When I tried to utilize the NPU or HTP or DSP as they call it using snpe framework following the guide from rubikpi documentation page and every step on host machine worked fine, but when I tried to run the model on rubikpi using this command, it threw error as:
The selected runtime is not available on this platform. Continue anyway to observe the failure at network creation time.
error_code=15001; error_message= QnnBackend_DeviceCreate() failed; QNN_DEVICE_ERROR_INVALID_CONFIG: Invalid config values; error_component=QNN; line_no=2489; thread_id=547649191968
Segmentation fault
Then I also tried qnn-net-run, and similar to above steps, on host machine worked fine but when I tried to run command to run inference on rubikpi as mentioned here in the documentation, it threw error as:
qnn-net-run pid:40936
qnn-net-run build version: v2.36.0.250627101419_123260
qnn-net-run log level is : QNN_LOG_LEVEL_ERROR
<W> Initializing HtpProvider
Processing inference input(s):
cropped/chairs.raw
cropped/notice_sign.raw
cropped/trash_bin.raw
cropped/plastic_cup.raw
2.5ms [ ERROR ] <E> createUnsignedPD unsigned PD or DSPRPC_GET_DSP_INFO not supported by HTP
2.6ms [ ERROR ] <E> DspTransport.openSession qnn_open failed, 0x00000072, prio 100
2.6ms [ ERROR ] <E> DspTransport.getHandle failed, error 0x00000007
2.6ms [ ERROR ] <E> createDspTransportInstance failed to config transport object
2.6ms [ ERROR ] <E> error in creation of transport instance
2.6ms [ ERROR ] <E> Failed to create transport for device, error: 1002
2.6ms [ ERROR ] <E> Failed to load skel, error: 1002
2.6ms [ ERROR ] <E> Transport layer setup failed: 14001
2.6ms [ ERROR ] <E> Failed to parse default platform info: 14001
2.6ms [ ERROR ] <E> Failed to load default platform info: 14001
2.6ms [ ERROR ] <E> Failed to parse platform config: 14001
Device Creation failure
<W> Backend 1 free cleanup called during process exit
I double checked all paths and libraries were set, except just one .cat file named libqnnhtpv68.cat was missing which I couldn’t find anywhere or at the place ${QNN_SDK_ROOT}/lib/hexagon-v68/unsigned where it was mentioned to be present at in the official documentation by rubikpi
Please check and let me know how to fix it, or if there is any missing step in documentation please update it.
As I had mentioned above the link of step where the command execution failed, the command failed while doing this step, specifically running ./snpe-net-run --container inception_v3_quantized_with_htp_cache.dlc --input_list input_list.txt --output_dir output_htp --use_dsp
and the 2nd error occurred while I was replicating this step, specifically running ./qnn-net-run --model libinception_v3_quantized.so --backend libQnnHtp.so --input_list input_list.txt --output_dir output_htp command.
my system’s info if that is required:
$ uname -a
Linux RUBIKPi 6.6.38 #1 SMP PREEMPT Wed Mar 26 11:40:28 UTC 2025 aarch64 GNU/Linux
$ cat /etc/os-release
PRETTY_NAME="Debian GNU/Linux 13 (trixie)"
NAME="Debian GNU/Linux"
VERSION_ID="13"
VERSION="13 (trixie)"
VERSION_CODENAME=trixie
DEBIAN_VERSION_FULL=13.0
ID=debian
HOME_URL="https://www.debian.org/"
SUPPORT_URL="https://www.debian.org/support"
BUG_REPORT_URL="https://bugs.debian.org/"
few errors regarding use of htp vanishes when I use the latest .so libraries instead of the version mentioned in the docs on rubikpi website, but some new error related to unable to use fastrpc occurs. I am not sure other errors didn’t occur because this error comes before and crashes the program before it could throw those error or new libraries solve the error. Please check and let me know the sdk versions which should be used and the correct steps to use them, if possible upload a compiled model and the script / code that can be easily replicated on rubik pi as demo on docs or github.
Dear customer,
some new error related to unable to use fastrpc occurs
–>Would you be able to share the .so libraries you are utilizing and the specific error messages you are encountering?
copied all the libraries mentioned anywhere in the docs, couldn’t keep track of what all I copied . The error was like below perhaps with qnn-net-run and with tflite delegate also:
/prj/qct/webtech_scratch20/mlg_user_admin/qaisw_source_repo/rel/2.31.0/release/snpe_src/avante-tools/prebuilt/dsp/hexagon-sdk-4.2.0/ipc/fastrpc/rpcmem/src/rpcmem_android.c:38:dummy call to rpcmem_init, rpcmem APIs will be used from libxdsprpc
ERROR: Restored original execution plan after delegate application failure.
could there be some issue with FASTRpc with debian image, as that might be natively written for android platform?
. The error was like below perhaps with qnn-net-run and with tflite delegate also: