fumiama/Retrieval-based-Voice-Conversion-WebUI
1
0
Fork 0
mirror of https://github.com/fumiama/Retrieval-based-Voice-Conversion-WebUI.git synced 2026-06-05 09:10:25 +08:00
Code Activity

optimize(train): combine extract f0 together

Browse Source
This commit is contained in:
源文雨
2024-11-28 18:03:17 +09:00
parent d3add81469
commit 7befbd10d9
10 changed files with 280 additions and 691 deletions
Download Patch File Download Diff File Expand all files Collapse all files

175
infer/modules/train/extract/extract_f0_print.py
View File

@@ -1,175 +0,0 @@
import os
import sys
import traceback
import parselmouth
now_dir = os.getcwd()
sys.path.append(now_dir)
import logging
import numpy as np
import pyworld
from infer.lib.audio import load_audio
logging.getLogger("numba").setLevel(logging.WARNING)
from multiprocessing import Process
exp_dir = sys.argv[1]
f = open("%s/extract_f0_feature.log" % exp_dir, "a+")
def printt(strr):
print(strr)
f.write("%s\n" % strr)
f.flush()
n_p = int(sys.argv[2])
f0method = sys.argv[3]
class FeatureInput(object):
def __init__(self, samplerate=16000, hop_size=160):
self.fs = samplerate
self.hop = hop_size
self.f0_bin = 256
self.f0_max = 1100.0
self.f0_min = 50.0
self.f0_mel_min = 1127 * np.log(1 + self.f0_min / 700)
self.f0_mel_max = 1127 * np.log(1 + self.f0_max / 700)
def compute_f0(self, path, f0_method):
x = load_audio(path, self.fs)
p_len = x.shape[0] // self.hop
if f0_method == "pm":
time_step = 160 / 16000 * 1000
f0_min = 50
f0_max = 1100
f0 = (
parselmouth.Sound(x, self.fs)
.to_pitch_ac(
time_step=time_step / 1000,
voicing_threshold=0.6,
pitch_floor=f0_min,
pitch_ceiling=f0_max,
)
.selected_array["frequency"]
)
pad_size = (p_len - len(f0) + 1) // 2
if pad_size > 0 or p_len - len(f0) - pad_size > 0:
f0 = np.pad(
f0, [[pad_size, p_len - len(f0) - pad_size]], mode="constant"
)
elif f0_method == "harvest":
f0, t = pyworld.harvest(
x.astype(np.double),
fs=self.fs,
f0_ceil=self.f0_max,
f0_floor=self.f0_min,
frame_period=1000 * self.hop / self.fs,
)
f0 = pyworld.stonemask(x.astype(np.double), f0, t, self.fs)
elif f0_method == "dio":
f0, t = pyworld.dio(
x.astype(np.double),
fs=self.fs,
f0_ceil=self.f0_max,
f0_floor=self.f0_min,
frame_period=1000 * self.hop / self.fs,
)
f0 = pyworld.stonemask(x.astype(np.double), f0, t, self.fs)
elif f0_method == "rmvpe":
if hasattr(self, "model_rmvpe") == False:
from rvc.f0.rmvpe import RMVPE
print("Loading rmvpe model")
self.model_rmvpe = RMVPE(
"assets/rmvpe/rmvpe.pt", is_half=False, device="cpu"
)
f0 = self.model_rmvpe.compute_f0(x, filter_radius=0.03)
return f0
def coarse_f0(self, f0):
f0_mel = 1127 * np.log(1 + f0 / 700)
f0_mel[f0_mel > 0] = (f0_mel[f0_mel > 0] - self.f0_mel_min) * (
self.f0_bin - 2
) / (self.f0_mel_max - self.f0_mel_min) + 1
# use 0 or 1
f0_mel[f0_mel <= 1] = 1
f0_mel[f0_mel > self.f0_bin - 1] = self.f0_bin - 1
