tuna2134/sbv2-v2
1
0
Fork 0
Code Issues Pull Requests 1 Actions Packages Projects Releases Wiki Activity

Revert "Refactor: No preloading of BERT models to avoid unnecessary GPU VRAM consumption during training in the Web UI"

Browse Source 
This reverts commit e8a76e547b.
This commit is contained in:
tsukumi
2024-03-12 19:01:45 +00:00
parent c4d6a8cdb7
commit fedd019c04
2 changed files with 9 additions and 12 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
style_bert_vits2/tts_model.py
View File

@@ -29,7 +29,6 @@ from style_bert_vits2.models.models import SynthesizerTrn
from style_bert_vits2.models.models_jp_extra import ( from style_bert_vits2.models.models_jp_extra import (
SynthesizerTrn as SynthesizerTrnJPExtra, SynthesizerTrn as SynthesizerTrnJPExtra,
) )
from style_bert_vits2.nlp import bert_models
from style_bert_vits2.voice import adjust_voice from style_bert_vits2.voice import adjust_voice
@@ -380,13 +379,6 @@ class TTSModelHolder:
def get_model_for_gradio( def get_model_for_gradio(
self, model_name: str, model_path_str: str self, model_name: str, model_path_str: str
) -> tuple[gr.Dropdown, gr.Button, gr.Dropdown]: ) -> tuple[gr.Dropdown, gr.Button, gr.Dropdown]:
bert_models.load_model(Languages.JP)
bert_models.load_tokenizer(Languages.JP)
bert_models.load_model(Languages.EN)
bert_models.load_tokenizer(Languages.EN)
bert_models.load_model(Languages.ZH)
bert_models.load_tokenizer(Languages.ZH)
model_path = Path(model_path_str) model_path = Path(model_path_str)
if model_name not in self.model_files_dict: if model_name not in self.model_files_dict:
raise ValueError(f"Model `{model_name}` is not found") raise ValueError(f"Model `{model_name}` is not found")

13
webui/inference.py
View File

@@ -19,6 +19,7 @@ from style_bert_vits2.constants import (
) )
from style_bert_vits2.logging import logger from style_bert_vits2.logging import logger
from style_bert_vits2.models.infer import InvalidToneError from style_bert_vits2.models.infer import InvalidToneError
from style_bert_vits2.nlp import bert_models
from style_bert_vits2.nlp.japanese import pyopenjtalk_worker as pyopenjtalk from style_bert_vits2.nlp.japanese import pyopenjtalk_worker as pyopenjtalk
from style_bert_vits2.nlp.japanese.g2p_utils import g2kata_tone, kata_tone2phone_tone from style_bert_vits2.nlp.japanese.g2p_utils import g2kata_tone, kata_tone2phone_tone
from style_bert_vits2.nlp.japanese.normalizer import normalize_text from style_bert_vits2.nlp.japanese.normalizer import normalize_text
@@ -29,10 +30,14 @@ from style_bert_vits2.tts_model import TTSModelHolder
## pyopenjtalk_worker は TCP ソケットサーバーのため、ここで起動する ## pyopenjtalk_worker は TCP ソケットサーバーのため、ここで起動する
pyopenjtalk.initialize_worker() pyopenjtalk.initialize_worker()
# Web UI での学習時の無駄な GPU VRAM 消費を避けるため、あえてここでは BERT モデルの事前ロードを行わない # 事前に BERT モデル/トークナイザーをロードしておく
# データセットの BERT 特徴量は事前に bert_gen.py により抽出されているため、学習時に BERT モデルをロードしておく必要はない ## ここでロードしなくても必要になった際に自動ロードされるが、時間がかかるため事前にロードしておいた方が体験が良い
# BERT モデルの事前ロードは「ロード」ボタン押下時に実行される TTSModelHolder.get_model_for_gradio() 内で行われる bert_models.load_model(Languages.JP)
# Web UI での学習時、音声合成タブの「ロード」ボタンを押さなければ、BERT モデルが VRAM にロードされていない状態で学習を開始できる bert_models.load_tokenizer(Languages.JP)
bert_models.load_model(Languages.EN)
bert_models.load_tokenizer(Languages.EN)
bert_models.load_model(Languages.ZH)
bert_models.load_tokenizer(Languages.ZH)
languages = [lang.value for lang in Languages] languages = [lang.value for lang in Languages]