# MagicONNX

**Repository Path**: echo_lin/MagicONNX

## Basic Information

- **Project Name**: MagicONNX
- **Description**: No description available
- **Primary Language**: Python
- **License**: Apache-2.0
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 20
- **Created**: 2022-10-27
- **Last Updated**: 2022-10-27

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

![logo](./image/logo.png)

- [1. MagicONNX简介](#1-magiconnx简介)
- [2. 安装](#2-安装)
- [3. 使用方法](#3-使用方法)

# 1. MagicONNX简介

可以借鉴的开源项目
onnxconverter-common
onnx_transformers
<https://github.com/BBuf/onnx_learn>

# 2. 安装

```shell
git clone https://gitee.com/Ronnie_zheng/MagicONNX.git
cd MagicONNX
pip install .
```

# 3. 使用方法

[学习教程](./docs/tutorials.md)

[API说明](./docs/operations.md)

![动画演示](./image/create.gif)

```python
from magiconnx import OnnxGraph
graph = OnnxGraph('layernorm.onnx')

# 原onnx中只有一个input输入，现在增加一个 dummy_input 输入结点
ph = graph.add_placeholder('dummy_input', 'int32', [2, 3, 4])

# 增加一个add节点
add = graph.add_node('dummy_add', 'Add')      # add_node默认单输入单输出，需要手动修改节点输入输出信息

# 设置add节点的输入输出【连边】
init = graph.add_initializer('dummy_init', np.array([[2, 3, 4]]))
add.inputs = ['dummy_input', 'dummy_init']
add.outputs = ['add_out']

# 在add前面插入argmax节点，属性用dict传参
argmax = graph.add_node('dummy_ArgMax',
                      'ArgMax',
                      {'axis': 0, 'keepdims': 1, 'select_last_index': 0})
graph.insert_node('dummy_add', argmax, mode='before')

# 保存修改好的onnx图
graph.save('case2.onnx')
```