Refactor Project to 3 parts: Models, Control, Data

Need readme

.gitignore

@@ -14,8 +14,8 @@
 *.bcf
 *.toc
 *.sh
-*/saved_models
-!vocoder/saved_models/pretrained/**
-!encoder/saved_models/pretrained.pt
+data/ckpt
+!data/ckpt/vocoder/pretrained/**
+!data/ckpt/encoder/pretrained.pt
 wavs
 log

archived_untest_files/demo_cli.py

@@ -1,9 +1,9 @@
-from encoder.params_model import model_embedding_size as speaker_embedding_size
+from models.encoder.params_model import model_embedding_size as speaker_embedding_size
 from utils.argutils import print_args
 from utils.modelutils import check_model_paths
-from synthesizer.inference import Synthesizer
-from encoder import inference as encoder
-from vocoder import inference as vocoder
+from models.synthesizer.inference import Synthesizer
+from models.encoder import inference as encoder
+from models.vocoder import inference as vocoder
 from pathlib import Path
 import numpy as np
 import soundfile as sf

encoder_preprocess.py → control/cli/encoder_preprocess.py

@@ -1,7 +1,10 @@
-from encoder.preprocess import preprocess_librispeech, preprocess_voxceleb1, preprocess_voxceleb2, preprocess_aidatatang_200zh
-from utils.argutils import print_args
-from pathlib import Path
 import argparse
+from pathlib import Path
+
+from models.encoder.preprocess import (preprocess_aidatatang_200zh,
+                                preprocess_librispeech, preprocess_voxceleb1,
+                                preprocess_voxceleb2)
+from utils.argutils import print_args
 
 if __name__ == "__main__":
     class MyFormatter(argparse.ArgumentDefaultsHelpFormatter, argparse.RawDescriptionHelpFormatter):

encoder_train.py → control/cli/encoder_train.py

@@ -1,5 +1,5 @@
 from utils.argutils import print_args
-from encoder.train import train
+from models.encoder.train import train
 from pathlib import Path
 import argparse
 

ppg2mel_train.py → control/cli/ppg2mel_train.py

@@ -3,7 +3,7 @@
 import argparse
 import numpy as np
 from utils.load_yaml import HpsYaml
-from ppg2mel.train.train_linglf02mel_seq2seq_oneshotvc import Solver
+from models.ppg2mel.train.train_linglf02mel_seq2seq_oneshotvc import Solver
 
 # For reproducibility, comment these may speed up training
 torch.backends.cudnn.deterministic = True

pre4ppg.py → control/cli/pre4ppg.py

@@ -1,7 +1,7 @@
 from pathlib import Path
 import argparse
 
-from ppg2mel.preprocess import preprocess_dataset
+from models.ppg2mel.preprocess import preprocess_dataset
 from pathlib import Path
 import argparse
 

synthesizer_preprocess_audio.py → control/cli/synthesizer_preprocess_audio.py

@@ -1,5 +1,5 @@
-from synthesizer.preprocess import preprocess_dataset
-from synthesizer.hparams import hparams
+from models.synthesizer.preprocess import preprocess_dataset
+from models.synthesizer.hparams import hparams
 from utils.argutils import print_args
 from pathlib import Path
 import argparse

synthesizer_preprocess_embeds.py → control/cli/synthesizer_preprocess_embeds.py

@@ -1,4 +1,4 @@
-from synthesizer.preprocess import create_embeddings
+from models.synthesizer.preprocess import create_embeddings
 from utils.argutils import print_args
 from pathlib import Path
 import argparse

synthesizer_train.py → control/cli/synthesizer_train.py

@@ -1,5 +1,5 @@
-from synthesizer.hparams import hparams
-from synthesizer.train import train
+from models.synthesizer.hparams import hparams
+from models.synthesizer.train import train
 from utils.argutils import print_args
 import argparse
 

vocoder_preprocess.py → control/cli/vocoder_preprocess.py

@@ -1,5 +1,5 @@
-from synthesizer.synthesize import run_synthesis
-from synthesizer.hparams import hparams
+from models.synthesizer.synthesize import run_synthesis
+from models.synthesizer.hparams import hparams
 from utils.argutils import print_args
 import argparse
 import os

vocoder_train.py → control/cli/vocoder_train.py

@@ -1,7 +1,7 @@
 from utils.argutils import print_args
-from vocoder.wavernn.train import train
-from vocoder.hifigan.train import train as train_hifigan
-from vocoder.fregan.train import train as train_fregan
+from models.vocoder.wavernn.train import train
+from models.vocoder.hifigan.train import train as train_hifigan
+from models.vocoder.fregan.train import train as train_fregan
 from utils.util import AttrDict
 from pathlib import Path
 import argparse

mkgui/app.py → control/mkgui/app.py

@@ -2,22 +2,22 @@
 import os
 from pathlib import Path
 from enum import Enum
-from encoder import inference as encoder
+from models.encoder import inference as encoder
 import librosa
 from scipy.io.wavfile import write
 import re
 import numpy as np
-from mkgui.base.components.types import FileContent
-from vocoder.hifigan import inference as gan_vocoder
-from synthesizer.inference import Synthesizer
+from control.mkgui.base.components.types import FileContent
+from models.vocoder.hifigan import inference as gan_vocoder
+from models.synthesizer.inference import Synthesizer
 from typing import Any, Tuple
 import matplotlib.pyplot as plt
 
