提交 e1d8b55b authored 作者: caifuming's avatar caifuming

下载语音识别文字和文字转语音模型,进行对话识别和回答播报语音

上级 073bfdf2
# RAGFlow 启动问题解决方案
## 问题总结
RAGFlow 服务启动脚本已创建并修复,但**无法完全启动**,主要原因如下:
### ✅ 已解决的问题
1. **缺少 quart-auth 0.12.0** - 已升级安装
2. **缺少 quart-schema 0.23.0** - 已安装
3. **缺少 .env 配置文件** - 已从示例文件创建
4. **启动脚本缺少环境检查** - 已添加完善的检查逻辑
### ❌ 当前阻塞问题
**RAGFlow 依赖的基础服务未运行:**
| 服务 | 端口 | 状态 | 说明 |
|------|------|------|------|
| MySQL | 3306 | ❌ 未运行 | 数据库服务 |
| Redis | 6379 | ❌ 未运行 | 缓存服务 |
| Elasticsearch | 1200 | ❌ 未运行 | 搜索引擎 |
| MinIO | 9000 | ❌ 未运行 | 对象存储 |
**根本原因:** 系统中**没有安装 Docker**,无法使用 `docker-compose` 启动基础服务。
---
## 解决方案
你有以下几种选择:
### 方案 1: 安装 Docker 并启动基础服务 (推荐)
这是最简单和官方推荐的方式。
#### 步骤 1: 安装 Docker
```bash
# 安装 Docker
sudo apt update
sudo apt install docker.io docker-compose -y
# 启动 Docker 服务
sudo systemctl start docker
sudo systemctl enable docker
# 将当前用户添加到 docker 组(避免每次都用 sudo)
sudo usermod -aG docker $USER
# 重新登录或执行以下命令使组权限生效
newgrp docker
```
#### 步骤 2: 启动 RAGFlow 基础服务
```bash
cd /datacore/ragflow
docker compose -f docker/docker-compose-base.yml up -d
```
#### 步骤 3: 检查服务状态
```bash
docker compose -f docker/docker-compose-base.yml ps
```
#### 步骤 4: 启动 RAGFlow
```bash
cd /datacore/ragflow
./scripts/ragflow_service.sh start
```
---
### 方案 2: 手动安装并配置基础服务
如果不想使用 Docker,需要手动安装和配置所有服务。
#### 安装 MySQL
```bash
sudo apt install mysql-server -y
sudo systemctl start mysql
sudo systemctl enable mysql
# 创建 RAGFlow 数据库和用户
sudo mysql -e "CREATE DATABASE ragflow CHARACTER SET utf8mb4 COLLATE utf8mb4_unicode_ci;"
sudo mysql -e "CREATE USER 'ragflow'@'localhost' IDENTIFIED BY 'ragflow_password';"
sudo mysql -e "GRANT ALL PRIVILEGES ON ragflow.* TO 'ragflow'@'localhost';"
sudo mysql -e "FLUSH PRIVILEGES;"
```
#### 安装 Redis
```bash
sudo apt install redis-server -y
sudo systemctl start redis-server
sudo systemctl enable redis-server
```
#### 安装 Elasticsearch
```bash
# 安装 Java (Elasticsearch 依赖)
sudo apt install openjdk-11-jdk -y
# 安装 Elasticsearch
wget -qO - https://artifacts.elastic.co/GPG-KEY-elasticsearch | sudo apt-key add -
echo "deb https://artifacts.elastic.co/packages/8.x/apt stable main" | sudo tee /etc/apt/sources.list.d/elastic-8.x.list
sudo apt update
sudo apt install elasticsearch -y
# 配置 Elasticsearch (监听 1200 端口)
sudo nano /etc/elasticsearch/elasticsearch.yml
# 修改: http.port: 1200
sudo systemctl start elasticsearch
sudo systemctl enable elasticsearch
```
#### 安装 MinIO
```bash
# 下载 MinIO
wget https://dl.min.io/server/minio/release/linux-amd64/minio
chmod +x minio
sudo mv minio /usr/local/bin/
# 创建数据目录
sudo mkdir -p /datacore/minio/data
# 创建 MinIO 服务
sudo tee /etc/systemd/system/minio.service > /dev/null <<EOF
[Unit]
Description=MinIO
After=network.target
[Service]
Type=simple
User=admin2
Group=admin2
ExecStart=/usr/local/bin/minio server /datacore/minio/data --address :9000 --console-address :9001
Restart=always
Environment="MINIO_ROOT_USER=admin"
Environment="MINIO_ROOT_PASSWORD=minioadmin"
[Install]
WantedBy=multi-user.target
EOF