f0_coarse = np.rint(f0_mel).astype(int)
assert f0_coarse.max() <= 255 and f0_coarse.min() >= 1, (
f0_coarse.max(),
f0_coarse.min(),
)
return f0_coarse
def go(self, paths, f0_method):
if len(paths) == 0:
printt("no-f0-todo")
else:
printt("todo-f0-%s" % len(paths))
n = max(len(paths) // 5, 1) # 每个进程最多打印5条
for idx, (inp_path, opt_path1, opt_path2) in enumerate(paths):
try:
if idx % n == 0:
printt("f0ing,now-%s,all-%s,-%s" % (idx, len(paths), inp_path))
if (
os.path.exists(opt_path1 + ".npy") == True
and os.path.exists(opt_path2 + ".npy") == True
):
continue
featur_pit = self.compute_f0(inp_path, f0_method)
np.save(
opt_path2,
featur_pit,
allow_pickle=False,
) # nsf
coarse_pit = self.coarse_f0(featur_pit)
np.save(
opt_path1,
coarse_pit,
allow_pickle=False,
) # ori
except:
printt("f0fail-%s-%s-%s" % (idx, inp_path, traceback.format_exc()))
if __name__ == "__main__":
# exp_dir=r"E:\codes\py39\dataset\mi-test"
# n_p=16
# f = open("%s/log_extract_f0.log"%exp_dir, "w")
printt(" ".join(sys.argv))
featureInput = FeatureInput()
paths = []
inp_root = "%s/1_16k_wavs" % (exp_dir)
opt_root1 = "%s/2a_f0" % (exp_dir)
opt_root2 = "%s/2b-f0nsf" % (exp_dir)
os.makedirs(opt_root1, exist_ok=True)
os.makedirs(opt_root2, exist_ok=True)
for name in sorted(list(os.listdir(inp_root))):
inp_path = "%s/%s" % (inp_root, name)
if "spec" in inp_path:
continue
opt_path1 = "%s/%s" % (opt_root1, name)
opt_path2 = "%s/%s" % (opt_root2, name)
paths.append([inp_path, opt_path1, opt_path2])
ps = []
for i in range(n_p):
p = Process(
target=featureInput.go,
args=(
paths[i::n_p],
f0method,
),
)
ps.append(p)
p.start()
for i in range(n_p):
ps[i].join()

141
infer/modules/train/extract/extract_f0_rmvpe.py
View File

@@ -1,141 +0,0 @@
import os
import sys
import traceback
import parselmouth
now_dir = os.getcwd()
sys.path.append(now_dir)
import logging
import numpy as np
import pyworld
from infer.lib.audio import load_audio
logging.getLogger("numba").setLevel(logging.WARNING)
n_part = int(sys.argv[1])
i_part = int(sys.argv[2])
i_gpu = sys.argv[3]
os.environ["CUDA_VISIBLE_DEVICES"] = str(i_gpu)
exp_dir = sys.argv[4]
is_half = sys.argv[5]
f = open("%s/extract_f0_feature.log" % exp_dir, "a+")
def printt(strr):
print(strr)
f.write("%s\n" % strr)
f.flush()
class FeatureInput(object):
def __init__(self, samplerate=16000, hop_size=160):
self.fs = samplerate
self.hop = hop_size
self.f0_bin = 256
self.f0_max = 1100.0
self.f0_min = 50.0
self.f0_mel_min = 1127 * np.log(1 + self.f0_min / 700)
self.f0_mel_max = 1127 * np.log(1 + self.f0_max / 700)
def compute_f0(self, path, f0_method):
x = load_audio(path, self.fs)
# p_len = x.shape[0] // self.hop
if f0_method == "rmvpe":
if hasattr(self, "model_rmvpe") == False:
from rvc.f0.rmvpe import RMVPE
print("Loading rmvpe model")
self.model_rmvpe = RMVPE(
"assets/rmvpe/rmvpe.pt", is_half=is_half, device="cuda"
