From 8004011d29f9ecb5bffcdc388f131ac3418cd136 Mon Sep 17 00:00:00 2001 From: Brandon Thomas Date: Mon, 22 Jun 2020 05:26:23 -0400 Subject: [PATCH] change model toml files --- tts_service/deployed_models.toml | 165 ++++++++----------------------- tts_service/models.toml | 19 +++- 2 files changed, 57 insertions(+), 127 deletions(-) diff --git a/tts_service/deployed_models.toml b/tts_service/deployed_models.toml index f3d67430be..8c71195896 100644 --- a/tts_service/deployed_models.toml +++ b/tts_service/deployed_models.toml @@ -1,28 +1,54 @@ -# --------------- Speakers --------------- +# --------------- Speakers: Trumped.com --------------- [[speakers]] name = "Donald Trump" slug = "trump" -#model_pipeline = "ArpabetTacotronMelgan" -#tacotron = "tacotron2_trump_txlearn_ljs_arpabet_2020-05-25_chkpt26500.jit" model_pipeline = "ArpabetGlowTtsMelgan" melgan = "melgan_trump_txlearn_2020-05-30_steno_chkpt9800.torchjit" -#glow_tts = "glow_tts_ljs_txlearn_arpabet_2020-06-01_G_1007.torchjit" -#glow_tts = "glow_tts_ljs_txlearn_arpabet_2020-06-01_G_1689.torchjit" glow_tts = "glow_tts_ljs_txlearn_arpabet_2020-06-01_G_1689-2.torchjit" +# --------------- Speakers: Vo.codes --------------- [[speakers]] -name = "High-pitched Linda Johnson" -slug = "ljsh" -model_pipeline = "ArpabetTacotronMelgan" -tacotron = "tacotron2_ljs_high_pitch_arpabet_2020-03-28.jit" -melgan = "melgan_ljs-high-pitch_container.pt" +name = "Donald Trump (Multi-speaker model)" +slug = "glow-multi-trump" +model_pipeline = "ArpabetGlowTtsMultiSpeakerMelgan" +glow_tts_multi_speaker = "glow_tts_multispeaker_8-speaker_256-gin_ljs_txlearn_2020.06.19_experiment-5_chkpt-7466.torchjit" +melgan = "melgan_vocodes1.0_multispeaker_2020.06.19_12d5838_experiment-2_chkpt-10400.torchjit" +speaker_id = 0 + +[[speakers]] +name = "John Oliver (Multi-speaker model)" +slug = "glow-multi-oliver" +model_pipeline = "ArpabetGlowTtsMultiSpeakerMelgan" +glow_tts_multi_speaker = "glow_tts_multispeaker_8-speaker_256-gin_ljs_txlearn_2020.06.19_experiment-5_chkpt-7466.torchjit" +melgan = "melgan_vocodes1.0_multispeaker_2020.06.19_12d5838_experiment-2_chkpt-10400.torchjit" +speaker_id = 1 + +[[speakers]] +name = "Bill Gates (Multi-speaker model)" +slug = "glow-multi-gates" +model_pipeline = "ArpabetGlowTtsMultiSpeakerMelgan" +glow_tts_multi_speaker = "glow_tts_multispeaker_8-speaker_256-gin_ljs_txlearn_2020.06.19_experiment-5_chkpt-7466.torchjit" +melgan = "melgan_vocodes1.0_multispeaker_2020.06.19_12d5838_experiment-2_chkpt-10400.torchjit" +speaker_id = 2 + +[[speakers]] +name = "Arnold Schwarzenegger (Multi-speaker model)" +slug = "glow-multi-arnold" +model_pipeline = "ArpabetGlowTtsMultiSpeakerMelgan" +glow_tts_multi_speaker = "glow_tts_multispeaker_8-speaker_256-gin_ljs_txlearn_2020.06.19_experiment-5_chkpt-7466.torchjit" +melgan = "melgan_vocodes1.0_multispeaker_2020.06.19_12d5838_experiment-2_chkpt-10400.torchjit" +speaker_id = 3 + +[[speakers]] +name = "Ronald Reagan (Multi-speaker model)" +slug = "glow-multi-reagan" +model_pipeline = "ArpabetGlowTtsMultiSpeakerMelgan" +glow_tts_multi_speaker = "glow_tts_multispeaker_8-speaker_256-gin_ljs_txlearn_2020.06.19_experiment-5_chkpt-7466.torchjit" +melgan = "melgan_vocodes1.0_multispeaker_2020.06.19_12d5838_experiment-2_chkpt-10400.torchjit" +speaker_id = 4 # --------------- Model configs --------------- -[[model_locations]] -model_type = "ArpabetTacotron" -base_directory = "/data/downloads/downloaded/models/tacotron2-nvidia/" - [[model_locations]] model_type = "ArpabetGlowTts" base_directory = "/data/downloads/downloaded/models/glow-tts-jaywalnut/" @@ -31,124 +57,19 @@ base_directory = "/data/downloads/downloaded/models/glow-tts-jaywalnut/" model_type = "Melgan" base_directory = "/data/downloads/downloaded/models/melgan-swpark/" -# --------------- Tacotron --------------- - -# TODO: Get a high-quality LJS Tacotron - -[[tacotron]] -file_path = "tacotron2_trump_txlearn_ljs_arpabet_2020-02-06_ckpt30000.jit" -description = "First Trump txlearn with Arpabet. Trained 30k iters and sounds great!" - -[[tacotron]] -file_path = "tacotron2_trump_txlearn_ljs_arpabet_2020-05-14_ckpt7000.jit" -description = "Trump txlearn with Arpabet 2020-05-14" - -[[tacotron]] -file_path = "tacotron2_trump_txlearn_ljs_arpabet_2020-05-14_ckpt11500.jit" -description = "(see previous)" - -[[tacotron]] -file_path = "tacotron2_trump_txlearn_ljs_arpabet_2020-05-14_ckpt17000.jit" -description = "(see previous)" - -[[tacotron]] -file_path = "tacotron2_trump_txlearn_ljs_arpabet_2020-05-14_ckpt27000.jit" -description = "(see previous)" - -[[tacotron]] -file_path = "tacotron2_trump_txlearn_ljs_arpabet_2020-05-14_ckpt27000-18000.jit" -description = "(see previous)" - -[[tacotron]] -file_path = "tacotron2_trump_txlearn_ljs_arpabet_2020-05-14_ckpt53500.jit" -description = "(see previous)" - -[[tacotron]] -file_path = "tacotron2_trump_txlearn_ljs_arpabet_2020-05-14_chkpt112500.jit" -description = "(see previous)" - -[[tacotron]] -file_path = "tacotron2_trump_txlearn_ljs_arpabet_2020-05-23_chkpt8000.jit" -description = "(see previous)" - -[[tacotron]] -file_path = "tacotron2_trump_txlearn_ljs_arpabet_2020-05-23_chkpt13000.jit" -description = "(see previous)" - -[[tacotron]] -file_path = "tacotron2_trump_txlearn_ljs_arpabet_2020-05-24_chkpt2000.jit" -description = "Trump txlearn with Arpabet 2020-05-24; adds 400 new training files from steno" - -[[tacotron]] -file_path = "tacotron2_trump_txlearn_ljs_arpabet_2020-05-24_chkpt16000.jit" -description = "(see previous)" - -[[tacotron]] -file_path = "tacotron2_trump_txlearn_ljs_arpabet_2020-05-25_chkpt7000.jit" -description = "Trump txlearn with Arpabet 2020-05-25; adds lots from steno, disabled one bad quality source" - -[[tacotron]] -file_path = "tacotron2_trump_txlearn_ljs_arpabet_2020-05-25_chkpt17000.jit" -description = "(see previous)" - - -# ======== OLDER RUNS ====== - -[[tacotron]] -# OLD NAME, perhaps I can get the epoch: file_path = "/data/models/tacotron2-nvidia/tacotron_arpabet_1.pt" -file_path = "tacotron2_ljs_high_pitch_arpabet_2020-03-28.jit" -description = "High-quality high-pitched LJS; arpabet encodings." - -# [[tacotron]] -# # TODO: Remove. This blows up. -# file_path = "/data/models/tacotron2-nvidia/tacotron_container.pt" -# description = "Unknown (broken)" -# -# [[tacotron]] -# # TODO: Remove. This hangs forever. -# file_path = "/data/models/tacotron2-nvidia/tacotron_jit_model_voder_c0cac635.pt" -# description = "Unknown (broken)" - # ---------------- Glow TTS---------------- -[[glow_tts]] -file_path = "glow_tts_ljs_txlearn_arpabet_2020-06-01_G_1007.torchjit" -description = "First experiments" - -[[glow_tts]] -file_path = "glow_tts_ljs_txlearn_arpabet_2020-06-01_G_1689.torchjit" -description = "First experiments (continued)" - [[glow_tts]] file_path = "glow_tts_ljs_txlearn_arpabet_2020-06-01_G_1689-2.torchjit" -description = "First experiments (continued); longer JIT tracing input" +description = "First experiments in glow-tts (continued); the '-2' is longer JIT tracing input" # ---------------- Melgan ---------------- -# TODO: Get a high-quality LJS Melgan - -[[melgan]] -file_path = "melgan_trump-txlearn-2020.05.05_7525.jit" -description = "Latest LJS transfer learned Trump." - -[[melgan]] -file_path = "melgan_trump-txlearn-2020.05.05_13675.jit" -description = "Latest LJS transfer learned Trump." - [[melgan]] file_path = "melgan_trump_txlearn_2020-05-30_steno_chkpt9800.torchjit" description = "Trump steno data" [[melgan]] -file_path = "melgan_ljs-high-pitch_container.pt" -description = "High-pitched LJS" - -[[melgan]] -file_path = "shared_melgan_container-2020.01.04.pt" -description = "Unknown" - -[[melgan]] -file_path = "shared_melgan_container-2020.01.05.pt" -description = "Unknown" - +file_path = "melgan_vocodes1.0_multispeaker_2020.06.19_12d5838_experiment-2_chkpt-10400.torchjit" +description = "First round of multi-speaker experiments for glow-tts" diff --git a/tts_service/models.toml b/tts_service/models.toml index 260ad46dce..47e61f5f2c 100644 --- a/tts_service/models.toml +++ b/tts_service/models.toml @@ -187,7 +187,8 @@ slug = "multi_speaker_G309_0" name = "LJS Dense Multi Speaker Donald Trump (G309)" model_pipeline = "ArpabetGlowTtsMultiSpeakerMelgan" #glow_tts_multi_speaker = "glow_ljs_base_dense3.torchjit" -glow_tts_multi_speaker = "glow_ljs_multispeaker_1.torchjit" +#glow_tts_multi_speaker = "glow_ljs_multispeaker_1.torchjit" +glow_tts_multi_speaker = "glow_tts_multispeaker_8-speaker_256-gin_ljs_txlearn_2020.06.19_experiment-5_chkpt-7466.torchjit" #melgan = "melgan_trump_txlearn_2020-05-30_steno_chkpt9800.torchjit" #melgan = "melgan_multivoice-1.0_2020.06.15_12d5838_6725.torchjit" #melgan = "melgan_vocodes1.0_multispeaker_2020_06_19_12d5838_7125.torchjit" @@ -200,7 +201,8 @@ slug = "multi_speaker_G309_1" name = "LJS Dense Multi Speaker John Oliver (G309)" model_pipeline = "ArpabetGlowTtsMultiSpeakerMelgan" #glow_tts_multi_speaker = "glow_ljs_base_dense3.torchjit" -glow_tts_multi_speaker = "glow_ljs_multispeaker_1.torchjit" +#glow_tts_multi_speaker = "glow_ljs_multispeaker_1.torchjit" +glow_tts_multi_speaker = "glow_tts_multispeaker_8-speaker_256-gin_ljs_txlearn_2020.06.19_experiment-5_chkpt-7466.torchjit" #melgan = "melgan_trump_txlearn_2020-05-30_steno_chkpt9800.torchjit" #melgan = "melgan_multivoice-1.0_2020.06.15_12d5838_6725.torchjit" #melgan = "melgan_vocodes1.0_multispeaker_2020_06_19_12d5838_7400.torchjit" @@ -212,7 +214,8 @@ slug = "multi_speaker_G309_2" name = "LJS Dense Multi Speaker Bill Gates (G309)" model_pipeline = "ArpabetGlowTtsMultiSpeakerMelgan" #glow_tts_multi_speaker = "glow_ljs_base_dense3.torchjit" -glow_tts_multi_speaker = "glow_ljs_multispeaker_1.torchjit" +#glow_tts_multi_speaker = "glow_ljs_multispeaker_1.torchjit" +glow_tts_multi_speaker = "glow_tts_multispeaker_8-speaker_256-gin_ljs_txlearn_2020.06.19_experiment-5_chkpt-7466.torchjit" #melgan = "melgan_trump_txlearn_2020-05-30_steno_chkpt9800.torchjit" #melgan = "melgan_multivoice-1.0_2020.06.15_12d5838_6725.torchjit" #melgan = "melgan_vocodes1.0_multispeaker_2020_06_19_12d5838_7400.torchjit" @@ -224,7 +227,8 @@ slug = "multi_speaker_G309_3" name = "LJS Dense Multi Speaker Arnold (G309)" model_pipeline = "ArpabetGlowTtsMultiSpeakerMelgan" #glow_tts_multi_speaker = "glow_ljs_base_dense3.torchjit" -glow_tts_multi_speaker = "glow_ljs_multispeaker_1.torchjit" +#glow_tts_multi_speaker = "glow_ljs_multispeaker_1.torchjit" +glow_tts_multi_speaker = "glow_tts_multispeaker_8-speaker_256-gin_ljs_txlearn_2020.06.19_experiment-5_chkpt-7466.torchjit" #melgan = "melgan_trump_txlearn_2020-05-30_steno_chkpt9800.torchjit" #melgan = "melgan_multivoice-1.0_2020.06.15_12d5838_6725.torchjit" #melgan = "melgan_vocodes1.0_multispeaker_2020_06_19_12d5838_7400.torchjit" @@ -236,7 +240,8 @@ slug = "multi_speaker_G309_4" name = "LJS Dense Multi Speaker Reagan (G309)" model_pipeline = "ArpabetGlowTtsMultiSpeakerMelgan" #glow_tts_multi_speaker = "glow_ljs_base_dense3.torchjit" -glow_tts_multi_speaker = "glow_ljs_multispeaker_1.torchjit" +#glow_tts_multi_speaker = "glow_ljs_multispeaker_1.torchjit" +glow_tts_multi_speaker = "glow_tts_multispeaker_8-speaker_256-gin_ljs_txlearn_2020.06.19_experiment-5_chkpt-7466.torchjit" #melgan = "melgan_trump_txlearn_2020-05-30_steno_chkpt9800.torchjit" #melgan = "melgan_multivoice-1.0_2020.06.15_12d5838_6725.torchjit" #melgan = "melgan_vocodes1.0_multispeaker_2020_06_19_12d5838_7400.torchjit" @@ -408,6 +413,10 @@ description = "Denser multi-speaker model. 8 speakers (train on a Tesla GPU)" file_path = "glow_ljs_multispeaker_1.torchjit" description = "Multispeaker now..." +[[glow_tts_multi_speaker]] +file_path = "glow_tts_multispeaker_8-speaker_256-gin_ljs_txlearn_2020.06.19_experiment-5_chkpt-7466.torchjit" +description = "Multispeaker now... last checkpoint before going back to single-speaker for a bit" + # ---------------- Melgan ---------------- # TODO: Get a high-quality LJS Melgan