Hi,
I have a problem about the * or mul operation in snpe.
I conduct attention block in my neural network that require * operation, i.e. input [24,4,60,60] (channel,batch,h,w) * attention [24,4,1,1].
When I convert the pytorch onnx to the snpe platform, it seems have problem about the mul operation.
Is there any idea to solve this?
Many thanks.
Did you solve the issue?
Could you kindly provide more details about the architecture that you were using? I'm porting a LLM on-device atm :)
Thanks!