From 1f4ec804b6b0cf8345f96397ff4ea7dc21245010 Mon Sep 17 00:00:00 2001 From: Eren Golge Date: Sat, 21 Sep 2019 09:58:58 +0200 Subject: [PATCH] compute and add style tokens in gst --- .compute | 1 + models/tacotrongst.py | 2 +- 2 files changed, 2 insertions(+), 1 deletion(-) diff --git a/.compute b/.compute index 3e009cae..24578189 100644 --- a/.compute +++ b/.compute @@ -13,4 +13,5 @@ python3 setup.py develop # python3 distribute.py --config_path config.json --data_path /data/ro/shared/data/keithito/LJSpeech-1.1/ # cp -R ${USER_DIR}/Mozilla_22050 ../tmp/ # python3 distribute.py --config_path config_tacotron_gst.json --data_path ../tmp/Mozilla_22050/ +python3 distribute.py --config_path config.json --data_path /data/rw/home/LibriTTS/train-clean-360 while true; do sleep 1000000; done diff --git a/models/tacotrongst.py b/models/tacotrongst.py index 0a9a7385..9819ec53 100644 --- a/models/tacotrongst.py +++ b/models/tacotrongst.py @@ -40,7 +40,7 @@ class TacotronGST(Tacotron): location_attn, separate_stopnet) gst_embedding_dim = 256 - decoder_dim = 512 + gst_embedding_dim if num_speakers > 1 else 256 + gst_embedding_dim + decoder_dim = 512 if num_speakers > 1 else 256 proj_speaker_dim = 80 if num_speakers > 1 else 0 self.decoder = Decoder(decoder_dim, mel_dim, r, memory_size, attn_win, attn_norm, prenet_type, prenet_dropout,