style update

TTS/bin/train_encoder.py

@@ -16,16 +16,11 @@ from TTS.speaker_encoder.model import SpeakerEncoder
 from TTS.speaker_encoder.utils.io import save_best_model, save_checkpoint
 from TTS.speaker_encoder.utils.visual import plot_embeddings
 from TTS.tts.datasets.preprocess import load_meta_data
+from TTS.utils.arguments import init_training
 from TTS.utils.audio import AudioProcessor
-from TTS.utils.generic_utils import (
+from TTS.utils.generic_utils import count_parameters, remove_experiment_folder, set_init_dict
-    count_parameters,
-    remove_experiment_folder,
-    set_init_dict,
-)
 from TTS.utils.radam import RAdam
 from TTS.utils.training import NoamLR, check_update
-from TTS.utils.arguments import init_training
-
 
 torch.backends.cudnn.enabled = True
 torch.backends.cudnn.benchmark = True

TTS/config/__init__.py

@@ -1,6 +1,7 @@
 import json
 import os
 import re
+
 import yaml
 
 from TTS.config.shared_configs import *
@@ -34,7 +35,7 @@ def _search_configs(model_name):
 
 def _process_model_name(config_dict):
     model_name = config_dict["model"] if "model" in config_dict else config_dict["generator_model"]
-    model_name = model_name.replace('_generator', '').replace('_discriminator', '')
+    model_name = model_name.replace("_generator", "").replace("_discriminator", "")
     return model_name
 
 

TTS/speaker_encoder/speaker_encoder_config.py

@@ -1,7 +1,9 @@
-from coqpit import MISSING
+from dataclasses import asdict, dataclass, field
-from dataclasses import dataclass, field, asdict
 from typing import List
-from TTS.config.shared_configs import BaseTrainingConfig, BaseAudioConfig, BaseDatasetConfig
+
+from coqpit import MISSING
+
+from TTS.config.shared_configs import BaseAudioConfig, BaseDatasetConfig, BaseTrainingConfig
 
 
 @dataclass
@@ -13,23 +15,27 @@ class SpeakerEncoderConfig(BaseTrainingConfig):
     datasets: List[BaseDatasetConfig] = field(default_factory=lambda: [BaseDatasetConfig()])
 
     # model params
-    model_params: dict = field(default_factory=lambda: {
+    model_params: dict = field(
-        "input_dim": 40,
+        default_factory=lambda: {
-        "proj_dim": 256,
+            "input_dim": 40,
-        "lstm_dim": 768,
+            "proj_dim": 256,
-        "num_lstm_layers": 3,
+            "lstm_dim": 768,
-        "use_lstm_with_projection": True
+            "num_lstm_layers": 3,
-    })
+            "use_lstm_with_projection": True,
+        }
+    )
 
-    storage: dict = field(default_factory=lambda:{
+    storage: dict = field(
-        "sample_from_storage_p": 0.66,  # the probability with which we'll sample from the DataSet in-memory storage
+        default_factory=lambda: {
-        "storage_size": 15,   # the size of the in-memory storage with respect to a single batch
+            "sample_from_storage_p": 0.66,  # the probability with which we'll sample from the DataSet in-memory storage
-        "additive_noise": 1e-5   # add very small gaussian noise to the data in order to increase robustness
+            "storage_size": 15,  # the size of the in-memory storage with respect to a single batch
-    })
+            "additive_noise": 1e-5,  # add very small gaussian noise to the data in order to increase robustness
+        }
+    )
 
     # training params
     max_train_step: int = 1000  # end training when number of training steps reaches this value.
-    loss: str = 'angleproto'
+    loss: str = "angleproto"
     grad_clip: float = 3.0
     lr: float = 0.0001
     lr_decay: bool = False
@@ -51,4 +57,6 @@ class SpeakerEncoderConfig(BaseTrainingConfig):
     def check_values(self):
         super().check_values()
         c = asdict(self)
-        assert c['model_params']['input_dim'] == self.audio.num_mels, " [!] model input dimendion must be equal to melspectrogram dimension."
+        assert (
+            c["model_params"]["input_dim"] == self.audio.num_mels
+        ), " [!] model input dimendion must be equal to melspectrogram dimension."

