VeriSilicon/tflite-vx-delegate

Conv3d operation support

michaelnguyen11 opened this issue · 2 comments

michaelnguyen11 commented

Dear supporters,

First of all, I would like to thank you for your great work. With TFLite VX Delegate, our model could run faster than 10-15%, compare to TFLite NNAPI Delegate on i.MX 8MP.

Now I'm trying to deploy a model which use Conv3d, but it's failed to prepare because of Conv3d.

root@imx8mpevk:~/demo# 
ERROR: Encountered unresolved custom op: Conv3D.
ERROR: Node number 446 (Conv3D) failed to prepare.
Failed to allocate tensors

Therefore, I would like you ask you a question : Do you have plan to delivery Conv3d operation in the near future ? and when will we can use that operation ?

I'm looking forward to hearing from you. Many thanks in advance.

Regards,
Hiep

sunshinemyson commented

@michaelnguyen11 ,

Yes, we have plan to support Conv3D. Actually, we already kick-off evaluation internally. We planned it in the early of 2022 Q1.

michaelnguyen11 commented

Thanks for the information. Hopefully I can use Conv3D with VX Delegate soon.