[TFLITE][frontend]Support for new operators for tflite frontend

Development
#1

Hi, this is my second post on this community.

I would like to use TVM to optimise a tensorflow model that uses tensorflow 2.2 operators that are currently not being supported by the tensorflow frontend.

I then convert the tensorflow model to tensorflow lite model and tried to use the tflite frontend:

mod, params = relay.frontend.from_tflite(tflite_model,
                                         shape_dict={input_tensor: input_shape},
                                         dtype_dict={input_tensor: input_dtype})

However, I got the following error information:

OpNotImplemented: The following operators are not supported in frontend TFLite: 'NON_MAX_SUPPRESSION_V5'

I looked into the source code of tflite and it seems to a bit learning curve on implement ‘NON_MAX_SUPPRESSION_V5’. I wonder if there is a plan to implement this or any advce on how I can implement and test it by myself. Finally I would like to contribute it back to the community.

Cheers.

1 Like
#2

Hi @cam145 have you found a workaround for this issue? I am currently experiencing the same, trying to import a model from TFLite