Forums - Tf-ResNet50-V1 cannot run in DSP

3 posts / 0 new
Last post
Tf-ResNet50-V1 cannot run in DSP
lvxiangxiang
Join Date: 28 Dec 18
Posts: 4
Posted: Wed, 2019-02-27 01:46

Hello,

I converted a tensorflow model resnet50-v1,it can run in CPU and GPU,but failed in DSP. Logs:

Quote:
A/DEBUG: *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** ***

2019-02-27 17:20:28.624 7892-7892/? A/DEBUG: Build fingerprint: 'Xiaomi/polaris/polaris:9/PKQ1.180729.001/V10.0.7.0.PDGCNFH:user/release-keys'
2019-02-27 17:20:28.624 7892-7892/? A/DEBUG: Revision: '0'
2019-02-27 17:20:28.624 7892-7892/? A/DEBUG: ABI: 'arm64'
2019-02-27 17:20:28.624 7892-7892/? A/DEBUG: pid: 7204, tid: 7889, name: pool-1-thread-2  >>> com.baidu.ai.easyaimobile.demo <<<
2019-02-27 17:20:28.624 7892-7892/? A/DEBUG: signal 6 (SIGABRT), code -6 (SI_TKILL), fault addr --------
2019-02-27 17:20:28.625 7892-7892/? A/DEBUG: Abort message: 'terminating with uncaught exception of type DlSystem::RuntimeException: Failed to execute network.  Error code: -6.'
2019-02-27 17:20:28.625 7892-7892/? A/DEBUG:     x0  0000000000000000  x1  0000000000001ed1  x2  0000000000000006  x3  0000000000000008
2019-02-27 17:20:28.625 7892-7892/? A/DEBUG:     x4  fefeff77ff2d352c  x5  fefeff77ff2d352c  x6  fefeff77ff2d352c  x7  7f7f7f7f7f7f7f7f
2019-02-27 17:20:28.626 7892-7892/? A/DEBUG:     x8  0000000000000083  x9  513f3cac977be403  x10 0000000000000000  x11 fffffffc7ffffbdf
2019-02-27 17:20:28.626 7892-7892/? A/DEBUG:     x12 0000000000000001  x13 000000005c76565c  x14 001d909a129191d5  x15 000098283ad45c0a
2019-02-27 17:20:28.626 7892-7892/? A/DEBUG:     x16 0000007a019af2c8  x17 0000007a018ed544  x18 0000000000000001  x19 0000000000001c24
2019-02-27 17:20:28.626 7892-7892/? A/DEBUG:     x20 0000000000001ed1  x21 0000007957bfb4b8  x22 00000000ffffff80  x23 00000000ffffffc8
2019-02-27 17:20:28.626 7892-7892/? A/DEBUG:     x24 0000007957bfb580  x25 0000007957bfb450  x26 0000007957bfb490  x27 0000000000000020
2019-02-27 17:20:28.626 7892-7892/? A/DEBUG:     x28 000000796121e000  x29 0000007957bfb3c0
2019-02-27 17:20:28.626 7892-7892/? A/DEBUG:     sp  0000007957bfb380  lr  0000007a018e1f0c  pc  0000007a018e1f34
2019-02-27 17:20:28.628 7892-7892/? A/DEBUG: backtrace:
2019-02-27 17:20:28.628 7892-7892/? A/DEBUG:     #00 pc 0000000000021f34  /system/lib64/libc.so (abort+116)
2019-02-27 17:20:28.628 7892-7892/? A/DEBUG:     #01 pc 00000000000a1ce4  /data/app/com.baidu.ai.easyaimobile.demo-xxe50G7Iw29wMbf_Vxxgkw==/lib/arm64/libc++_shared.so
2019-02-27 17:20:28.628 7892-7892/? A/DEBUG:     #02 pc 00000000000a1ee0  /data/app/com.baidu.ai.easyaimobile.demo-xxe50G7Iw29wMbf_Vxxgkw==/lib/arm64/libc++_shared.so
2019-02-27 17:20:28.628 7892-7892/? A/DEBUG:     #03 pc 00000000000c8960  /data/app/com.baidu.ai.easyaimobile.demo-xxe50G7Iw29wMbf_Vxxgkw==/lib/arm64/libc++_shared.so
2019-02-27 17:20:28.628 7892-7892/? A/DEBUG:     #04 pc 00000000000c8028  /data/app/com.baidu.ai.easyaimobile.demo-xxe50G7Iw29wMbf_Vxxgkw==/lib/arm64/libc++_shared.so (__cxa_throw+128)
2019-02-27 17:20:28.628 7892-7892/? A/DEBUG:     #05 pc 00000000001fda18  /data/app/com.baidu.ai.easyaimobile.demo-xxe50G7Iw29wMbf_Vxxgkw==/lib/arm64/libSNPE.so
2019-02-27 17:20:28.628 7892-7892/? A/DEBUG:     #06 pc 00000000001e86d0  /data/app/com.baidu.ai.easyaimobile.demo-xxe50G7Iw29wMbf_Vxxgkw==/lib/arm64/libSNPE.so
2019-02-27 17:20:28.628 7892-7892/? A/DEBUG:     #07 pc 00000000000fb778  /data/app/com.baidu.ai.easyaimobile.demo-xxe50G7Iw29wMbf_Vxxgkw==/lib/arm64/libSNPE.so
2019-02-27 17:20:28.628 7892-7892/? A/DEBUG:     #08 pc 00000000000f6ce4  /data/app/com.baidu.ai.easyaimobile.demo-xxe50G7Iw29wMbf_Vxxgkw==/lib/arm64/libSNPE.so (DnnRuntime::DnnRuntime::Execute(zdl::DlSystem::ITensor const*, std::__ndk1::map<std::__ndk1::basic_string<char, std::__ndk1::char_traits<char>, std::__ndk1::allocator<char>>, std::__ndk1::shared_ptr<zdl::DlSystem::ITensor>, std::__ndk1::less<std::__ndk1::basic_string<char, std::__ndk1::char_traits<char>, std::__ndk1::allocator<char>>>, std::__ndk1::allocator<std::__ndk1::pair<std::__ndk1::ba
2019-02-27 17:20:28.628 7892-7892/? A/DEBUG:     #09 pc 00000000000bdaf8  /data/app/com.baidu.ai.easyaimobile.demo-xxe50G7Iw29wMbf_Vxxgkw==/lib/arm64/libSNPE.so (zdl::SNPE::SnpeRuntime::Execute(zdl::DlSystem::ITensor const*, zdl::DlSystem::TensorMap&)+536)

