コードベースを大規模にリファクタリングし、ライブラリ (Python パッケージ) としてほかの Python コードから利用できるように改善

.gitignore

@@ -1,7 +1,8 @@
-.vscode/
-
 __pycache__/
 venv/
+.venv/
+dist/
+.coverage
 .ipynb_checkpoints/
 
 /*.yml

.vscode/extensions.json

@@ -0,0 +1,6 @@
+{
+    "recommendations": [
+        "ms-python.python",
+        "ms-python.vscode-pylance"
+    ]
+}

.vscode/settings.json

@@ -0,0 +1,22 @@
+{
+    // Pylance の Type Checking を有効化
+    "python.languageServer": "Pylance",
+    "python.analysis.typeCheckingMode": "strict",
+    // Pylance の Type Checking のうち、いくつかのエラー報告を抑制する
+    "python.analysis.diagnosticSeverityOverrides": {
+        "reportConstantRedefinition": "none",
+        "reportGeneralTypeIssues": "warning",
+        "reportMissingParameterType": "warning",
+        "reportMissingTypeStubs": "none",
+        "reportPrivateImportUsage": "none",
+        "reportPrivateUsage": "warning",
+        "reportShadowedImports": "none",
+        "reportUnnecessaryComparison": "none",
+        "reportUnknownArgumentType": "none",
+        "reportUnknownMemberType": "none",
+        "reportUnknownParameterType": "warning",
+        "reportUnknownVariableType": "none",
+        "reportUnusedFunction": "none",
+        "reportUnusedVariable": "information",
+    },
+}

app.py

@@ -5,8 +5,8 @@
 import torch
 import yaml
 
-from common.constants import GRADIO_THEME, LATEST_VERSION
-from common.tts_model import ModelHolder
+from style_bert_vits2.constants import GRADIO_THEME, VERSION
+from style_bert_vits2.tts_model import TTSModelHolder
 from webui import (
     create_dataset_app,
     create_inference_app,
@@ -15,6 +15,7 @@
     create_train_app,
 )
 
+
 # Get path settings
 with Path("configs/paths.yml").open("r", encoding="utf-8") as f:
     path_config: dict[str, str] = yaml.safe_load(f.read())
@@ -23,6 +24,8 @@
 
 parser = argparse.ArgumentParser()
 parser.add_argument("--device", type=str, default="cuda")
+parser.add_argument("--host", type=str, default="127.0.0.1")
+parser.add_argument("--port", type=int, default=7860)
 parser.add_argument("--no_autolaunch", action="store_true")
 parser.add_argument("--share", action="store_true")
 
@@ -31,10 +34,10 @@
 if device == "cuda" and not torch.cuda.is_available():
     device = "cpu"
 
-model_holder = ModelHolder(Path(assets_root), device)
+model_holder = TTSModelHolder(Path(assets_root), device)
 
 with gr.Blocks(theme=GRADIO_THEME) as app:
-    gr.Markdown(f"# Style-Bert-VITS2 WebUI (version {LATEST_VERSION})")
+    gr.Markdown(f"# Style-Bert-VITS2 WebUI (version {VERSION})")
     with gr.Tabs():
         with gr.Tab("音声合成"):
             create_inference_app(model_holder=model_holder)
@@ -48,4 +51,4 @@
             create_merge_app(model_holder=model_holder)
 
 
-app.launch(inbrowser=not args.no_autolaunch, share=args.share)
+app.launch(server_name=args.host, server_port=args.port, inbrowser=not args.no_autolaunch, share=args.share)

bert_gen.py

@@ -5,14 +5,12 @@
 import torch.multiprocessing as mp
 from tqdm import tqdm
 
-import commons
-from text.get_bert import get_bert
-import text.pyopenjtalk_worker as pyopenjtalk
-import utils
-from common.log import logger
-from common.stdout_wrapper import SAFE_STDOUT
 from config import config
-from text import cleaned_text_to_sequence
+from style_bert_vits2.logging import logger
+from style_bert_vits2.models import commons
+from style_bert_vits2.models.hyper_parameters import HyperParameters
+from style_bert_vits2.nlp import cleaned_text_to_sequence, extract_bert_feature
+from style_bert_vits2.utils.stdout_wrapper import SAFE_STDOUT
 
 
 def process_line(x):
@@ -47,7 +45,7 @@ def process_line(x):
         bert = torch.load(bert_path)
         assert bert.shape[-1] == len(phone)
     except Exception:
-        bert = get_bert(text, word2ph, language_str, device)
+        bert = extract_bert_feature(text, word2ph, language_str, device)
         assert bert.shape[-1] == len(phone)
         torch.save(bert, bert_path)
 
@@ -64,7 +62,7 @@ def process_line(x):
     )
     args, _ = parser.parse_known_args()
     config_path = args.config
-    hps = utils.get_hparams_from_file(config_path)
+    hps = HyperParameters.load_from_json(config_path)
     lines = []
     with open(hps.data.training_files, encoding="utf-8") as f:
         lines.extend(f.readlines())