update

.gitignore

@@ -1,5 +1,5 @@
 **/__pycache__
 /Model/
-/bert_vits2/monotonic_align/monotonic_align.egg-info/
-/bert_vits2/monotonic_align/dist/
-/bert_vits2/monotonic_align/build/
+/logs/
+/cache/
+/upload/

Dockerfile

@@ -5,31 +5,35 @@ WORKDIR /app
 
 ENV DEBIAN_FRONTEND=noninteractive
 
-COPY . /app
-
 RUN apt-get update && \
     apt-get install -yq build-essential espeak-ng cmake wget && \
     apt-get clean && \
     apt-get purge -y --auto-remove -o APT::AutoRemove::RecommendsImportant=false && \
-    rm -rf /var/lib/apt/lists/*
+    rm -rf /var/lib/apt/lists/* 
 
-RUN pip install --upgrade pip --no-cache-dir && \
-    pip install MarkupSafe==2.1.2 numpy==1.23.3 cython six==1.16.0 safetensors==0.3.2 --no-cache-dir
+# Install jemalloc
+RUN wget https://github.com/jemalloc/jemalloc/releases/download/5.3.0/jemalloc-5.3.0.tar.bz2 && \
+    tar -xvf jemalloc-5.3.0.tar.bz2 && \
+    cd jemalloc-5.3.0 && \
+    ./configure && \
+    make && \
+    make install && \
+    cd .. && \
+    rm -rf jemalloc-5.3.0* && \
+    ldconfig 
 
-RUN wget https://raw.githubusercontent.com/Artrajz/archived/main/openjtalk/openjtalk-0.3.0.dev2.tar.gz && \
-    tar -zxvf openjtalk-0.3.0.dev2.tar.gz && \
-    cd openjtalk-0.3.0.dev2 && \
-    rm -rf ./pyopenjtalk/open_jtalk_dic_utf_8-1.11 && \
-    python setup.py install && \
-    cd ../ && \
-    rm -f openjtalk-0.3.0.dev2.tar.gz && \
-    rm -rf openjtalk-0.3.0.dev2
+ENV LD_PRELOAD=/usr/local/lib/libjemalloc.so
 
 RUN pip install torch --index-url https://download.pytorch.org/whl/cpu --no-cache-dir
 
-RUN pip install -r requirements.txt --no-cache-dir
+COPY requirements.txt /app/
+RUN pip install --upgrade pip && \
+    pip install pyopenjtalk==0.3.2 -i https://pypi.artrajz.cn/simple --no-cache-dir && \
+    pip install gunicorn --no-cache-dir && \
+    pip install -r requirements.txt --no-cache-dir&& \
+    rm -rf /root/.cache/pip/*
 
-RUN pip install gunicorn --no-cache-dir
+COPY . /app
 
 EXPOSE 23456
 

Dockerfile_GPU

@@ -1,35 +1,39 @@
-FROM python:3.10.11-slim-bullseye
+FROM cnstark/pytorch:2.0.1-py3.10.11-cuda11.8.0-ubuntu22.04
 
 RUN mkdir -p /app
 WORKDIR /app
 
 ENV DEBIAN_FRONTEND=noninteractive
 
-COPY . /app
-
 RUN apt-get update && \
-    apt-get install -yq build-essential espeak-ng cmake wget && \
+    apt-get install -yq build-essential espeak-ng cmake wget ca-certificates && \
+    update-ca-certificates && \
     apt-get clean && \
     apt-get purge -y --auto-remove -o APT::AutoRemove::RecommendsImportant=false && \
-    rm -rf /var/lib/apt/lists/*
+    rm -rf /var/lib/apt/lists/* 
 
-RUN pip install --upgrade pip --no-cache-dir && \
-    pip install MarkupSafe==2.1.2 numpy==1.23.3 cython six==1.16.0  safetensors==0.3.2 --no-cache-dir
 