)
f0 = self.model_rmvpe.compute_f0(x, filter_radius=0.03)
return f0
def coarse_f0(self, f0):
f0_mel = 1127 * np.log(1 + f0 / 700)
f0_mel[f0_mel > 0] = (f0_mel[f0_mel > 0] - self.f0_mel_min) * (
self.f0_bin - 2
) / (self.f0_mel_max - self.f0_mel_min) + 1
# use 0 or 1
f0_mel[f0_mel <= 1] = 1
f0_mel[f0_mel > self.f0_bin - 1] = self.f0_bin - 1
f0_coarse = np.rint(f0_mel).astype(int)
assert f0_coarse.max() <= 255 and f0_coarse.min() >= 1, (
f0_coarse.max(),
f0_coarse.min(),
)
return f0_coarse
def go(self, paths, f0_method):
if len(paths) == 0:
printt("no-f0-todo")
else:
printt("todo-f0-%s" % len(paths))
n = max(len(paths) // 5, 1) # 每个进程最多打印5条
for idx, (inp_path, opt_path1, opt_path2) in enumerate(paths):
try:
if idx % n == 0:
printt("f0ing,now-%s,all-%s,-%s" % (idx, len(paths), inp_path))
if (
os.path.exists(opt_path1 + ".npy") == True
and os.path.exists(opt_path2 + ".npy") == True
):
continue
featur_pit = self.compute_f0(inp_path, f0_method)
np.save(
opt_path2,
featur_pit,
allow_pickle=False,
) # nsf
coarse_pit = self.coarse_f0(featur_pit)
np.save(
opt_path1,
coarse_pit,
allow_pickle=False,
) # ori
except:
printt("f0fail-%s-%s-%s" % (idx, inp_path, traceback.format_exc()))
if __name__ == "__main__":
# exp_dir=r"E:\codes\py39\dataset\mi-test"
# n_p=16
# f = open("%s/log_extract_f0.log"%exp_dir, "w")
printt(" ".join(sys.argv))
featureInput = FeatureInput()
paths = []
inp_root = "%s/1_16k_wavs" % (exp_dir)
opt_root1 = "%s/2a_f0" % (exp_dir)
opt_root2 = "%s/2b-f0nsf" % (exp_dir)
os.makedirs(opt_root1, exist_ok=True)
os.makedirs(opt_root2, exist_ok=True)
for name in sorted(list(os.listdir(inp_root))):
inp_path = "%s/%s" % (inp_root, name)
if "spec" in inp_path:
continue
opt_path1 = "%s/%s" % (opt_root1, name)
opt_path2 = "%s/%s" % (opt_root2, name)
paths.append([inp_path, opt_path1, opt_path2])
try:
featureInput.go(paths[i_part::n_part], "rmvpe")
except:
printt("f0_all_fail-%s" % (traceback.format_exc()))
# ps = []
# for i in range(n_p):
# p = Process(
# target=featureInput.go,
# args=(
# paths[i::n_p],
# f0method,
# ),
# )
# ps.append(p)
# p.start()
# for i in range(n_p):
# ps[i].join()

265
infer/modules/train/extract/extract_f0_rmvpe_dml.py → infer/modules/train/extract_f0_print.py
View File

@@ -1,139 +1,126 @@
import os
import sys
import traceback
import parselmouth
now_dir = os.getcwd()
sys.path.append(now_dir)
import logging
import numpy as np
import pyworld
from infer.lib.audio import load_audio
logging.getLogger("numba").setLevel(logging.WARNING)
exp_dir = sys.argv[1]
import torch_directml
device = torch_directml.device(torch_directml.default_device())
f = open("%s/extract_f0_feature.log" % exp_dir, "a+")
def printt(strr):
print(strr)
f.write("%s\n" % strr)
f.flush()
class FeatureInput(object):
def __init__(self, samplerate=16000, hop_size=160):
