Error 100 with a customized example
Hi,
We're evaluating the KL720 devkit, and we're trying to create a custom application and firmware to run inference on the device.
We have a compiled model (nef file) and followed the guide on creating a single model example from http://doc.kneron.com/docs/#plus/customized_api/create_single_model_example/
We made some modifications, since we're not using yolo, instead we're using a customized SSD mobilenet model with the bottom nodes removed (for compliance with Kneron supported operations).
We've using a color image with size 320x320 to feed as input to our model, and our model's output is a set vectors of features from the box predictor and the scores predictor. We uploaded a similar model in a previous post: https://www.kneron.com/forum/discussion/167/migrating-a-quantized-aware-model-from-tensorflow#latest
After the application reaches kp_customized_inference_receive() we get a return value of 100 with recv_size=16. From the documentation this seems to correlate to KP_FW_ERROR_UNKNOWN_APP.
What could be the cause of the issue?
Thanks.
Comments
Hi oded,
Can you provide your PLUS version / Firmware version ?
or you can follow here and try first : http://doc.kneron.com/docs/#plus/customized_api/create_single_model_example/
3. SCPU Firmware Development
5. Edit main.c
Maybe you should switch your job ID ?
Thanks.
Using PLUS 1.2.1
The log from the device during boot:
SPL v1.0.0-0
>> Start running KL720 KDP2 companion mode ...
==boot from External Flash==
Load NCPU from flash image1: 0x002a0000
SDK v1.4.1-build.000
The job ID is 1003 both on the FW side and on the application side.
We copied the example for KL520 and changed it to suit the KL720, since there were no examples for the KL720.
Hi Oded,
We could help you debug if you could provide the firmware and PLUS example files.