mirror of
https://github.com/storytold/artcraft.git
synced 2026-11-04 17:38:10 +00:00
greatly simplify model config
This commit is contained in:
Brandon Thomas committed
1 parent
97d9490f2f
commit
36c1c0f007
3 files changed
+35
-502
No files matched your search
+32
-467
@@ -1,501 +1,66 @@
|
||||
# --------------- Speakers ---------------
|
||||
# --------------- Experimental Speakers ---------------
|
||||
[[speakers]]
|
||||
name = "Glow LJS (Single Speaker)"
|
||||
slug = "glow-ljs-single"
|
||||
name = "Current Experiment"
|
||||
slug = "experiment"
|
||||
model_pipeline = "ArpabetGlowTtsMelgan"
|
||||
glow_tts = "glow_tts_single_speaker.torchjit"
|
||||
#glow_tts = "glow-tts_solo_zuckerberg_ljstx_2020.06.24_expr-1_chkpt-2280.torchjit"
|
||||
#melgan = "melgan_trump_txlearn_2020-05-30_steno_chkpt9800.torchjit"
|
||||
melgan = "shared_melgan_container-2020.01.04.pt"
|
||||
#melgan = "melgan_multivoice-1.0_2020.06.15_12d5838_6725.torchjit"
|
||||
melgan = "melgan-vocodes.torchjit"
|
||||
|
||||
# --------------- Stable Speakers ---------------
|
||||
[[speakers]]
|
||||
name = "Donald Trump"
|
||||
slug = "trump"
|
||||
model_pipeline = "ArpabetTacotronMelgan"
|
||||
tacotron = "tacotron2_trump_txlearn_ljs_arpabet_2020-05-25_chkpt26500.jit"
|
||||
#melgan = "melgan_trump-txlearn-2020.05.05_13675.jit"
|
||||
melgan = "melgan_trump_txlearn_2020-05-30_steno_chkpt9800.torchjit"
|
||||
|
||||
[[speakers]]
|
||||
name = "Donald Trump (Glow-TTS experiments)"
|
||||
slug = "trump2"
|
||||
model_pipeline = "ArpabetGlowTtsMelgan"
|
||||
#glow_tts = "glow-experimental.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"
|
||||
#glow_tts = "glow_tts_ljs_txlearn_arpabet_2020-06-01_G_2590.torchjit"
|
||||
glow_tts = "glow_tts_ljs_txlearn_arpabet_2020-06-01_G_3614.torchjit"
|
||||
melgan = "melgan_trump_txlearn_2020-05-30_steno_chkpt9800.torchjit"
|
||||
|
||||
[[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"
|
||||
|
||||
[[speakers]]
|
||||
name = "John F. Kennedy"
|
||||
slug = "jfk"
|
||||
model_pipeline = "ArpabetGlowTtsMultiSpeakerMelgan"
|
||||
glow_tts_multi_speaker = "glow_multispeaker.torchjit"
|
||||
#melgan = "melgan_trump_txlearn_2020-05-30_steno_chkpt9800.torchjit"
|
||||
melgan = "melgan_multivoice-1.0_2020.06.15_12d5838_6725.torchjit" # TODO: Wtf - noise
|
||||
speaker_id = 1
|
||||
|
||||
[[speakers]]
|
||||
name = "Arnold Schwarzenegger"
|
||||
slug = "arnold"
|
||||
model_pipeline = "ArpabetGlowTtsMultiSpeakerMelgan"
|
||||
glow_tts_multi_speaker = "glow_multispeaker.torchjit"
|
||||
#melgan = "melgan_trump_txlearn_2020-05-30_steno_chkpt9800.torchjit"
|
||||
melgan = "melgan_multivoice-1.0_2020.06.15_12d5838_6725.torchjit"
|
||||
speaker_id = 2
|
||||
|
||||
[[speakers]]
|
||||
name = "Bill Gates"
|
||||
slug = "gates"
|
||||
model_pipeline = "ArpabetGlowTtsMultiSpeakerMelgan"
|
||||
glow_tts_multi_speaker = "glow_multispeaker.torchjit"
|
||||
#melgan = "melgan_trump_txlearn_2020-05-30_steno_chkpt9800.torchjit"
|
||||
melgan = "melgan_multivoice-1.0_2020.06.15_12d5838_6725.torchjit"
|
||||
speaker_id = 3
|
||||
|
||||
