Initial commit: add Trading Studio voice-over pipeline for quant trading review videos.
Co-authored-by: Cursor <cursoragent@cursor.com>
This commit is contained in:
dekun
+46
@@ -0,0 +1,46 @@
|
||||
# Python 虚拟环境
|
||||
venv/
|
||||
.venv/
|
||||
env/
|
||||
|
||||
# 模型权重与音色文件（体积大，不入库）
|
||||
*.pt
|
||||
*.pth
|
||||
*.onnx
|
||||
*.bin
|
||||
*.safetensors
|
||||
|
||||
# 音频产物
|
||||
*.wav
|
||||
*.mp3
|
||||
*.flac
|
||||
*.ogg
|
||||
*.m4a
|
||||
|
||||
# 日志
|
||||
*.log
|
||||
|
||||
# 运行时目录
|
||||
uploads/
|
||||
outputs/
|
||||
__pycache__/
|
||||
*.py[cod]
|
||||
*$py.class
|
||||
.Python
|
||||
|
||||
# IDE
|
||||
.idea/
|
||||
.vscode/
|
||||
*.swp
|
||||
*.swo
|
||||
|
||||
# 系统文件
|
||||
.DS_Store
|
||||
Thumbs.db
|
||||
|
||||
# 环境变量与密钥
|
||||
.env
|
||||
.env.*
|
||||
|
||||
# Gradio 临时
|
||||
gradio_cached_examples/
|
||||
@@ -0,0 +1,488 @@
|
||||
# Trading Studio 部署指南 (DEPLOY.md)
|
||||
|
||||
本文档面向 **Ubuntu 物理服务器**（搭载 RTX 3060 Ti，已锁定 120W 功耗墙）的完整环境配置与 PM2 常驻部署流程。适用于首次安装或迁移重装场景。
|
||||
|
||||
**Git 仓库：** https://git.bz121.com/dekun/Trading_Studio.git
|
||||
|
||||
---
|
||||
|
||||
## 目录
|
||||
|
||||
1. [硬件与系统前提](#1-硬件与系统前提)
|
||||
2. [3060 Ti 120W 功耗墙配置](#2-3060-ti-120w-功耗墙配置)
|
||||
3. [NVIDIA 驱动与 CUDA](#3-nvidia-驱动与-cuda)
|
||||
4. [Python 虚拟环境](#4-python-虚拟环境)
|
||||
5. [PyTorch CUDA 12.1 安装](#5-pytorch-cuda-121-安装)
|
||||
6. [项目依赖安装](#6-项目依赖安装)
|
||||
7. [远程 Ollama 节点配置](#7-远程-ollama-节点配置)
|
||||
8. [首次运行与验证](#8-首次运行与验证)
|
||||
9. [PM2 进程守护](#9-pm2-进程守护)
|
||||
10. [迁移与故障排查](#10-迁移与故障排查)
|
||||
|
||||
---
|
||||
|
||||
## 1. 硬件与系统前提
|
||||
|
||||
| 项目 | 要求 |
|
||||
|------|------|
|
||||
| GPU | NVIDIA RTX 3060 Ti 8GB |
|
||||
| 功耗墙 | 120W（推荐锁定，见下文） |
|
||||
| 系统 | Ubuntu 22.04 / 24.04 LTS |
|
||||
| 内存 | ≥ 16GB |
|
||||
| 磁盘 | ≥ 30GB 可用（含模型缓存） |
|
||||
| 网络 | 局域网可访问 `192.168.8.64:11434` |
|
||||
|
||||
```bash
|
||||
# 基础工具
|
||||
sudo apt update && sudo apt upgrade -y
|
||||
sudo apt install -y git curl wget build-essential \
|
||||
python3 python3-venv python3-dev \
|
||||
ffmpeg libsndfile1 portaudio19-dev
|
||||
```
|
||||
|
||||
> 若 `python3-venv` 包名报错，使用 `python3-venv`。
|
||||
|
||||
---
|
||||
|
||||
## 2. 3060 Ti 120W 功耗墙配置
|
||||
|
||||
锁定 GPU 功耗有助于稳定 7×24 运行、降低散热压力，避免 Whisper + ChatTTS 并发时触发功耗波动。
|
||||
|
||||
### 2.1 安装 nvidia-smi 功耗管理工具
|
||||
|
||||
驱动安装后自带 `nvidia-smi`。确认 GPU 可见：
|
||||
|
||||
```bash
|
||||
nvidia-smi
|
||||
```
|
||||
|
||||
### 2.2 临时设置 120W 功耗上限
|
||||
|
||||
```bash
|
||||
# 查看支持的功耗范围
|
||||
nvidia-smi -q -d POWER | grep -A3 "Power Limit"
|
||||
|
||||
# 设置最大功耗为 120W（需 root）
|
||||
sudo nvidia-smi -pl 120
|
||||
```
|
||||
|
||||
### 2.3 开机持久化（推荐）
|
||||
|
||||
创建 systemd 服务，每次启动自动应用：
|
||||
|
||||
```bash
|
||||
sudo tee /etc/systemd/system/nvidia-powerlimit.service << 'EOF'
