Hello,
I installed snpe 2.6.0.4131 on Windows DevKit 2023 (8cx gen3).
snpe-net-run.exe --container=vgg16_q.dlc --input_list=input_list.txt --use_dsp
yields the output like:
error_code=401; error_message=Network creation has failed. error_code=401; error_message=Network creation has failed. QnnGraph_create() Failed: 6031; error_component=Dl Network; line_no=1809; thread_id=6012; error_component=Dl Network; line_no=311; thread_id=17488
I installed snpe 2.5.0.4852 for Linux on WSL2 and tried below in /home/polphit/snpe-2.5.0.4052/bin/aarch64-ubuntu-gcc7.5:
LD_LIBRARY_PATH=/home/polphit/snpe-2.5.0.4052/lib/aarch64-ubuntu-gcc7.5 ADSP_LIBRARY_PATH=/home/polphit/snpe-2.5.0.4052/lib/aarch64-ubuntu-gcc7.5/ ./snpe-platform-validator --runtime all
yields the output like:
SNPE is not supported on this SoC.
Do I have to do something to enable DSP on 8cx gen3?
Thank you.
Dear developer,
It's great to see you have tried to work on Gen3 already. Up to now, the tools you have checked needs to confirm internal and get back to you.
BTW, the tools you can try to on Windows devices with TShell related.
BR.
Wei
Hi, I also have the Windows DevKit and am trying to get SNPE working on it. Do you have an update on whether it is now supported?
Thank you!
I am also look for the support as the PC now did not show any difference from my other PC. I would like to get more support to make it as a real development kit.
Thanks.
Dear developer,
You could convert/quantize AI model on Linux SNPE SDK and deploy model Window SNPE SDK to device at end to end.
BR.
Wei
Is there any instruction how to install the Linux SNPE SDK? When I download 'Qualcomm Neural Process SDK' from Qualcomm Neural Processing SDK for Windows on Snapdragon - Qualcomm Developer Network, the linux file is a qik file extension. But I expected it was a zip file as tutorial said.
BTW, please help me how to install SNPE SDK on WSL2 on windows devkit 2023. And if possible, please let me know how to set up the tensorflow to get the hw acceleration on devkit 2023.
I hope there is detailed intruction how to set up the SNPE with tensorflow on WSL2 in devkit 2023.
Although qualcomm provides many software tools, I couldn't find a proper solution or documents. For instance, there is Package Manager tools 3 for linux which is only for X86 arch, but WSL2 in devkit-2023 is definetly is aarch64, therefore, it couldn't be installed.
And SNPE is qik extension file, it could not extact without Package Manager in WSL2 in devkit-2023 environment.
I'm so disappointed how No.1 SoC company doesn't provide enough informantions such for a long time.