I want to know the reason of the below error.
error_message=Layer parameters combination is invalid in GPU Layer MobilenetV2/expanded_conv_7/depthwise/depthwise: input depth must be divisible by 4 * number of groups; error_component=GPU Runtime;
I run snpe(1.24) on the snapdragon 855 and Qualcomm, Adreno (TM) 640.
If set up com.qualcomm.qti.snpe.NeuralNetwork.Runtime.GPU and dlc file's outputnode is argmax, I met the error!
But on com.qualcomm.qti.snpe.NeuralNetwork.Runtime.DSP, run well.
need your help!
I know argmax only running cpu.
Also, https://developer.qualcomm.com/docs/snpe/network_layers.html, this link declars that argmax is only used in cpu.
But, argmax is running on DSP.
I wonder that reason.
If you see the support from Neural Processing SDK for layer ArgMax of Tensorflow is provided only on CPU runtime.
Please check SNPE Docs for more information.
If you running it on GPU/DSP it will throw the error. Please make sure that you've set the GPU/DSP fallback to the CPU runtime in the code.
So whenever your code fails to run on GPU/DSP will switch to CPU automatically and complete the execution.