sudo systemctl daemon-reload
sudo systemctl start minio
sudo systemctl enable minio
```
#### 更新 RAGFlow 配置
编辑 `/datacore/ragflow/docker/.env` 文件,配置正确的连接信息:
```bash
# 数据库配置
export MYSQL_HOST=localhost
export MYSQL_PORT=3306
export MYSQL_USER=ragflow
export MYSQL_PASSWORD=ragflow_password
export MYSQL_DB=ragflow
# Redis 配置
export REDIS_HOST=localhost
export REDIS_PORT=6379
# Elasticsearch 配置
export ES_HOST=localhost
export ES_PORT=1200
# MinIO 配置
export MINIO_HOST=localhost:9000
export MINIO_USER=admin
export MINIO_PASSWORD=minioadmin
```
---
### 方案 3: 使用已有的远程服务
如果你的环境中已经有运行的 MySQL、Redis、Elasticsearch 和 MinIO 服务(可能在其他服务器上),可以直接配置连接信息。
编辑 `/datacore/ragflow/docker/.env`:
```bash
# 远程服务配置示例
export MYSQL_HOST=192.168.1.100
export MYSQL_PORT=3306
export MYSQL_USER=ragflow
export MYSQL_PASSWORD=your_password
export MYSQL_DB=ragflow
export REDIS_HOST=192.168.1.101
export REDIS_PORT=6379
export REDIS_PASSWORD=your_redis_password
export ES_HOST=192.168.1.102
export ES_PORT=9200
export MINIO_HOST=192.168.1.103:9000
export MINIO_USER=your_access_key
export MINIO_PASSWORD=your_secret_key
```
然后运行启动脚本:
```bash
cd /datacore/ragflow
./scripts/ragflow_service.sh start
```
---
## 启动脚本功能说明
已创建的启动脚本 `/datacore/ragflow/scripts/ragflow_service.sh` 包含以下功能:
### ✅ 环境检查
- 检查 conda 虚拟环境
- 检查配置文件
- 检查前端依赖
- **检查基础服务连接** (MySQL/Redis/Elasticsearch/MinIO)
### ✅ 服务管理
```bash
./scripts/ragflow_service.sh start # 启动服务
./scripts/ragflow_service.sh stop # 停止服务
./scripts/ragflow_service.sh restart # 重启服务
./scripts/ragflow_service.sh status # 查看状态
./scripts/ragflow_service.sh logs # 查看日志
```
### ✅ 错误提示
- 当基础服务未运行时,会友好提示并询问是否继续
- 显示缺失服务和推荐的解决方法
---
## 前端服务说明
前端服务也需要单独处理:
### 安装前端依赖
```bash
cd /datacore/ragflow/web
# 首次运行需要安装 node_modules
npm install
```
### 注意事项
- 前端需要 **Node.js** 环境
- 检查是否已安装: `node --version` 和 `npm --version`
- 如果未安装,需要先安装 Node.js:
```bash
# 安装 Node.js (推荐使用 nvm)
curl -o- https://raw.githubusercontent.com/nvm-sh/nvm/v0.39.0/install.sh | bash
source ~/.bashrc
nvm install 18
nvm use 18
```
---
## 完整启动流程
假设选择**方案 1**(推荐),完整流程如下:
```bash
# 1. 安装 Docker
sudo apt update && sudo apt install docker.io docker-compose -y
sudo systemctl start docker
sudo usermod -aG docker $USER
newgrp docker
# 2. 启动基础服务
cd /datacore/ragflow
docker compose -f docker/docker-compose-base.yml up -d
# 3. 等待服务启动(约 30 秒)
sleep 30
# 4. 检查服务状态
docker compose -f docker/docker-compose-base.yml ps
# 5. 安装前端依赖(首次)
cd /datacore/ragflow/web
npm install
# 6. 启动 RAGFlow
cd /datacore/ragflow
./scripts/ragflow_service.sh start
# 7. 查看日志
./scripts/ragflow_service.sh logs
```
---
## 故障排查
### 检查基础服务是否运行
```bash
# 检查 MySQL
nc -zv localhost 3306
# 检查 Redis
nc -zv localhost 6379
# 检查 Elasticsearch
nc -zv localhost 1200
# 检查 MinIO
nc -zv localhost 9000
```
### 查看 Docker 服务日志
```bash
cd /datacore/ragflow
docker compose -f docker/docker-compose-base.yml logs -f
```
### 查看 RAGFlow 日志
```bash
# 后端日志
tail -f /datacore/ragflow/logs/backend.log
# 前端日志
tail -f /datacore/ragflow/logs/frontend.log
