# Qwen3-Embedding-0.6B.axera

**Repository Path**: axera-opensource/Qwen3-Embedding-0.6B.axera

## Basic Information

- **Project Name**: Qwen3-Embedding-0.6B.axera
- **Description**: No description available
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: main
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2025-09-14
- **Last Updated**: 2025-09-14

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# Qwen3-Embedding-0.6B.axera

> HuggingFaceTB Qwen3-Embedding-0.6B DEMO on Axera.

- 目前支持 `Python` 语言, `C++` 代码在开发中.
- 预编译模型可以从[百度网盘](https://pan.baidu.com/s/1DvhH3neOfdRLtTuVn53sxw?pwd=fd1a)下载.
- 如需自行导出编译 `axmodel` 模型请参考 [模型转换](/model_convert/README.md).

## 支持平台

- [x] AX650N
- [ ] AX630C

## Git Clone

首先使用如下命令 `clone` 本项目, 然后进入 `python` 文件夹:

```bash
$ git clone git@github.com:AXERA-TECH/Qwen3-Embedding-0.6B.axera.git
$ cd Qwen3-Embedding-0.6B.axera
```

之后在开发板上下载或安装以下支持库:

- 从 `huggingface` 下载 `Qwen3-Embedding-0.6B` 模型.

    ```bash
    $ git clone https://huggingface.co/Qwen/Qwen3-Embedding-0.6B
    ```

- 在开发板上安装配置 `pyaxengine`, [点击跳转下载链接](https://github.com/AXERA-TECH/pyaxengine/releases). 注意板端 `SDK` 最低版本要求:

    - AX650 SDK >= 2.18
    - AX620E SDK >= 3.12
    - 执行 `pip3 install axengine-x.x.x-py3-none-any.whl` 安装

将下载后的预编译模型解压到当前文件夹[🔔可选], 默认文件夹排布如下:

```bash
(lerobot) ➜  python git:(master) ✗ tree -L 2 .
.
├── infer_axmodel.py
├── infer_torch.py
├── Qwen3-Embedding-0.6B
│   ├── 1_Pooling
│   ├── config.json
│   ├── config_sentence_transformers.json
│   ├── generation_config.json
│   ├── merges.txt
│   ├── model.safetensors
│   ├── modules.json
│   ├── README.md
│   ├── tokenizer_config.json
│   ├── tokenizer.json
│   └── vocab.json
├── Qwen3-Embedding-0.6B_axmodel
│   ├── model.embed_tokens.weight.bfloat16.bin
│   ├── model.embed_tokens.weight.float32.bin
│   ├── model.embed_tokens.weight.npy
│   ├── qwen3_p128_l0_together.axmodel
│   ├── qwen3_p128_l10_together.axmodel
│   ├── qwen3_p128_l11_together.axmodel
│   ├── qwen3_p128_l12_together.axmodel
│   ├── qwen3_p128_l13_together.axmodel
│   ├── qwen3_p128_l14_together.axmodel
│   ├── qwen3_p128_l15_together.axmodel
│   ├── qwen3_p128_l16_together.axmodel
│   ├── qwen3_p128_l17_together.axmodel
│   ├── qwen3_p128_l18_together.axmodel
│   ├── qwen3_p128_l19_together.axmodel
│   ├── qwen3_p128_l1_together.axmodel
│   ├── qwen3_p128_l20_together.axmodel
│   ├── qwen3_p128_l21_together.axmodel
│   ├── qwen3_p128_l22_together.axmodel
│   ├── qwen3_p128_l23_together.axmodel
│   ├── qwen3_p128_l24_together.axmodel
│   ├── qwen3_p128_l25_together.axmodel
│   ├── qwen3_p128_l26_together.axmodel
│   ├── qwen3_p128_l27_together.axmodel
│   ├── qwen3_p128_l2_together.axmodel
│   ├── qwen3_p128_l3_together.axmodel
│   ├── qwen3_p128_l4_together.axmodel
│   ├── qwen3_p128_l5_together.axmodel
│   ├── qwen3_p128_l6_together.axmodel
│   ├── qwen3_p128_l7_together.axmodel
│   ├── qwen3_p128_l8_together.axmodel
│   ├── qwen3_p128_l9_together.axmodel
│   └── qwen3_post.axmodel
├── README.md
└── utils
    ├── infer_func.py
```

## 上板部署

- `AX650N` 的设备已预装 `Ubuntu 22.04`
- 以 `root` 权限登陆 `AX650N` 的板卡设备
- 接入互联网, 确保 `AX650N` 的设备能正常执行 `apt install`, `pip install` 等指令
- 已验证设备: `AX650N DEMO Board`、`爱芯派Pro(AX650N)`

### Python API 运行

#### Requirements

```bash
$ mkdir /opt/site-packages
$ cd python
$ pip3 install -r requirements.txt --prefix=/opt/site-packages
``` 

#### 添加环境变量

将以下两行添加到 `/root/.bashrc`(实际添加的路径需要自行检查)后, 重新连接终端或者执行 `source ~/.bashrc`

```bash
$ export PYTHONPATH=$PYTHONPATH:/opt/site-packages/local/lib/python3.10/dist-packages  
$ export PATH=$PATH:/opt/site-packages/local/bin
``` 

#### 运行

在 `Axera 开发板` 上运行以下命令开启图像理解功能:

```sh
$ cd Qwen3-Embedding-0.6B.axera
$ python3 infer_axmodel.py
```

输出日志如下:

```sh
Model loaded successfully!
slice_indices: [0]
Slice prefill done: 0
slice_indices: [0]
Slice prefill done: 0
slice_indices: [0]
Slice prefill done: 0
slice_indices: [0]
Slice prefill done: 0
[[0.7555467486381531, 0.1756950318813324], [0.4137178063392639, 0.4459586441516876]]
```

## 技术讨论

- Github issues
- QQ 群: 139953715