fumiama/Retrieval-based-Voice-Conversion-WebUI
1
0
Fork 0
mirror of https://github.com/fumiama/Retrieval-based-Voice-Conversion-WebUI.git synced 2026-06-10 21:24:16 +08:00
Code Activity

optimize(rvc.onnx): add types defs

Browse Source
This commit is contained in:
源文雨
2024-06-06 01:01:59 +09:00
parent f60bebe89c
commit 6e8feb9028
6 changed files with 123 additions and 214 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
rvc/onnx/f0predictor/__init__.py
View File

@@ -1,3 +1,4 @@
from .dio import DioF0Predictor
from .harvest import HarvestF0Predictor
from .pm import PMF0Predictor
from .f0 import F0Predictor

63
rvc/onnx/f0predictor/dio.py
View File

@@ -1,66 +1,15 @@
import numpy as np
import pyworld
import typing
from .f0 import F0Predictor
class DioF0Predictor(F0Predictor):
def __init__(self, hop_length=512, f0_min=50, f0_max=1100, sampling_rate=44100):
self.hop_length = hop_length
self.f0_min = f0_min
self.f0_max = f0_max
self.sampling_rate = sampling_rate
super().__init__(hop_length, f0_min, f0_max, sampling_rate)
def interpolate_f0(self, f0):
"""
对F0进行插值处理
"""
data = np.reshape(f0, (f0.size, 1))
vuv_vector = np.zeros((data.size, 1), dtype=np.float32)
vuv_vector[data > 0.0] = 1.0
vuv_vector[data <= 0.0] = 0.0
ip_data = data
frame_number = data.size
last_value = 0.0
for i in range(frame_number):
if data[i] <= 0.0:
j = i + 1
for j in range(i + 1, frame_number):
if data[j] > 0.0:
break
if j < frame_number - 1:
if last_value > 0.0:
step = (data[j] - data[i - 1]) / float(j - i)
for k in range(i, j):
ip_data[k] = data[i - 1] + step * (k - i + 1)
else:
for k in range(i, j):
ip_data[k] = data[j]
else:
for k in range(i, frame_number):
ip_data[k] = last_value
else:
ip_data[i] = data[i] # 这里可能存在一个没有必要的拷贝
last_value = data[i]
return ip_data[:, 0], vuv_vector[:, 0]
def resize_f0(self, x, target_len):
source = np.array(x)
source[source < 0.001] = np.nan
target = np.interp(
np.arange(0, len(source) * target_len, len(source)) / target_len,
np.arange(0, len(source)),
source,
)
res = np.nan_to_num(target)
return res
def compute_f0(self, wav, p_len=None):
def compute_f0(self, wav: np.ndarray[typing.Any, np.dtype], p_len: int | None = None):
if p_len is None:
p_len = wav.shape[0] // self.hop_length
f0, t = pyworld.dio(
@@ -73,9 +22,9 @@ class DioF0Predictor(F0Predictor):
f0 = pyworld.stonemask(wav.astype(np.double), f0, t, self.sampling_rate)
for index, pitch in enumerate(f0):
f0[index] = round(pitch, 1)
return self.interpolate_f0(self.resize_f0(f0, p_len))[0]
return self.__interpolate_f0(self.__resize_f0(f0, p_len))[0]
def compute_f0_uv(self, wav, p_len=None):
def compute_f0_uv(self, wav: np.ndarray[typing.Any, np.dtype], p_len: int | None = None):
if p_len is None:
p_len = wav.shape[0] // self.hop_length
f0, t = pyworld.dio(
@@ -88,4 +37,4 @@ class DioF0Predictor(F0Predictor):
f0 = pyworld.stonemask(wav.astype(np.double), f0, t, self.sampling_rate)
for index, pitch in enumerate(f0):
f0[index] = round(pitch, 1)
return self.interpolate_f0(self.resize_f0(f0, p_len))
return self.__interpolate_f0(self.__resize_f0(f0, p_len))

