Forums - Is tf.image.resize_bilinear really supported in SNPE?

1 post / 0 new
or Register
Is tf.image.resize_bilinear really supported in SNPE?
cf.hsieh
Join Date: 3 Jan 18
Posts: 2
Posted: Sat, 2018-03-10 00:15
Top

Hi ,

my system configurations are:

Ubuntu 16.04

Python-2.7

Tensorflow-1.6 gpu version

snpe-1.13.0

-----------------------------------------------

The doc in snpe-1.13.0 say that  "tf.image.resize_bilinear" is supported .

So I make a simple net as following and save it into .pb file:

#First convolutional layer

with tf.name_scope('Conv1'):
    conv1=tf.layers.conv2d(Input,Filters_layer1,Kernelsize,activation=activationList[ActivationFunc], padding='SAME', name='Conv1')#convolution
    s = tf.shape(conv1)
    h = s[1]
    w = s[2]
    upscale = tf.image.resize_bilinear(conv1, [h * 2, w * 2], align_corners=False, name="upscale")
    # upscale =  tf.image.resize_images(conv1, [h * 2, w * 2])
    OutNode = tf.nn.relu(upscale, name="OutNode")
but when I use convert to generate .dlc file, it show error msg as following:
2018-03-10 16:02:30.034842: I tensorflow/core/platform/cpu_feature_guard.cc:140] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2 FMA
2018-03-10 16:02:30,067 - 364 - WARNING - WARNING_TF_SCOPE_OP_NOT_CONSUMED: Operation (Conv1/mul_1) not consumed by converter: Mul.
2018-03-10 16:02:30,067 - 364 - WARNING - WARNING_TF_SCOPE_OP_NOT_CONSUMED: Operation (Conv1/upscale/size) not consumed by converter: Pack.
2018-03-10 16:02:30,067 - 364 - WARNING - WARNING_TF_SCOPE_OP_NOT_CONSUMED: Operation (Conv1/strided_slice_1) not consumed by converter: StridedSlice.
2018-03-10 16:02:30,067 - 364 - WARNING - WARNING_TF_SCOPE_OP_NOT_CONSUMED: Operation (Conv1/strided_slice) not consumed by converter: StridedSlice.
2018-03-10 16:02:30,067 - 364 - WARNING - WARNING_TF_SCOPE_OP_NOT_CONSUMED: Operation (Conv1/mul) not consumed by converter: Mul.
2018-03-10 16:02:30,067 - 123 - ERROR - Conversion failed: ERROR_TF_OPERATION_NOT_MAPPED_TO_LAYER: Some operations in the Tensorflow graph were not resolved to a layer. You can use --allow_unconsumed_nodes for partial graph resolution!
 
the Deconvolution fail at StridedSlice with same error msg... 
 
Any suggestion? Is tf.image.resize_bilinear  really supported in SNPE?  Thanks~~~


 

  • 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.