Colab에서 RVC 모델로 Train 하기

이진나무·2023년 9월 20일

AI 노래 Cover

목록 보기

3/5

Intro

Google Colab에서 무료 이용자에게 gradio UI 사용을 내용으로 새로운 제한사항을 추가하였다. 따라서 Easy GUI 또한 이에 영향을 받아 유료 사용자가 아닌 이상 graio UI 기반으로 사용할 수 없게 되었다. 이 변동사항에 대응하여 colab에서 수동으로 코드를 직접 실행할 수 있게끔 하는 노트북을 추가하였다.

처음에는 당황했는데 오히려 UI 뒤에 가려져있던 전체 코드가 깔끔한 형식으로 공개되어 RVC 모델 공부에 도움이 되었다.

기존 EasyGUI
https://colab.research.google.com/drive/1r4IRL0UA7JEoZ0ZK8PKfMyTIBHKpyhcw#scrollTo=FFfC9x239kC1

Colab 기반으로 RVC 실행할 수 있게 해놓은 또다른 기여자

https://github.com/ardha27/AI-Song-Cover-RVC

00.Setup

drive.mount는 찝찝하므로 주석처리하고
if not os.path.exists('/content/drive'): print("Your drive is not mounted. Creating Fake Drive.")
조건문 타고 들어가도록 처리
온갖 설치가 이뤄지는 단계. 엄청나게 긴 로그가 3분가량 줄줄 출력되다가 Success 버튼이 뿅 나타난다.

%cd /content
#from google.colab import drive
#drive.mount('/content/drive')
from IPython.display import clear_output
from ipywidgets import Button
import os
if not os.path.exists('/content/drive'):
    print("Your drive is not mounted. Creating Fake Drive.")
    os.makedirs('/content/drive/MyDrive')
if not os.path.exists('/content/drive/MyDrive/project-main'):
    !wget https://huggingface.co/Rejekts/project/resolve/main/project-main.zip -O '/content/project-main.zip' && unzip 'project-main.zip' -d /content/drive/MyDrive
!cd '/content/drive/MyDrive/project-main' && python download_files.py && pip install -r 'requirements-safe.txt'
!rm /content/project-main.zip
!rm -r /content/sample_data
!mkdir -p /content/dataset
clear_output()
Button(description="\u2714 Success", button_style="success")

01.Preprocess Data

model_name 입력해준다
/content/dataset/ 에 학습용 데이터 업로드한다
/content/drive/MyDrive/project-main Starting... 로그 한줄 뜨고 동일하게 Success 버튼 뿅 나타난다

%cd /content/drive/MyDrive/project-main
model_name = 'My-Voice' #@param {type:"string"}
#@markdown <small> Enter the path to your dataset, or if you want just upload the audios using the File Manager into the 'dataset' folder.
dataset_folder = '/content/dataset' #@param {type:"string"}
while len(os.listdir(dataset_folder)) < 1:
    input("Your dataset folder is empty.")
!mkdir -p ./logs/{model_name}
with open(f'./logs/{model_name}/preprocess.log','w') as f:
    print("Starting...")
!python infer/modules/train/preprocess.py {dataset_folder} 40000 2 ./logs/{model_name} False 3.0 > /dev/null 2>&1
with open(f'./logs/{model_name}/preprocess.log','r') as f:
    if 'end preprocess' in f.read():
        clear_output()
        display(Button(description="\u2714 Success", button_style="success"))
    else:
        print("Error preprocessing data... Make sure your dataset folder is correct.")

2.Extract Features

"pm", "harvest", "rmvpe", "rmvpe_gpu" 중에서 하나 선택
동일하게 Success 버튼 뿅 나타난다

f0method = "rmvpe_gpu" # @param ["pm", "harvest", "rmvpe", "rmvpe_gpu"]
%cd /content/drive/MyDrive/project-main
with open(f'./logs/{model_name}/extract_f0_feature.log','w') as f:
    print("Starting...")
if f0method != "rmvpe_gpu":
    !python infer/modules/train/extract/extract_f0_print.py ./logs/{model_name} 2 {f0method}
else:
    !python infer/modules/train/extract/extract_f0_rmvpe.py 1 0 0 ./logs/{model_name} True
!python infer/modules/train/extract_feature_print.py cuda:0 1 0 0 ./logs/{model_name} v2
with open(f'./logs/{model_name}/extract_f0_feature.log','r') as f:
    if 'all-feature-done' in f.read():
        clear_output()
        display(Button(description="\u2714 Success", button_style="success"))
    else:
        print("Error preprocessing data... Make sure your data was preprocessed.")