-RUN wget https://raw.githubusercontent.com/Artrajz/archived/main/openjtalk/openjtalk-0.3.0.dev2.tar.gz && \
-    tar -zxvf openjtalk-0.3.0.dev2.tar.gz && \
-    cd openjtalk-0.3.0.dev2 && \
-    rm -rf ./pyopenjtalk/open_jtalk_dic_utf_8-1.11 && \
-    python setup.py install && \
-    cd ../ && \
-    rm -f openjtalk-0.3.0.dev2.tar.gz && \
-    rm -rf openjtalk-0.3.0.dev2
+# Install jemalloc
+RUN wget https://github.com/jemalloc/jemalloc/releases/download/5.3.0/jemalloc-5.3.0.tar.bz2 && \
+    tar -xvf jemalloc-5.3.0.tar.bz2 && \
+    cd jemalloc-5.3.0 && \
+    ./configure && \
+    make && \
+    make install && \
+    cd .. && \
+    rm -rf jemalloc-5.3.0* && \
+    ldconfig 
 
-RUN pip install torch --index-url https://download.pytorch.org/whl/cu117 --no-cache-dir
+ENV LD_PRELOAD=/usr/local/lib/libjemalloc.so
 
-RUN pip install -r requirements.txt --no-cache-dir
+COPY requirements.txt /app/
+RUN pip install --upgrade pip && \
+    pip install pyopenjtalk==0.3.2 fasttext -i https://pypi.artrajz.cn/simple --no-cache-dir && \
+    pip install gunicorn --no-cache-dir && \
+    pip install -r requirements.txt --no-cache-dir&& \
+    rm -rf /root/.cache/pip/*
 
-RUN pip install gunicorn --no-cache-dir
+COPY . /app
 
 EXPOSE 23456
 

bert_vits2/bert_vits2.py

@@ -1,3 +1,5 @@
+import re
+
 import numpy as np
 import torch
 

bert_vits2/models.py

@@ -643,7 +643,7 @@ class SynthesizerTrn(nn.Module):
         self.sdp = StochasticDurationPredictor(hidden_channels, 192, 3, 0.5, 4, gin_channels=gin_channels)
         self.dp = DurationPredictor(hidden_channels, 256, 3, 0.5, gin_channels=gin_channels)
 
-        if n_speakers > 1:
+        if n_speakers >= 1:
             self.emb_g = nn.Embedding(n_speakers, gin_channels)
         else:
             self.ref_enc = ReferenceEncoder(spec_channels, gin_channels)

docker-compose-gpu.yaml

@@ -1,4 +1,5 @@
-version: '3.4'
+version: '3.8'
+
 services:
   vits:
     image: artrajz/vits-simple-api:latest-gpu
@@ -13,4 +14,8 @@ services:
       - ./config.py:/app/config.py # 挂载配置文件
       - ./logs:/app/logs # logging logs
       - ./gunicorn_config.py:/app/gunicorn_config.py # gunicorn configuration
-      - ./bert_vits2/bert/chinese-roberta-wwm-ext-large:/app/bert_vits2/bert/chinese-roberta-wwm-ext-large
+      - ./vits/bert:/app/vits/bert # vits_chinese
+      - ./bert_vits2/bert/chinese-roberta-wwm-ext-large:/app/bert_vits2/bert/chinese-roberta-wwm-ext-large # Bert-vits2
+      - ./pyopenjtalk/open_jtalk_dic_utf_8-1.11:/usr/local/lib/python3.10/site-packages/pyopenjtalk/open_jtalk_dic_utf_8-1.11 #pyopentjalk
+    devices:
+      - "/dev/nvidia0:/dev/nvidia0"