TTS/speaker_encoder/utils/generic_utils.py

@@ -9,6 +9,10 @@ def to_camel(text):
 
 
 def setup_model(c):
-    model = SpeakerEncoder(c.model_params["input_dim"], c.model_params["proj_dim"],
+    model = SpeakerEncoder(
-                           c.model_params["lstm_dim"], c.model_params["num_lstm_layers"])
+        c.model_params["input_dim"],
-    return model
+        c.model_params["proj_dim"],
+        c.model_params["lstm_dim"],
+        c.model_params["num_lstm_layers"],
+    )
+    return model

TTS/speaker_encoder/utils/io.py

@@ -1,5 +1,6 @@
-import os
 import datetime
+import os
+
 import torch
 
 
@@ -19,8 +20,7 @@ def save_checkpoint(model, optimizer, model_loss, out_path, current_step):
     torch.save(state, checkpoint_path)
 
 
-def save_best_model(model, optimizer, model_loss, best_loss, out_path,
+def save_best_model(model, optimizer, model_loss, best_loss, out_path, current_step):
-                    current_step):
     if model_loss < best_loss:
         new_state_dict = model.state_dict()
         state = {
@@ -33,7 +33,6 @@ def save_best_model(model, optimizer, model_loss, best_loss, out_path,
         best_loss = model_loss
         bestmodel_path = "best_model.pth.tar"
         bestmodel_path = os.path.join(out_path, bestmodel_path)
-        print("\n > BEST MODEL ({0:.5f}) : {1:}".format(
+        print("\n > BEST MODEL ({0:.5f}) : {1:}".format(model_loss, bestmodel_path))
-            model_loss, bestmodel_path))
         torch.save(state, bestmodel_path)
     return best_loss

TTS/tts/configs/shared_configs.py

@@ -41,7 +41,7 @@ class CharactersConfig(Coqpit):
     characters: str = None
     punctuations: str = None
     phonemes: str = None
-    unique: bool = True     # for backwards compatibility of models trained with char sets with duplicates
+    unique: bool = True  # for backwards compatibility of models trained with char sets with duplicates
 
     def check_values(
         self,

TTS/tts/utils/speakers.py

@@ -6,9 +6,9 @@ from typing import Union
 import numpy as np
 import torch
 
+from TTS.config import load_config
 from TTS.speaker_encoder.utils.generic_utils import setup_model
 from TTS.utils.audio import AudioProcessor
-from TTS.config import load_config
 
 
 def make_speakers_json_path(out_path):

TTS/utils/generic_utils.py

@@ -7,10 +7,11 @@ import re
 import shutil
 import subprocess
 import sys
-import torch
 from pathlib import Path
 from typing import List
 
+import torch
+
 
 def get_cuda():
     use_cuda = torch.cuda.is_available()

TTS/utils/manage.py

@@ -8,8 +8,8 @@ from shutil import copyfile
 import gdown
 import requests
 
-from TTS.utils.generic_utils import get_user_data_dir
 from TTS.config import load_config
+from TTS.utils.generic_utils import get_user_data_dir
 
 
 class ModelManager(object):

TTS/utils/synthesizer.py

@@ -5,6 +5,7 @@ import numpy as np
 import pysbd
 import torch
 
+from TTS.config import load_config
 from TTS.tts.utils.generic_utils import setup_model
 from TTS.tts.utils.speakers import SpeakerManager
 
@@ -13,7 +14,6 @@ from TTS.tts.utils.speakers import SpeakerManager
 from TTS.tts.utils.synthesis import synthesis, trim_silence
 from TTS.tts.utils.text import make_symbols, phonemes, symbols
 from TTS.utils.audio import AudioProcessor
-from TTS.config import load_config
 from TTS.vocoder.utils.generic_utils import interpolate_vocoder_input, setup_generator
 
 
@@ -117,7 +117,7 @@ class Synthesizer(object):
         self.use_phonemes = self.tts_config.use_phonemes
         self.ap = AudioProcessor(verbose=False, **self.tts_config.audio)
 