 # Constants
-AUDIO_SAMPLES_DIR = f"samples{os.sep}"
-SYN_MODELS_DIRT = f"synthesizer{os.sep}saved_models"
-ENC_MODELS_DIRT = f"encoder{os.sep}saved_models"
-VOC_MODELS_DIRT = f"vocoder{os.sep}saved_models"
+AUDIO_SAMPLES_DIR = f"data{os.sep}samples{os.sep}"
+SYN_MODELS_DIRT = f"data{os.sep}ckpt{os.sep}synthesizer"
+ENC_MODELS_DIRT = f"data{os.sep}ckpt{os.sep}encoder"
+VOC_MODELS_DIRT = f"data{os.sep}ckpt{os.sep}vocoder"
 TEMP_SOURCE_AUDIO = f"wavs{os.sep}temp_source.wav"
 TEMP_RESULT_AUDIO = f"wavs{os.sep}temp_result.wav"
 if not os.path.isdir("wavs"):
@@ -31,7 +31,7 @@
     synthesizers =  Enum('synthesizers', list((file.name, file) for file in Path(SYN_MODELS_DIRT).glob("**/*.pt")))
     print("Loaded synthesizer models: " + str(len(synthesizers)))
 else:
-    raise Exception(f"Model folder {SYN_MODELS_DIRT} doesn't exist.")
+    raise Exception(f"Model folder {SYN_MODELS_DIRT} doesn't exist. 请将模型文件位置移动到上述位置中进行重试！")
 
 if os.path.isdir(ENC_MODELS_DIRT):    
     encoders =  Enum('encoders', list((file.name, file) for file in Path(ENC_MODELS_DIRT).glob("**/*.pt")))

mkgui/app_vc.py → control/mkgui/app_vc.py

@@ -1,27 +1,26 @@
-from synthesizer.inference import Synthesizer
-from pydantic import BaseModel, Field
-from encoder import inference as speacker_encoder
-import torch
 import os
-from pathlib import Path
 from enum import Enum
-import ppg_extractor as Extractor
-import ppg2mel as Convertor
-import librosa
-from scipy.io.wavfile import write
-import re
-import numpy as np
-from mkgui.base.components.types import FileContent
-from vocoder.hifigan import inference as gan_vocoder
+from pathlib import Path
 from typing import Any, Tuple
+
+import librosa
 import matplotlib.pyplot as plt
+import torch
+from pydantic import BaseModel, Field
+from scipy.io.wavfile import write
 
+import models.ppg2mel as Convertor
+import models.ppg_extractor as Extractor
+from control.mkgui.base.components.types import FileContent
+from models.encoder import inference as speacker_encoder
+from models.synthesizer.inference import Synthesizer
+from models.vocoder.hifigan import inference as gan_vocoder
 
 # Constants
-AUDIO_SAMPLES_DIR = f'samples{os.sep}'
-EXT_MODELS_DIRT = f'ppg_extractor{os.sep}saved_models'
-CONV_MODELS_DIRT = f'ppg2mel{os.sep}saved_models'
-VOC_MODELS_DIRT = f'vocoder{os.sep}saved_models'
+AUDIO_SAMPLES_DIR = f'data{os.sep}samples{os.sep}'
+EXT_MODELS_DIRT = f'data{os.sep}ckpt{os.sep}ppg_extractor'
+CONV_MODELS_DIRT = f'data{os.sep}ckpt{os.sep}ppg2mel'
+VOC_MODELS_DIRT = f'data{os.sep}ckpt{os.sep}vocoder'
 TEMP_SOURCE_AUDIO = f'wavs{os.sep}temp_source.wav'
 TEMP_TARGET_AUDIO = f'wavs{os.sep}temp_target.wav'
 TEMP_RESULT_AUDIO = f'wavs{os.sep}temp_result.wav'
@@ -132,9 +131,10 @@ def convert(input: Input) -> Output:
 