72
rvc/onnx/f0predictor/f0.py
View File

@@ -1,16 +1,62 @@
class F0Predictor(object):
def compute_f0(self, wav, p_len):
"""
input: wav:[signal_length]
p_len:int
output: f0:[signal_length//hop_length]
"""
pass
import numpy as np
import typing
def compute_f0_uv(self, wav, p_len):
class F0Predictor(object):
def __init__(self, hop_length=512, f0_min=50, f0_max=1100, sampling_rate=44100):
self.hop_length = hop_length
self.f0_min = f0_min
self.f0_max = f0_max
self.sampling_rate = sampling_rate
def compute_f0(self, wav: np.ndarray[typing.Any, np.dtype], p_len: int | None = None): ...
def compute_f0_uv(self, wav: np.ndarray[typing.Any, np.dtype], p_len: int | None = None): ...
def __interpolate_f0(self, f0: np.ndarray[typing.Any, np.dtype]):
"""
input: wav:[signal_length]
p_len:int
output: f0:[signal_length//hop_length],uv:[signal_length//hop_length]
对F0进行插值处理
"""
pass
data = np.reshape(f0, (f0.size, 1))
vuv_vector = np.zeros((data.size, 1), dtype=np.float32)
vuv_vector[data > 0.0] = 1.0
vuv_vector[data <= 0.0] = 0.0
ip_data = data
frame_number = data.size
last_value = 0.0
for i in range(frame_number):
if data[i] <= 0.0:
j = i + 1
for j in range(i + 1, frame_number):
if data[j] > 0.0:
break
if j < frame_number - 1:
if last_value > 0.0:
step = (data[j] - data[i - 1]) / float(j - i)
for k in range(i, j):
ip_data[k] = data[i - 1] + step * (k - i + 1)
else:
for k in range(i, j):
ip_data[k] = data[j]
else:
for k in range(i, frame_number):
ip_data[k] = last_value
else:
ip_data[i] = data[i] # 这里可能存在一个没有必要的拷贝
last_value = data[i]
return ip_data[:, 0], vuv_vector[:, 0]
def __resize_f0(self, x: np.ndarray[typing.Any, np.dtype], target_len: int):
source = np.array(x)
source[source < 0.001] = np.nan
target = np.interp(
np.arange(0, len(source) * target_len, len(source)) / target_len,
np.arange(0, len(source)),
source,
)
res = np.nan_to_num(target)
return res

63
rvc/onnx/f0predictor/harvest.py
View File

@@ -1,66 +1,15 @@
import numpy as np
import pyworld
import typing
from .f0 import F0Predictor
class HarvestF0Predictor(F0Predictor):
def __init__(self, hop_length=512, f0_min=50, f0_max=1100, sampling_rate=44100):
self.hop_length = hop_length
self.f0_min = f0_min
self.f0_max = f0_max
self.sampling_rate = sampling_rate
super().__init__(hop_length, f0_min, f0_max, sampling_rate)
def interpolate_f0(self, f0):
"""
对F0进行插值处理
"""
data = np.reshape(f0, (f0.size, 1))
vuv_vector = np.zeros((data.size, 1), dtype=np.float32)
vuv_vector[data > 0.0] = 1.0
vuv_vector[data <= 0.0] = 0.0
ip_data = data
frame_number = data.size
last_value = 0.0
for i in range(frame_number):
if data[i] <= 0.0:
j = i + 1
for j in range(i + 1, frame_number):
if data[j] > 0.0:
break
if j < frame_number - 1:
if last_value > 0.0:
step = (data[j] - data[i - 1]) / float(j - i)
for k in range(i, j):
ip_data[k] = data[i - 1] + step * (k - i + 1)
else:
for k in range(i, j):
ip_data[k] = data[j]
else:
for k in range(i, frame_number):
ip_data[k] = last_value
else:
ip_data[i] = data[i] # 这里可能存在一个没有必要的拷贝
last_value = data[i]
return ip_data[:, 0], vuv_vector[:, 0]
def resize_f0(self, x, target_len):
source = np.array(x)
source[source < 0.001] = np.nan
target = np.interp(
np.arange(0, len(source) * target_len, len(source)) / target_len,
np.arange(0, len(source)),
source,
)
res = np.nan_to_num(target)
return res
def compute_f0(self, wav, p_len=None):
def compute_f0(self, wav: np.ndarray[typing.Any, np.dtype], p_len: int | None = None):
if p_len is None:
p_len = wav.shape[0] // self.hop_length
f0, t = pyworld.harvest(
@@ -71,9 +20,9 @@ class HarvestF0Predictor(F0Predictor):
frame_period=1000 * self.hop_length / self.sampling_rate,
)
f0 = pyworld.stonemask(wav.astype(np.double), f0, t, self.fs)
return self.interpolate_f0(self.resize_f0(f0, p_len))[0]
return self.__interpolate_f0(self.__resize_f0(f0, p_len))[0]
def compute_f0_uv(self, wav, p_len=None):
def compute_f0_uv(self, wav: np.ndarray[typing.Any, np.dtype], p_len: int | None = None):
if p_len is None:
p_len = wav.shape[0] // self.hop_length
f0, t = pyworld.harvest(
@@ -84,4 +33,4 @@ class HarvestF0Predictor(F0Predictor):
frame_period=1000 * self.hop_length / self.sampling_rate,
)
f0 = pyworld.stonemask(wav.astype(np.double), f0, t, self.sampling_rate)
return self.interpolate_f0(self.resize_f0(f0, p_len))
return self.__interpolate_f0(self.__resize_f0(f0, p_len))

