:stuck_out_tongue_closed_eyes: TensorFlowTTS: Real-Time State-of-the-art Speech Synthesis for Tensorflow 2 (supported including English, French, Korean, Chinese, German and Easy to adapt for other languages)
In the CPP tflite example it says the inference function need delete Dropout before converting tflite model.
I have opened the colab after cloning the repository in tensorflow_tts/models/fastspeech2.py I changed training to false on both lines but the runtime crashes, I have also tried deleting the dropout from here like the example describes but it still needs tf.lite.OpsSet.SELECT_TF_OPS
@twister11211 the last time i tried this model on tensorflow lite was successfully (just need to remove dropout layer). Maybe there is a bug come from tflite itself rather than the code from this repo.
In the CPP tflite example it says the inference function need delete Dropout before converting tflite model. I have opened the colab after cloning the repository in tensorflow_tts/models/fastspeech2.py I changed training to false on both lines but the runtime crashes, I have also tried deleting the dropout from here like the example describes but it still needs tf.lite.OpsSet.SELECT_TF_OPS
f0_embedding = self.f0_dropout( self.f0_embeddings(tf.expand_dims(f0_outputs, 2)), training=False ) energy_embedding = self.energy_dropout( self.energy_embeddings(tf.expand_dims(energy_outputs, 2)), training=False )
f0_embedding = self.f0_embeddings(tf.expand_dims(f0_outputs, 2)) energy_embedding = self.energy_embeddings(tf.expand_dims(energy_outputs, 2))
Is this the correct way to do this or do I have it wrong