[[speakers]]
|
||||
name = "Mark Zuckerberg"
|
||||
slug = "zuckerberg"
|
||||
model_pipeline = "ArpabetGlowTtsMultiSpeakerMelgan"
|
||||
glow_tts_multi_speaker = "glow_multispeaker.torchjit"
|
||||
#melgan = "melgan_trump_txlearn_2020-05-30_steno_chkpt9800.torchjit"
|
||||
melgan = "melgan_multivoice-1.0_2020.06.15_12d5838_6725.torchjit"
|
||||
speaker_id = 4
|
||||
glow_tts = "glow_tts_ljs_txlearn_arpabet_2020-06-01_G_1689-2.torchjit"
|
||||
|
||||
[[speakers]]
|
||||
name = "Ronald Reagan"
|
||||
slug = "reagan"
|
||||
model_pipeline = "ArpabetGlowTtsMultiSpeakerMelgan"
|
||||
glow_tts_multi_speaker = "glow_multispeaker.torchjit"
|
||||
#melgan = "melgan_trump_txlearn_2020-05-30_steno_chkpt9800.torchjit"
|
||||
melgan = "melgan_multivoice-1.0_2020.06.15_12d5838_6725.torchjit"
|
||||
speaker_id = 5
|
||||
slug = "glow-solo-reagan"
|
||||
model_pipeline = "ArpabetGlowTtsMelgan"
|
||||
glow_tts = "glow-tts_solo_reagan_ljstx_2020.06.22_expr-1_chkpt-1406.torchjit"
|
||||
melgan = "melgan_vocodes1.0_multispeaker_2020.06.19_12d5838_experiment-2_chkpt-10400.torchjit"
|
||||
|
||||
[[speakers]]
|
||||
name = "Queen Elizabeth II"
|
||||
slug = "queen-elizabeth"
|
||||
model_pipeline = "ArpabetGlowTtsMultiSpeakerMelgan"
|
||||
glow_tts_multi_speaker = "glow_multispeaker.torchjit"
|
||||
#melgan = "melgan_trump_txlearn_2020-05-30_steno_chkpt9800.torchjit"
|
||||
melgan = "melgan_multivoice-1.0_2020.06.15_12d5838_6725.torchjit"
|
||||
speaker_id = 6
|
||||
name = "John Oliver"
|
||||
slug = "glow-solo-oliver"
|
||||
model_pipeline = "ArpabetGlowTtsMelgan"
|
||||
glow_tts = "glow-tts_solo_oliver_ljstx_2020.06.22_expr-1_chkpt-1189.torchjit"
|
||||
melgan = "melgan_vocodes1.0_multispeaker_2020.06.19_12d5838_experiment-2_chkpt-10400.torchjit"
|
||||
|
||||
[[speakers]]
|
||||
name = "Mario"
|
||||
slug = "mario"
|
||||
model_pipeline = "ArpabetGlowTtsMultiSpeakerMelgan"
|
||||
glow_tts_multi_speaker = "glow_multispeaker.torchjit"
|
||||
#melgan = "melgan_trump_txlearn_2020-05-30_steno_chkpt9800.torchjit"
|
||||
melgan = "melgan_multivoice-1.0_2020.06.15_12d5838_6725.torchjit"
|
||||
speaker_id = 7
|
||||
name = "Bill Gates"
|
||||
slug = "glow-solo-billgates"
|
||||
model_pipeline = "ArpabetGlowTtsMelgan"
|
||||
glow_tts = "glow-tts_solo_billgates_ljstx_2020.06.24_expr-1_chkpt-2011.torchjit"
|
||||
melgan = "melgan_vocodes1.0_multispeaker_2020.06.19_12d5838_experiment-2_chkpt-10400.torchjit"
|
||||
|
||||
[[speakers]]
|
||||
name = "Zapp Brannigan"
|
||||
slug = "zapp"
|
||||
model_pipeline = "ArpabetGlowTtsMultiSpeakerMelgan"
|
||||
glow_tts_multi_speaker = "glow_multispeaker.torchjit"
|
||||
#melgan = "melgan_trump_txlearn_2020-05-30_steno_chkpt9800.torchjit"
|
||||
melgan = "melgan_multivoice-1.0_2020.06.15_12d5838_6725.torchjit"
|
||||
speaker_id = 8
|
||||
name = "Mark Zuckerberg"
|
||||
slug = "glow-solo-zuckerberg"
|
||||
model_pipeline = "ArpabetGlowTtsMelgan"
|
||||
glow_tts = "glow-tts_solo_zuckerberg_ljstx_2020.06.24_expr-1_chkpt-2280.torchjit"
|
||||
melgan = "melgan_vocodes1.0_multispeaker_2020.06.19_12d5838_experiment-2_chkpt-10400.torchjit"