docker-compose.yaml

@@ -13,4 +13,6 @@ services:
       - ./config.py:/app/config.py # 挂载配置文件
       - ./logs:/app/logs # logging logs
       - ./gunicorn_config.py:/app/gunicorn_config.py # gunicorn configuration
-      - ./bert_vits2/bert/chinese-roberta-wwm-ext-large:/app/bert_vits2/bert/chinese-roberta-wwm-ext-large
+      - ./vits/bert:/app/vits/bert # vits_chinese
+      - ./bert_vits2/bert/chinese-roberta-wwm-ext-large:/app/bert_vits2/bert/chinese-roberta-wwm-ext-large # Bert-vits2
+      - ./pyopenjtalk/open_jtalk_dic_utf_8-1.11:/usr/local/lib/python3.10/site-packages/pyopenjtalk/open_jtalk_dic_utf_8-1.11 #pyopentjalk

gunicorn_config.py

@@ -1,4 +1,19 @@
+import gc
 import multiprocessing
 
 bind = "0.0.0.0:23456"
-workers = multiprocessing.cpu_count()
+# workers = multiprocessing.cpu_count()
+workers = 1
+preload_app = True
+
+# disable GC in master as early as possible
+gc.disable()
+
+def when_ready(server):
+    # freeze objects after preloading app
+    gc.freeze()
+    print("Objects frozen in perm gen: ", gc.get_freeze_count())
+
+def post_fork(server, worker):
+    # reenable GC on worker
+    gc.enable()

utils/nlp.py

@@ -3,6 +3,9 @@ import config
 from .utils import check_is_none
 from logger import logger
 
+# 读取配置选择语种识别库
+clf = getattr(config, "LANGUAGE_IDENTIFICATION_LIBRARY", "fastlid")
+
 
 def clasify_lang(text, speaker_lang):
     pattern = r'[\!\"\#\$\%\&\'\(\)\*\+\,\-\.\/\:\;\<\>\=\?\@\[\]\{\}\\\\\^\_\`' \
@@ -12,22 +15,23 @@ def clasify_lang(text, speaker_lang):
 
     pre = ""
     p = 0
+
+    if clf.upper() == "FASTLID" or clf.upper() == "FASTTEXT":
+        from fastlid import fastlid
+        detect = fastlid
+        if speaker_lang != None: fastlid.set_languages = speaker_lang
+    elif clf.upper() == "LANGID":
+        import langid
+        detect = langid.classify
+        if speaker_lang != None: langid.set_languages(speaker_lang)
+    else:
+        raise ValueError(f"Wrong LANGUAGE_IDENTIFICATION_LIBRARY in config.py")
+
     for word in words:
 
         if check_is_none(word): continue
 
-        # 读取配置选择语种识别库
-        clf = getattr(config, "LANGUAGE_IDENTIFICATION_LIBRARY", "fastlid")
-        if clf.upper() == "FASTLID" or clf.upper() == "FASTTEXT":
-            from fastlid import fastlid
-            lang = fastlid(word)[0]
-            if speaker_lang != None: fastlid.set_languages = speaker_lang
-        elif clf.upper() == "LANGID":
-            import langid
-            lang = langid.classify(word)[0]
-            if speaker_lang != None: langid.set_languages(speaker_lang)
-        else:
-            raise ValueError(f"Wrong LANGUAGE_IDENTIFICATION_LIBRARY in config.py")
+        lang = detect(word)[0]
 
         if pre == "":
             text = text[:p] + text[p:].replace(word, f'[{lang.upper()}]' + word, 1)

vits-simple-api-installer-latest.sh

@@ -5,48 +5,280 @@ GREEN='\033[0;32m'
 YELLOW='\033[0;33m'
 PLAIN='\033[0m'
 
