Forums - Failed to load the model on 8gen1 using SNPE.

3 posts / 0 new
or Register
Last post
Failed to load the model on 8gen1 using SNPE.
zzy913
Join Date: 10 Dec 18
Posts: 3
Posted: Thu, 2022-12-01 18:06
Top

After the quantization model is converted using SDK 1.66, loading on 8gen1 fails. Has anyone encountered and solved this problem?  PS:--use_float_io has been used during model conversion.

  • Up0
  • Down0
Re: Failed to load the model on 8gen1 using SNPE.
weihuan
Join Date: 12 Apr 20
Posts: 270
Posted: Sun, 2022-12-04 00:07
Top

Dear customer,

Could you please help to share your converson and execution commands to us?

Also, you need to quantize your model if you hope to run on cDSP runtime.

unsigedPD:OFF is required after 1.60 or later.

BR.

Wei

  • Up0
  • Down0
Re: Failed to load the model on 8gen1 using SNPE.
zzy913
Join Date: 10 Dec 18
Posts: 3
Posted: Tue, 2022-12-06 05:51
Top
The model conversion command is:
./bin/x86_64-linux-clang/snpe-dlc-quantize --input_dlc ./resnet50_tf_b1_fp16.dlc --input_list ./QuantFiles/1_224_224_3/file_list.txt --output_dlc ./resnet50_tf_b1_int8.dlc --enable_htp --use_float_io
I'm confused about "unsigedPD:OFF", do you need to modify "RuntimeCheckOption_t = UNSIGNEDPD_CHECK" in the code? I have now done this and still the model fails to load. I have made this modification, but still the model fails to load on the DSP.
  • Up0
  • Down0
or Register

Opinions expressed in the content posted here are the personal opinions of the original authors, and do not necessarily reflect those of Qualcomm Incorporated or its subsidiaries (“Qualcomm”). The content is provided for informational purposes only and is not meant to be an endorsement or representation by Qualcomm or any other party. This site may also provide links or references to non-Qualcomm sites and resources. Qualcomm makes no representations, warranties, or other commitments whatsoever about any non-Qualcomm sites or third-party resources that may be referenced, accessible from, or linked to this site.