52
rvc/onnx/f0predictor/pm.py
View File

@@ -1,55 +1,15 @@
import numpy as np
import parselmouth
import typing
from .f0 import F0Predictor
class PMF0Predictor(F0Predictor):
def __init__(self, hop_length=512, f0_min=50, f0_max=1100, sampling_rate=44100):
self.hop_length = hop_length
self.f0_min = f0_min
self.f0_max = f0_max
self.sampling_rate = sampling_rate
super().__init__(hop_length, f0_min, f0_max, sampling_rate)
def interpolate_f0(self, f0):
"""
对F0进行插值处理
"""
data = np.reshape(f0, (f0.size, 1))
vuv_vector = np.zeros((data.size, 1), dtype=np.float32)
vuv_vector[data > 0.0] = 1.0
vuv_vector[data <= 0.0] = 0.0
ip_data = data
frame_number = data.size
last_value = 0.0
for i in range(frame_number):
if data[i] <= 0.0:
j = i + 1
for j in range(i + 1, frame_number):
if data[j] > 0.0:
break
if j < frame_number - 1:
if last_value > 0.0:
step = (data[j] - data[i - 1]) / float(j - i)
for k in range(i, j):
ip_data[k] = data[i - 1] + step * (k - i + 1)
else:
for k in range(i, j):
ip_data[k] = data[j]
else:
for k in range(i, frame_number):
ip_data[k] = last_value
else:
ip_data[i] = data[i] # 这里可能存在一个没有必要的拷贝
last_value = data[i]
return ip_data[:, 0], vuv_vector[:, 0]
def compute_f0(self, wav, p_len=None):
def compute_f0(self, wav: np.ndarray[typing.Any, np.dtype], p_len: int | None = None):
x = wav
if p_len is None:
p_len = x.shape[0] // self.hop_length
@@ -70,10 +30,10 @@ class PMF0Predictor(F0Predictor):
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")
f0, uv = self.interpolate_f0(f0)
f0, uv = self.__interpolate_f0(f0)
return f0
def compute_f0_uv(self, wav, p_len=None):
def compute_f0_uv(self, wav: np.ndarray[typing.Any, np.dtype], p_len: int | None = None):
x = wav
if p_len is None:
p_len = x.shape[0] // self.hop_length
@@ -94,5 +54,5 @@ class PMF0Predictor(F0Predictor):
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")
f0, uv = self.interpolate_f0(f0)
f0, uv = self.__interpolate_f0(f0)
return f0, uv