CoverGen-RVC / src /rvc.py
Politrees's picture
Update src/rvc.py
9e66ad7 verified
raw
history blame
5.69 kB
from multiprocessing import cpu_count
from pathlib import Path
import torch
from fairseq import checkpoint_utils
from scipy.io import wavfile
from infer_pack.models import (
SynthesizerTrnMs256NSFsid,
SynthesizerTrnMs256NSFsid_nono,
SynthesizerTrnMs768NSFsid,
SynthesizerTrnMs768NSFsid_nono,
)
from my_utils import load_audio
from vc_infer_pipeline import VC
BASE_DIR = Path(__file__).resolve().parent.parent
class Config:
def __init__(self, device, is_half):
self.device = device
self.is_half = is_half
self.n_cpu = 0
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:
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(BASE_DIR / "src" / "configs" / config_file, "r") as f:
strr = f.read().replace("true", "false")
with open(BASE_DIR / "src" / "configs" / config_file, "w") as f:
f.write(strr)
with open(BASE_DIR / "src" / "trainset_preprocess_pipeline_print.py", "r") as f:
strr = f.read().replace("3.7", "3.0")
with open(BASE_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(BASE_DIR / "src" / "trainset_preprocess_pipeline_print.py", "r") as f:
strr = f.read().replace("3.7", "3.0")
with open(BASE_DIR / "src" / "trainset_preprocess_pipeline_print.py", "w") as f:
f.write(strr)
elif torch.backends.mps.is_available():
print("No supported N-card found, use MPS for inference")
self.device = "mps"
else:
print("No supported N-card found, use CPU for inference")
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
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
return x_pad, x_query, x_center, x_max
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)
if is_half:
hubert = hubert.half()
else:
hubert = hubert.float()
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.')
tgt_sr = cpt["config"][-1]
cpt["config"][-3] = cpt["weight"]["emb_g.weight"].shape[0]
if_f0 = cpt.get("f0", 1)
version = cpt.get("version", "v1")
if version == "v1":
if if_f0 == 1:
net_g = SynthesizerTrnMs256NSFsid(*cpt["config"], is_half=is_half)
else:
net_g = SynthesizerTrnMs256NSFsid_nono(*cpt["config"])
elif version == "v2":
if if_f0 == 1:
net_g = SynthesizerTrnMs768NSFsid(*cpt["config"], is_half=is_half)
else:
net_g = SynthesizerTrnMs768NSFsid_nono(*cpt["config"])
del net_g.enc_q
print(net_g.load_state_dict(cpt["weight"], strict=False))
net_g.eval().to(device)
if is_half:
net_g = net_g.half()
else:
net_g = net_g.float()
vc = VC(tgt_sr, config)
return cpt, version, net_g, tgt_sr, vc
def rvc_infer(
index_path,
index_rate,
input_path,
output_path,
pitch_change,
f0_method,
cpt,
version,
net_g,
filter_radius,
tgt_sr,
rms_mix_rate,
protect,
crepe_hop_length,
vc,
hubert_model,
f0_min=50,
f0_max=1100
):
audio = load_audio(input_path, 16000)
times = [0, 0, 0]
if_f0 = cpt.get('f0', 1)
audio_opt = vc.pipeline(
hubert_model,
net_g,
0,
audio,
input_path,
times,
pitch_change,
f0_method,
index_path,
index_rate,
if_f0,
filter_radius,
tgt_sr,
0,
rms_mix_rate,
version,
protect,
crepe_hop_length,
f0_min=f0_min,
f0_max=f0_max
)
wavfile.write(output_path, tgt_sr, audio_opt)