self.fs = samplerate
self.hop = hop_size
self.f0_bin = 256
self.f0_max = 1100.0
self.f0_min = 50.0
self.f0_mel_min = 1127 * np.log(1 + self.f0_min / 700)
self.f0_mel_max = 1127 * np.log(1 + self.f0_max / 700)
def compute_f0(self, path, f0_method):
x = load_audio(path, self.fs)
# p_len = x.shape[0] // self.hop
if f0_method == "rmvpe":
if hasattr(self, "model_rmvpe") == False:
from rvc.f0.rmvpe import RMVPE
print("Loading rmvpe model")
self.model_rmvpe = RMVPE(
"assets/rmvpe/rmvpe.pt", is_half=False, device=device
)
f0 = self.model_rmvpe.compute_f0(x, filter_radius=0.03)
return f0
def coarse_f0(self, f0):
f0_mel = 1127 * np.log(1 + f0 / 700)
f0_mel[f0_mel > 0] = (f0_mel[f0_mel > 0] - self.f0_mel_min) * (
self.f0_bin - 2
) / (self.f0_mel_max - self.f0_mel_min) + 1
# use 0 or 1
f0_mel[f0_mel <= 1] = 1
f0_mel[f0_mel > self.f0_bin - 1] = self.f0_bin - 1
f0_coarse = np.rint(f0_mel).astype(int)
assert f0_coarse.max() <= 255 and f0_coarse.min() >= 1, (
f0_coarse.max(),
f0_coarse.min(),
)
return f0_coarse
def go(self, paths, f0_method):
if len(paths) == 0:
printt("no-f0-todo")
else:
printt("todo-f0-%s" % len(paths))
n = max(len(paths) // 5, 1) # 每个进程最多打印5条
for idx, (inp_path, opt_path1, opt_path2) in enumerate(paths):
try:
if idx % n == 0:
printt("f0ing,now-%s,all-%s,-%s" % (idx, len(paths), inp_path))
if (
os.path.exists(opt_path1 + ".npy") == True
and os.path.exists(opt_path2 + ".npy") == True
):
continue
featur_pit = self.compute_f0(inp_path, f0_method)
np.save(
opt_path2,
featur_pit,
allow_pickle=False,
) # nsf
coarse_pit = self.coarse_f0(featur_pit)
np.save(
opt_path1,
coarse_pit,
allow_pickle=False,
) # ori
except:
printt("f0fail-%s-%s-%s" % (idx, inp_path, traceback.format_exc()))
if __name__ == "__main__":
# exp_dir=r"E:\codes\py39\dataset\mi-test"
# n_p=16
# f = open("%s/log_extract_f0.log"%exp_dir, "w")
printt(" ".join(sys.argv))
featureInput = FeatureInput()
paths = []
inp_root = "%s/1_16k_wavs" % (exp_dir)
opt_root1 = "%s/2a_f0" % (exp_dir)
opt_root2 = "%s/2b-f0nsf" % (exp_dir)
os.makedirs(opt_root1, exist_ok=True)
os.makedirs(opt_root2, exist_ok=True)
for name in sorted(list(os.listdir(inp_root))):
inp_path = "%s/%s" % (inp_root, name)
if "spec" in inp_path:
continue
opt_path1 = "%s/%s" % (opt_root1, name)
opt_path2 = "%s/%s" % (opt_root2, name)
paths.append([inp_path, opt_path1, opt_path2])
try:
featureInput.go(paths, "rmvpe")
except:
printt("f0_all_fail-%s" % (traceback.format_exc()))
# ps = []
# for i in range(n_p):
# p = Process(
# target=featureInput.go,
# args=(
# paths[i::n_p],
# f0method,
# ),
# )
# ps.append(p)
# p.start()
# for i in range(n_p):
# ps[i].join()
import os
import sys
import traceback
from pathlib import Path
from dotenv import load_dotenv
now_dir = os.getcwd()
sys.path.append(now_dir)
load_dotenv()
load_dotenv("sha256.env")
now_dir = os.getcwd()
sys.path.append(now_dir)