03.Train Index

동일하게 Success 버튼 뿅 나타난다

import numpy as np
import faiss
%cd /content/drive/MyDrive/project-main
def train_index(exp_dir1, version19):
   exp_dir = "logs/%s" % (exp_dir1)
   os.makedirs(exp_dir, exist_ok=True)
   feature_dir = (
       "%s/3_feature256" % (exp_dir)
       if version19 == "v1"
       else "%s/3_feature768" % (exp_dir)
   )
   if not os.path.exists(feature_dir):
       return "请先进行特征提取!"
   listdir_res = list(os.listdir(feature_dir))
   if len(listdir_res) == 0:
       return "请先进行特征提取！"
   infos = []
   npys = []
   for name in sorted(listdir_res):
       phone = np.load("%s/%s" % (feature_dir, name))
       npys.append(phone)
   big_npy = np.concatenate(npys, 0)
   big_npy_idx = np.arange(big_npy.shape[0])
   np.random.shuffle(big_npy_idx)
   big_npy = big_npy[big_npy_idx]
   if big_npy.shape[0] > 2e5:
       infos.append("Trying doing kmeans %s shape to 10k centers." % big_npy.shape[0])
       yield "\n".join(infos)
       try:
           big_npy = (
               MiniBatchKMeans(
                   n_clusters=10000,
                   verbose=True,
                   batch_size=256 * config.n_cpu,
                   compute_labels=False,
                   init="random",
               )
               .fit(big_npy)
               .cluster_centers_
           )
       except:
           info = traceback.format_exc()
           logger.info(info)
           infos.append(info)
           yield "\n".join(infos)

   np.save("%s/total_fea.npy" % exp_dir, big_npy)
   n_ivf = min(int(16 * np.sqrt(big_npy.shape[0])), big_npy.shape[0] // 39)
   infos.append("%s,%s" % (big_npy.shape, n_ivf))
   yield "\n".join(infos)
   index = faiss.index_factory(256 if version19 == "v1" else 768, "IVF%s,Flat" % n_ivf)
   infos.append("training")
   yield "\n".join(infos)
   index_ivf = faiss.extract_index_ivf(index)  #
   index_ivf.nprobe = 1
   index.train(big_npy)
   faiss.write_index(
       index,
       "%s/trained_IVF%s_Flat_nprobe_%s_%s_%s.index"
       % (exp_dir, n_ivf, index_ivf.nprobe, exp_dir1, version19),
   )

   infos.append("adding")
   yield "\n".join(infos)
   batch_size_add = 8192
   for i in range(0, big_npy.shape[0], batch_size_add):
       index.add(big_npy[i : i + batch_size_add])
   faiss.write_index(
       index,
       "%s/added_IVF%s_Flat_nprobe_%s_%s_%s.index"
       % (exp_dir, n_ivf, index_ivf.nprobe, exp_dir1, version19),
   )
   infos.append(
       "成功构建索引，added_IVF%s_Flat_nprobe_%s_%s_%s.index"
       % (n_ivf, index_ivf.nprobe, exp_dir1, version19)
   )
training_log = train_index(model_name, 'v2')
for line in training_log:
   print(line)
   if 'adding' in line:
       clear_output()
       display(Button(description="\u2714 Success", button_style="success")) ```

04. Train Model

맨처음 입력했던 것과 동일한 model_name 입력
아래처럼 epoch 로그 출력되면 Train 시작된 것

INFO:yuri:Train Epoch: 1 [0%] INFO:yuri:[0, 0.0001] INFO:yuri:loss_disc=4.686, loss_gen=2.666, loss_fm=14.291,loss_mel=40.714, loss_kl=9.000 INFO:yuri:====> Epoch: 1 [2023-09-20 12:36:45] | (0:00:34.700875) INFO:yuri:====> Epoch: 2 [2023-09-20 12:36:50] | (0:00:05.268260)

%cd /content/drive/MyDrive/project-main
from random import shuffle
import json
import os
import pathlib
from subprocess import Popen, PIPE, STDOUT
now_dir=os.getcwd()
#@markdown <small> Enter the name of your model again. It must be the same you chose before.
model_name = ''#@param {type:"string"}
#@markdown <small> Choose how often to save the model and how much training you want it to have.
save_frequency = 25 # @param {type:"slider", min:5, max:50, step:5}
epochs = 200 # @param {type:"slider", min:10, max:1000, step:10}
#@markdown <small> ONLY cache datasets under 10 minutes long. Otherwise leave this unchecked.
cache = False #@param {type:"boolean"}
# Remove the logging setup

