TF Lite Custom Ops for ResizeArea and Cross op

General Discussion
,
#1

Hi,

I am trying to convert a Tensorflow SavedModel to TF-Lite. However, while converting I get an error due to unsupported ops: ResizeArea and Cross. Hence, the solution was to write custom ops for these two. I tried to see how it is implemented in TF Core, but I am unable to understand how to write the correct implementation of the kernels for these two in TFLite. If someone can provide a tutorial for how to write and register it… I tried to follow the one on Tensorflow website, but it’s not detailed and I am stuck.

Thanks!

#4

Please refer to the custom operators guide explaining the end-to-end example of running a model with a custom operator tf.sin, which is supported in TensorFlow, but unsupported in TensorFlow Lite.

Thank you!

#6

I also have this same question and i cannot find any proper answers on the internet and also here. need solution.

#7

@Amin_Jabari,

Have you checked the end to end example to create and register Custom operators  |  TensorFlow Lite ?

If you have any blockers kindly share the details here.

Thank you!