-
Notifications
You must be signed in to change notification settings - Fork 37
Atlas 200 DK Setup and Update Guide
Official Atlas 200 DK Developer Kit
Ascend Samples
Ascend ModelZoo
Ascend Wiki C++ ENV Preparation Guide
Ascend Wiki Python ENV Preparation Guide
Ascend Wiki C++ ENV & Software Installation
Ascend Wiki Python ENV & Software Installation
0. Setup Hardware
1. Development and Operating Environment Setup
2. Install Software Dependencies (C++/Python)
Check Version
Build from new SD Card
Build from Docker Image
Setup Virtual Environment
To setup Atlas 200 DK hardware, please refer to Official Atlas 200 DK Developer Kit
Development Environment: the environment for model conversion or application development, which can run on CPU and does NOT rely on the Ascend processor. It requires Ubuntu 18.04
OS. It can be installed on PC or Atlas 200 DK. MindStudio IDE can be installed.
Operating Environment: the environment with Ascend processor to run Ascend AI model/applications. It requires Atlas 200 DK.
To setup the development and operating environment, you can refer to `Section 5 : Setup Development and Operating Environments' of the 'Ascend AI Processor and Software Stack Basics' Slides (Download) on Ascend community. Alternative methods can be found at Ascend Wiki - Setting Up the Development and Operating Environments
To configure the C++ environment and install dependencies, refer to the Ascend Wiki C++ ENV Guide
To configure the Python environment and install dependencies (such as: pip3
, Pillow
, OpenCV-Python3
, FFmpeg
), refer to the Ascend Wiki Python ENV Guide
Check Version
Flash & Upgrade SD Card
Setup Virtual Environment
Before we begin upgrading the Atlas 200 DK, we need to check the software version of the device. Depending on its current version, we will either upgrade by downloading software components or flashing the SD card.
-
SSH to the Atlas 200 DK as the HwHiAiUser user
-
Switch to the root user and check software versions of the device (Driver, Firmware, AI CPU, ACLlib)
cat /var/davinci/driver/version.info cat /usr/local/Ascend/firmware/version.info cat /var/davinci/aicpu_kernels/version.info cat /home/HwHiAiUser/Ascend/acllib/version.info
NOTE
If the system components of the Atlas 200 DK are of the 1.73.xx.xx or earlier version, you will need to flash the SD card with an updated image. Otherwise you may refer to the official Atlas 200 DK developer guide and upgrade via downloading software components.
- <= 1.73.xx.xx - refer Flash & Upgrade SD Card
- > 1.73.xx.xx - refer to Official Atlas 200 DK Developer Kit
We will create an updated system boot disk for the Atlas 200 DK by preparing an SD card. Having a SD card reader will make this step easier, otherwise you will need to connect the pins of the Atlas 200 DK with a jumper cap/wire before connecting the DK and running the SD preparation script.
You may also follow the Official Ascend Wiki Guide
If you have a SD card reader: The following procedure is done in Window's environment
- Download Etcher for Windows
- Download and extract image for the SD card (obtain
.iso
images from the URLs in the Official Ascend Wiki Guide. We recommend going with versions at least 3.3.0.alpha001) - Insert the SD card to the card reader, connect it to the machine via USB, open Etcher and select the corresponding image and SD card. Hit Flash. Note: this can take up to 20-30mins
- Once the flashing is complete, remove the SD card from the card reader and insert it to the Atlas 200 DK. Power up the board and wait for it to upgrade the software automatically. Wait until the 4 LEDs to be fully lit up (this may take around 5-10mins).
NOTE: Do not power off the board while it is upgrading else it can corrupt the image and you will have to restart.
Aside from flashing and upgrading the SD card, you may also build the Ubuntu environment from a docker image with MindStudio and Ascend Toolkit installed. Follow the Ascend docker image usage guide
- Install Docker
- Configure image acceleration
- Pull the docker image
- Create and configure the docker environment
NOTE: Only version 3.3.0.alpha001 is released for Docker images as of now. If you are looking for CANN verisons newer than 3.3.0 then please manually download the drivers and upgrade by following the Official Atlas 200 DK Developer Kit or refer to the Flash and Upgrade SD Card section.
Install miniconda and create virtual environments whenever you start to work on different projects to avoid package dependency conflicts.
Visit the official link (Miniconda installers) and choose Miniconda3 Linux-aarch64 64-bit
to install the compatible version on the 200 DK
Alternatively you may create virtual environments using Python.
python3 -m venv <venv_name>
Include
--system-site-packages
to include system level packages
To activate and use the venv
source <venv_name>/bin/activate