-        if self.tts_config.has('characters') and self.tts_config.characters:
+        if self.tts_config.has("characters") and self.tts_config.characters:
             symbols, phonemes = make_symbols(**self.tts_config.characters)
 
         if self.use_phonemes:

tests/test_audio.py

@@ -2,8 +2,8 @@ import os
 import unittest
 
 from tests import get_tests_input_path, get_tests_output_path, get_tests_path
-from TTS.utils.audio import AudioProcessor
 from TTS.config import BaseAudioConfig
+from TTS.utils.audio import AudioProcessor
 
 TESTS_PATH = get_tests_path()
 OUT_PATH = os.path.join(get_tests_output_path(), "audio_tests")

tests/test_glow_tts.py

@@ -6,10 +6,10 @@ import torch
 from torch import optim
 
 from tests import get_tests_input_path
+from TTS.tts.configs import GlowTTSConfig
 from TTS.tts.layers.losses import GlowTTSLoss
 from TTS.tts.models.glow_tts import GlowTTS
 from TTS.utils.audio import AudioProcessor
-from TTS.tts.configs import GlowTTSConfig
 
 # pylint: disable=unused-variable
 

tests/test_loader.py

@@ -7,10 +7,10 @@ import torch
 from torch.utils.data import DataLoader
 
 from tests import get_tests_input_path, get_tests_output_path
+from TTS.tts.configs import BaseTTSConfig
 from TTS.tts.datasets import TTSDataset
 from TTS.tts.datasets.preprocess import ljspeech
 from TTS.utils.audio import AudioProcessor
-from TTS.tts.configs import BaseTTSConfig
 
 # pylint: disable=unused-variable
 
@@ -18,7 +18,7 @@ OUTPATH = os.path.join(get_tests_output_path(), "loader_tests/")
 os.makedirs(OUTPATH, exist_ok=True)
 
 # create a dummy config for testing data loaders.
-c = BaseTTSConfig(text_cleaner='english_cleaners', num_loader_workers=0, batch_size=2)
+c = BaseTTSConfig(text_cleaner="english_cleaners", num_loader_workers=0, batch_size=2)
 c.r = 5
 c.data_path = "tests/data/ljspeech/"
 ok_ljspeech = os.path.exists(c.data_path)

tests/test_speaker_encoder_train.py

@@ -2,10 +2,9 @@ import glob
 import os
 import shutil
 
-from tests import get_tests_output_path, run_cli, get_device_id
+from tests import get_device_id, get_tests_output_path, run_cli
-
-from TTS.speaker_encoder.speaker_encoder_config import SpeakerEncoderConfig
 from TTS.config.shared_configs import BaseAudioConfig
+from TTS.speaker_encoder.speaker_encoder_config import SpeakerEncoderConfig
 
 config_path = os.path.join(get_tests_output_path(), "test_model_config.json")
 output_path = os.path.join(get_tests_output_path(), "train_outputs")
@@ -20,7 +19,7 @@ config = SpeakerEncoderConfig(
     print_step=1,
     save_step=1,
     print_eval=True,
-    audio=BaseAudioConfig(num_mels=40)
+    audio=BaseAudioConfig(num_mels=40),
 )
 config.audio.do_trim_silence = True
 config.audio.trim_db = 60
@@ -42,6 +41,8 @@ run_cli(command_train)
 continue_path = max(glob.glob(os.path.join(output_path, "*/")), key=os.path.getmtime)
 
 # restore the model and continue training for one more epoch
-command_train = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_encoder.py --continue_path {continue_path} "
+command_train = (
+    f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_encoder.py --continue_path {continue_path} "
+)
 run_cli(command_train)
 shutil.rmtree(continue_path)

tests/test_speaker_manager.py

@@ -5,10 +5,10 @@ import numpy as np
 import torch
 
 from tests import get_tests_input_path
-from TTS.speaker_encoder.model import SpeakerEncoder
-from TTS.speaker_encoder.utils.io import save_checkpoint
-from TTS.speaker_encoder.speaker_encoder_config import SpeakerEncoderConfig
 from TTS.config import load_config
+from TTS.speaker_encoder.model import SpeakerEncoder
+from TTS.speaker_encoder.speaker_encoder_config import SpeakerEncoderConfig
+from TTS.speaker_encoder.utils.io import save_checkpoint
 from TTS.tts.utils.speakers import SpeakerManager
 from TTS.utils.audio import AudioProcessor
 

