Khadas Docs

Amazing Khadas, always amazes you!

User Tools

Site Tools

Home / products / sbc / edge2 / npu / npu-convert
products:sbc:edge2:npu:npu-convert

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revision Previous revision
Next revision 		Previous revision
products:sbc:edge2:npu:npu-convert [2023/07/04 05:18]
louis 		products:sbc:edge2:npu:npu-convert [2024/06/06 21:57] (current)
nick
Line 1: Line 1:
-====== NPU Model Convert ======+{{indexmenu_n>3}}
  
-===== Introduction =====+====== Edge2 NPU Model Convert ======
  
-===== Virtual Environment =====+===== Build Virtual Environment =====
  
 The SDK only supports **python3.6** or **python3.8**, here is an example of creating a virtual environment for **python3.8**. The SDK only supports **python3.6** or **python3.8**, here is an example of creating a virtual environment for **python3.8**.
Line 10: Line 10:
  
 ```shell ```shell
-sudo apt update +$ sudo apt update 
-sudo apt install python3-dev python3-numpy+$ sudo apt install python3-dev python3-numpy
 ``` ```
  
Line 19: Line 19:
  
 ```shell ```shell
-conda create -n npu-env python=3.8 +$ conda create -n npu-env python=3.8 
-conda activate npu-env     #activate +$ conda activate npu-env     #activate 
-conda deactivate           #deactivate+$ conda deactivate           #deactivate
 ``` ```
  
-After creating the virtual environment, start the virtual environment and start the next step.+===== Get Convert Tool =====
  
-===== Get SDK ===== +Download Tool from [[gh>rockchip-linux/rknn-toolkit2]].
- +
-Clone SDK from [[https://github.com/rockchip-linux/rknn-toolkit2.git|Rockchip Github]].+
  
 ```shell ```shell
-git clone https://github.com/rockchip-linux/rknn-toolkit2.git +$ git clone https://github.com/rockchip-linux/rknn-toolkit2.git 
-git checkout 9ad79343fae625f4910242e370035fcbc40cc31a+$ git checkout 9ad79343fae625f4910242e370035fcbc40cc31a
 ``` ```
- 
-===== Convert ===== 
  
 Install dependences and RKNN toolkit2 packages, Install dependences and RKNN toolkit2 packages,
  
 ```shell ```shell
-cd rknn-toolkit2 +$ cd rknn-toolkit2 
-sudo apt-get install python3 python3-dev python3-pip +$ sudo apt-get install python3 python3-dev python3-pip 
-sudo apt-get install libxslt1-dev zlib1g-dev libglib2.0 libsm6 libgl1-mesa-glx libprotobuf-dev gcc cmake +$ sudo apt-get install libxslt1-dev zlib1g-dev libglib2.0 libsm6 libgl1-mesa-glx libprotobuf-dev gcc cmake 
-pip3 install -r doc/requirements_cp38-*.txt +$ pip3 install -r doc/requirements_cp38-*.txt 
-pip3 install packages/rknn_toolkit2-*-cp38-cp38-linux_x86_64.whl+$ pip3 install packages/rknn_toolkit2-*-cp38-cp38-linux_x86_64.whl
 ``` ```
  
-Choose ''RK3588S'',+===== Convert Model =====
  
-```diff patch +Converting model has five main steps. Create RKNN object, pre-process config, load model, build model and export RKNN model. Here, take ''yolov5'' ''onnx'' as an example.
-diff --git a/examples/onnx/yolov5/test.py b/examples/onnx/yolov5/test.py +
-index a1c9988..f7ce11e 100644 +
---- a/examples/onnx/yolov5/test.py +
-+++ b/examples/onnx/yolov5/test.py +
-@@ -240,7 +240,7 @@ if __name__ == '__main__': +
-  +
-     # pre-process config +
-     print('--> Config model') +
--    rknn.config(mean_values=[[0, 0, 0]], std_values=[[255, 255, 255]]) +
-+    rknn.config(mean_values=[[0, 0, 0]], std_values=[[255, 255, 255]], target_platform='rk3588') +
-     print('done') +
-  +
- ```+
  
-At present, we support loading and converting Caffe, TensorFlow, TensorFlow Lite, ONNX, DarkNet, PyTorch model. Each platform has its own interface.+Create RKNN object. 
  
-Caffe+```python 
 +# Create RKNN object 
 +rknn = RKNN(verbose=True)  
 +```
  
-```shell +Pre-process config. 
-ret = rknn.load_caffe(model='./mobilenet_v2.prototxt',  + 
-                      blobs='./mobilenet_v2.caffemodel')+```python 
 +# pre-process config 
 +print('--> Config model') 
 +rknn.config(mean_values=[[0, 0, 0]], std_values=[[255, 255, 255]], target_platform='rk3588') 
 +print('done')
 ``` ```
  
-Tensorflow+  * **mean_values** - The mean of normalization parameter. 
 +  * **std_values** - The variance of normalization parameter. 
 +model input = (image – mean_values) / std_values 
 +  * **target_platform** - Chooses ''rk3588''.
  