```
---
## 当前状态
| 项目 | 状态 | 说明 |
|------|------|------|
| Python 环境 | ✅ 正常 | conda 环境 ragflow,Python 3.13.15 |
| Python 依赖 | ✅ 已安装 | 483 个包,包括 quart-auth 0.12.0, quart-schema 0.23.0 |
| 启动脚本 | ✅ 已创建 | `/datacore/ragflow/scripts/ragflow_service.sh` |
| 配置文件 | ✅ 已创建 | `/datacore/ragflow/docker/.env` |
| MySQL | ❌ 未运行 | 需要安装或启动 |
| Redis | ❌ 未运行 | 需要安装或启动 |
| Elasticsearch | ❌ 未运行 | 需要安装或启动 |
| MinIO | ❌ 未运行 | 需要安装或启动 |
| Docker | ❌ 未安装 | 需要安装 |
| 前端依赖 | ❌ 未安装 | 需要运行 npm install |
---
## 推荐行动
**立即执行(推荐顺序):**
1. ✅ 安装 Docker: `sudo apt install docker.io docker-compose -y`
2. ✅ 启动 Docker: `sudo systemctl start docker`
3. ✅ 启动基础服务: `docker compose -f docker/docker-compose-base.yml up -d`
4. ✅ 安装前端依赖: `cd web && npm install`
5. ✅ 启动服务: `./scripts/ragflow_service.sh start`
---
## 联系支持
如遇到其他问题,请查看:
- RAGFlow 官方文档: https://github.com/infiniflow/ragflow
- 查看日志文件: `/datacore/ragflow/logs/`
- 检查配置文件: `/datacore/ragflow/docker/.env`
......@@ -71,11 +71,22 @@ file_syncer:
max_concurrent_syncs: 5
user_default_llm:
default_models:
embedding_model:
name: 'bge-m3'
factory: 'xxxx'
api_key: 'xxx'
base_url: 'http://localhost:6380'
# ASR model configuration (Whisper via faster-whisper)
# Service: python3 scripts/deploy_whisper_server.py --model base --port 7777 --device cpu
asr_model:
name: 'whisper-base'
factory: 'FunASR'
api_key: 'dummy'
base_url: 'http://localhost:7777/v1'
model: 'Systran/faster-whisper-base'
# TTS model configuration (CosyVoice)
# Service: python3 scripts/deploy_cosyvoice_server.py --gpu 6 --port 7787
tts_model:
name: 'CosyVoice2-0.5B'
factory: 'FishAudio'
api_key: 'dummy'
base_url: 'http://localhost:7787/v1'
voice: 'default'
gcs:
bucket: '${GCS_BUCKET:-ragflow}'
prefix_path: '${GCS_PREFIX_PATH:-}'
......
......@@ -73,12 +73,43 @@ ingestor:
file_syncer:
max_concurrent_syncs: 5
user_default_llm:
# Default LLM factory for new tenants (e.g., BAAI, Tongyi-Qianwen)
factory: 'BAAI'
api_key: 'dummy'
base_url: 'http://localhost:8000/v1'
default_models:
# Chat model configuration
chat_model:
name: 'qwen2.5-7b-instruct'
factory: 'Qwen'
api_key: 'dummy'
base_url: 'http://localhost:8000/v1'
# Embedding model configuration
embedding_model:
name: 'bge-m3'
factory: 'xxxx'
api_key: 'xxx'
factory: 'BAAI'
api_key: 'dummy'
base_url: 'http://localhost:6380'
# Rerank model configuration
rerank_model:
name: 'bge-reranker-v2-m3'
factory: 'BAAI'
# ASR model configuration (Whisper via faster-whisper)
# Service: python3 scripts/deploy_whisper_server.py --model base --port 7777
asr_model:
name: 'whisper-base'
factory: 'FunASR'
api_key: 'dummy'
base_url: 'http://localhost:7777/v1'
model: 'Systran/faster-whisper-base'
# TTS model configuration (CosyVoice)
# Service: python3 scripts/deploy_cosyvoice_server.py --gpu 6 --port 7787
tts_model:
name: 'CosyVoice2-0.5B'
factory: 'FishAudio'
api_key: 'dummy'
base_url: 'http://localhost:7787/v1'
voice: 'default'
gcs:
bucket: '${GCS_BUCKET:-ragflow}'
prefix_path: '${GCS_PREFIX_PATH:-}'
......