tests/test_synthesize.py

@@ -1,7 +1,6 @@
 import os
 
-from tests import get_tests_output_path, run_cli, get_device_id
+from tests import get_device_id, get_tests_output_path, run_cli
-
 
 
 def test_synthesize():

tests/test_synthesizer.py

@@ -2,10 +2,10 @@ import os
 import unittest
 
 from tests import get_tests_input_path, get_tests_output_path
+from TTS.config import load_config
 from TTS.tts.utils.generic_utils import setup_model
 from TTS.tts.utils.io import save_checkpoint
 from TTS.tts.utils.text.symbols import make_symbols, phonemes, symbols
-from TTS.config import load_config
 from TTS.utils.synthesizer import Synthesizer
 
 
@@ -15,7 +15,7 @@ class SynthesizerTest(unittest.TestCase):
         # pylint: disable=global-statement
         global symbols, phonemes
         config = load_config(os.path.join(get_tests_output_path(), "dummy_model_config.json"))
-        if config.has('characters') and config.characters:
+        if config.has("characters") and config.characters:
             symbols, phonemes = make_symbols(**config.characters.to_dict())
 
         num_chars = len(phonemes) if config.use_phonemes else len(symbols)
@@ -26,8 +26,8 @@ class SynthesizerTest(unittest.TestCase):
     def test_in_out(self):
         self._create_random_model()
         tts_root_path = get_tests_output_path()
-        tts_checkpoint = os.path.join(tts_root_path, 'checkpoint_10.pth.tar')
+        tts_checkpoint = os.path.join(tts_root_path, "checkpoint_10.pth.tar")
-        tts_config = os.path.join(tts_root_path, 'dummy_model_config.json')
+        tts_config = os.path.join(tts_root_path, "dummy_model_config.json")
         synthesizer = Synthesizer(tts_checkpoint, tts_config, None, None)
         synthesizer.tts("Better this test works!!")
 

tests/test_tacotron2_model.py

@@ -6,10 +6,10 @@ import torch
 from torch import nn, optim
 
 from tests import get_tests_input_path
+from TTS.tts.configs import Tacotron2Config
 from TTS.tts.layers.losses import MSELossMasked
 from TTS.tts.models.tacotron2 import Tacotron2
 from TTS.utils.audio import AudioProcessor
-from TTS.tts.configs import Tacotron2Config
 
 # pylint: disable=unused-variable
 
@@ -148,13 +148,7 @@ class TacotronGSTTrainTest(unittest.TestCase):
 
         criterion = MSELossMasked(seq_len_norm=False).to(device)
         criterion_st = nn.BCEWithLogitsLoss().to(device)
-        model = Tacotron2(
+        model = Tacotron2(num_chars=24, r=c.r, num_speakers=5, use_gst=True, gst=c.gst).to(device)
-            num_chars=24,
-            r=c.r,
-            num_speakers=5,
-            use_gst=True,
-            gst=c.gst
-        ).to(device)
         model.train()
         model_ref = copy.deepcopy(model)
         count = 0
@@ -210,13 +204,7 @@ class TacotronGSTTrainTest(unittest.TestCase):
 
         criterion = MSELossMasked(seq_len_norm=False).to(device)
         criterion_st = nn.BCEWithLogitsLoss().to(device)
-        model = Tacotron2(
+        model = Tacotron2(num_chars=24, r=c.r, num_speakers=5, use_gst=True, gst=c.gst).to(device)
-            num_chars=24,
-            r=c.r,
-            num_speakers=5,
-            use_gst=True,
-            gst =c.gst
-        ).to(device)
         model.train()
         model_ref = copy.deepcopy(model)
         count = 0
@@ -271,14 +259,9 @@ class SCGSTMultiSpeakeTacotronTrainTest(unittest.TestCase):
         stop_targets = (stop_targets.sum(2) > 0.0).unsqueeze(2).float().squeeze()
         criterion = MSELossMasked(seq_len_norm=False).to(device)
         criterion_st = nn.BCEWithLogitsLoss().to(device)
-        model = Tacotron2(
+        model = Tacotron2(num_chars=24, r=c.r, num_speakers=5, speaker_embedding_dim=55, use_gst=True, gst=c.gst).to(
-            num_chars=24,
+            device
-            r=c.r,
+        )
-            num_speakers=5,
-            speaker_embedding_dim=55,
-            use_gst=True,
-            gst=c.gst
-        ).to(device)
         model.train()
         model_ref = copy.deepcopy(model)
         count = 0

