 25.56 | 4.14\nresnet101 | 44.55 | 7.87\nresnet152 | 60.19 | 11.61\nwide_resnet101_2 | 126.89 | 22.84\nwide_resnet50_2 | 68.88 | 11.46\n\n</td>\n<td>\n\nModel | Params(M) | MACs(G)\n---|---|---\nresnext50_32x4d | 25.03 | 4.29\nresnext101_32x8d | 88.79 | 16.54\ndensenet121 | 7.98 | 2.90\ndensenet161 | 28.68 | 7.85\ndensenet169 | 14.15 | 3.44\ndensenet201 | 20.01 | 4.39\nsqueezenet1_0 | 1.25 | 0.82\nsqueezenet1_1 | 1.24 | 0.35\nmnasnet0_5 | 2.22 | 0.14\nmnasnet0_75 | 3.17 | 0.24\nmnasnet1_0 | 4.38 | 0.34\nmnasnet1_3 | 6.28 | 0.53\nmobilenet_v2 | 3.50 | 0.33\nshufflenet_v2_x0_5 | 1.37 | 0.05\nshufflenet_v2_x1_0 | 2.28 | 0.15\nshufflenet_v2_x1_5 | 3.50 | 0.31\nshufflenet_v2_x2_0 | 7.39 | 0.60\ninception_v3 | 27.16 | 5.75\n\n</td>\n</tr>\n</p>\n\n\n",
"bugtrack_url": null,
"license": "MIT",
"summary": "A tool to count the FLOPs of PyTorch model.",
"version": "0.1.1.post2209072238",
"split_keywords": [],
"urls": [
{
"comment_text": "",
"digests": {
"md5": "269834966ac96990646cb15dac973af7",
"sha256": "01473c225231927d2ad718351f78ebf7cffe6af3bed464c4f1ba1ef0f7cdda27"
},
"downloads": -1,
"filename": "thop-0.1.1.post2209072238-py3-none-any.whl",
"has_sig": false,
"md5_digest": "269834966ac96990646cb15dac973af7",
"packagetype": "bdist_wheel",
"python_version": "py3",
"requires_python": null,
"size": 15443,
"upload_time": "2022-09-07T14:38:37",
"upload_time_iso_8601": "2022-09-07T14:38:37.211729Z",
"url": "https://files.pythonhosted.org/packages/bb/0f/72beeab4ff5221dc47127c80f8834b4bcd0cb36f6ba91c0b1d04a1233403/thop-0.1.1.post2209072238-py3-none-any.whl",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2022-09-07 14:38:37",
"github": true,
"gitlab": false,
"bitbucket": false,
"github_user": "Lyken17",
"github_project": "pytorch-OpCounter",
"travis_ci": false,
"coveralls": false,
"github_actions": true,
"requirements": [],
"lcname": "thop"
}
JSON
