Skip to content

Commit

Permalink
Update rvc.py
Browse files Browse the repository at this point in the history
  • Loading branch information
@Bebra777228
Bebra777228 authored Aug 2, 2024
1 parent a6147f5 commit 7253b1c
Showing 1 changed file with 73 additions and 100 deletions.
173 changes: 73 additions & 100 deletions src/rvc.py
View file
Original file line number Diff line number Diff line change
@@ -1,12 +1,12 @@
import logging
import os
from multiprocessing import cpu_count
from pathlib import Path

import os
import torch
from fairseq import checkpoint_utils
from scipy.io import wavfile

now_dir = os.getcwd()
now_dir = Path(os.getcwd())

from src.infer_pack.models import (
SynthesizerTrnMs256NSFsid,
Expand All @@ -17,89 +17,67 @@
from src.my_utils import load_audio
from src.vc_infer_pipeline import VC

logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)


class Config:
def __init__(self, device, is_half):
self.device = device
self.is_half = is_half
self.n_cpu = 0
self.n_cpu = cpu_count()
self.gpu_name = None
self.gpu_mem = None
self.x_pad, self.x_query, self.x_center, self.x_max = self.device_config()

def device_config(self) -> tuple:
def device_config(self):
if torch.cuda.is_available():
i_device = int(self.device.split(":")[-1])
self.gpu_name = torch.cuda.get_device_name(i_device)
if (
("16" in self.gpu_name and "V100" not in self.gpu_name.upper())
or "P40" in self.gpu_name.upper()
or "1060" in self.gpu_name
or "1070" in self.gpu_name
or "1080" in self.gpu_name
):
print("16 series/10 series P40 forced single precision")
self.is_half = False
for config_file in ["32k.json", "40k.json", "48k.json"]:
with open(now_dir / "src" / "configs" / config_file, "r") as f:
strr = f.read().replace("true", "false")
with open(now_dir / "src" / "configs" / config_file, "w") as f:
f.write(strr)
with open(now_dir / "src" / "trainset_preprocess_pipeline_print.py", "r") as f:
strr = f.read().replace("3.7", "3.0")
with open(now_dir / "src" / "trainset_preprocess_pipeline_print.py", "w") as f:
f.write(strr)
else:
self.gpu_name = None
self.gpu_mem = int(
torch.cuda.get_device_properties(i_device).total_memory
/ 1024
/ 1024
/ 1024
+ 0.4
)
if self.gpu_mem <= 4:
with open(now_dir / "src" / "trainset_preprocess_pipeline_print.py", "r") as f:
strr = f.read().replace("3.7", "3.0")
with open(now_dir / "src" / "trainset_preprocess_pipeline_print.py", "w") as f:
f.write(strr)
self._configure_gpu()
elif torch.backends.mps.is_available():
print("No supported N-card found, use MPS for inference")
logger.info("Не обнаружена поддерживаемая N-карта, используйте MPS для вывода")
self.device = "mps"
else:
print("No supported N-card found, use CPU for inference")
logger.info("Не обнаружена поддерживаемая N-карта, используйте CPU для вывода")
self.device = "cpu"
self.is_half = True

if self.n_cpu == 0:
self.n_cpu = cpu_count()

if self.is_half:
# 6G memory config
x_pad = 3
x_query = 10
x_center = 60
x_max = 65
x_pad, x_query, x_center, x_max = 3, 10, 60, 65
else:
# 5G memory config
x_pad = 1
x_query = 6
x_center = 38
x_max = 41

if self.gpu_mem != None and self.gpu_mem <= 4:
x_pad = 1
x_query = 5
x_center = 30
x_max = 32
x_pad, x_query, x_center, x_max = 1, 6, 38, 41

if self.gpu_mem is not None and self.gpu_mem <= 4:
x_pad, x_query, x_center, x_max = 1, 5, 30, 32

return x_pad, x_query, x_center, x_max

def _configure_gpu(self):
self.gpu_name = torch.cuda.get_device_name(self.device)
if "16" in self.gpu_name and "V100" not in self.gpu_name.upper() or "P40" in self.gpu_name.upper() or "1060" in self.gpu_name or "1070" in self.gpu_name or "1080" in self.gpu_name:
logger.info("16 серия/10 серия P40 принудительно используется одинарная точность")
self.is_half = False
self._update_config_files()
self.gpu_mem = int(torch.cuda.get_device_properties(self.device).total_memory / 1024 / 1024 / 1024 + 0.4)
if self.gpu_mem <= 4:
self._update_config_files()

def _update_config_files(self):
for config_file in ["32k.json", "40k.json", "48k.json"]:
config_path = now_dir / "src" / "configs" / config_file
self._replace_in_file(config_path, "true", "false")
trainset_path = now_dir / "src" / "trainset_preprocess_pipeline_print.py"
self._replace_in_file(trainset_path, "3.7", "3.0")