|
||||
|
||||
[[speakers]]
|
||||
name = "Morgan Freeman"
|
||||
slug = "morgan"
|
||||
name = "Arnold Schwarzenegger (Multi-speaker model)"
|
||||
slug = "glow-multi-arnold"
|
||||
model_pipeline = "ArpabetGlowTtsMultiSpeakerMelgan"
|
||||
glow_tts_multi_speaker = "glow_multispeaker.torchjit"
|
||||
#melgan = "melgan_trump_txlearn_2020-05-30_steno_chkpt9800.torchjit"
|
||||
#melgan = "shared_melgan_container-2020.01.05.pt"
|
||||
#melgan = "melgan_ljs-high-pitch_container.pt"
|
||||
melgan = "melgan_multivoice-1.0_2020.06.15_12d5838_6725.torchjit"
|
||||
speaker_id = 9
|
||||
|
||||
[[speakers]]
|
||||
name = "LJS Multi Base 1"
|
||||
slug = "multi1"
|
||||
model_pipeline = "ArpabetGlowTtsMultiSpeakerMelgan"
|
||||
glow_tts_multi_speaker = "glow_ljs_base.torchjit"
|
||||
melgan = "melgan_trump_txlearn_2020-05-30_steno_chkpt9800.torchjit"
|
||||
speaker_id = 1
|
||||
|
||||
[[speakers]]
|
||||
name = "LJS Multi Base 2"
|
||||
slug = "multi1"
|
||||
model_pipeline = "ArpabetGlowTtsMultiSpeakerMelgan"
|
||||
glow_tts_multi_speaker = "glow_ljs_base.torchjit"
|
||||
melgan = "melgan_trump_txlearn_2020-05-30_steno_chkpt9800.torchjit"
|
||||
speaker_id = 2
|
||||
|
||||
[[speakers]]
|
||||
name = "LJS Multi Base 50"
|
||||
slug = "multi1"
|
||||
model_pipeline = "ArpabetGlowTtsMultiSpeakerMelgan"
|
||||
glow_tts_multi_speaker = "glow_ljs_base.torchjit"
|
||||
melgan = "melgan_trump_txlearn_2020-05-30_steno_chkpt9800.torchjit"
|
||||
speaker_id = 50
|
||||
|
||||
[[speakers]]
|
||||
slug = "multi_dense1_g44"
|
||||
name = "LJS Dense Multi Base 1 (G44)"
|
||||
model_pipeline = "ArpabetGlowTtsMultiSpeakerMelgan"
|
||||
glow_tts_multi_speaker = "glow_ljs_base_dense.torchjit"
|
||||
melgan = "melgan_trump_txlearn_2020-05-30_steno_chkpt9800.torchjit"
|
||||
speaker_id = 1
|
||||
|
||||
[[speakers]]
|
||||
slug = "multi_dense2_g44"
|
||||
name = "LJS Dense Multi Base 2 (G44)"
|
||||
model_pipeline = "ArpabetGlowTtsMultiSpeakerMelgan"
|
||||
glow_tts_multi_speaker = "glow_ljs_base_dense.torchjit"
|
||||
melgan = "melgan_trump_txlearn_2020-05-30_steno_chkpt9800.torchjit"
|
||||
speaker_id = 2
|
||||
|
||||
[[speakers]]
|
||||
slug = "multi_dense1_g142"
|
||||
name = "LJS Dense Multi Base 1 (G142)"
|
||||
model_pipeline = "ArpabetGlowTtsMultiSpeakerMelgan"
|
||||
glow_tts_multi_speaker = "glow_ljs_base_dense2.torchjit"
|
||||
melgan = "melgan_trump_txlearn_2020-05-30_steno_chkpt9800.torchjit"
|
||||
speaker_id = 1
|
||||
|
||||
[[speakers]]
|
||||
slug = "multi_dense2_g142"
|
||||
name = "LJS Dense Multi Base 2 (G142)"
|
||||
model_pipeline = "ArpabetGlowTtsMultiSpeakerMelgan"
|
||||
glow_tts_multi_speaker = "glow_ljs_base_dense2.torchjit"
|
||||
melgan = "melgan_trump_txlearn_2020-05-30_steno_chkpt9800.torchjit"
|
||||
speaker_id = 2
|
||||
|
||||
[[speakers]]
|
||||
slug = "multi_dense1_g264"
|
||||
name = "LJS Dense Multi Base 1 (G264)"
|
||||
model_pipeline = "ArpabetGlowTtsMultiSpeakerMelgan"
|
||||
glow_tts_multi_speaker = "glow_ljs_base_dense3.torchjit"
|
||||
melgan = "melgan_trump_txlearn_2020-05-30_steno_chkpt9800.torchjit"
|
||||