def click_train(
    exp_dir1,
    sr2,
    if_f0_3,
    spk_id5,
    save_epoch10,
    total_epoch11,
    batch_size12,
    if_save_latest13,
    pretrained_G14,
    pretrained_D15,
    gpus16,
    if_cache_gpu17,
    if_save_every_weights18,
    version19,
):
    # 生成filelist
    exp_dir = "%s/logs/%s" % (now_dir, exp_dir1)
    os.makedirs(exp_dir, exist_ok=True)
    gt_wavs_dir = "%s/0_gt_wavs" % (exp_dir)
    feature_dir = (
        "%s/3_feature256" % (exp_dir)
        if version19 == "v1"
        else "%s/3_feature768" % (exp_dir)
    )
    if if_f0_3:
        f0_dir = "%s/2a_f0" % (exp_dir)
        f0nsf_dir = "%s/2b-f0nsf" % (exp_dir)
        names = (
            set([name.split(".")[0] for name in os.listdir(gt_wavs_dir)])
            & set([name.split(".")[0] for name in os.listdir(feature_dir)])
            & set([name.split(".")[0] for name in os.listdir(f0_dir)])
            & set([name.split(".")[0] for name in os.listdir(f0nsf_dir)])
        )
    else:
        names = set([name.split(".")[0] for name in os.listdir(gt_wavs_dir)]) & set(
            [name.split(".")[0] for name in os.listdir(feature_dir)]
        )
    opt = []
    for name in names:
        if if_f0_3:
            opt.append(
                "%s/%s.wav|%s/%s.npy|%s/%s.wav.npy|%s/%s.wav.npy|%s"
                % (
                    gt_wavs_dir.replace("\\", "\\\\"),
                    name,
                    feature_dir.replace("\\", "\\\\"),
                    name,
                    f0_dir.replace("\\", "\\\\"),
                    name,
                    f0nsf_dir.replace("\\", "\\\\"),
                    name,
                    spk_id5,
                )
            )
        else:
            opt.append(
                "%s/%s.wav|%s/%s.npy|%s"
                % (
                    gt_wavs_dir.replace("\\", "\\\\"),
                    name,
                    feature_dir.replace("\\", "\\\\"),
                    name,
                    spk_id5,
                )
            )
    fea_dim = 256 if version19 == "v1" else 768
    if if_f0_3:
        for _ in range(2):
            opt.append(
                "%s/logs/mute/0_gt_wavs/mute%s.wav|%s/logs/mute/3_feature%s/mute.npy|%s/logs/mute/2a_f0/mute.wav.npy|%s/logs/mute/2b-f0nsf/mute.wav.npy|%s"
                % (now_dir, sr2, now_dir, fea_dim, now_dir, now_dir, spk_id5)
            )
    else:
        for _ in range(2):
            opt.append(
                "%s/logs/mute/0_gt_wavs/mute%s.wav|%s/logs/mute/3_feature%s/mute.npy|%s"
                % (now_dir, sr2, now_dir, fea_dim, spk_id5)
            )
    shuffle(opt)
    with open("%s/filelist.txt" % exp_dir, "w") as f:
        f.write("\n".join(opt))

    # Replace logger.debug, logger.info with print statements
    print("Write filelist done")
    print("Use gpus:", str(gpus16))
    if pretrained_G14 == "":
        print("No pretrained Generator")
    if pretrained_D15 == "":
        print("No pretrained Discriminator")
    if version19 == "v1" or sr2 == "40k":
        config_path = "configs/v1/%s.json" % sr2
    else:
        config_path = "configs/v2/%s.json" % sr2
    config_save_path = os.path.join(exp_dir, "config.json")
    if not pathlib.Path(config_save_path).exists():
        with open(config_save_path, "w", encoding="utf-8") as f:
            with open(config_path, "r") as config_file:
                config_data = json.load(config_file)
                json.dump(
                    config_data,
                    f,
                    ensure_ascii=False,
                    indent=4,
                    sort_keys=True,
                )
            f.write("\n")

    cmd = (
        'python infer/modules/train/train.py -e "%s" -sr %s -f0 %s -bs %s -g %s -te %s -se %s %s %s -l %s -c %s -sw %s -v %s'
        % (
            exp_dir1,
            sr2,
            1 if if_f0_3 else 0,
            batch_size12,
            gpus16,
            total_epoch11,
            save_epoch10,
            "-pg %s" % pretrained_G14 if pretrained_G14 != "" else "",
            "-pd %s" % pretrained_D15 if pretrained_D15 != "" else "",
            1 if if_save_latest13 == True else 0,
            1 if if_cache_gpu17 == True else 0,
            1 if if_save_every_weights18 == True else 0,
            version19,
        )
    )
    # Use PIPE to capture the output and error streams
    p = Popen(cmd, shell=True, cwd=now_dir, stdout=PIPE, stderr=STDOUT, bufsize=1, universal_newlines=True)

    # Print the command's output as it runs
    for line in p.stdout:
        print(line.strip())

    # Wait for the process to finish
    p.wait()
    return "训练结束, 您可查看控制台训练日志或实验文件夹下的train.log"
%load_ext tensorboard
%tensorboard --logdir ./logs
training_log = click_train(
    model_name,
    '40k',
    True,
    0,
    save_frequency,
    epochs,
    7,
    True,
    'assets/pretrained_v2/f0G40k.pth',
    'assets/pretrained_v2/f0D40k.pth',
    0,
    cache,
    True,
    'v2',
)
print(training_log)