Page MenuHomePhabricator
Files F5021462

svc_online.py
No OneTemporary
Actions

svc_online.py
View Options

# -*- coding: UTF-8 -*-
"""
SVC的核心处理逻辑
"""
import os
import time
import socket
import shutil
import hashlib
from AIMeiSheng.meisheng_svc_final import load_model, process_svc_online
from AIMeiSheng.cos_similar_ui_zoom import cos_similar
from AIMeiSheng.meisheng_env_preparex import meisheng_env_prepare, demucs_env_prepare
from AIMeiSheng.voice_classification.online.voice_class_online_fang import VoiceClass, download_volume_balanced
from AIMeiSheng.docker_demo.common import *
# from AIMeiSheng.demucs_process_one import demucs_process_one
from AIMeiSheng.separate_demucs import main_seperator
import logging
hostname = socket.gethostname()
log_file_name = f"{os.path.dirname(os.path.abspath(__file__))}/av_meisheng_{hostname}.log"
# 设置logger
svc_offline_logger = logging.getLogger("svc_offline")
file_handler = logging.FileHandler(log_file_name)
file_handler.setLevel(logging.INFO)
formatter = logging.Formatter('%(asctime)s %(levelname)s %(message)s', datefmt='%Y-%m-%d %I:%M:%S')
file_handler.setFormatter(formatter)
# if gs_prod:
# svc_offline_logger.addHandler(file_handler)
if os.path.exists(gs_tmp_dir):
shutil.rmtree(gs_tmp_dir)
os.makedirs(gs_model_dir, exist_ok=True)
os.makedirs(gs_resource_cache_dir, exist_ok=True)
# 预设参数
gs_gender_models_url = "https://av-audit-sync-sg-1256122840.cos.ap-singapore.myqcloud.com/hub/voice_classification/models.zip"
gs_volume_bin_url = "https://av-audit-sync-sg-1256122840.cos.ap-singapore.myqcloud.com/dataset/AIMeiSheng/ebur128_tool/v1/ebur128_tool"
class GSWorkerAttr:
def __init__(self, input_data):
# 取出输入资源
vocal_url = input_data["record_song_url"]
target_url = input_data["target_url"]
start = input_data["start"] # 单位是ms
end = input_data["end"] # 单位是ms
vocal_loudness = input_data["vocal_loudness"]
female_recording_url = input_data["female_recording_url"]
male_recording_url = input_data["male_recording_url"]
self.distinct_id = hashlib.md5(vocal_url.encode()).hexdigest()
self.tmp_dir = os.path.join(gs_tmp_dir, self.distinct_id)
if os.path.exists(self.tmp_dir):
shutil.rmtree(self.tmp_dir)
os.makedirs(self.tmp_dir)
self.vocal_url = vocal_url
self.target_url = target_url
ext = vocal_url.split(".")[-1]
self.vocal_path = os.path.join(self.tmp_dir, self.distinct_id + f"_in.{ext}")
self.vocal_demucs_path = os.path.join(self.tmp_dir, self.distinct_id + f"_in_demucs.wav")
self.demucs_flg = False
self.target_wav_path = os.path.join(self.tmp_dir, self.distinct_id + "_out.wav")
self.target_wav_ad_path = os.path.join(self.tmp_dir, self.distinct_id + "_out_ad.wav")
self.target_path = os.path.join(self.tmp_dir, self.distinct_id + "_out.m4a")
self.female_svc_source_url = female_recording_url
self.male_svc_source_url = male_recording_url
ext = female_recording_url.split(".")[-1]
self.female_svc_source_path = os.path.join(self.tmp_dir, self.distinct_id + f"_female.{ext}")
ext = male_recording_url.split(".")[-1]
self.male_svc_source_path = os.path.join(self.tmp_dir, self.distinct_id + f"_male.{ext}")
# self.female_svc_source_path = os.path.join(gs_resource_cache_dir,
# hashlib.md5(female_recording_url.encode()).hexdigest() + "." + ext)
# ext = male_recording_url.split(".")[-1]
# self.male_svc_source_path = os.path.join(gs_resource_cache_dir,
# hashlib.md5(male_recording_url.encode()).hexdigest() + "." + ext)
self.st_tm = start
self.ed_tm = end
self.target_loudness = vocal_loudness
def log_info_name(self):
return f"d_id={self.distinct_id}, vocal_url={self.vocal_url}"
def rm_cache(self):
if os.path.exists(self.tmp_dir):
shutil.rmtree(self.tmp_dir)
def init_gender_model():
"""
下载模型
:return:
"""
dst_model_dir = os.path.join(gs_model_dir, "voice_classification")
if not os.path.exists(dst_model_dir):
dst_zip_path = os.path.join(gs_model_dir, "models.zip")
if not download2disk(gs_gender_models_url, dst_zip_path):
svc_offline_logger.fatal(f"download gender_model err={gs_gender_models_url}")
cmd = f"cd {gs_model_dir}; unzip {dst_zip_path}; mv models voice_classification; rm -f {dst_zip_path}"
os.system(cmd)
if not os.path.exists(dst_model_dir):
svc_offline_logger.fatal(f"unzip {dst_zip_path} err")
music_voice_pure_model = os.path.join(dst_model_dir, "voice_005_rec_v5.pth")
music_voice_no_pure_model = os.path.join(dst_model_dir, "voice_10_v5.pth")