     ppg = extractor.extract_from_wav(src_wav)
     # Import necessary dependency of Voice Conversion
-    from utils.f0_utils import compute_f0, f02lf0, compute_mean_std, get_converted_lf0uv   
+    from utils.f0_utils import (compute_f0, compute_mean_std, f02lf0,
+                                get_converted_lf0uv)   
     ref_lf0_mean, ref_lf0_std = compute_mean_std(f02lf0(compute_f0(ref_wav)))
-    speacker_encoder.load_model(Path(f"encoder{os.sep}saved_models{os.sep}pretrained_bak_5805000.pt"))
+    speacker_encoder.load_model(Path(f"data{os.sep}ckpt{os.sep}encoder{os.sep}pretrained_bak_5805000.pt"))
     embed = speacker_encoder.embed_utterance(ref_wav)
     lf0_uv = get_converted_lf0uv(src_wav, ref_lf0_mean, ref_lf0_std, convert=True)
     min_len = min(ppg.shape[1], len(lf0_uv))

mkgui/base/ui/streamlit_ui.py → control/mkgui/base/ui/streamlit_ui.py

@@ -14,14 +14,13 @@
 from loguru import logger
 from pydantic import BaseModel, ValidationError, parse_obj_as
 
-from mkgui.base import Opyrator
-from mkgui.base.core import name_to_title
-from mkgui.base.ui import schema_utils
-from mkgui.base.ui.streamlit_utils import CUSTOM_STREAMLIT_CSS
+from control.mkgui.base import Opyrator
+from control.mkgui.base.core import name_to_title
+from . import schema_utils
+from .streamlit_utils import CUSTOM_STREAMLIT_CSS
 
 STREAMLIT_RUNNER_SNIPPET = """
-from mkgui.base.ui import render_streamlit_ui
-from mkgui.base import Opyrator
+from control.mkgui.base.ui import render_streamlit_ui
 
 import streamlit as st
 
@@ -807,18 +806,18 @@ def _render_list_output(self, streamlit: st, output_data: List) -> None:
 
 def getOpyrator(mode: str) -> Opyrator:
     if mode == None or mode.startswith('VC'):
-        from mkgui.app_vc import convert
+        from control.mkgui.app_vc import convert
         return  Opyrator(convert)
     if mode == None or mode.startswith('预处理'):
-        from mkgui.preprocess import preprocess
+        from control.mkgui.preprocess import preprocess
         return  Opyrator(preprocess)
     if mode == None or mode.startswith('模型训练'):
-        from mkgui.train import train
+        from control.mkgui.train import train
         return  Opyrator(train)
     if mode == None or mode.startswith('模型训练(VC)'):
-        from mkgui.train_vc import train_vc
+        from control.mkgui.train_vc import train_vc
         return  Opyrator(train_vc)
-    from mkgui.app import synthesize
+    from control.mkgui.app import synthesize
     return Opyrator(synthesize)
 
 
@@ -845,7 +844,7 @@ def render_streamlit_ui() -> None:
     col2.title(title)
     col2.markdown("欢迎使用MockingBird Web 2")
 
-    image = Image.open(path.join('mkgui', 'static', 'mb.png'))
+    image = Image.open(path.join('control','mkgui', 'static', 'mb.png'))
     col1.image(image)
 
     st.markdown("---")

mkgui/preprocess.py → control/mkgui/preprocess.py

@@ -6,8 +6,8 @@
 
 
 # Constants
-EXT_MODELS_DIRT = f"ppg_extractor{os.sep}saved_models"
-ENC_MODELS_DIRT = f"encoder{os.sep}saved_models"
+EXT_MODELS_DIRT = f"data{os.sep}ckpt{os.sep}ppg_extractor"
+ENC_MODELS_DIRT = f"data{os.sep}ckpt{os.sep}encoder"
 
 
 if os.path.isdir(EXT_MODELS_DIRT):    
@@ -83,7 +83,7 @@ def preprocess(input: Input) -> Output:
     """Preprocess(预处理)"""
     finished = 0
     if input.model == Model.VC_PPG2MEL:
-        from ppg2mel.preprocess import preprocess_dataset
+        from models.ppg2mel.preprocess import preprocess_dataset
         finished = preprocess_dataset(
             datasets_root=Path(input.datasets_root),
             dataset=input.dataset,

mkgui/train.py → control/mkgui/train.py

@@ -3,17 +3,17 @@
 from pathlib import Path
 from enum import Enum
 from typing import Any
-from synthesizer.hparams import hparams
-from synthesizer.train import train as synt_train
+from models.synthesizer.hparams import hparams
+from models.synthesizer.train import train as synt_train
 
 # Constants
-SYN_MODELS_DIRT = f"synthesizer{os.sep}saved_models"
-ENC_MODELS_DIRT = f"encoder{os.sep}saved_models"
+SYN_MODELS_DIRT = f"data{os.sep}ckpt{os.sep}synthesizer"
+ENC_MODELS_DIRT = f"data{os.sep}ckpt{os.sep}encoder"
 
