bacow/GPT-SoVITS
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Chores (#1422)

Browse Source 
Make Gradio Great Again
附带vs2017下载链接
增加了各种数据集检查,若缺失会弹出warning
修复了warning乱弹的bug
修复了参考音频混合只能上传一条的bug
修复了下载G2PW Model出现import错误的bug
修复了Windows训练无N卡在预处理阶段的报错
This commit is contained in:
XXXXRT666
2024-08-13 16:31:12 +08:00
committed by GitHub
parent 7cfe578968
commit ea56b814bd
27 changed files with 447 additions and 181 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
GPT_SoVITS/download.py
View File

@@ -1,5 +1,5 @@
import os, sys
now_dir = os.getcwd()
sys.path.insert(0, now_dir)
from .text.g2pw import G2PWPinyin
from text.g2pw import G2PWPinyin
g2pw = G2PWPinyin(model_dir="GPT_SoVITS/text/G2PWModel",model_source="GPT_SoVITS/pretrained_models/chinese-roberta-wwm-ext-large",v_to_u=False, neutral_tone_with_five=True)

13
GPT_SoVITS/inference_webui.py
View File

@@ -695,13 +695,14 @@ with gr.Blocks(title="GPT-SoVITS WebUI") as app:
with gr.Row():
inp_ref = gr.Audio(label=i18n("请上传3~10秒内参考音频，超过会报错！"), type="filepath", scale=13)
with gr.Column(scale=13):
ref_text_free = gr.Checkbox(label=i18n("开启无参考文本模式。不填参考文本亦相当于开启。"), value=False, interactive=True, show_label=True)
ref_text_free = gr.Checkbox(label=i18n("开启无参考文本模式。不填参考文本亦相当于开启。"), value=False, interactive=True, show_label=True,scale=1)
gr.Markdown(html_left(i18n("使用无参考文本模式时建议使用微调的GPT，听不清参考音频说的啥(不晓得写啥)可以开。<br>开启后无视填写的参考文本。")))
prompt_text = gr.Textbox(label=i18n("参考音频的文本"), value="", lines=3, max_lines=3)
prompt_language = gr.Dropdown(
label=i18n("参考音频的语种"), choices=list(dict_language.keys()), value=i18n("中文"), scale=14
)
inp_refs = gr.File(label=i18n("可选项：通过拖拽多个文件上传多个参考音频（建议同性），平均融合他们的音色。如不填写此项，音色由左侧单个参考音频控制。如是微调模型，建议参考音频全部在微调训练集音色内，底模不用管。"),file_count="file_count",scale=13)
prompt_text = gr.Textbox(label=i18n("参考音频的文本"), value="", lines=5, max_lines=5,scale=1)
with gr.Column(scale=14):
prompt_language = gr.Dropdown(
label=i18n("参考音频的语种"), choices=list(dict_language.keys()), value=i18n("中文"),
)
inp_refs = gr.File(label=i18n("可选项：通过拖拽多个文件上传多个参考音频（建议同性），平均融合他们的音色。如不填写此项，音色由左侧单个参考音频控制。如是微调模型，建议参考音频全部在微调训练集音色内，底模不用管。"),file_count="multiple")
gr.Markdown(html_center(i18n("*请填写需要合成的目标文本和语种模式"),'h3'))
with gr.Row():
with gr.Column(scale=13):

3
GPT_SoVITS/prepare_datasets/1-get-text.py
View File

@@ -7,7 +7,8 @@ inp_wav_dir = os.environ.get("inp_wav_dir")
exp_name = os.environ.get("exp_name")
i_part = os.environ.get("i_part")
all_parts = os.environ.get("all_parts")
os.environ["CUDA_VISIBLE_DEVICES"] = os.environ.get("_CUDA_VISIBLE_DEVICES")
if "_CUDA_VISIBLE_DEVICES" in os.environ:
os.environ["CUDA_VISIBLE_DEVICES"] = os.environ["_CUDA_VISIBLE_DEVICES"]
opt_dir = os.environ.get("opt_dir")
bert_pretrained_dir = os.environ.get("bert_pretrained_dir")
import torch

3
GPT_SoVITS/prepare_datasets/2-get-hubert-wav32k.py
View File

@@ -6,7 +6,8 @@ inp_wav_dir= os.environ.get("inp_wav_dir")
exp_name= os.environ.get("exp_name")
i_part= os.environ.get("i_part")
all_parts= os.environ.get("all_parts")
os.environ["CUDA_VISIBLE_DEVICES"]= os.environ.get("_CUDA_VISIBLE_DEVICES")
if "_CUDA_VISIBLE_DEVICES" in os.environ:
os.environ["CUDA_VISIBLE_DEVICES"] = os.environ["_CUDA_VISIBLE_DEVICES"]
from feature_extractor import cnhubert
opt_dir= os.environ.get("opt_dir")
cnhubert.cnhubert_base_path= os.environ.get("cnhubert_base_dir")

3
GPT_SoVITS/prepare_datasets/3-get-semantic.py
View File

@@ -4,7 +4,8 @@ inp_text = os.environ.get("inp_text")
exp_name = os.environ.get("exp_name")
i_part = os.environ.get("i_part")
all_parts = os.environ.get("all_parts")
os.environ["CUDA_VISIBLE_DEVICES"] = os.environ.get("_CUDA_VISIBLE_DEVICES")
if "_CUDA_VISIBLE_DEVICES" in os.environ:
os.environ["CUDA_VISIBLE_DEVICES"] = os.environ["_CUDA_VISIBLE_DEVICES"]
opt_dir = os.environ.get("opt_dir")
pretrained_s2G = os.environ.get("pretrained_s2G")
s2config_path = os.environ.get("s2config_path")

6
GPT_SoVITS/utils.py
View File

@@ -18,7 +18,7 @@ logging.getLogger("matplotlib").setLevel(logging.ERROR)
MATPLOTLIB_FLAG = False
logging.basicConfig(stream=sys.stdout, level=logging.DEBUG)
logging.basicConfig(stream=sys.stdout, level=logging.ERROR)
logger = logging
@@ -319,13 +319,13 @@ def check_git_hash(model_dir):
def get_logger(model_dir, filename="train.log"):
global logger
logger = logging.getLogger(os.path.basename(model_dir))
logger.setLevel(logging.DEBUG)
logger.setLevel(logging.ERROR)
formatter = logging.Formatter("%(asctime)s\t%(name)s\t%(levelname)s\t%(message)s")
if not os.path.exists(model_dir):
os.makedirs(model_dir)
h = logging.FileHandler(os.path.join(model_dir, filename))
h.setLevel(logging.DEBUG)
h.setLevel(logging.ERROR)
h.setFormatter(formatter)
logger.addHandler(h)
return logger