2019-02-27 17:20:28.628 7892-7892/? A/DEBUG:     #10 pc 00000000000bb0b8  /data/app/com.baidu.ai.easyaimobile.demo-xxe50G7Iw29wMbf_Vxxgkw==/lib/arm64/libSNPE.so (zdl::SNPE::SNPE::execute(zdl::DlSystem::ITensor const*, zdl::DlSystem::TensorMap&)+16)

BTW,

Caffe version has no such problem.

Other tf models have no such problem so far.

SNPE version:1.22

  • Up0
  • Down0
KaiJ
Join Date: 4 Apr 18
Posts: 4
Posted: Wed, 2019-02-27 04:21

Hi.

How many output nodes do you have all together in ouput layers?

I'm asking this because I get that same error (Failed to execute network.  Error code: -6) if I have more than two output nodes. This happens only with DSP runtime. I'm running Mobilenet SSD and the output layer unfortunately has 3 nodes, so it does not work with DSP runtime, but it works if I change the output layers so that there are only two output nodes.

Is it possible for you to change the output layers just to test if it works with different ouput layers or numbers of output nodes? And if you can it would be interesting to hear your results.

 

  • Up0
  • Down0
zhaoyangstar
Join Date: 14 Apr 19
Posts: 23
Posted: Thu, 2019-05-23 06:25

Hi KaiJ,

I run a PVANET-LITE-Faster-RCNN. The network has 3 output layers. I also met the same error when running on DSP. I just delete one output layer in order to verify what you said is true or not.

After delete one output layer there are only 2 output layers. But it still get that error.

By the way, the network has 2 input layers ( one called data, the other called im_info). I think DSP runtime should support multiple input layers.

Do you have any advice on solving this error? Thanks in advance!

  • 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.