# =============================================================================
# Dockerfile for CosyVoice TTS Service
# =============================================================================
# This image provides a CosyVoice based TTS service with OpenAI-compatible API.
# Runs on a specific GPU (configured via CUDA_VISIBLE_DEVICES environment variable).
# =============================================================================
FROM python:3.10-slim
WORKDIR /app
# Install system dependencies
RUN apt-get update && apt-get install -y \
build-essential \
curl \
git \
libsndfile1 \
ffmpeg \
&& rm -rf /var/lib/apt/lists/*
# Install Python dependencies
COPY requirements-cosyvoice.txt /app/
RUN pip install --no-cache-dir -r requirements-cosyvoice.txt
# Copy application code
COPY deploy_cosyvoice_server.py /app/
# Create cache directories
RUN mkdir -p /root/.cache/huggingface \
&& mkdir -p /root/.cache/modelscope \
&& mkdir -p /app/models
# Expose port
EXPOSE 8000
# Environment variables
ENV PYTHONUNBUFFERED=1
ENV MODEL_NAME=CosyVoice2-0.5B
ENV PORT=8000
ENV CUDA_VISIBLE_DEVICES=6
# Run the server
CMD ["python", "-u", "deploy_cosyvoice_server.py", "--gpu", "6", "--port", "7787", "--host", "127.0.0.1"]
# =============================================================================
# Dockerfile for Whisper ASR Service (faster-whisper)
# =============================================================================
# This image provides a faster-whisper based ASR service with OpenAI-compatible API.
# =============================================================================
FROM python:3.10-slim
WORKDIR /app
# Install system dependencies
RUN apt-get update && apt-get install -y \
build-essential \
curl \
git \
&& rm -rf /var/lib/apt/lists/*
# Install Python dependencies
COPY requirements-whisper.txt /app/
RUN pip install --no-cache-dir -r requirements-whisper.txt
# Copy application code
COPY deploy_whisper_server.py /app/
# Create cache directory
RUN mkdir -p /root/.cache/huggingface
# Expose port
EXPOSE 8000
# Environment variables
ENV PYTHONUNBUFFERED=1
ENV MODEL_NAME=Systran/faster-whisper-base
ENV PORT=8000
ENV DEVICE=cuda
ENV COMPUTE_TYPE=float16
# Run the server
CMD ["python", "-u", "deploy_whisper_server.py", "--model", "base", "--port", "7777", "--host", "127.0.0.1"]
# =============================================================================
# RAGFlow ASR/TTS Services Docker Compose
# =============================================================================
# Deploys Whisper (ASR) and CosyVoice (TTS) services with OpenAI-compatible APIs.
#
# Usage:
# docker compose -f docker/docker-compose-asr-tts.yml up -d
#
# Ports:
# - Whisper ASR: 127.0.0.1:7777 (localhost only)
# - CosyVoice TTS: 127.0.0.1:7787 (localhost only)
# =============================================================================
services:
# -------------------------------------------------------------------
# Whisper ASR Service (faster-whisper with OpenAI-compatible API)
# -------------------------------------------------------------------
whisper-asr:
build:
context: .
dockerfile: Dockerfile.whisper
container_name: whisper-asr-service
image: whisper-asr:latest
profiles:
- whisper
ports:
- "127.0.0.1:7777:8000" # Bind to localhost only
environment:
- MODEL_NAME=Systran/faster-whisper-base
- PORT=8000
- DEVICE=cuda
- COMPUTE_TYPE=float16
volumes:
- whisper_cache:/root/.cache/huggingface
deploy:
resources:
reservations:
devices:
- driver: nvidia
count: 1
capabilities: [gpu]
networks:
- ragflow
healthcheck:
test: ["CMD", "curl", "-f", "http://localhost:8000/health"]
interval: 30s
timeout: 10s
retries: 3
start_period: 60s
restart: unless-stopped
# -------------------------------------------------------------------
# CosyVoice TTS Service (running on GPU 6)
# -------------------------------------------------------------------
cosyvoice-tts:
build:
context: .