tests/test_tacotron2_tf_model.py

@@ -5,10 +5,9 @@ import numpy as np
 import tensorflow as tf
 import torch
 
+from TTS.tts.configs import Tacotron2Config
 from TTS.tts.tf.models.tacotron2 import Tacotron2
 from TTS.tts.tf.utils.tflite import convert_tacotron2_to_tflite, load_tflite_model
-from TTS.tts.configs import Tacotron2Config
-
 
 tf.get_logger().setLevel("INFO")
 

tests/test_tacotron_model.py

@@ -6,11 +6,10 @@ import torch
 from torch import nn, optim
 
 from tests import get_tests_input_path
+from TTS.tts.configs import TacotronConfig
 from TTS.tts.layers.losses import L1LossMasked
 from TTS.tts.models.tacotron import Tacotron
 from TTS.utils.audio import AudioProcessor
-from TTS.tts.configs import TacotronConfig
-
 
 # pylint: disable=unused-variable
 

tests/test_text_processing.py

@@ -4,9 +4,8 @@ import os
 # pylint: disable=wildcard-import
 # pylint: disable=unused-import
 from tests import get_tests_input_path, get_tests_path
-from TTS.tts.utils.text import *
 from TTS.tts.configs import TacotronConfig
-
+from TTS.tts.utils.text import *
 
 conf = TacotronConfig()
 

tests/tts_tests/test_align_tts_train.py

@@ -2,8 +2,7 @@ import glob
 import os
 import shutil
 
-from tests import get_tests_output_path, run_cli, get_device_id
+from tests import get_device_id, get_tests_output_path, run_cli
-
 from TTS.tts.configs import AlignTTSConfig
 
 config_path = os.path.join(get_tests_output_path(), "test_model_config.json")
@@ -44,6 +43,8 @@ run_cli(command_train)
 continue_path = max(glob.glob(os.path.join(output_path, "*/")), key=os.path.getmtime)
 
 # restore the model and continue training for one more epoch
-command_train = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_align_tts.py --continue_path {continue_path} "
+command_train = (
+    f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_align_tts.py --continue_path {continue_path} "
+)
 run_cli(command_train)
 shutil.rmtree(continue_path)

tests/tts_tests/test_glow_tts_train.py

@@ -2,8 +2,7 @@ import glob
 import os
 import shutil
 
-from tests import get_tests_output_path, run_cli, get_device_id
+from tests import get_device_id, get_tests_output_path, run_cli
-
 from TTS.tts.configs import GlowTTSConfig
 
 config_path = os.path.join(get_tests_output_path(), "test_model_config.json")
@@ -45,6 +44,8 @@ run_cli(command_train)
 continue_path = max(glob.glob(os.path.join(output_path, "*/")), key=os.path.getmtime)
 
 # restore the model and continue training for one more epoch
-command_train = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_glow_tts.py --continue_path {continue_path} "
+command_train = (
+    f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_glow_tts.py --continue_path {continue_path} "
+)
 run_cli(command_train)
 shutil.rmtree(continue_path)

tests/tts_tests/test_speedy_speech_train.py

@@ -2,8 +2,7 @@ import glob
 import os
 import shutil
 
-from tests import get_tests_output_path, run_cli, get_device_id
+from tests import get_device_id, get_tests_output_path, run_cli
-
 from TTS.tts.configs import SpeedySpeechConfig
 
 config_path = os.path.join(get_tests_output_path(), "test_speedy_speech_config.json")
@@ -45,6 +44,8 @@ run_cli(command_train)
 continue_path = max(glob.glob(os.path.join(output_path, "*/")), key=os.path.getmtime)
 
