I executed my ai model through the C++ apis. I've tested all runtimes, CPU,GPU,DSP,AIP. CPU,GPU&DSP just worked fine.
When I execute the quantized model with --enable_hta in AIP runtim, snpe->execute(inputtensor.get(), outputTensorMap) returned false.
I didn't find any error through the logcat console:
E/npu_user_driver: NPU_Stats: npu_load_network(): 95.17 ms
E/npu_user_driver: NPU_Stats: npu_alloc_buffer_v2(): 0.08 ms sts=0
...
E/npu_user_driver: npu_set_property status: 0
Net Execute Failed //this is a custom print code when snpe->execute returns false
...
E/npu_user_driver: * NPU_Stats: npu_free_buffer_v2(): 0.01 ms
E/npu_user_driver: * NPU_Stats: npu_unload_network(): 1607.50 ms
E/npu_user_driver: npu_set_property status: 0
Please help me debug this issue.
Dear customer,
Could you please help to reply below confusion to us?
1 Which platform you used?
2 What's SNPE version you tested?
3 What's SNPE conversion and quantization commands you executed?
BR.
Wei