Write Model To Flash

Note: KneronDFUT supports 3 platforms - Windows 10 (x86_64 64-bit), Ubuntu 18.04 (x86_64 64-bit), and Raspberry Pi OS - Buster (armv7l 32-bit)

Note: If you are not using the 3 platforms, you may use the DFUT_console provided in Kneron PLUS. Please refer Build with DFUT_console

Note: Please use the latest version of KneronDFUT to avoid problems caused by incompatibility.

1. Introduction

The inference model must be loaded into Kneron AI device before the inference process.

There are two ways to load models:

• Upload Model via USB

  • The model file (.nef) can be uploaded to Kneron AI device using kp_load_model_from_file(), a KP API, before inference.

  • For Python users, the model file (.nef) can be uploaded to Kneron AI device using kp.core.load_model_from_file().

  • For the usage, please refer examples related to inference.

  • In this method, the size of the model in device DDR memory (larger than NEF file size) must be below 35 MB for KL520, and 75 MB for KL720.

• Load Model from Flash

  • The model file (.nef) is written in flash, and it can be loaded using kp_load_model_from_flash(), a KP API, before inference.

  • For Python users, the model file (.nef) is written in flash, and it can be loaded using kp.core.load_model_from_flash().

  • The GUI or command line of KneronDFUT can be used for writing the model file into flash.

  • For the usage, please refer the example kl520_demo_generic_inference_flash_model or kl720_demo_generic_inference_flash_model.

  • For Python users, please refer the example KL520DemoGenericInferenceFlashModel.py or KL720DemoGenericInferenceFlashModel.py.

  • In this method, the size of the model file must be below 32 MB for KL520, and 70 MB for KL720.

Note: Only one model file (.nef) can be loaded, no matter it was uploaded via USB or loaded from flash. If you want to change the model, please reboot the Kneron AI device.

Note: Upload model via USB can be directly used without writing model into flash.

2. Download Kneron DFUT

Download the KneronDFUT_ubuntu.zip into Ubuntu from https://www.kneron.com/tw/support/developers/. It is located at Kneron PLUS section.

$ unzip KneronDFUT_ubuntu.zip
$ cd Kneron_DFUT/bin/

Show help message

$ sudo ./KneronDFUT --help

3. Write Model Into KL520

3.1 Use GUI to Write Model into AI Device

$ sudo ./KneronDFUT

1. Select KL520 Tab.

2. Select the KL520 devices to write model into.

3. Select Update Model to Flash

4. Manually choose Model file.

5. Push Run button.

   

3.2 Use Command Line to Write Model into AI Device

1. List all devices

   $ sudo ./KneronDFUT --list
   

   ===========================================
   Index:          1
   Port Id:        133
   Kn Number:      0x270A265C
   Device Type:    KL520
   FW Type:        KDP
   Usb Speed:      High-Speed
   Connectable:    true
   ===========================================
   

2. Write model into the selected KL520 devices using the port id

   $ sudo ./KneronDFUT --model-to-flash ${MODEL_FILE_PATH} --port 133 -- type KL520
   

   Start Update Model to Device with Port Id 133
   
   ==== Update Model to Device with Port Id: 133 Succeeded ====
   

4. Write Model Into KL720

4.1 Use GUI to Write Model into AI Device

$ sudo ./KneronDFUT

1. Select KL720 Tab.

2. Select the KL720 devices to write model into.

3. Select Update Model to Flash

4. Manually choose Model file.

5. Push Run button.

   

4.2 Use Command Line to Write Model into AI Device

1. List all devices

   $ sudo ./KneronDFUT --list
   

   ===========================================
   Index:          1
   Port Id:        262
   Kn Number:      0x2004142C
   Device Type:    KL720
   FW Type:        KDP
   Usb Speed:      Super-Speed
   Connectable:    true
   ===========================================
   

2. Write model into the selected KL720 devices using the port id

   $ sudo ./KneronDFUT --model-to-flash ${MODEL_FILE_PATH} --port 262 --type KL720
   

   Start Update Model to Device with Port Id 262
   
   ==== Update Model to Device with Port Id: 262 Succeeded ====