 # restore the model and continue training for one more epoch
-command_train = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_speedy_speech.py --continue_path {continue_path} "
+command_train = (
+    f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_speedy_speech.py --continue_path {continue_path} "
+)
 run_cli(command_train)
 shutil.rmtree(continue_path)

tests/tts_tests/test_tacotron2_train.py

@@ -2,8 +2,7 @@ import glob
 import os
 import shutil
 
-from tests import get_tests_output_path, run_cli, get_device_id
+from tests import get_device_id, get_tests_output_path, run_cli
-
 from TTS.tts.configs import Tacotron2Config
 
 config_path = os.path.join(get_tests_output_path(), "test_model_config.json")
@@ -45,6 +44,8 @@ run_cli(command_train)
 continue_path = max(glob.glob(os.path.join(output_path, "*/")), key=os.path.getmtime)
 
 # restore the model and continue training for one more epoch
-command_train = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_tacotron.py --continue_path {continue_path} "
+command_train = (
+    f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_tacotron.py --continue_path {continue_path} "
+)
 run_cli(command_train)
 shutil.rmtree(continue_path)

tests/tts_tests/test_tacotron_train.py

@@ -2,8 +2,7 @@ import glob
 import os
 import shutil
 
-from tests import get_tests_output_path, run_cli, get_device_id
+from tests import get_device_id, get_tests_output_path, run_cli
-
 from TTS.tts.configs import TacotronConfig
 
 config_path = os.path.join(get_tests_output_path(), "test_model_config.json")
@@ -44,6 +43,8 @@ run_cli(command_train)
 continue_path = max(glob.glob(os.path.join(output_path, "*/")), key=os.path.getmtime)
 
 # restore the model and continue training for one more epoch
-command_train = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_tacotron.py --continue_path {continue_path} "
+command_train = (
+    f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_tacotron.py --continue_path {continue_path} "
+)
 run_cli(command_train)
 shutil.rmtree(continue_path)

tests/vocoder_tests/test_fullband_melgan_train.py

@@ -2,8 +2,7 @@ import glob
 import os
 import shutil
 
-from tests import get_tests_output_path, run_cli, get_device_id
+from tests import get_device_id, get_tests_output_path, run_cli
-
 from TTS.vocoder.configs import FullbandMelganConfig
 
 config_path = os.path.join(get_tests_output_path(), "test_vocoder_config.json")
@@ -29,13 +28,17 @@ config.audio.trim_db = 60
 config.save_json(config_path)
 
 # train the model for one epoch
-command_train = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_vocoder_gan.py --config_path {config_path} "
+command_train = (
+    f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_vocoder_gan.py --config_path {config_path} "
+)
 run_cli(command_train)
 
 # Find latest folder
 continue_path = max(glob.glob(os.path.join(output_path, "*/")), key=os.path.getmtime)
 
 # restore the model and continue training for one more epoch
-command_train = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_vocoder_gan.py --continue_path {continue_path} "
+command_train = (
+    f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_vocoder_gan.py --continue_path {continue_path} "
+)
 run_cli(command_train)
 shutil.rmtree(continue_path)

tests/vocoder_tests/test_hifigan_train.py

@@ -2,8 +2,7 @@ import glob
 import os
 import shutil
 
-from tests import get_tests_output_path, run_cli, get_device_id
+from tests import get_device_id, get_tests_output_path, run_cli
-
 from TTS.vocoder.configs import HifiganConfig
 
 config_path = os.path.join(get_tests_output_path(), "test_vocoder_config.json")
@@ -30,13 +29,17 @@ config.audio.trim_db = 60
 config.save_json(config_path)
 
 # train the model for one epoch
-command_train = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_vocoder_gan.py --config_path {config_path} "
+command_train = (
+    f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_vocoder_gan.py --config_path {config_path} "
+)
 run_cli(command_train)
 
 # Find latest folder
 continue_path = max(glob.glob(os.path.join(output_path, "*/")), key=os.path.getmtime)
 
 # restore the model and continue training for one more epoch
-command_train = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_vocoder_gan.py --continue_path {continue_path} "
+command_train = (
+    f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_vocoder_gan.py --continue_path {continue_path} "
+)
 run_cli(command_train)
 shutil.rmtree(continue_path)

tests/vocoder_tests/test_melgan_train.py

@@ -2,10 +2,9 @@ import glob
 import os
 import shutil
 
-from tests import get_tests_output_path, run_cli, get_device_id
+from tests import get_device_id, get_tests_output_path, run_cli
 from TTS.vocoder.configs import MelganConfig
 