dockerfile: Dockerfile.cosyvoice
container_name: cosyvoice-tts-service
image: cosyvoice-tts:latest
profiles:
- cosyvoice
ports:
- "127.0.0.1:7787:8000" # Bind to localhost only
environment:
- MODEL_NAME=CosyVoice2-0.5B
- PORT=8000
- CUDA_VISIBLE_DEVICES=6
volumes:
- cosyvoice_cache:/root/.cache/huggingface
- cosyvoice_models:/app/models
deploy:
resources:
reservations:
devices:
- driver: nvidia
capabilities: [gpu]
# Optionally specify specific GPU
# count: 1
networks:
- ragflow
healthcheck:
test: ["CMD", "curl", "-f", "http://localhost:8000/health"]
interval: 30s
timeout: 10s
retries: 3
start_period: 120s
restart: unless-stopped
# -------------------------------------------------------------------
# Combined ASR + TTS Service (all-in-one)
# -------------------------------------------------------------------
audio-services:
build:
context: .
dockerfile: Dockerfile.audio
container_name: audio-services
image: audio-services:latest
profiles:
- audio
ports:
- "127.0.0.1:7777:8000" # Whisper ASR - localhost only
- "127.0.0.1:7787:8090" # CosyVoice TTS - localhost only
environment:
- CUDA_VISIBLE_DEVICES=6
volumes:
- whisper_cache:/root/.cache/huggingface
- cosyvoice_cache:/root/.cache/modelscope
- cosyvoice_models:/app/models
deploy:
resources:
reservations:
devices:
- driver: nvidia
count: 1
capabilities: [gpu]
networks:
- ragflow
healthcheck:
test: ["CMD", "curl", "-f", "http://localhost:8000/health"]
interval: 30s
timeout: 10s
retries: 3
start_period: 120s
restart: unless-stopped
volumes:
whisper_cache:
cosyvoice_cache:
cosyvoice_models:
networks:
ragflow:
name: ragflow
external: true
# Requirements for CosyVoice TTS Service
# Note: CosyVoice may have specific version requirements
# Check https://github.com/modelscope/CosyVoice for latest requirements
torch>=2.0.0
torchaudio>=2.0.0
fastapi>=0.100.0
uvicorn>=0.23.0
pydantic>=2.0.0
numpy>=1.24.0
soundfile>=0.12.0
# CosyVoice specific dependencies (uncomment when available)
# cosyvoice>=0.1.0
# WeChatKernel>=1.0.0
# Requirements for Whisper ASR Service
faster-whisper>=1.0.0
torch>=2.0.0
torchaudio>=2.0.0
fastapi>=0.100.0
uvicorn>=0.23.0
pydantic>=2.0.0
numpy>=1.24.0
soundfile>=0.12.0
# ASR/TTS 服务部署指南
本文档介绍如何在 RAGFlow 中部署和使用本地 Whisper ASR 和 CosyVoice TTS 服务。
## 概述
RAGFlow 支持本地部署语音识别(ASR)和语音合成(TTS)服务,提供更好的隐私保护和性能:
- **ASR (语音识别)**: 使用 faster-whisper 提供 OpenAI 兼容的 API
- **TTS (语音合成)**: 使用 CosyVoice 在指定 GPU 上运行
## 快速开始
### 1. 安装依赖
```bash
# 安装 Python 依赖
pip install faster-whisper torch torchaudio fastapi uvicorn pydantic numpy soundfile
```
### 2. 启动服务
```bash
# 部署 ASR + TTS 服务
bash scripts/deploy_asr_tts.sh --all
# 或者单独启动
bash scripts/deploy_asr_tts.sh --whisper # 仅 ASR
bash scripts/deploy_asr_tts.sh --cosyvoice # 仅 TTS
# 查看状态
bash scripts/deploy_asr_tts.sh --status
# 停止服务
bash scripts/deploy_asr_tts.sh --stop
```
### 3. 服务地址
- **Whisper ASR**: http://localhost:8001
- **CosyVoice TTS**: http://localhost:8090
- **API 文档**: http://localhost:8001/docs 和 http://localhost:8090/docs
## 手动部署
### Whisper ASR
```bash
# 使用 Python 脚本启动
python3 scripts/deploy_whisper_server.py --model base --port 7777
# 可用模型: tiny, base, small, medium, large-v3
python3 scripts/deploy_whisper_server.py --model large-v3 --port 7777
