Merge branch 'dev' of github.com:litagin02/Style-Bert-VITS2

This commit is contained in:
tsukumi
2024-03-08 05:49:52 +00:00
19 changed files with 1411 additions and 1070 deletions

View File

@@ -3,7 +3,7 @@ from pathlib import Path
# Style-Bert-VITS2 のバージョン
VERSION = "2.3.1"
VERSION = "2.4"
# Style-Bert-VITS2 のベースディレクトリ
BASE_DIR = Path(__file__).parent.parent

View File

@@ -1,13 +1,15 @@
import pyopenjtalk
import re
from style_bert_vits2.constants import Languages
from style_bert_vits2.logging import logger
from style_bert_vits2.text_processing import bert_models
from style_bert_vits2.text_processing.japanese import pyopenjtalk_worker as pyopenjtalk
from style_bert_vits2.text_processing.japanese.mora_list import MORA_KATA_TO_MORA_PHONEMES
from style_bert_vits2.text_processing.japanese.normalizer import replace_punctuation
from style_bert_vits2.text_processing.symbols import PUNCTUATIONS
pyopenjtalk.initialize()
def g2p(
norm_text: str,

View File

@@ -0,0 +1,126 @@
"""
Run the pyopenjtalk worker in a separate process
to avoid user dictionary access error
"""
from typing import Any, Optional
from style_bert_vits2.logging import logger
from style_bert_vits2.text_processing.japanese.pyopenjtalk_worker.worker_client import WorkerClient
from style_bert_vits2.text_processing.japanese.pyopenjtalk_worker.worker_common import WORKER_PORT
WORKER_CLIENT: Optional[WorkerClient] = None
# pyopenjtalk interface
# g2p(): not used
def run_frontend(text: str) -> list[dict[str, Any]]:
assert WORKER_CLIENT
ret = WORKER_CLIENT.dispatch_pyopenjtalk("run_frontend", text)
assert isinstance(ret, list)
return ret
def make_label(njd_features: Any) -> list[str]:
assert WORKER_CLIENT
ret = WORKER_CLIENT.dispatch_pyopenjtalk("make_label", njd_features)
assert isinstance(ret, list)
return ret
def mecab_dict_index(path: str, out_path: str, dn_mecab: Optional[str] = None):
assert WORKER_CLIENT
WORKER_CLIENT.dispatch_pyopenjtalk("mecab_dict_index", path, out_path, dn_mecab)
def update_global_jtalk_with_user_dict(path: str):
assert WORKER_CLIENT
WORKER_CLIENT.dispatch_pyopenjtalk("update_global_jtalk_with_user_dict", path)
def unset_user_dict():
assert WORKER_CLIENT
WORKER_CLIENT.dispatch_pyopenjtalk("unset_user_dict")
# initialize module when imported
def initialize(port: int = WORKER_PORT) -> None:
import time
import socket
import sys
import atexit
import signal
logger.debug("initialize")
global WORKER_CLIENT
if WORKER_CLIENT:
return
client = None
try:
client = WorkerClient(port)
except (socket.timeout, socket.error):
logger.debug("try starting pyopenjtalk worker server")
import os
import subprocess
worker_pkg_path = os.path.relpath(
os.path.dirname(__file__), os.getcwd()
).replace(os.sep, ".")
args = [sys.executable, "-m", worker_pkg_path, "--port", str(port)]
# new session, new process group
if sys.platform.startswith("win"):
cf = subprocess.CREATE_NEW_CONSOLE | subprocess.CREATE_NEW_PROCESS_GROUP # type: ignore
si = subprocess.STARTUPINFO() # type: ignore
si.dwFlags |= subprocess.STARTF_USESHOWWINDOW # type: ignore
si.wShowWindow = subprocess.SW_HIDE # type: ignore
subprocess.Popen(args, creationflags=cf, startupinfo=si)
else:
# align with Windows behavior
# start_new_session is same as specifying setsid in preexec_fn
subprocess.Popen(args, stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL, start_new_session=True) # type: ignore
# wait until server listening
count = 0
while True:
try:
client = WorkerClient(port)
break
except socket.error:
time.sleep(1)
count += 1
# 10: max number of retries
if count == 10:
raise TimeoutError("サーバーに接続できませんでした")
WORKER_CLIENT = client
atexit.register(terminate)
# when the process is killed
def signal_handler(signum: int, frame: Any):
terminate()
signal.signal(signal.SIGTERM, signal_handler)
# top-level declaration
def terminate() -> None:
logger.debug("terminate")
global WORKER_CLIENT
if not WORKER_CLIENT:
return
# repare for unexpected errors
try:
if WORKER_CLIENT.status() == 1:
WORKER_CLIENT.quit_server()
except Exception as e:
logger.error(e)
WORKER_CLIENT.close()
WORKER_CLIENT = None