-
 config_path = os.path.join(get_tests_output_path(), "test_vocoder_config.json")
 output_path = os.path.join(get_tests_output_path(), "train_outputs")
 
@@ -29,13 +28,17 @@ config.audio.trim_db = 60
 config.save_json(config_path)
 
 # train the model for one epoch
-command_train = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_vocoder_gan.py --config_path {config_path} "
+command_train = (
+    f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_vocoder_gan.py --config_path {config_path} "
+)
 run_cli(command_train)
 
 # Find latest folder
 continue_path = max(glob.glob(os.path.join(output_path, "*/")), key=os.path.getmtime)
 
 # restore the model and continue training for one more epoch
-command_train = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_vocoder_gan.py --continue_path {continue_path} "
+command_train = (
+    f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_vocoder_gan.py --continue_path {continue_path} "
+)
 run_cli(command_train)
 shutil.rmtree(continue_path)

tests/vocoder_tests/test_multiband_melgan_train.py

@@ -2,8 +2,7 @@ import glob
 import os
 import shutil
 
-from tests import get_tests_output_path, run_cli, get_device_id
+from tests import get_device_id, get_tests_output_path, run_cli
-
 from TTS.vocoder.configs import MultibandMelganConfig
 
 config_path = os.path.join(get_tests_output_path(), "test_vocoder_config.json")
@@ -29,13 +28,17 @@ config.audio.trim_db = 60
 config.save_json(config_path)
 
 # train the model for one epoch
-command_train = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_vocoder_gan.py --config_path {config_path} "
+command_train = (
+    f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_vocoder_gan.py --config_path {config_path} "
+)
 run_cli(command_train)
 
 # Find latest folder
 continue_path = max(glob.glob(os.path.join(output_path, "*/")), key=os.path.getmtime)
 
 # restore the model and continue training for one more epoch
-command_train = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_vocoder_gan.py --continue_path {continue_path} "
+command_train = (
+    f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_vocoder_gan.py --continue_path {continue_path} "
+)
 run_cli(command_train)
 shutil.rmtree(continue_path)

tests/vocoder_tests/test_parallel_wavegan_train.py

@@ -2,8 +2,7 @@ import glob
 import os
 import shutil
 
-from tests import get_tests_output_path, run_cli, get_device_id
+from tests import get_device_id, get_tests_output_path, run_cli
-
 from TTS.vocoder.configs import ParallelWaveganConfig
 
 config_path = os.path.join(get_tests_output_path(), "test_vocoder_config.json")
@@ -29,13 +28,17 @@ config.audio.trim_db = 60
 config.save_json(config_path)
 
 # train the model for one epoch
-command_train = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_vocoder_gan.py --config_path {config_path} "
+command_train = (
+    f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_vocoder_gan.py --config_path {config_path} "
+)
 run_cli(command_train)
 
 # Find latest folder
 continue_path = max(glob.glob(os.path.join(output_path, "*/")), key=os.path.getmtime)
 
 # restore the model and continue training for one more epoch
-command_train = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_vocoder_gan.py --continue_path {continue_path} "
+command_train = (
+    f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_vocoder_gan.py --continue_path {continue_path} "
+)
 run_cli(command_train)
 shutil.rmtree(continue_path)

tests/vocoder_tests/test_vocoder_gan_datasets.py

@@ -5,9 +5,9 @@ from torch.utils.data import DataLoader
 
 from tests import get_tests_output_path, get_tests_path
 from TTS.utils.audio import AudioProcessor
+from TTS.vocoder.configs import BaseGANVocoderConfig
 from TTS.vocoder.datasets.gan_dataset import GANDataset
 from TTS.vocoder.datasets.preprocess import load_wav_data
-from TTS.vocoder.configs import BaseGANVocoderConfig
 
 file_path = os.path.dirname(os.path.realpath(__file__))
 OUTPATH = os.path.join(get_tests_output_path(), "loader_tests/")

tests/vocoder_tests/test_vocoder_losses.py

@@ -3,8 +3,8 @@ import os
 import torch
 
 from tests import get_tests_input_path, get_tests_output_path, get_tests_path
-from TTS.utils.audio import AudioProcessor
 from TTS.config import BaseAudioConfig
+from TTS.utils.audio import AudioProcessor
 from TTS.vocoder.layers.losses import MelganFeatureLoss, MultiScaleSTFTLoss, STFTLoss, TorchSTFT
 
