CLIPImageEncoder.so qnn-net-run run on HTP backend failed
#1
by
finalf0
- opened
I test CLIPImageEncoder.so using qnn-net-run on Xiaomi 14 Pro( Snapdragon 8 Gen 3)
here is my script:
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/data/local/tmp/libs/aarch64-android:/data/local/tmp/libs/hexagon-v75/unsigned
model=CLIPImageEncoder.so
./qnn-net-run \
--model ${model} \
--input_list raw_list.txt \
--backend libQnnHtp.so \
--profiling_level basic \
--perf_profile high_performance
Failed with error:
qnn-net-run pid:28507
Initialization failure
run with cpu and gpu backend is ok, Did I missed something?
At the moment, our Open AI Clip doesn't work with QNN HTP. So the behavior observed is expected.