View File

@@ -0,0 +1,16 @@
import argparse
from style_bert_vits2.text_processing.japanese.pyopenjtalk_worker.worker_common import WORKER_PORT
from style_bert_vits2.text_processing.japanese.pyopenjtalk_worker.worker_server import WorkerServer
def main() -> None:
parser = argparse.ArgumentParser()
parser.add_argument("--port", type=int, default=WORKER_PORT)
args = parser.parse_args()
server = WorkerServer()
server.start_server(port=args.port)
if __name__ == "__main__":
main()

View File

@@ -0,0 +1,63 @@
import socket
from typing import Any, cast
from style_bert_vits2.logging import logger
from style_bert_vits2.text_processing.japanese.pyopenjtalk_worker.worker_common import RequestType, receive_data, send_data
class WorkerClient:
""" pyopenjtalk worker client """
def __init__(self, port: int) -> None:
sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
# 5: timeout
sock.settimeout(5)
sock.connect((socket.gethostname(), port))
self.sock = sock
def __enter__(self) -> "WorkerClient":
return self
def __exit__(self, exc_type: Any, exc_value: Any, traceback: Any) -> None:
self.close()
def close(self) -> None:
self.sock.close()
def dispatch_pyopenjtalk(self, func: str, *args: Any, **kwargs: Any) -> Any:
data = {
"request-type": RequestType.PYOPENJTALK,
"func": func,
"args": args,
"kwargs": kwargs,
}
logger.trace(f"client sends request: {data}")
send_data(self.sock, data)
logger.trace("client sent request successfully")
response = receive_data(self.sock)
logger.trace(f"client received response: {response}")
return response.get("return")
def status(self) -> int:
data = {"request-type": RequestType.STATUS}
logger.trace(f"client sends request: {data}")
send_data(self.sock, data)
logger.trace("client sent request successfully")
response = receive_data(self.sock)
logger.trace(f"client received response: {response}")
return cast(int, response.get("client-count"))
def quit_server(self) -> None:
data = {"request-type": RequestType.QUIT_SERVER}
logger.trace(f"client sends request: {data}")
send_data(self.sock, data)
logger.trace("client sent request successfully")
response = receive_data(self.sock)
logger.trace(f"client received response: {response}")

View File

@@ -0,0 +1,45 @@
import json
import socket
from enum import IntEnum, auto
from typing import Any, Final
WORKER_PORT: Final[int] = 7861
HEADER_SIZE: Final[int] = 4
class RequestType(IntEnum):
STATUS = auto()
QUIT_SERVER = auto()
PYOPENJTALK = auto()
class ConnectionClosedException(Exception):
pass
# socket communication
def send_data(sock: socket.socket, data: dict[str, Any]):
json_data = json.dumps(data).encode()
header = len(json_data).to_bytes(HEADER_SIZE, byteorder="big")
sock.sendall(header + json_data)
def __receive_until(sock: socket.socket, size: int):
data = b""
while len(data) < size:
part = sock.recv(size - len(data))
if part == b"":
raise ConnectionClosedException("接続が閉じられました")
data += part
return data
def receive_data(sock: socket.socket) -> dict[str, Any]:
header = __receive_until(sock, HEADER_SIZE)
data_length = int.from_bytes(header, byteorder="big")
body = __receive_until(sock, data_length)
return json.loads(body.decode())