+declare -A EN_MESSAGES
+declare -A ZH_MESSAGES
+
+EN_MESSAGES=(
+  ["ATTEMPT_DOWNLOAD"]="Attempting to download"
+  ["FROM"]="from"
+  ["DOWNLOAD_FAIL"]="Failed to download"
+  ["FROM_ALL_URLS"]="from all provided URLs."
+  ["DOWNLOADING"]="Downloading..."
+  ["VERIFYING"]="Verifying..."
+  ["UNZIPPING"]="Unzipping..."
+  ["CHOOSE_VERSION"]="Which version of docker-compose.yaml do you want to download?"
+  ["DOCKER_CPU"]="docker-compose.yaml (CPU version)"
+  ["DOCKER_GPU"]="docker-compose-gpu.yaml (GPU version)"
+  ["ENTER_CHOICE"]="Enter your choice (1 or 2): "
+  ["INVALID_CHOICE"]="Invalid choice. Please enter 1 or 2."
+  ["DOWNLOAD_CONFIG"]="Downloading configuration file shortly..."
+  ["PULL_IMAGE"]="Do you want to start pulling the image? Enter 1 for yes or 2 for no"
+  ["DOWNLOAD_DICT"]="Do you want to download the pyopenjtalk dictionary file? Enter 1 for yes or 2 for no"
+  ["MUST_DOWNLOAD_JP"]="Japanese model must be downloaded."
+  ["DOWNLOAD_VITS_CHINESE"]="Do you want to download the bert model for vits_chinese? Enter 1 for yes, 2 for no."
+  ["MUST_DOWNLOAD_VITS_CHINESE"]="Using vits_chinese requires downloading these models, which will take up about 410MB."
+  ["DOWNLOAD_BERT_VITS2"]="Do you want to download chinese-roberta-wwm-ext-large? Enter 1 for yes or 2 for no"
+  ["MUST_DOWNLOAD_BERT_VITS2"]="To use Bert-VITS2, you must download these models, which will take up about 3.64GB."
+  ["DOWNLOADED"]="File is downloaded correctly."
+  ["CORRUPTED"]="File is corrupted or incomplete."
+  ["INSTALL_COMPLETE"]="The upgrade or installation has been completed."
+  ["CONFIG_DIR"]="The configuration file directory is"
+  ["IMPORT_NOTICE"]="If the vits model is not imported, it cannot be used. Import the model in the configuration file directory."
+  ["RESTART_NOTICE"]="After modifying the configuration file, restart the docker container for the modification to take effect."
+  ["ISSUE_NOTICE"]="If you have any questions, please put them in the issues."
+  ["GITHUB_LINK"]="https://github.com/Artrajz/vits-simple-api"
+)
+
+ZH_MESSAGES=(
+  ["ATTEMPT_DOWNLOAD"]="正在尝试下载"
+  ["FROM"]="从"
+  ["DOWNLOAD_FAIL"]="都下载失败"
+  ["FROM_ALL_URLS"]="从所有提供的URLs"
+  ["DOWNLOADING"]="正在下载..."
+  ["VERIFYING"]="正在校验"
+  ["UNZIPPING"]="正在解压..."
+  ["CHOOSE_VERSION"]="你想下载哪个版本的docker-compose.yaml？"
+  ["DOCKER_CPU"]="docker-compose.yaml (CPU版本)"
+  ["DOCKER_GPU"]="docker-compose-gpu.yaml (GPU版本)"
+  ["ENTER_CHOICE"]="请输入您的选择 (1 或 2): "
+  ["INVALID_CHOICE"]="无效选择。 请重新输入 1 或 2。"
+  ["DOWNLOAD_CONFIG"]="即将下载配置文件..."
+  ["PULL_IMAGE"]="是否要开始拉取镜像？输入1表示是，2表示否。"
+  ["DOWNLOAD_DICT"]="是否要下载pyopenjtalk的词典文件？输入1表示是，2表示否。"
+  ["MUST_DOWNLOAD_JP"]="使用日语模型必须下载该词典文件，将占用大约102MB。"
+  ["DOWNLOAD_VITS_CHINESE"]="是否要下载vits_chinese的bert模型？输入1表示是，2表示否。"
+  ["MUST_DOWNLOAD_VITS_CHINESE"]="使用vits_chinese必须下载这些模型，将占用大约410MB。"
+  ["DOWNLOAD_BERT_VITS2"]="是否要下载chinese-roberta-wwm-ext-large？输入1表示是，2表示否。"
+  ["MUST_DOWNLOAD_BERT_VITS2"]="使用Bert-VITS2必须下载这些模型，将占用大约3.64GB。"
+  ["DOWNLOADED"]="文件已正确下载。"
+  ["CORRUPTED"]="文件已损坏或不完整。"
+  ["INSTALL_COMPLETE"]="更新或安装已完成。"
+  ["CONFIG_DIR"]="配置文件目录是"
+  ["IMPORT_NOTICE"]="如果vits模型没有被导入，它是无法使用的。请在配置文件目录中导入模型。"
+  ["RESTART_NOTICE"]="修改配置文件后，请重启docker容器以使修改生效。"
+  ["ISSUE_NOTICE"]="如果你有任何问题，请在issues中提出，或者加入q群提问。"
