src/rvc.py

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)