Hello,
I’ve converted a tensorflow model to a tensorflow lite model and I’ve specified these supported operators:
converter.target_spec.supported_ops = [
tf.lite.OpsSet.TFLITE_BUILTINS,
tf.lite.OpsSet.SELECT_TF_OPS
]
I’m trying to use it on Unity but whenever I execute it, this error pops up.
I’ve read that in order to make this work I would have to rebuild the library but I’m having trouble with it. Is there any other way to fix this error?
Hi @lleirell
Please provide some info on how you are trying to use it? Are you adding a dependency somewhere for TensorFlow Lite Interpreter?
For example inside an android project the common case is to use:
implementation "org.tensorflow:tensorflow-lite:2.9.0"
and when you have special operators that the native library cannot support you have to add the dependency:
implementation 'org.tensorflow:tensorflow-lite-select-tf-ops:2.9.0'
So how are you importing the library?
Regards
Hi @George_Soloupis ,
Thanks for your response. I added the packages through the package manager in Unity. In the manifest.json file I have these two lines:
"dependencies": {
...
"com.github.asus4.tflite": "2.10.0",
"com.github.asus4.tflite.common": "2.10.0"
}
I tried to add a dependency related to select-tf-ops writing it in the manifest.json but it didn’t work.
HI @lleirell
I saw this issue. I think Unity does not natively support this and you should open an issue at the github repo to get some specific answers.
Regards
one other option is to try to change the model so you don’t need selected ops. I know this can be hard but it’s a possibility
Do you know which ops are you missing?