+  ["GITHUB_LINK"]="https://github.com/Artrajz/vits-simple-api"
+)
+
+echo -e "${PLAIN}${GREEN}Choose a language/选择语言: ${PLAIN}"
+echo "1. English"
+echo "2. 中文"
+read -p "Enter your choice (1 or 2): " choice_language
+
+declare -A MESSAGES
+if [ "$choice_language" -eq 1 ]; then
+  for key in "${!EN_MESSAGES[@]}"; do
+    MESSAGES["$key"]="${EN_MESSAGES[$key]}"
+  done
+else
+  for key in "${!ZH_MESSAGES[@]}"; do
+    MESSAGES["$key"]="${ZH_MESSAGES[$key]}"
+  done
+fi
+
 mkdir -p $INSTALL_DIR
 cd $INSTALL_DIR
+
+download_with_fallback() {
+  local filename=$1
+  shift # Shift arguments to the left to handle URLs
+
+  local success=0
+  local url
+  for url in "$@"; do
+    echo -e "${YELLOW}${MESSAGES["ATTEMPT_DOWNLOAD"]} $filename ${MESSAGES["FROM"]} $url\n${PLAIN}"
+    if wget -O "$INSTALL_DIR/$filename" "$url"; then
+      success=1
+      break
+    fi
+  done
+
+  if [ "$success" -ne 1 ]; then
+    echo -e "${RED} $filename ${MESSAGES["FROM_ALL_URLS"]} ${MESSAGES["DOWNLOAD_FAIL"]}${PLAIN}"
+    exit 1
+  fi
+}
+
+version_gt() {
+  test "$(echo "$@" | tr " " "\n" | sort -V | head -n 1)" != "$1"
+}
+
+while true; do
+  echo -e "${GREEN}${MESSAGES["CHOOSE_VERSION"]}${PLAIN}"
+  echo -e "1. ${MESSAGES["DOCKER_CPU"]}"
+  echo -e "2. ${MESSAGES["DOCKER_GPU"]}"
+  read -p "${MESSAGES["ENTER_CHOICE"]}" choice_gpu
+  case $choice_gpu in
+  1)
+    echo -e "${MESSAGES["DOWNLOADING"]}"
+    download_with_fallback docker-compose.yaml \
+      "https://raw.githubusercontent.com/Artrajz/vits-simple-api/main/docker-compose.yaml" \
+      "https://ghproxy.com/https://raw.githubusercontent.com/Artrajz/vits-simple-api/main/docker-compose.yaml"
+    break
+    ;;
+  2)
+    echo -e "${MESSAGES["DOWNLOADING"]}"
+    download_with_fallback docker-compose.yaml \
+      "https://raw.githubusercontent.com/Artrajz/vits-simple-api/main/docker-compose-gpu.yaml" \
+      "https://ghproxy.com/https://raw.githubusercontent.com/Artrajz/vits-simple-api/main/docker-compose-gpu.yaml"
+    break
+    ;;
+  *)
+    echo -e "${RED}${MESSAGES["INVALID_CHOICE"]}${PLAIN}"
+    ;;
+  esac
+done
+
+if [ "$choice_gpu" -eq 2 ]; then
+  DOCKER_VERSION=$(docker version --format '{{.Server.Version}}')
+  MIN_DOCKER_VERSION="19.03"
+
+  if version_gt $MIN_DOCKER_VERSION $DOCKER_VERSION; then
+    echo -e "${RED}Your Docker version ($DOCKER_VERSION) does not support GPU. You need at least version $MIN_DOCKER_VERSION.${PLAIN}"
+    exit 1
+  fi
+fi
+
+if ! command -v docker-compose &>/dev/null; then
+  echo -e "${RED}docker-compose could not be found.${PLAIN}"
+  exit 1
+fi
+
+echo -e "${GREEN}${MESSAGES["PULL_IMAGE"]}${PLAIN}"
+read -p "${MESSAGES["ENTER_CHOICE"]}" choice_pull
+
+if [ "$choice_pull" -eq 1 ]; then
+  docker compose pull
+  docker compose up -d
+fi
+
+echo -e "${YELLOW}${MESSAGES["DOWNLOAD_CONFIG"]}${PLAIN}"
+
 if [ ! -f config.py ]; then
-    echo -e "${YELLOW}download config.py\n${PLAIN}"
-    wget -O $INSTALL_DIR/config.py https://raw.githubusercontent.com/Artrajz/vits-simple-api/main/config.py
+  download_with_fallback config.py \
+    "https://raw.githubusercontent.com/Artrajz/vits-simple-api/main/config.py" \
+    "https://ghproxy.com/https://raw.githubusercontent.com/Artrajz/vits-simple-api/main/config.py"
 fi
 