 
-# EXT_MODELS_DIRT = f"ppg_extractor{os.sep}saved_models"
-# CONV_MODELS_DIRT = f"ppg2mel{os.sep}saved_models"
-# ENC_MODELS_DIRT = f"encoder{os.sep}saved_models"
+# EXT_MODELS_DIRT = f"data{os.sep}ckpt{os.sep}ppg_extractor"
+# CONV_MODELS_DIRT = f"data{os.sep}ckpt{os.sep}ppg2mel"
+# ENC_MODELS_DIRT = f"data{os.sep}ckpt{os.sep}encoder"
 
 # Pre-Load models
 if os.path.isdir(SYN_MODELS_DIRT):    
@@ -96,7 +96,7 @@ def train(input: Input) -> Output:
     synt_train(
         input.run_id, 
         input.input_root, 
-        f"synthesizer{os.sep}saved_models", 
+        f"data{os.sep}ckpt{os.sep}synthesizer", 
         input.save_every, 
         input.backup_every, 
         input.log_every, 

mkgui/train_vc.py → control/mkgui/train_vc.py

@@ -9,9 +9,9 @@
 import torch
 
 # Constants
-EXT_MODELS_DIRT = f"ppg_extractor{os.sep}saved_models"
-CONV_MODELS_DIRT = f"ppg2mel{os.sep}saved_models"
-ENC_MODELS_DIRT = f"encoder{os.sep}saved_models"
+EXT_MODELS_DIRT = f"data{os.sep}ckpt{os.sep}ppg_extractor"
+CONV_MODELS_DIRT = f"data{os.sep}ckpt{os.sep}ppg2mel"
+ENC_MODELS_DIRT = f"data{os.sep}ckpt{os.sep}encoder"
 
 
 if os.path.isdir(EXT_MODELS_DIRT):    
@@ -144,7 +144,7 @@ def train_vc(input: Input) -> Output:
     if torch.cuda.is_available():
         torch.cuda.manual_seed_all(input.seed)
     mode = "train"
-    from ppg2mel.train.train_linglf02mel_seq2seq_oneshotvc import Solver
+    from models.ppg2mel.train.train_linglf02mel_seq2seq_oneshotvc import Solver
     solver = Solver(config, params, mode)
     solver.load_data()
     solver.set_model()

toolbox/__init__.py → control/toolbox/__init__.py

@@ -1,12 +1,12 @@
-from toolbox.ui import UI
-from encoder import inference as encoder
-from synthesizer.inference import Synthesizer
-from vocoder.wavernn import inference as rnn_vocoder
-from vocoder.hifigan import inference as gan_vocoder
-from vocoder.fregan import inference as fgan_vocoder
+from control.toolbox.ui import UI
+from models.encoder import inference as encoder
+from models.synthesizer.inference import Synthesizer
+from models.vocoder.wavernn import inference as rnn_vocoder
+from models.vocoder.hifigan import inference as gan_vocoder
+from models.vocoder.fregan import inference as fgan_vocoder
 from pathlib import Path
 from time import perf_counter as timer
-from toolbox.utterance import Utterance
+from control.toolbox.utterance import Utterance
 import numpy as np
 import traceback
 import sys
@@ -396,7 +396,7 @@ def init_extractor(self):
         self.ui.log("Loading the extractor %s... " % model_fpath)
         self.ui.set_loading(1)
         start = timer()
-        import ppg_extractor as extractor
+        import models.ppg_extractor as extractor
         self.extractor = extractor.load_model(model_fpath)
         self.ui.log("Done (%dms)." % int(1000 * (timer() - start)), "append")
         self.ui.set_loading(0)
@@ -408,7 +408,7 @@ def init_convertor(self):
         self.ui.log("Loading the convertor %s... " % model_fpath)
         self.ui.set_loading(1)
         start = timer()
-        import ppg2mel as convertor
+        import models.ppg2mel as convertor
         self.convertor = convertor.load_model( model_fpath)
         self.ui.log("Done (%dms)." % int(1000 * (timer() - start)), "append")
         self.ui.set_loading(0)

toolbox/ui.py → control/toolbox/ui.py

@@ -4,8 +4,8 @@
 import matplotlib.pyplot as plt
 from matplotlib.backends.backend_qt5agg import FigureCanvasQTAgg as FigureCanvas
 from matplotlib.figure import Figure
-from encoder.inference import plot_embedding_as_heatmap
-from toolbox.utterance import Utterance
+from models.encoder.inference import plot_embedding_as_heatmap
+from control.toolbox.utterance import Utterance
 from pathlib import Path
 from typing import List, Set
 import sounddevice as sd

demo_toolbox.py

@@ -1,5 +1,5 @@
 from pathlib import Path
-from toolbox import Toolbox
+from control.toolbox import Toolbox
 from utils.argutils import print_args
 from utils.modelutils import check_model_paths
 import argparse