 TESTS_PATH = get_tests_path()

tests/vocoder_tests/test_wavegrad_train.py

@@ -2,42 +2,40 @@ import glob
 import os
 import shutil
 
-from tests import get_tests_output_path, run_cli, get_device_id
+from tests import get_device_id, get_tests_output_path, run_cli
-
 from TTS.vocoder.configs import WavegradConfig
 
 config_path = os.path.join(get_tests_output_path(), "test_vocoder_config.json")
 output_path = os.path.join(get_tests_output_path(), "train_outputs")
 
-config = WavegradConfig(batch_size=8,
+config = WavegradConfig(
-                        eval_batch_size=8,
+    batch_size=8,
-                        num_loader_workers=0,
+    eval_batch_size=8,
-                        num_val_loader_workers=0,
+    num_loader_workers=0,
-                        run_eval=True,
+    num_val_loader_workers=0,
-                        test_delay_epochs=-1,
+    run_eval=True,
-                        epochs=1,
+    test_delay_epochs=-1,
-                        seq_len=8192,
+    epochs=1,
-                        eval_split_size=1,
+    seq_len=8192,
-                        print_step=1,
+    eval_split_size=1,
-                        print_eval=True,
+    print_step=1,
-                        data_path="tests/data/ljspeech",
+    print_eval=True,
-                        output_path=output_path,
+    data_path="tests/data/ljspeech",
-                        test_noise_schedule={
+    output_path=output_path,
-                            "min_val": 1e-6,
+    test_noise_schedule={"min_val": 1e-6, "max_val": 1e-2, "num_steps": 2},
-                            "max_val": 1e-2,
+)
-                            "num_steps": 2
-                        })
 config.audio.do_trim_silence = True
 config.audio.trim_db = 60
 config.save_json(config_path)
 
 # train the model for one epoch
-command_train = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_vocoder_wavegrad.py --config_path {config_path} "
+command_train = (
+    f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_vocoder_wavegrad.py --config_path {config_path} "
+)
 run_cli(command_train)
 
 # Find latest folder
-continue_path = max(glob.glob(os.path.join(output_path, "*/")),
+continue_path = max(glob.glob(os.path.join(output_path, "*/")), key=os.path.getmtime)
-                    key=os.path.getmtime)
 
 # restore the model and continue training for one more epoch
 command_train = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_vocoder_wavegrad.py --continue_path {continue_path} "

tests/vocoder_tests/test_wavernn_train.py

@@ -2,8 +2,7 @@ import glob
 import os
 import shutil
 
-from tests import get_tests_output_path, run_cli, get_device_id
+from tests import get_device_id, get_tests_output_path, run_cli
-
 from TTS.vocoder.configs import WavernnConfig
 
 config_path = os.path.join(get_tests_output_path(), "test_vocoder_config.json")
@@ -29,14 +28,17 @@ config.audio.trim_db = 60
 config.save_json(config_path)
 
 # train the model for one epoch
-command_train = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_vocoder_wavernn.py --config_path {config_path} "
+command_train = (
+    f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_vocoder_wavernn.py --config_path {config_path} "
+)
 run_cli(command_train)
 
 # Find latest folder
-continue_path = max(glob.glob(os.path.join(output_path, "*/")),
+continue_path = max(glob.glob(os.path.join(output_path, "*/")), key=os.path.getmtime)
-                    key=os.path.getmtime)
 
 # restore the model and continue training for one more epoch
-command_train = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_vocoder_wavernn.py --continue_path {continue_path} "
+command_train = (
+    f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_vocoder_wavernn.py --continue_path {continue_path} "
+)
 run_cli(command_train)
 shutil.rmtree(continue_path)