 if [ ! -f gunicorn_config.py ]; then
-    echo -e "${YELLOW}download config.py\n${PLAIN}"
-    wget -O $INSTALL_DIR/gunicorn_config.py https://raw.githubusercontent.com/Artrajz/vits-simple-api/main/gunicorn_config.py
+  download_with_fallback gunicorn_config.py \
+    "https://raw.githubusercontent.com/Artrajz/vits-simple-api/main/gunicorn_config.py" \
+    "https://ghproxy.com/https://raw.githubusercontent.com/Artrajz/vits-simple-api/main/gunicorn_config.py"
 fi
 
-while true; do
-    echo -e "${GREEN}Which version of docker-compose.yaml do you want to download?"
-    echo -e "1. docker-compose.yaml (CPU version)"
-    echo -e "2. docker-compose-gpu.yaml (GPU version)"
-    read -p "Enter your choice (1 or 2): " choice
-    case $choice in
-        1)
-            echo -e "${YELLOW}Downloading docker-compose.yaml (CPU version)\n${PLAIN}"
-            wget -O $INSTALL_DIR/docker-compose.yaml https://raw.githubusercontent.com/Artrajz/vits-simple-api/main/docker-compose.yaml
-            break
-            ;;
-        2)
-            echo -e "${YELLOW}Downloading docker-compose-gpu.yaml (GPU version)\n${PLAIN}"
-            wget -O $INSTALL_DIR/docker-compose.yaml https://raw.githubusercontent.com/Artrajz/vits-simple-api/main/docker-compose-gpu.yaml
-            break
-            ;;
-        *)
-            echo -e "${RED}Invalid choice. Please enter 1 or 2.${PLAIN}"
-            ;;
-    esac
-done
+download_with_fallback config.example.py \
+  "https://raw.githubusercontent.com/Artrajz/vits-simple-api/main/config.py" \
+  "https://ghproxy.com/https://raw.githubusercontent.com/Artrajz/vits-simple-api/main/config.py"
+
+download_with_fallback gunicorn_config.example.py \
+  "https://raw.githubusercontent.com/Artrajz/vits-simple-api/main/gunicorn_config.py" \
+  "https://ghproxy.com/https://raw.githubusercontent.com/Artrajz/vits-simple-api/main/gunicorn_config.py"
+
+echo -e "${GREEN}${MESSAGES["DOWNLOAD_DICT"]}${PLAIN}"
+echo -e "${GREEN}${MESSAGES["MUST_DOWNLOAD_JP"]}${PLAIN}"
+read -p "${MESSAGES["ENTER_CHOICE"]}" choice_download_pyopenjtalk
+
+if [ "$choice_download_pyopenjtalk" -eq 1 ]; then
+  mkdir -p pyopenjtalk
+  echo -e "${MESSAGES["DOWNLOADING"]}"
+  download_with_fallback open_jtalk_dic_utf_8-1.11.tar.gz \
