Hi,
I am trying to convert my tensorflow model into DLC using this command.
snpe-tensorflow-to-dlc --graph=onet_landmark.pb --input_dim "onet/input" 48,48,3 --out_node onet/prob1 --out_node onet/conv6-2/conv6-2 --out_node onet/conv6-3/conv6-3 --dlc=onet_landmark.dlc --verbose
But some of the nodes in graph are not consumed by the DLC converter.
Building Network
==============================================================
2018-05-30 08:51:24,149 - 327 - INFO - INFO_TF_BUILDING_INPUT_LAYER: Building layer (INPUT) with node: onet/input, shape [48, 48, 3]
2018-05-30 08:51:24,213 - 365 - WARNING - WARNING_TF_SCOPE_OP_NOT_CONSUMED: Operation (onet/conv6-1/conv6-1) not consumed by converter: BiasAdd.
2018-05-30 08:51:24,213 - 365 - WARNING - WARNING_TF_SCOPE_OP_NOT_CONSUMED: Operation (onet/prelu4/Neg_1) not consumed by converter: Neg.
2018-05-30 08:51:24,213 - 365 - WARNING - WARNING_TF_SCOPE_OP_NOT_CONSUMED: Operation (onet/Max) not consumed by converter: Max.
2018-05-30 08:51:24,213 - 365 - WARNING - WARNING_TF_SCOPE_OP_NOT_CONSUMED: Operation (onet/sub) not consumed by converter: Sub.
2018-05-30 08:51:24,213 - 365 - WARNING - WARNING_TF_SCOPE_OP_NOT_CONSUMED: Operation (onet/Exp) not consumed by converter: Exp.
2018-05-30 08:51:24,213 - 365 - WARNING - WARNING_TF_SCOPE_OP_NOT_CONSUMED: Operation (onet/prelu1/Relu_1) not consumed by converter: Relu.
2018-05-30 08:51:24,213 - 365 - WARNING - WARNING_TF_SCOPE_OP_NOT_CONSUMED: Operation (onet/Sum) not consumed by converter: Sum.
2018-05-30 08:51:24,213 - 365 - WARNING - WARNING_TF_SCOPE_OP_NOT_CONSUMED: Operation (onet/conv6-1/conv6-1/MatMul) not consumed by converter: MatMul.
2018-05-30 08:51:24,213 - 365 - WARNING - WARNING_TF_SCOPE_OP_NOT_CONSUMED: Operation (onet/prelu5/Neg_1) not consumed by converter: Neg.
2018-05-30 08:51:24,214 - 365 - WARNING - WARNING_TF_SCOPE_OP_NOT_CONSUMED: Operation (onet/prelu2/Neg) not consumed by converter: Neg.
2018-05-30 08:51:24,214 - 365 - WARNING - WARNING_TF_SCOPE_OP_NOT_CONSUMED: Operation (onet/prelu2/Relu_1) not consumed by converter: Relu.
2018-05-30 08:51:24,214 - 365 - WARNING - WARNING_TF_SCOPE_OP_NOT_CONSUMED: Operation (onet/prelu2/Neg_1) not consumed by converter: Neg.
2018-05-30 08:51:24,214 - 365 - WARNING - WARNING_TF_SCOPE_OP_NOT_CONSUMED: Operation (onet/prelu1/Neg) not consumed by converter: Neg.
2018-05-30 08:51:24,214 - 365 - WARNING - WARNING_TF_SCOPE_OP_NOT_CONSUMED: Operation (onet/prelu3/Relu_1) not consumed by converter: Relu.
2018-05-30 08:51:24,214 - 365 - WARNING - WARNING_TF_SCOPE_OP_NOT_CONSUMED: Operation (onet/prelu3/Neg) not consumed by converter: Neg.
2018-05-30 08:51:24,214 - 365 - WARNING - WARNING_TF_SCOPE_OP_NOT_CONSUMED: Operation (onet/prelu1/Neg_1) not consumed by converter: Neg.
2018-05-30 08:51:24,214 - 365 - WARNING - WARNING_TF_SCOPE_OP_NOT_CONSUMED: Operation (onet/prelu4/Relu_1) not consumed by converter: Relu.
2018-05-30 08:51:24,214 - 365 - WARNING - WARNING_TF_SCOPE_OP_NOT_CONSUMED: Operation (onet/prob1) not consumed by converter: RealDiv.
2018-05-30 08:51:24,214 - 365 - WARNING - WARNING_TF_SCOPE_OP_NOT_CONSUMED: Operation (onet/prelu4/Neg) not consumed by converter: Neg.
2018-05-30 08:51:24,214 - 365 - WARNING - WARNING_TF_SCOPE_OP_NOT_CONSUMED: Operation (onet/prelu5/Relu_1) not consumed by converter: Relu.
2018-05-30 08:51:24,214 - 365 - WARNING - WARNING_TF_SCOPE_OP_NOT_CONSUMED: Operation (onet/prelu3/Neg_1) not consumed by converter: Neg.
2018-05-30 08:51:24,214 - 365 - WARNING - WARNING_TF_SCOPE_OP_NOT_CONSUMED: Operation (onet/prelu5/Neg) not consumed by converter: Neg.
2018-05-30 08:51:24,214 - 123 - ERROR - Conversion failed: ERROR_TF_OPERATION_NOT_MAPPED_TO_LAYER: Some operations in the Tensorflow graph were not resolved to a layer.
Please clarify Is these layers are not supported by SNPE or something that I missed that gave me wrong output?
Thanks in Advance,
Kshama Vete
Hi
You can go through the below link which lists out all the layers and activation functions supported by SNPE,
https://developer.qualcomm.com/docs/snpe/network_layers.html
Using the below command you can get the DLC model file excluding the unsupported layers. Check if the model works as expected,
snpe-tensorflow-to-dlc --graph=onet_landmark.pb --input_dim "onet/input" 48,48,3 --out_node onet/prob1 --out_node onet/conv6-2/conv6-2 --out_node onet/conv6-3/conv6-3 --dlc=onet_landmark.dlc --verbose --allow_unconsumed_nodes
To give more details for your error, kindly provide us the architecture of the model you are working on.