0e12fe12ea2f001ea3e4bf4d6cb5d6c6cab2a8cb,open_seq2seq/models/text2speech.py,Text2Speech,finalize_inference,#Text2Speech#Any#Any#,365

Before Change

          if "magnitude" in self.get_data_layer().params["output_type"]:
            predicted_final_spectrogram_sample = self.get_data_layer().denormalize(predicted_final_spectrogram_sample)
            predicted_final_spectrogram_sample = np.exp(predicted_final_spectrogram_sample)
            save_audio(predicted_final_spectrogram_sample, self.params["logdir"], 0, mode="infer",number= i*batch_size+j)
          elif "mel" in self.get_data_layer().params["output_type"]:
            save_audio(predicted_final_spectrogram_sample, self.params["logdir"], 0, mode="infer",number= i*batch_size+j)
          // if "spectrogram" in self.get_data_layer().params["output_type"]:
          //   save_audio(predicted_final_spectrogram_sample, self.params["logdir"], 0, mode="infer", number=i*batch_size+j)

After Change

        // print(predicted_final_spectrogram_sample.shape)
        if audio_length > 2:
          predicted_final_spectrogram_sample = predicted_final_spectrogram_sample[:audio_length-1,:]
          predicted_final_spectrogram_sample = self.get_data_layer().get_magnitude_spec(predicted_final_spectrogram_sample)
          wav_summary = save_audio(predicted_final_spectrogram_sample, self.params["logdir"], 0, mode="infer", 
            number= i*batch_size+j, save_to_tensorboard = False)

Instances