I am working on converting the model to the .rknn format, but I get a problem of converting the 'Frame_position_embedding:Any' layer. I need to find an alternative layer that will be supported when converting to .rknn format. Maybe someone has encountered a similar situation or knows about possible options for replacing this layer? I will be grateful for any help or advice!
Found a "tensorflow patcher" on github, that replacing other unsupported layers in rknn format (FusedBatchNormV3 and AddV2)