speaker_id = 1
|
||||
|
||||
[[speakers]]
|
||||
slug = "multi_dense2_g264"
|
||||
name = "LJS Dense Multi Base 2 (G264)"
|
||||
model_pipeline = "ArpabetGlowTtsMultiSpeakerMelgan"
|
||||
glow_tts_multi_speaker = "glow_ljs_base_dense3.torchjit"
|
||||
melgan = "melgan_trump_txlearn_2020-05-30_steno_chkpt9800.torchjit"
|
||||
speaker_id = 2
|
||||
|
||||
# -------
|
||||
|
||||
[[speakers]]
|
||||
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_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"
|
||||
#melgan = "melgan_trump_txlearn_2020-05-30_steno_chkpt9800.torchjit"
|
||||
melgan = "melgan-vocodes.torchjit"
|
||||
speaker_id = 0
|
||||
|
||||
[[speakers]]
|
||||
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_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"
|
||||
melgan = "melgan-vocodes.torchjit"
|
||||
speaker_id = 1
|
||||
|
||||
[[speakers]]
|
||||
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_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"
|
||||
melgan = "melgan-vocodes.torchjit"
|
||||
speaker_id = 2
|
||||
|
||||
[[speakers]]
|
||||
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_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"
|
||||
melgan = "melgan-vocodes.torchjit"
|
||||
melgan = "melgan_vocodes1.0_multispeaker_2020.06.19_12d5838_experiment-2_chkpt-10400.torchjit"
|
||||
speaker_id = 3
|
||||
|
||||
[[speakers]]
|
||||
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_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"
|
||||
melgan = "melgan-vocodes.torchjit"
|
||||
speaker_id = 4
|
||||
|
||||
# --------------- Model configs ---------------
|
||||
|
||||
[[model_locations]]
|
||||
model_type = "ArpabetTacotron"
|
||||
base_directory = "/home/bt/dev/voder/tacotron_melgan"
|
||||
|
||||
[[model_locations]]
|
||||
model_type = "ArpabetGlowTts" # TODO: This is not the Arpabet I'm using
|
||||
#base_directory = "/home/bt/dev/jit-models/output/glow-tts-jaywalnut"
|
||||
base_directory = "/home/bt/dev/jit-models/glow-tts-jaywalnut"
|
||||
model_type = "ArpabetGlowTts"
|
||||
base_directory = "/home/bt/dev/jit-models/glow-tts-jaywalnut/OUTPUT"
|
||||
|
||||
[[model_locations]]
|
||||
model_type = "Melgan"
|
||||
base_directory = "/home/bt/dev/tacotron-melgan/"
|
||||
base_directory = "/home/bt/dev/jit-models/melgan-swpark/OUTPUT"
|
||||
|
||||
[[model_locations]]
|
||||
model_type = "ArpabetGlowTtsMultiSpeaker"
|
||||
base_directory = "/home/bt/dev/jit-models/glow-tts-jaywalnut-multispeaker"
|
||||
|
||||
# --------------- 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 = "acotron2_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)"
|
||||
|
||||
[[tacotron]]
|
||||
# This is the best version of the Trump Tacotron network so far.
|
||||
# Later iterations lose quality. You can see it in the training graph.
|
||||
# It looks like ~28.62k is the lowest, and ~29.14k is the last before the
|
||||
# first of two major peaks upwards.
|
||||
file_path = "tacotron2_trump_txlearn_ljs_arpabet_2020-05-25_chkpt26500.jit"
|
||||
description = "BEST TACOTRON FOR TRUMP THUS FAR!"