gender_pure_model = os.path.join(dst_model_dir, "gender_8k_ratev5_v6_adam.pth")
gender_no_pure_model = os.path.join(dst_model_dir, "gender_8k_v6_adam.pth")
vc = VoiceClass(music_voice_pure_model, music_voice_no_pure_model, gender_pure_model, gender_no_pure_model)
return vc
def init_demucs_seperator():
demucs_env_prepare(logging, gs_demucs_model_path)
seperator = main_seperator()
return seperator
def init_svc_model():
meisheng_env_prepare(logging, gs_model_dir)
embed_model, hubert_model = load_model()
cs_sim = cos_similar()
return embed_model, hubert_model, cs_sim
def download_volume_adjustment():
"""
下载音量调整工具
:return:
"""
volume_bin_path = os.path.join(gs_model_dir, "ebur128_tool")
if not os.path.exists(volume_bin_path):
if not download2disk(gs_volume_bin_url, volume_bin_path):
svc_offline_logger.fatal(f"download volume_bin err={gs_volume_bin_url}")
os.system(f"chmod +x {volume_bin_path}")
def volume_adjustment(wav_path, target_loudness, out_path):
"""
音量调整
:param wav_path:
:param target_loudness:
:param out_path:
:return:
"""
volume_bin_path = os.path.join(gs_model_dir, "ebur128_tool")
cmd = f"{volume_bin_path} {wav_path} {target_loudness} {out_path}"
os.system(cmd)
class SVCOnline:
def __init__(self):
st = time.time()
self.gender_model = init_gender_model()
self.embed_model, self.hubert_model, self.cs_sim = init_svc_model()
self.demucs_seprator = init_demucs_seperator()
download_volume_adjustment()
download_volume_balanced()
svc_offline_logger.info(f"svc init finished, sp = {time.time() - st}")
def gender_process(self, worker_attr):
st = time.time()
gender, female_rate, is_pure = self.gender_model.process(worker_attr.vocal_path)
svc_offline_logger.info(
f"{worker_attr.vocal_url}, gender={gender}, female_rate={female_rate}, is_pure={is_pure}, "
f"gender_process sp = {time.time() - st}")
if gender == 0:
gender = 'female'
elif gender == 1:
gender = 'male'
elif female_rate == None:
gender = 'male'
return gender, gs_err_code_gender_classify
elif female_rate > 0.5:
gender = 'female'
else:
gender = 'male'
self.demucs_flg = False
if gender == 'female':
if self.gender_model.vocal_ratio <= 1.0:#0.5:
print(f"@@@ vocal_ratio: {self.gender_model.vocal_ratio}, gender : {gender}, gs_err_code_vocal_ratio : {gs_err_code_vocal_ratio}")
# if not demucs_process_one(worker_attr.vocal_path, worker_attr.vocal_demucs_path):
self.demucs_flg = True
if not self.demucs_seprator.process_one(worker_attr.vocal_path, worker_attr.vocal_demucs_path):
self.demucs_flg = False
# return gender, gs_err_code_vocal_ratio
svc_offline_logger.info(f" demucs fail and gender={gender}")
else:
if self.gender_model.vocal_ratio <= 1.0:#0.6:
print(f"@@@ vocal_ratio: {self.gender_model.vocal_ratio}, gender : {gender}, gs_err_code_vocal_ratio : {gs_err_code_vocal_ratio}")
# if not demucs_process_one(worker_attr.vocal_path, worker_attr.vocal_demucs_path):
self.demucs_flg = True
if not self.demucs_seprator.process_one(worker_attr.vocal_path, worker_attr.vocal_demucs_path):
self.demucs_flg = False
# return gender, gs_err_code_vocal_ratio
svc_offline_logger.info(f" demucs fail and gender={gender}")
svc_offline_logger.info(f"{worker_attr.vocal_url}, modified gender={gender}")
# err = gs_err_code_success
# if female_rate == -1:
# err = gs_err_code_target_silence
return gender, gs_err_code_success
def process(self, worker_attr):
gender, err = self.gender_process(worker_attr)
if err != gs_err_code_success:
return gender, err
song_path = worker_attr.female_svc_source_path
if gender == "male":
song_path = worker_attr.male_svc_source_path
params = {'gender': gender, 'tst': worker_attr.st_tm, "tnd": worker_attr.ed_tm, 'delay': 0, 'song_path': None}
st = time.time()
if os.path.exists(worker_attr.vocal_demucs_path) and worker_attr.demucs_flg:
vocal_tmp_path = worker_attr.vocal_demucs_path
else:
vocal_tmp_path = worker_attr.vocal_path
err_code = process_svc_online(song_path,vocal_tmp_path, worker_attr.target_wav_path, self.embed_model,
self.hubert_model, self.cs_sim, params)
svc_offline_logger.info(f"{worker_attr.vocal_url}, err_code={err_code} process svc sp = {time.time() - st}")
return gender, err_code

File Metadata

Mime Type
text/x-python
Expires
Mon, May 19, 02:30 (1 d, 59 m)
Storage Engine
blob
Storage Format
Raw Data
Storage Handle
1433513
Default Alt Text
svc_online.py (9 KB)

Event Timeline