Which Khadas SBC do you use?
VIM3
Which system do you use? Android, Ubuntu, OOWOW or others?
Ubuntu
Which version of system do you use? Khadas official images, self built images, or others?
vim3-ubuntu-20.04-server-linux-4.9-fenix-1.5-230425-emmc.img
Please describe your issue below:
The ksnn demos models from before March work fine, but the new version does not. The latest models with “_uint8” fail on nn_inference.
My converted model failed with the latest version but ran fine with the ./convert using the previous version.
I have not been able to catch an error message.
I assume this is because of my kernel version is old or something similar, but what specific package should I update so that I make the least changes to our system?
Hello @BrandonTetra ,
Have you updated KSNN demo codes? The latest KSNN should do normalization before nn_inference.
If still fail, please add argument –level 2 when you run demo and provide the log.
I have updated to 1.4. using "ksnn-1.4-py3-none-any.whl.
I set the level to 1 and then 2. The script “hangs” on nn_inference call.
|—+ KSNN Version: v1.4 ±–|
Start init neural network …
#productname=VIPNano-QI, pid=0x88
Create Neural Network: 87ms or 87310us
Done.
Get input data …
Done.
Start inference …
resize input pictures error !!!
Set nn inputs error !!!
zonecommand@Khadas:~/ksnn/examples/yolov7_tiny$
The above is from the examples
Hello @BrandonTetra ,
Your ksnn-1.4-py3-none-any.whl is not the latest. The latest version removes resize input image. So the error resize input pictures error !!! also is removed.
Please check your KSNN whether the latest. The latest version.