+    "https://github.com/r9y9/open_jtalk/releases/download/v1.11.1/open_jtalk_dic_utf_8-1.11.tar.gz" \
+    "https://ghproxy.com/https://github.com/r9y9/open_jtalk/releases/download/v1.11.1/open_jtalk_dic_utf_8-1.11.tar.gz"
+  echo -e "${MESSAGES["UNZIPPING"]}"
+  tar -xzvf open_jtalk_dic_utf_8-1.11.tar.gz -C pyopenjtalk/
+  rm open_jtalk_dic_utf_8-1.11.tar.gz
+fi
+
+echo -e "${GREEN}${MESSAGES["DOWNLOAD_VITS_CHINESE"]}${PLAIN}"
+echo -e "${GREEN}${MESSAGES["MUST_DOWNLOAD_VITS_CHINESE"]}${PLAIN}"
+read -p "${MESSAGES["ENTER_CHOICE"]}" choice_download_vits_chinese
+
+if [ "$choice_download_vits_chinese" -eq 1 ]; then
+  mkdir -p vits/bert
+
+  EXPECTED_MD5="dea78034433141adc8002404aa1b3184"
+  FILE_PATH="vits/bert/prosody_model.pt"
+  echo -e "${MESSAGES["VERIFYING"]}$FILE_PATH"
+  ACTUAL_MD5=$(md5sum $FILE_PATH | awk '{print $1}')
+
+  if [ "$EXPECTED_MD5" == "$ACTUAL_MD5" ]; then
+    echo "${MESSAGES["DOWNLOADED"]}"
+  else
+    echo "${MESSAGES["CORRUPTED"]}"
+    download_with_fallback vits/bert/prosody_model.pt \
+      "https://huggingface.co/spaces/maxmax20160403/vits_chinese/resolve/main/bert/prosody_model.pt"
+  fi
+
+fi
+
+echo -e "${GREEN}${MESSAGES["DOWNLOAD_BERT_VITS2"]}${PLAIN}"
+echo -e "${GREEN}${MESSAGES["MUST_DOWNLOAD_BERT_VITS2"]}${PLAIN}"
+read -p "${MESSAGES["ENTER_CHOICE"]}" choice_download_bert_vits2
+
+if [ "$choice_download_bert_vits2" -eq 1 ]; then
+  mkdir -p bert_vits2/bert/chinese-roberta-wwm-ext-large
+
+  EXPECTED_MD5="78ef42421495cb23372bef9d069a75f3"
+  FILE_PATH="bert_vits2/bert/chinese-roberta-wwm-ext-large/flax_model.msgpack"
+  echo -e "${MESSAGES["VERIFYING"]}$FILE_PATH"
+  ACTUAL_MD5=$(md5sum $FILE_PATH | awk '{print $1}')
+
+  if [ "$EXPECTED_MD5" == "$ACTUAL_MD5" ]; then
+    echo "${MESSAGES["DOWNLOADED"]}"
+  else
+    echo "${MESSAGES["CORRUPTED"]}"
+    download_with_fallback bert_vits2/bert/chinese-roberta-wwm-ext-large/flax_model.msgpack \
+      "https://huggingface.co/hfl/chinese-roberta-wwm-ext-large/resolve/main/flax_model.msgpack"
+  fi
+
+  EXPECTED_MD5="15d7435868fef1bd4222ff7820149a2a"
+  FILE_PATH="bert_vits2/bert/chinese-roberta-wwm-ext-large/pytorch_model.bin"
+  echo -e "${MESSAGES["VERIFYING"]}$FILE_PATH"
+  ACTUAL_MD5=$(md5sum $FILE_PATH | awk '{print $1}')
 