import logging
import numpy as np
from infer.lib.audio import load_audio
from rvc.f0 import Generator
logging.getLogger("numba").setLevel(logging.WARNING)
from multiprocessing import Process
exp_dir = sys.argv[1]
f = open("%s/extract_f0_feature.log" % exp_dir, "a+")
def printt(strr):
print(strr)
f.write("%s\n" % strr)
f.flush()
n_p = int(sys.argv[2])
f0method = sys.argv[3]
device = sys.argv[4]
is_half = sys.argv[5] == "True"
class FeatureInput(object):
def __init__(self, is_half: bool, device = "cpu", samplerate=16000, hop_size=160):
self.fs = samplerate
self.hop = hop_size
self.f0_bin = 256
self.f0_max = 1100.0
self.f0_min = 50.0
self.f0_mel_min = 1127 * np.log(1 + self.f0_min / 700)
self.f0_mel_max = 1127 * np.log(1 + self.f0_max / 700)
self.f0_gen = Generator(
Path(os.environ["rmvpe_root"]),
is_half,
0,
device,
hop_size,
samplerate,
)
def go(self, paths, f0_method):
if len(paths) == 0:
printt("no-f0-todo")
else:
printt("todo-f0-%s" % len(paths))
n = max(len(paths) // 5, 1) # 每个进程最多打印5条
for idx, (inp_path, opt_path1, opt_path2) in enumerate(paths):
try:
if idx % n == 0:
printt("f0ing,now-%s,all-%s,-%s" % (idx, len(paths), inp_path))
if (
os.path.exists(opt_path1 + ".npy") == True
and os.path.exists(opt_path2 + ".npy") == True
):
continue
x = load_audio(inp_path, self.fs)
coarse_pit, feature_pit = self.f0_gen.calculate(x, x.shape[0] // self.hop, 0, f0_method, None)
np.save(
opt_path2,
feature_pit,
allow_pickle=False,
) # nsf
np.save(
opt_path1,
coarse_pit,
allow_pickle=False,
) # ori
except:
printt("f0fail-%s-%s-%s" % (idx, inp_path, traceback.format_exc()))
if __name__ == "__main__":
# exp_dir=r"E:\codes\py39\dataset\mi-test"
# n_p=16
# f = open("%s/log_extract_f0.log"%exp_dir, "w")
printt(" ".join(sys.argv))
featureInput = FeatureInput(is_half, device)
paths = []
inp_root = "%s/1_16k_wavs" % (exp_dir)
opt_root1 = "%s/2a_f0" % (exp_dir)
opt_root2 = "%s/2b-f0nsf" % (exp_dir)
os.makedirs(opt_root1, exist_ok=True)
os.makedirs(opt_root2, exist_ok=True)
for name in sorted(list(os.listdir(inp_root))):
inp_path = "%s/%s" % (inp_root, name)
if "spec" in inp_path:
continue
opt_path1 = "%s/%s" % (opt_root1, name)
opt_path2 = "%s/%s" % (opt_root2, name)
paths.append([inp_path, opt_path1, opt_path2])
ps = []
for i in range(n_p):
p = Process(
target=featureInput.go,
args=(
paths[i::n_p],
f0method,
),
)
ps.append(p)
p.start()
for i in range(n_p):
ps[i].join()

101
infer/modules/vc/pipeline.py
View File

@@ -5,6 +5,7 @@ import logging
logger = logging.getLogger(__name__)
from pathlib import Path
from time import time
import faiss
@@ -14,7 +15,7 @@ import torch
import torch.nn.functional as F
from scipy import signal
from rvc.f0 import PM, Harvest, RMVPE, CRePE, Dio, FCPE
from rvc.f0 import Generator
now_dir = os.getcwd()
sys.path.append(now_dir)
@@ -63,95 +64,15 @@ class Pipeline(object):