@staticmethod
def _replace_in_file(file_path, old, new):
with open(file_path, "r") as f:
content = f.read().replace(old, new)
with open(file_path, "w") as f:
f.write(content)

def load_hubert(device, is_half, model_path):
models, saved_cfg, task = checkpoint_utils.load_model_ensemble_and_task([model_path], suffix='', )
hubert = models[0]
hubert = hubert.to(device)
models, saved_cfg, task = checkpoint_utils.load_model_ensemble_and_task([model_path], suffix='')
hubert = models[0].to(device)

if is_half:
hubert = hubert.half()
Expand All @@ -109,30 +87,23 @@ def load_hubert(device, is_half, model_path):
hubert.eval()
return hubert


def get_vc(device, is_half, config, model_path):
cpt = torch.load(model_path, map_location='cpu')
if "config" not in cpt or "weight" not in cpt:
raise ValueError(f'Incorrect format for {model_path}. Use a voice model trained using RVC v2 instead.')
raise ValueError(f'Некорректный формат для {model_path}. Используйте голосовую модель, обученную с использованием RVC v2.')

tgt_sr = cpt["config"][-1]
cpt["config"][-3] = cpt["weight"]["emb_g.weight"].shape[0]
pitch_guidance = cpt.get("f0", 1)
version = cpt.get("version", "v1")

if version == "v1":
if pitch_guidance == 1:
net_g = SynthesizerTrnMs256NSFsid(*cpt["config"], is_half=is_half)
else:
net_g = SynthesizerTrnMs256NSFsid_nono(*cpt["config"])
net_g = SynthesizerTrnMs256NSFsid(*cpt["config"], is_half=is_half) if pitch_guidance == 1 else SynthesizerTrnMs256NSFsid_nono(*cpt["config"])
elif version == "v2":
if pitch_guidance == 1:
net_g = SynthesizerTrnMs768NSFsid(*cpt["config"], is_half=is_half)
else:
net_g = SynthesizerTrnMs768NSFsid_nono(*cpt["config"])
net_g = SynthesizerTrnMs768NSFsid(*cpt["config"], is_half=is_half) if pitch_guidance == 1 else SynthesizerTrnMs768NSFsid_nono(*cpt["config"])

del net_g.enc_q
print(net_g.load_state_dict(cpt["weight"], strict=False))
logger.info(net_g.load_state_dict(cpt["weight"], strict=False))
net_g.eval().to(device)

if is_half:
Expand All @@ -143,7 +114,6 @@ def get_vc(device, is_half, config, model_path):
vc = VC(tgt_sr, config)
return cpt, version, net_g, tgt_sr, vc


def rvc_infer(
index_path,
index_rate,
Expand All @@ -164,28 +134,31 @@ def rvc_infer(
f0_min=50,
f0_max=1100
):
audio = load_audio(input_path, 16000)
pitch_guidance = cpt.get('f0', 1)
audio_opt = vc.pipeline(
hubert_model,
net_g,
0,
audio,
input_path,
pitch_change,
f0_method,
index_path,
index_rate,
pitch_guidance,
filter_radius,
tgt_sr,
0,
volume_envelope,
version,
protect,
hop_length,
f0_file=None,
f0_min=f0_min,
f0_max=f0_max
)
wavfile.write(output_path, tgt_sr, audio_opt)
try:
audio = load_audio(input_path, 16000)
pitch_guidance = cpt.get('f0', 1)
audio_opt = vc.pipeline(
hubert_model,
net_g,
0,
audio,
input_path,
pitch_change,
f0_method,
index_path,
index_rate,
pitch_guidance,
filter_radius,
tgt_sr,
0,
volume_envelope,
version,
protect,
hop_length,
f0_file=None,
f0_min=f0_min,
f0_max=f0_max
)
wavfile.write(output_path, tgt_sr, audio_opt)
except Exception as e:
logger.error(f"Ошибка во время вывода: {e}")

0 comments on commit 7253b1c

Please sign in to comment.