search

Kyubyong / dc_tts

A TensorFlow Implementation of DC-TTS: yet another text-to-speech model
Apache License 2.0
1.16k stars 370 forks source link

Rewriting Tensorflow 1.x code to tensorflow 2.x #95

Closed neuronx1 closed 3 years ago

neuronx1 commented 3 years ago

@Kyubyong Thanks for the great GitHub repository! Unfortunatly my GPU only supports tensorflow 2.x and not 1.x . For that reason I wanted to rewrite the Code to tensorflow 2.x . Everything worked fine until line 122 in data_load.py I haven't found a solution how to rewrite the following function: tf.contrib.training.bucket_by_sequence_length() I am glad for any idea how to solve this problem.

dmmagdal commented 3 years ago

I've been working on that implementation in Tensorflow 2. I have my current progress public on my Github. At the moment, models have been trained for about 7-800K steps on the LJ Speech dataset but I'm yet to produce any discernable speech samples. If you think you can improve upon the existing code, I more than welcome the contribution.

Only disclaimer is that I am in no way related to Kyubyong. I've never met the guy. I just like this repo and the project he's worked on with this.

neuronx1 commented 3 years ago

@dmmagdal : Thank you very much!