-```shell+Load model. 
 + 
 +```python 
 +# Load ONNX model 
 +print('--> Loading model') 
 +ret = rknn.load_onnx(model='./yolov5.onnx') 
 +if ret != 0: 
 +    print('Load model failed!') 
 +    exit(ret) 
 +print('done') 
 +``` 
 + 
 +  * **model** - The path of model. 
 + 
 +Load other platform model. 
 + 
 +```python 
 +# Load pytorch model 
 +print('--> Loading model') 
 +ret = rknn.load_pytorch(model='./resnet18.pt', input_size_list=[[1, 3, 224, 224]]) 
 +if ret != 0: 
 +    print('Load model failed!') 
 +    exit(ret) 
 +print('done') 
 + 
 +# Load tensorflow model 
 +print('--> Loading model')
 ret = rknn.load_tensorflow(tf_pb='./ssd_mobilenet_v1_coco_2017_11_17.pb', ret = rknn.load_tensorflow(tf_pb='./ssd_mobilenet_v1_coco_2017_11_17.pb',
                            inputs=['Preprocessor/sub'],                            inputs=['Preprocessor/sub'],
                            outputs=['concat', 'concat_1'],                            outputs=['concat', 'concat_1'],
                            input_size_list=[[300, 300, 3]])                            input_size_list=[[300, 300, 3]])
-```+if ret != 0: 
 +    print('Load model failed!') 
 +    exit(ret) 
 +print('done')
  
-TensorFlow Lite+# Load caffe model 
 +print('--> Loading model') 
 +ret = rknn.load_caffe(model='./mobilenet_v2.prototxt',  
 +                      blobs='./mobilenet_v2.caffemodel') 
 +if ret != 0: 
 +    print('Load model failed!') 
 +    exit(ret) 
 +print('done')
  
-```shell +# Load tensorflow lite model 
-ret = rknn.load_tflite(model = './mobilenet_v1.tflite')+print('--> Loading model') 
 +ret = rknn.load_tflite(model='./mobilenet_v1.tflite') 
 +if ret != 0: 
 +    print('Load model failed!') 
 +    exit(ret) 
 +print('done') 
 + 
 +# Load darknet model 
 +print('--> Loading model') 
 +ret = rknn.load_darknet(model='./yolov3-tiny.cfg', 
 +                        weight='./yolov3.weights') 
 +if ret != 0: 
 +    print('Load model failed!') 
 +    exit(ret) 
 +print('done')
 ``` ```
  
-Onnx+  * **inputs/outputs** - Only use in tensorflow model. It is the name of inputs/outputs. 
 +  * **input_size_list** - The size and channels of input.
  
-```shell +Build model 
-ret = rknn.load_onnx(model = './arcface.onnx')+ 
 +```python 
 +# Build model 
 +print('--> Building model') 
 +ret = rknn.build(do_quantization=True, dataset='./dataset.txt') 
 +if ret != 0: 
 +    print('Build model failed!') 
 +    exit(ret) 
 +print('done')
 ``` ```
  
-Darknet+  * **do_quantization** - Quantize model or not. 
 +  * **dataset** - The path of txt file which is written in image path.
  
-```shell +Export RKNN model 
-ret = rknn.load_darknet(model = './yolov3-tiny.cfg', + 
-                        weight= './yolov3.weights')+```python 
 +# Export RKNN model 
 +print('--> Export rknn model') 
 +ret = rknn.export_rknn(export_path='./yolov5_int8.rknn') 
 +if ret != 0: 
 +    print('Export rknn model failed!') 
 +    exit(ret) 
 +print('done')
 ``` ```
  
-Pytorch+  * **export_path** - The path of rknn model.
  
-```shell +All the above codes can be found in ''rknn-toolkit2/examples''. There are all platforms we support now. Choose ''rk3588'' in ''rknn-toolkit2/examples/onnx/yolov5/test.py'' and run the file to convert model.
-ret = rknn.load_pytorch(model = './resnet18.pt', +
-                         input_size_list=[[1,3,224,224]]) +
-```+
  
-Convert ''yolov5'' as a example,+```diff patch 
 +diff --git a/examples/onnx/yolov5/test.py b/examples/onnx/yolov5/test.py 
 +index a1c9988..f7ce11e 100644 
 +--- a/examples/onnx/yolov5/test.py 
 ++++ b/examples/onnx/yolov5/test.py 
 +@@ -240,7 +240,7 @@ if __name__ == '__main__':
  
-```shell +     # pre-process config 
-cd examples/onnx/yolov5/ +     print('--> Config model') 
-python3 test.py+-    rknn.config(mean_values=[[0, 0, 0]], std_values=[[255, 255, 255]]) 
 ++    rknn.config(mean_values=[[0, 0, 0]], std_values=[[255, 255, 255]], target_platform='rk3588') 
 +     print('done')
 ``` ```
  
-After convert, it will generate the rknn file ''yolov5s.rknn''.+Run ''test.py'' to generate rknn model.
  
 ```shell ```shell
-$ ls +$ python3 test.py
-bus.jpg  dataset.txt  onnx_yolov5_0.npy  onnx_yolov5_1.npy  onnx_yolov5_2.npy  test.py  yolov5s.onnx  yolov5s.rknn+
 ``` ```
 +
 +<WRAP tip >
 +In ''test.py'', there are inferring rknn model codes. You can refer it to infer rknn on PC.
 +</WRAP>
  
 ===== See Also ===== ===== See Also =====
Last modified: 2023/07/04 05:18 by louis

Page Tools

Khadas Site

Khadas Community

Khadas GitHub

Khadas Downloads