unboxing photos
burn image
Windows version of the imager tool download address:
https://ascend-repo.obs.cn-east-2.myhuaweicloud.com/Atlas%20200I%20DK%20A2/DevKit/tools/latest/Ascend-devkit-imager_latest_win-x86_64. exe
Prepare a MicroSD card, the recommended capacity is at least 64GB
The main interface of the one-key card making tool
The entire process of burning the image may take a long time, and you can do other things while burning. There will be a prompt after the burning is completed
The network card configuration information after burning is as follows:
eth1: 192.168.137.100
eth0: DHCP
After burning, insert the card into the SD slot of the development version, connect one end of the network cable to the computer, and the other end to the ETH1 interface. Then turn on the power
Enter the 200DK A2 environment
Use terminal software such as Mobax to enter the development environment
Login to 192.168.137.100 using SSH
Default username and password:
Username: root
Password: Mind@123
You can use the lscpu command to view hardware information
lscpu
Use the npu-smi info command to check the NPU information
npu-smi info
Being able to enter the system shows that the development board is working normally.
run a base instance
The jupyter lab software (visual code demonstration, data analysis tools) has been included in the image to provide users with a graphical operation interface
enter the directory
cd /home/HwHiAiUser/samples/notebooks
modify startup script
vi start_notebook.sh
startup script
./start_notebook.sh
On the echo information page, press and hold the "Ctrl" key on the keyboard and use the left mouse button to click the bold URL link in the echo in step 3 (for example: http://192.168.0.2:8888/lab?token=a046a76dc21f1504f271c16278ed62ed7fb014aaf38ee807), enter Jupyter lab interface, you can run the Python inference samples preset in the developer kit.
Of course, you can also reset the password for jupyter lab
jupyter lab password
Then execute the start_notebook.sh script to start the debugging environment
Enter the address in the browser, for example: https://192.168.137.100:8888 to enter the development environment
You can see that there are various cases in it, and you can experience them all. The experience method is the same as the usual notebook experience on the server or on a stand-alone machine
Here I randomly ran an image classification case under the 03-resnet folder and the results are as follows:
Recommended code warehouse resources
Huawei Ascend Developer Kit Atlas 200I DK A2 code warehouse address:
https://gitee.com/HUAWEI-ASCEND/ascend-devkit