Forums - snpe-tensorflow-to-dlc convert error

2 posts / 0 new
or Register
Last post
snpe-tensorflow-to-dlc convert error
konan91
Join Date: 2 Apr 19
Posts: 8
Posted: Thu, 2019-07-04 02:35
Top

I'm tring to convert tensorflow meta file to snpe dlc file using snpe-tensoflow-to-dlc, but it has been some error happened.Log is

E0704 17:19:32.484879 140150566917952 snpe-tensorflow-to-dlc:106] Conversion failed: ERROR_TF_CHANNEL_SHUFFLE_OUTPUT: Channel Shuffle layer requires the Output Shape to be same as Input Shape.
2019-07-04 17:19:32,484 - 106 - ERROR - Conversion failed: ERROR_TF_CHANNEL_SHUFFLE_OUTPUT: Channel Shuffle layer requires the Output Shape to be same as Input Shape.
 
it seems to be some shuffle node input output's dememsion is different.
This link is model's tensorboard image that conversion error happened
 
And the other question is 'Does snpe layer has only one output node'?
E0704 18:27:39.850524 140637585200960 snpe-tensorflow-to-dlc:106] Conversion failed: ERROR_TF_INPUT_OPERATION_NOT_FOUND: Input operation not found for network/maxpool5/MaxPool
2019-07-04 18:27:39,850 - 106 - ERROR - Conversion failed: ERROR_TF_INPUT_OPERATION_NOT_FOUND: Input operation not found for network/maxpool5/MaxPool
 
it seems to be  network/maxpool5/MaxPool layer's input node is not matched previous layer's output node.
in that case, the maxpool5's input node is 'convolutional13' but the convolutional13's output node is route1, not maxpool5.
This link is also model's tensorboard image that conversion error happened
 
I need some advice to convert this model to dlc file.

Plaese help. Thanks 

  • Up0
  • Down0
Re: snpe-tensorflow-to-dlc convert error
gesqdn-forum
Join Date: 4 Nov 18
Posts: 184
Posted: Tue, 2019-07-23 05:43
Top

Hi,
The error message describes the required input layer shape should be same as output layer shape.
Kindly check the input shape passed should match with the output while conversion of model.

It seems like the network layer you mentioned "network/maxpool5/MaxPool" is not supported by SNPE in Tensorflow framework.
Only "average_pooling2d", "max_pooling2d" layers are suppoeted for Tensorflow in SNPE.
 
You can find the supporetd SNPE network layers from https://developer.qualcomm.com/docs/snpe/network_layers.html
 
  • 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.