diff --git a/train.py b/train.py index 75b109ed..4fa925ef 100644 --- a/train.py +++ b/train.py @@ -399,7 +399,7 @@ def evaluate(model, criterion, criterion_st, ap, current_step, epoch): test_audios = {} test_figures = {} print(" | > Synthesizing test sentences") - speaker_id = 0 if c.num_speakers > 1 else None + speaker_id = 0 if c.use_speaker_embedding else None for idx, test_sentence in enumerate(test_sentences): try: wav, alignment, decoder_output, postnet_output, stop_tokens = synthesis( diff --git a/utils/synthesis.py b/utils/synthesis.py index 78ae365e..c782b4c1 100644 --- a/utils/synthesis.py +++ b/utils/synthesis.py @@ -109,7 +109,7 @@ def synthesis(model, inputs = text_to_seqvec(text, CONFIG, use_cuda) speaker_id = id_to_torch(speaker_id) if use_cuda: - speaker_id.cuda() + speaker_id = speaker_id.cuda() # synthesize voice decoder_output, postnet_output, alignments, stop_tokens = run_model( model, inputs, CONFIG, truncated, speaker_id, style_mel)