Forums - ERROR - Conversion failed: Cannot resolve BatchNorm layer due to missing variance value.

3 posts / 0 new
Last post
ERROR - Conversion failed: Cannot resolve BatchNorm layer due to missing variance value.
yuan.wenhua
Join Date: 14 Sep 17
Posts: 8
Posted: Mon, 2017-11-06 18:40

Hi, When I convert a tf model to dlc, error occured:" ERROR - Conversion failed: Cannot resolve BatchNorm layer due to missing variance value.

I am sure that snpe support BatchNorm layer, and the tf model runs normal on android. So what the problem probably is

Thanks!

  • Up0
  • Down0
scott.creager
Join Date: 6 Sep 17
Posts: 7
Posted: Wed, 2018-02-07 08:20

Are you using tf.slim batchnorm? the SNPE batchnorm script expects variance to either be an "identity" or a "Const".
If you aren't properly converting your batchnorm vars to consts (either by an is_training flag or by having in-place updates of variance) this will cause this error.

  • Up0
  • Down0
yang.xu
Join Date: 15 Oct 19
Posts: 1
Posted: Fri, 2019-10-18 00:04

have you known how to solve this problem?

  • 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.