self.t_max = self.sr * self.x_max # 免查询时长阈值
self.device = config.device
def get_f0(
self,
x,
p_len,
f0_up_key,
f0_method,
filter_radius,
inp_f0=None,
):
f0_min = 50
f0_max = 1100
f0_mel_min = 1127 * np.log(1 + f0_min / 700)
f0_mel_max = 1127 * np.log(1 + f0_max / 700)
if f0_method == "pm":
if not hasattr(self, "pm"):
self.pm = PM(self.window, f0_min, f0_max, self.sr)
f0 = self.pm.compute_f0(x, p_len=p_len)
if f0_method == "dio":
if not hasattr(self, "dio"):
self.dio = Dio(self.window, f0_min, f0_max, self.sr)
f0 = self.dio.compute_f0(x, p_len=p_len)
elif f0_method == "harvest":
if not hasattr(self, "harvest"):
self.harvest = Harvest(self.window, f0_min, f0_max, self.sr)
f0 = self.harvest.compute_f0(x, p_len=p_len, filter_radius=filter_radius)
elif f0_method == "crepe":
if not hasattr(self, "crepe"):
self.crepe = CRePE(
self.window,
f0_min,
f0_max,
self.sr,
self.device,
)
f0 = self.crepe.compute_f0(x, p_len=p_len)
elif f0_method == "rmvpe":
if not hasattr(self, "rmvpe"):
logger.info(
"Loading rmvpe model %s" % "%s/rmvpe.pt" % os.environ["rmvpe_root"]
)
self.rmvpe = RMVPE(
"%s/rmvpe.pt" % os.environ["rmvpe_root"],
is_half=self.is_half,
device=self.device,
# use_jit=self.config.use_jit,
)
f0 = self.rmvpe.compute_f0(x, p_len=p_len, filter_radius=0.03)
self.f0_gen = Generator(
Path(os.environ["rmvpe_root"]),
self.is_half,
self.x_pad,
self.device,
self.window,
self.sr,
)
if "privateuseone" in str(self.device): # clean ortruntime memory
del self.rmvpe.model
del self.rmvpe
logger.info("Cleaning ortruntime memory")
elif f0_method == "fcpe":
if not hasattr(self, "model_fcpe"):
logger.info("Loading fcpe model")
self.model_fcpe = FCPE(
self.window,
f0_min,
f0_max,
self.sr,
self.device,
)
f0 = self.model_fcpe.compute_f0(x, p_len=p_len)
f0 *= pow(2, f0_up_key / 12)
# with open("test.txt","w")as f:f.write("\n".join([str(i)for i in f0.tolist()]))
tf0 = self.sr // self.window # 每秒f0点数
if inp_f0 is not None:
delta_t = np.round(
(inp_f0[:, 0].max() - inp_f0[:, 0].min()) * tf0 + 1
).astype("int16")
replace_f0 = np.interp(
list(range(delta_t)), inp_f0[:, 0] * 100, inp_f0[:, 1]
)
shape = f0[self.x_pad * tf0 : self.x_pad * tf0 + len(replace_f0)].shape[0]
f0[self.x_pad * tf0 : self.x_pad * tf0 + len(replace_f0)] = replace_f0[
:shape
]
# with open("test_opt.txt","w")as f:f.write("\n".join([str(i)for i in f0.tolist()]))
f0bak = f0.copy()
f0_mel = 1127 * np.log(1 + f0 / 700)
f0_mel[f0_mel > 0] = (f0_mel[f0_mel > 0] - f0_mel_min) * 254 / (
f0_mel_max - f0_mel_min
) + 1
f0_mel[f0_mel <= 1] = 1
f0_mel[f0_mel > 255] = 255
f0_coarse = np.rint(f0_mel).astype(np.int32)
return f0_coarse, f0bak # 1-0
def vc(
self,
@@ -337,7 +258,7 @@ class Pipeline(object):
pitch, pitchf = None, None
if if_f0:
if if_f0 == 1:
pitch, pitchf = self.get_f0(
pitch, pitchf = self.f0_gen.calculate(
audio_pad,
p_len,
f0_up_key,