|
||||
[Unit]
|
||||
Description=Set NVIDIA GPU Power Limit to 120W
|
||||
After=multi-user.target
|
||||
|
||||
[Service]
|
||||
Type=oneshot
|
||||
ExecStart=/usr/bin/nvidia-smi -pl 120
|
||||
RemainAfterExit=yes
|
||||
|
||||
[Install]
|
||||
WantedBy=multi-user.target
|
||||
EOF
|
||||
|
||||
sudo systemctl daemon-reload
|
||||
sudo systemctl enable nvidia-powerlimit.service
|
||||
sudo systemctl start nvidia-powerlimit.service
|
||||
|
||||
# 验证
|
||||
nvidia-smi --query-gpu=power.limit --format=csv
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 3. NVIDIA 驱动与 CUDA
|
||||
|
||||
### 3.1 安装驱动（推荐 535+ 或 550+）
|
||||
|
||||
```bash
|
||||
# Ubuntu 自动安装推荐驱动
|
||||
sudo ubuntu-drivers devices
|
||||
sudo ubuntu-drivers autoinstall
|
||||
# 或指定版本: sudo apt install nvidia-driver-550
|
||||
|
||||
sudo reboot
|
||||
```
|
||||
|
||||
重启后验证：
|
||||
|
||||
```bash
|
||||
nvidia-smi
|
||||
nvcc --version # 若未安装 nvcc 不影响 PyTorch，可选
|
||||
```
|
||||
|
||||
### 3.2 cuDNN（Faster-Whisper / PyTorch 需要）
|
||||
|
||||
PyTorch cu121 wheel 通常自带运行时库。若 Whisper 报 cuDNN 错误：
|
||||
|
||||
```bash
|
||||
# 参考 NVIDIA 官方文档安装 cuDNN for CUDA 12.x
|
||||
# https://developer.nvidia.com/cudnn
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 4. Python 虚拟环境
|
||||
|
||||
```bash
|
||||
# 克隆项目
|
||||
cd ~
|
||||
git clone https://git.bz121.com/dekun/Trading_Studio.git
|
||||
cd Trading_Studio
|
||||
|
||||
# 创建虚拟环境（必须使用 venv，与 PM2 interpreter 路径一致）
|
||||
python3 -m venv venv
|
||||
|
||||
# 激活
|
||||
source venv/bin/activate
|
||||
|
||||
# 升级 pip
|
||||
pip install --upgrade pip setuptools wheel
|
||||
```
|
||||
|
||||
**重要：** PM2 配置中 `interpreter` 指向 `./venv/bin/python`，请确保在项目根目录创建 `venv/`。
|
||||
|
||||
---
|
||||
|
||||
## 5. PyTorch CUDA 12.1 安装
|
||||
|
||||
**必须先于其他 GPU 依赖安装**，避免 pip 拉取 CPU 版 torch。
|
||||
|
||||
```bash
|
||||
source venv/bin/activate
|
||||
|
||||
pip install torch torchvision torchaudio \
|
||||
--index-url https://download.pytorch.org/whl/cu121
|
||||
```
|
||||
|
||||
验证 CUDA 可用：
|
||||
|
||||
```bash
|
||||
python -c "
|
||||
import torch
|
||||
print('PyTorch:', torch.__version__)
|
||||
print('CUDA available:', torch.cuda.is_available())
|
||||
print('GPU:', torch.cuda.get_device_name(0) if torch.cuda.is_available() else 'N/A')
|
||||
"
|
||||
```
|
||||
|
||||
期望输出类似：
|
||||
|
||||
```
|
||||
PyTorch: 2.x.x+cu121
|
||||
CUDA available: True
|
||||
GPU: NVIDIA GeForce RTX 3060 Ti
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 6. 项目依赖安装
|
||||
|
||||
```bash
|
||||
source venv/bin/activate
|
||||
cd ~/Trading_Studio
|
||||
|
||||
# 安装其余依赖
|
||||
pip install -r requirements.txt
|
||||
```
|
||||
|
||||
### 6.1 Faster-Whisper
|
||||
|
||||
随 `requirements.txt` 安装。首次运行会自动下载 `small` 模型（约 500MB）至 HuggingFace 缓存。