|
||||
|
||||
[[tacotron]]
|
||||
# Oh my god, this suddenly lost the ability to speak words.
|
||||
# This is a serious loss of quality.
|
||||
file_path = "tacotron2_trump_txlearn_ljs_arpabet_2020-05-25_chkpt40000.jit"
|
||||
description = "(see previous)"
|
||||
|
||||
[[tacotron]]
|
||||
# The syllables are back, but the words are inexact.
|
||||
# It sounds like Trump.
|
||||
# Step 26500 still sounds much better than this.
|
||||
file_path = "tacotron2_trump_txlearn_ljs_arpabet_2020-05-25_chkpt50000.jit"
|
||||
description = "(see previous)"
|
||||
|
||||
# ======== OLDER RUNS ======
|
||||
|
||||
[[tacotron]]
|
||||
# OLD NAME, perhaps I can get the epoch: file_path = "/home/bt/dev/voder/tacotron_melgan/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 = "tacotron_container.pt"
|
||||
description = "Unknown (broken)"
|
||||
|
||||
[[tacotron]]
|
||||
# TODO: Remove. This hangs forever.
|
||||
file_path = "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); modified with longer input to JIT"
|
||||
|
||||
[[glow_tts]]
|
||||
file_path = "glow_tts_ljs_txlearn_arpabet_2020-06-01_G_2590.torchjit"
|
||||
description = "First experiments (continued); modified with longer input to JIT"
|
||||
|
||||
[[glow_tts]]
|
||||
file_path = "glow_tts_ljs_txlearn_arpabet_2020-06-01_G_3614.torchjit"
|
||||
description = "First experiments (continued); modified with longer input to JIT"
|
||||
|
||||
[[glow_tts]]
|
||||
file_path = "glow_tts_single_speaker.torchjit"
|
||||
description = "Single speaker training again (fixed arpabet; 2020-06-22)"
|
||||
|
||||
[[glow_tts]]
|
||||
file_path = "glow-tts_solo_zuckerberg_ljstx_2020.06.24_expr-1_chkpt-2280.torchjit"
|
||||
description = ""
|
||||
|
||||
|
||||
# ---------------- Glow TTS Multi-speaker ----------------
|
||||
[[glow_tts_multi_speaker]]
|
||||
file_path = "glow_multispeaker.torchjit"
|
||||
description = "Mutli-speaker garbage"
|
||||
|
||||
[[glow_tts_multi_speaker]]
|
||||
file_path = "glow_ljs_base.torchjit"
|
||||
description = "Mutli-speaker base model (LambdaLabs expensive Tesla GPU)"
|
||||
|
||||
[[glow_tts_multi_speaker]]
|
||||
file_path = "glow_ljs_base_dense.torchjit"
|
||||
description = "Denser multi-speaker model. 8 speakers (train on a Tesla GPU)"
|
||||
|
||||
[[glow_tts_multi_speaker]]
|
||||
file_path = "glow_ljs_base_dense2.torchjit"
|
||||
description = "Denser multi-speaker model. 8 speakers (train on a Tesla GPU)"
|
||||
|
||||
[[glow_tts_multi_speaker]]
|
||||
file_path = "glow_ljs_base_dense3.torchjit"
|
||||
description = "Denser multi-speaker model. 8 speakers (train on a Tesla GPU)"
|
||||
|
||||
[[glow_tts_multi_speaker]]
|
||||
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
|
||||
|
||||
[[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"
|
||||
|
||||
[[melgan]]
|
||||
file_path = "melgan_multivoice-1.0_2020.06.15_12d5838_6275.torchjit"
|
||||
description = "first multivoice (txlearned)"
|
||||
|
||||
[[melgan]]
|
||||
file_path = "melgan_multivoice-1.0_2020.06.15_12d5838_6300.torchjit"
|
||||
description = "first multivoice (txlearned)"
|
||||
|
||||
[[melgan]]
|
||||
file_path = "melgan_multivoice-1.0_2020.06.15_12d5838_6475.torchjit"
|
||||
description = "first multivoice (txlearned)"
|
||||
|
||||
[[melgan]]
|
||||
file_path = "melgan_multivoice-1.0_2020.06.15_12d5838_6725.torchjit"
|
||||