View File

@@ -0,0 +1,126 @@
import select
import socket
import time
from typing import Any, cast
import pyopenjtalk
from style_bert_vits2.logging import logger
from style_bert_vits2.text_processing.japanese.pyopenjtalk_worker.worker_common import (
ConnectionClosedException,
RequestType,
receive_data,
send_data,
)
# To make it as fast as possible
# Probably faster than calling getattr every time
__PYOPENJTALK_FUNC_DICT = {
"run_frontend": pyopenjtalk.run_frontend,
"make_label": pyopenjtalk.make_label,
"mecab_dict_index": pyopenjtalk.mecab_dict_index,
"update_global_jtalk_with_user_dict": pyopenjtalk.update_global_jtalk_with_user_dict,
"unset_user_dict": pyopenjtalk.unset_user_dict,
}
class WorkerServer:
""" pyopenjtalk worker server """
def __init__(self) -> None:
self.client_count: int = 0
self.quit: bool = False
def handle_request(self, request: dict[str, Any]) -> dict[str, Any]:
request_type = None
try:
request_type = RequestType(cast(int, request.get("request-type")))
except Exception:
return {
"success": False,
"reason": "request-type is invalid",
}
response: dict[str, Any] = {}
if request_type:
if request_type == RequestType.STATUS:
response = {
"success": True,
"client-count": self.client_count,
}
elif request_type == RequestType.QUIT_SERVER:
self.quit = True
response = {"success": True}
elif request_type == RequestType.PYOPENJTALK:
func_name = request.get("func")
assert isinstance(func_name, str)
func = __PYOPENJTALK_FUNC_DICT[func_name]
args = request.get("args")
kwargs = request.get("kwargs")
assert isinstance(args, list)
assert isinstance(kwargs, dict)
ret = func(*args, **kwargs)
response = {"success": True, "return": ret}
else:
# NOT REACHED
response = request
return response
def start_server(self, port: int, no_client_timeout: int = 30) -> None:
logger.info("start pyopenjtalk worker server")
with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as server_socket:
server_socket.bind((socket.gethostname(), port))
server_socket.listen()
sockets = [server_socket]
no_client_since = time.time()
while True:
if self.client_count == 0:
if no_client_since is None:
no_client_since = time.time()
elif (time.time() - no_client_since) > no_client_timeout:
logger.info("quit because there is no client")
return
else:
no_client_since = None
ready_sockets, _, _ = select.select(sockets, [], [], 0.1)
for sock in ready_sockets:
if sock is server_socket:
logger.info("new client connected")
client_socket, _ = server_socket.accept()
sockets.append(client_socket)
self.client_count += 1
else:
# client
try:
request = receive_data(sock)
except Exception as e:
sock.close()
sockets.remove(sock)
self.client_count -= 1
# unexpected disconnections
if not isinstance(e, ConnectionClosedException):
logger.error(e)
logger.info("close connection")
continue
logger.trace(f"server received request: {request}")
response = self.handle_request(request)
logger.trace(f"server sends response: {response}")
try:
send_data(sock, response)
logger.trace("server sent response successfully")
except Exception:
logger.warning(
"an exception occurred during sending responce"
)
if self.quit:
logger.info("quit pyopenjtalk worker server")
return

View File

@@ -13,15 +13,15 @@ from typing import Dict, List, Optional
from uuid import UUID, uuid4
import numpy as np
import pyopenjtalk
from fastapi import HTTPException
from style_bert_vits2.constants import DEFAULT_USER_DICT_DIR
from style_bert_vits2.text_processing.japanese import pyopenjtalk_worker as pyopenjtalk
from style_bert_vits2.text_processing.japanese.user_dict.word_model import UserDictWord, WordTypes
# from ..utility.mutex_utility import mutex_wrapper
# from ..utility.path_utility import engine_root, get_save_dir
from style_bert_vits2.text_processing.japanese.user_dict.part_of_speech_data import MAX_PRIORITY, MIN_PRIORITY, part_of_speech_data
pyopenjtalk.initialize()
# root_dir = engine_root()
# save_dir = get_save_dir()