Hi. We are trying to run a UNET type model on the NPU. The NNAPI delegate for TFLite rejects the model for lack of transpose convolutions. Is that a limitation in the NNAPI delegate or does the NPU not support transpose convolutions?
WARNING: Operator TRANSPOSE_CONV (v3) refused by NNAPI delegate: OP Version different from 1
Hi Christian,
We need to look deeper into this, but NNAPI delegate support is maintained by NXP, It would be possible to have some limitations, you can contact NXP support for more details.
https://community.nxp.com/t5/eIQ-Machine-Learning-Software/bd-p/eiq
If still cannot find help, please let us know. we will try to look into this
