https://github.com/daquexian/onnx-simplifier
Simplify your onnx model
https://github.com/daquexian/onnx-simplifier
deep-learning onnx pytorch
Last synced: about 1 year ago
JSON representation
Simplify your onnx model
- Host: GitHub
- URL: https://github.com/daquexian/onnx-simplifier
- Owner: daquexian
- License: apache-2.0
- Created: 2019-04-01T10:12:34.000Z (about 7 years ago)
- Default Branch: master
- Last Pushed: 2024-09-03T22:43:22.000Z (almost 2 years ago)
- Last Synced: 2025-05-06T21:11:58.264Z (about 1 year ago)
- Topics: deep-learning, onnx, pytorch
- Language: C++
- Homepage:
- Size: 420 KB
- Stars: 4,067
- Watchers: 48
- Forks: 396
- Open Issues: 169
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
- awesome-ncnn - **ONNX Simplifier**
README
# ONNX Simplifier
[](https://pypi.python.org/pypi/onnx-simplifier/)
[](https://pypi.python.org/pypi/onnx-simplifier/)
[](https://pypi.python.org/pypi/onnx-simplifier/)
[](https://github.com/daquexian/onnx-simplifier/pulls)
_ONNX is great, but sometimes too complicated._
## Background
One day I wanted to export the following simple reshape operation to ONNX:
```python
import torch
class JustReshape(torch.nn.Module):
def __init__(self):
super(JustReshape, self).__init__()
def forward(self, x):
return x.view((x.shape[0], x.shape[1], x.shape[3], x.shape[2]))
net = JustReshape()
model_name = 'just_reshape.onnx'
dummy_input = torch.randn(2, 3, 4, 5)
torch.onnx.export(net, dummy_input, model_name, input_names=['input'], output_names=['output'])
```
The input shape in this model is static, so what I expected is
However, I got the following complicated model instead:
## Our solution
ONNX Simplifier is presented to simplify the ONNX model. It infers the whole computation graph
and then replaces the redundant operators with their constant outputs (a.k.a. constant folding).
### Web version
We have published ONNX Simplifier on [convertmodel.com](https://www.convertmodel.com/#input=onnx&output=onnx). It works out of the box and **doesn't need any installation**. Note that it runs in the browser locally and your model is completely safe.
### Python version
```
pip3 install -U pip && pip3 install onnxsim
```
Then
```
onnxsim input_onnx_model output_onnx_model
```
For more advanced features, try the following command for help message
```
onnxsim -h
```
## Demonstration
An overall comparison between
[a complicated model](https://github.com/JDAI-CV/DNNLibrary/issues/17#issuecomment-455934190)
and its simplified version:
## In-script workflow
If you would like to embed ONNX simplifier python package in another script, it is just that simple.
```python
import onnx
from onnxsim import simplify
# load your predefined ONNX model
model = onnx.load(filename)
# convert model
model_simp, check = simplify(model)
assert check, "Simplified ONNX model could not be validated"
# use model_simp as a standard ONNX model object
```
You can see more details of the API in [onnxsim/onnx_simplifier.py](onnxsim/onnx_simplifier.py)
## Projects Using ONNX Simplifier
* [MXNet](https://mxnet.apache.org/versions/1.9.1/api/python/docs/tutorials/deploy/export/onnx.html#Simplify-the-exported-ONNX-model)
* [MMDetection](https://github.com/open-mmlab/mmdetection)
* [YOLOv5](https://github.com/ultralytics/yolov5)
* [ncnn](https://github.com/Tencent/ncnn)
* ...
## Chat
We created a Chinese QQ group for ONNX!
ONNX QQ Group (Chinese): 1021964010, verification code: nndab. Welcome to join!
For English users, I'm active on the [ONNX Slack](https://github.com/onnx/onnx#discuss). You can find and chat with me (daquexian) there.