Forums - Error in converting retrained inception_v3 model to .dlc

1 post / 0 new
or Register
Error in converting retrained inception_v3 model to .dlc
saikrish
Join Date: 21 Sep 17
Posts: 3
Posted: Tue, 2018-04-10 12:46
Top

I attempted to run a retrained Inception_v3 model (followed the procedure found here in tensorflow's inception_v3 retraining tutorial). The issue was that all the layers of the graph were not supported by the tensorflow to dlc converter. 

    Two sample warnings are shown below: 
  1. WARNING - WARNING_TF_SCOPE_OP_NOT_CONSUMED: Operation (module_apply_default/InceptionV3/InceptionV3/Mixed_6d/Branch_1/Conv2d_0b_1x7/Conv2D) not consumed by converter: Conv2D.
  2. WARNING - WARNING_TF_SCOPE_OP_NOT_CONSUMED: Operation (module_apply_default/InceptionV3/InceptionV3/Mixed_6b/Branch_2/Conv2d_0c_1x7/Relu) not consumed by converter: Relu.
    I went through the tensorflow graph compatibility chapter and these layers by themselves are supported by SNPE. Is there an issue with the way they're being defined? 
 
    Here's some additional information which might help:
  1. My SNPE version: 1.12.0
  2. To make use of Tensorflow_hub (which is used by the tutorial), I had to build tensorflow 1.7.0 from sources on my Ubuntu 14.04 VM. 
Thank you,
Sai Krishnan
  • Up0
  • Down0

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.