-echo -e "${YELLOW}Pulling the image might take a while, so why not grab a cup of java first?\n${PLAIN}"
+  if [ "$EXPECTED_MD5" == "$ACTUAL_MD5" ]; then
+    echo "${MESSAGES["DOWNLOADED"]}"
+  else
+    echo ${MESSAGES["CORRUPTED"]}
+    download_with_fallback bert_vits2/bert/chinese-roberta-wwm-ext-large/pytorch_model.bin \
+      "https://huggingface.co/hfl/chinese-roberta-wwm-ext-large/resolve/main/pytorch_model.bin"
+  fi
 
-docker compose pull
-docker compose up -d
+  EXPECTED_MD5="d15991416bd4a86fa127c70d3c0f4779"
+  FILE_PATH="bert_vits2/bert/chinese-roberta-wwm-ext-large/tf_model.h5"
+  echo -e "${MESSAGES["VERIFYING"]}$FILE_PATH"
+  ACTUAL_MD5=$(md5sum $FILE_PATH | awk '{print $1}')
+
+  if [ "$EXPECTED_MD5" == "$ACTUAL_MD5" ]; then
+    echo "${MESSAGES["DOWNLOADED"]}"
+  else
+    echo "${MESSAGES["CORRUPTED"]}"
+    download_with_fallback bert_vits2/bert/chinese-roberta-wwm-ext-large/tf_model.h5 \
+      "https://huggingface.co/hfl/chinese-roberta-wwm-ext-large/resolve/main/tf_model.h5"
+  fi
+
+fi
+
+if [ "$choice_gpu" -eq 2 ]; then
+  if ! docker run --gpus all artrajz/vits-simple-api:latest-gpu nvidia-smi &>/dev/null; then
+    echo -e "${RED}Your Docker does not seem to support GPU or NVIDIA Docker is not installed properly.${PLAIN}"
+    exit 1
+  fi
+fi
 
-echo -e "\nThe upgrade or installation has been completed."
-echo -e "The configuration file directory is $(realpath $INSTALL_DIR)"
-echo -e "${YELLOW}If the vits model is not imported, it cannot be used. Import the model in the configuration file directory.${PLAIN}"
-echo -e "After modifying the configuration file, restart the docker container for the modification to take effect."
-echo -e "${YELLOW}If you have any questions, please put them in the issues.${PLAIN}"
-echo -e "https://github.com/Artrajz/vits-simple-api"
+echo -e "\n${MESSAGES["INSTALL_COMPLETE"]}"
+echo -e "${MESSAGES["CONFIG_DIR"]} $(realpath $INSTALL_DIR)"
+echo -e "${YELLOW}${MESSAGES["IMPORT_NOTICE"]}${PLAIN}"
+echo -e "${YELLOW}${MESSAGES["RESTART_NOTICE"]}${PLAIN}"
+echo -e "${MESSAGES["ISSUE_NOTICE"]}"
+echo -e "${MESSAGES["GITHUB_LINK"]}"

vits/text/cantonese.py

@@ -3,7 +3,7 @@ import cn2an
 import opencc
 import config
 
-converter = opencc.OpenCC(config.ABS_PATH + '/chinese_dialect_lexicons/jyutjyu_2')
+converter = opencc.OpenCC(config.ABS_PATH + '/vits/text/chinese_dialect_lexicons/jyutjyu_2')
 
 # List of (Latin alphabet, ipa) pairs:
 _latin_to_ipa = [(re.compile('%s' % x[0]), x[1]) for x in [

vits/text/shanghainese.py

@@ -3,7 +3,7 @@ import cn2an
 import opencc
 import config
 
-converter = opencc.OpenCC(config.ABS_PATH + '/chinese_dialect_lexicons/zaonhe')
+converter = opencc.OpenCC(config.ABS_PATH + '/vits/text/chinese_dialect_lexicons/zaonhe')
 
 # List of (Latin alphabet, ipa) pairs:
 _latin_to_ipa = [(re.compile('%s' % x[0]), x[1]) for x in [