```
### CosyVoice TTS (GPU 6)
```bash
# 在指定 GPU 上启动
python3 scripts/deploy_cosyvoice_server.py --gpu 6 --port 7787
# 使用其他模型
python3 scripts/deploy_cosyvoice_server.py --model CosyVoice2-0.5B-S-Omni --gpu 6 --port 7787
```
## 配置说明
### 后端配置
编辑 `conf/service_conf.yaml` 配置默认模型:
```yaml
user_default_llm:
default_models:
asr_model:
name: 'whisper-base'
factory: 'FunASR'
api_key: 'dummy'
base_url: 'http://localhost:7777/v1'
tts_model:
name: 'CosyVoice2-0.5B'
factory: 'FishAudio'
api_key: 'dummy'
base_url: 'http://localhost:7787/v1'
```
### 前端配置
前端默认隐藏 ASR/TTS 设置项,直接使用后端配置的默认值。
如需显示设置项,修改 `web/src/pages/user-setting/setting-model/layout/system-setting.tsx`:
```tsx
const HIDE_ASR_TTS_SETTINGS = false; // 改为 false 显示设置项
```
## Docker 部署
### 使用 Docker Compose
```bash
# 启动服务
docker compose -f docker/docker-compose-asr-tts.yml up -d
# 查看日志
docker logs -f whisper-asr-service
docker logs -f cosyvoice-tts-service
# 停止服务
docker compose -f docker/docker-compose-asr-tts.yml down
```
### 使用特定 profile
```bash
# 仅启动 Whisper
docker compose -f docker/docker-compose-asr-tts.yml --profile whisper up -d
# 仅启动 CosyVoice
docker compose -f docker/docker-compose-asr-tts.yml --profile cosyvoice up -d
```
## API 使用
### ASR API
```bash
# 语音转文字
curl -X POST http://localhost:7777/v1/audio/transcriptions \
-F file=@audio.wav \
-F model=whisper-base
```
### TTS API
```bash
# 文字转语音
curl -X POST http://localhost:7787/v1/audio/speech \
-H "Content-Type: application/json" \
-d '{
"input": "你好,世界!",
"voice": "default",
"response_format": "mp3"
}' \
--output speech.mp3
```
## 为现有租户设置默认模型
```bash
# 干跑模式(不实际修改)
python3 scripts/setup_default_asr_tts.py --dry-run
# 实际执行
python3 scripts/setup_default_asr_tts.py
```
## GPU 配置
### 指定 GPU 运行
CosyVoice 默认使用 GPU 6。如需更改:
```bash
# 方法1: 环境变量
export CUDA_VISIBLE_DEVICES=0
python3 scripts/deploy_cosyvoice_server.py --gpu 0 --port 7787
# 方法2: 启动参数
python3 scripts/deploy_cosyvoice_server.py --gpu 1 --port 7787
```
### 查看可用 GPU
```bash
nvidia-smi
```
## 故障排除
### 服务启动失败
1. 检查日志文件:
- `/datacore/ragflow/logs/whisper.log`
- `/datacore/ragflow/logs/cosyvoice.log`
2. 验证端口占用:
```bash
lsof -i :8001 # ASR
lsof -i :8090 # TTS
```
3. 检查 CUDA 是否可用:
```bash
python3 -c "import torch; print(torch.cuda.is_available())"
```
### 模型下载失败
手动下载模型:
```bash
# Whisper
python3 -c "from faster_whisper import WhisperModel; WhisperModel('Systran/faster-whisper-base')"
# CosyVoice (需要从 ModelScope 下载)
git lfs install
git clone https://www.modelscope.cn/iic/CosyVoice2-0.5B.git
```
## 性能优化
### Whisper ASR
- 使用更大的模型获得更好的识别效果
- 启用 GPU 加速 (`--device cuda`)
- 使用 float16 计算类型减少内存使用
### CosyVoice TTS
- 确保使用 GPU 加速
- 调整语速参数 (`speed`)
- 使用参考音频获得更好的音色匹配
## 相关文件
- `scripts/deploy_whisper_server.py` - Whisper ASR 服务
- `scripts/deploy_cosyvoice_server.py` - CosyVoice TTS 服务
- `scripts/deploy_asr_tts.sh` - 一键部署脚本
- `scripts/setup_default_asr_tts.py` - 为现有租户设置默认模型
- `conf/service_conf.yaml` - 后端配置
- `docker/docker-compose-asr-tts.yml` - Docker 部署配置
This source diff could not be displayed because it is too large. You can view the blob instead.
差异被折叠。
差异被折叠。
差异被折叠。
差异被折叠。
差异被折叠。
差异被折叠。
差异被折叠。
差异被折叠。
Markdown 格式
0% 或
您添加了 0 人 到此讨论。请谨慎行事。
请先完成此评论的编辑!
请 注册 或者 后发表评论