Say we have a deconv layer defined as follows,
inputs = tf.placeholder(shape=(1,10,10,1), dtype=tf.float32,name='input')
deconv = tf.layers.conv2d_transpose(inputs,1,5,strides=2,padding='SAME')
tf.identity(deconv, name='output')
When we freeze such a model and apply `snpe-tensorflow-to-dlc`, it reports a warning like,
2019-09-18 16:52:51,512 - 372 - WARNING - WARNING_UNSUPPORTED_OPS_FOUND: Some Operations are not supported by converter. Printing the list of operations:
2019-09-18 16:52:51,512 - 374 - WARNING - WARNING_TF_OP_NOT_SUPPORTED: Operation (conv2d_transpose/strided_slice_2) of type (StridedSlice) is not supported by converter.
2019-09-18 16:52:51,512 - 374 - WARNING - WARNING_TF_OP_NOT_SUPPORTED: Operation (conv2d_transpose/strided_slice_1) of type (StridedSlice) is not supported by converter.
2019-09-18 16:52:51,512 - 106 - 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!
Note: We can't simply use --allow_unconsumed_nodes to ignore this layer. Could anyone provide a walk around for such deconv layer? Many thanks!