description = "first multivoice (txlearned)"
|
||||
|
||||
[[melgan]]
|
||||
file_path = "vocodes1.0_multispeaker_2020_05_19_12d5838_6675.torchjit"
|
||||
description = "second multivoice; more non-trump samples (txlearned)"
|
||||
|
||||
[[melgan]]
|
||||
file_path = "vocodes1.0_multispeaker_2020_05_19_12d5838_6900.torchjit"
|
||||
description = "second multivoice; more non-trump samples (txlearned)"
|
||||
|
||||
[[melgan]]
|
||||
file_path = "vocodes1.0_multispeaker_2020_05_19_12d5838_7125.torchjit"
|
||||
description = "second multivoice; more non-trump samples (txlearned)"
|
||||
|
||||
[[melgan]]
|
||||
file_path = "vocodes1.0_multispeaker_2020_05_19_12d5838_7400.torchjit"
|
||||
description = "second multivoice; more non-trump samples (txlearned)"
|
||||
|
||||
[[melgan]]
|
||||
file_path = "melgan-vocodes.torchjit" # Stop having to create entries while training ongoing
|
||||
description = "second multivoice; more non-trump samples (txlearned)"
|
||||
base_directory = "/home/bt/dev/jit-models/glow-tts-jaywalnut-multispeaker/OUTPUT"
|
||||
|
||||
@@ -197,6 +197,7 @@ async fn run_server(app_state: AppState, server_args: ServerArgs) -> std::io::Re
|
||||
.wrap(Cors::new()
|
||||
.allowed_origin("http://localhost:12345")
|
||||
.allowed_origin("http://localhost:3000")
|
||||
.allowed_origin("http://localhost:5555")
|
||||
.allowed_origin("http://localhost:8080")
|
||||
.allowed_origin("http://localhost:8000")
|
||||
.allowed_origin("http://localhost:7000")
|
||||
|
||||
@@ -6,10 +6,6 @@ use std::path::Path;
|
||||
pub struct ModelConfigs {
|
||||
pub speakers: Vec<Speaker>,
|
||||
pub model_locations: Vec<ModelLocation>,
|
||||
tacotron: Vec<ModelDetails>,
|
||||
glow_tts: Vec<ModelDetails>,
|
||||
glow_tts_multi_speaker: Vec<ModelDetails>,
|
||||
melgan: Vec<ModelDetails>,
|
||||
}
|
||||
|
||||
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq)]
|
||||
@@ -60,41 +56,12 @@ pub enum ModelPipeline {
|
||||
ArpabetGlowTtsMultiSpeakerMelgan,
|
||||
}
|
||||
|
||||
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq)]
|
||||
pub struct ModelDetails {
|
||||
pub file_path: String,
|
||||
pub description: String,
|
||||
|
||||
/// Just the file name without a path
|
||||
#[serde(skip_deserializing)]
|
||||
pub base_name: Option<String>,
|
||||
}
|
||||
|
||||
impl ModelConfigs {
|
||||
pub fn load_from_file(filename: &str) -> Self {
|
||||
let contents = fs::read_to_string(filename)
|
||||
.expect("Couldn't read file");
|
||||
let mut model_configs : ModelConfigs = toml::from_str(&contents)
|
||||
.expect("Couldn't parse toml");
|
||||
|
||||
model_configs.update_base_names();
|
||||
model_configs
|
||||
}
|
||||
|
||||
fn update_base_names(&mut self) {
|
||||
for mut details in self.tacotron.iter_mut() {
|
||||
let path = Path::new(&details.file_path);
|
||||
details.base_name = path.file_name()
|
||||
.and_then(|name| name.to_str())
|
||||
.map(|s| s.to_string());
|
||||
}
|
||||
|
||||
for mut details in self.melgan.iter_mut() {
|
||||
let path = Path::new(&details.file_path);
|
||||
details.base_name = path.file_name()
|
||||
.and_then(|name| name.to_str())
|
||||
.map(|s| s.to_string());
|
||||
}
|
||||
toml::from_str(&contents)
|
||||
.expect("Couldn't parse toml")
|
||||
}
|
||||
|
||||
pub fn find_speaker_by_slug(&self, slug: &str) -> Option<&Speaker> {
|
||||
|
||||
Reference in new issue
Block a user