|
||||
|
||||
### 6.2 ChatTTS
|
||||
|
||||
从 GitHub 源码安装（已在 requirements.txt 中指定）：
|
||||
|
||||
```bash
|
||||
pip install ChatTTS @ git+https://github.com/2noise/ChatTTS.git
|
||||
```
|
||||
|
||||
首次 `save_fixed_speaker` 或 `generate_voice` 时会下载模型权重（数 GB），请确保网络畅通或提前配置 HuggingFace 镜像：
|
||||
|
||||
```bash
|
||||
export HF_ENDPOINT=https://hf-mirror.com # 可选，国内加速
|
||||
```
|
||||
|
||||
### 6.3 Gradio
|
||||
|
||||
```bash
|
||||
pip install gradio>=4.44.0
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 7. 远程 Ollama 节点配置
|
||||
|
||||
Trading Studio 的 LLM 润色模块连接局域网 Ollama，**不在本机运行大模型**。
|
||||
|
||||
| 配置项 | 值 |
|
||||
|--------|-----|
|
||||
| 地址 | `http://192.168.8.64:11434` |
|
||||
| API | `POST /api/chat` |
|
||||
| 模型 | `huihui_ai/gemma-4-abliterated:e4b` |
|
||||
| 流式 | `stream: false` |
|
||||
|
||||
### 7.1 在 Ollama 节点（192.168.8.64）上
|
||||
|
||||
```bash
|
||||
# 安装 Ollama（若未安装）
|
||||
curl -fsSL https://ollama.com/install.sh | sh
|
||||
|
||||
# 拉取模型
|
||||
ollama pull huihui_ai/gemma-4-abliterated:e4b
|
||||
|
||||
# 允许局域网访问（编辑 systemd 或环境变量）
|
||||
sudo systemctl edit ollama
|
||||
```
|
||||
|
||||
添加：
|
||||
|
||||
```ini
|
||||
[Service]
|
||||
Environment="OLLAMA_HOST=0.0.0.0:11434"
|
||||
```
|
||||
|
||||
```bash
|
||||
sudo systemctl daemon-reload
|
||||
sudo systemctl restart ollama
|
||||
```
|
||||
|
||||
### 7.2 在本机（Trading Studio 服务器）验证
|
||||
|
||||
```bash
|
||||
curl http://192.168.8.64:11434/api/tags
|
||||
|
||||
curl http://192.168.8.64:11434/api/chat -d '{
|
||||
"model": "huihui_ai/gemma-4-abliterated:e4b",
|
||||
"messages": [{"role": "user", "content": "ping"}],
|
||||
"stream": false
|
||||
}'
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 8. 首次运行与验证
|
||||
|
||||
```bash
|
||||
source venv/bin/activate
|
||||
cd ~/Trading_Studio
|
||||
|
||||
# 前台启动（调试）
|
||||
python app.py
|
||||
```
|
||||
|
||||
浏览器访问：
|
||||
|
||||
```
|
||||
http://<本机局域网IP>:5683
|
||||
```
|
||||
|
||||
### 8.1 验证清单
|
||||
|
||||
- [ ] 页面加载，Ollama 状态显示在线
|
||||
- [ ] 上传 10-30s 参考人声 → 音色锁定成功，生成 `speaker_emb.pt`
|
||||
- [ ] 上传复盘录音 → Whisper 识别出中文文本
|
||||
- [ ] 点击润色 → 返回 Gemma4 处理后的文稿
|
||||
- [ ] 点击合成 → `outputs/` 下生成 24kHz wav
|
||||
|
||||
### 8.2 日志位置
|
||||
|
||||
- 应用日志：`trading_studio.log`（项目根目录）
|
||||
- PM2 日志：`logs/pm2-out.log`、`logs/pm2-error.log`
|
||||
|
||||
```bash
|
||||
mkdir -p logs
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 9. PM2 进程守护
|
||||
|
||||
Trading Studio 原生支持 PM2 常驻管理，确保 Gradio 服务崩溃后自动重启、开机自启。
|
||||
|
||||
### 9.1 安装 Node.js 与 PM2
|
||||
|
||||
```bash
|
||||
# 安装 Node.js 20 LTS
|
||||
curl -fsSL https://deb.nodesource.com/setup_20.x | sudo -E bash -
|
||||
sudo apt install -y nodejs
|
||||
|
||||
# 全局安装 PM2
|
||||
sudo npm install -g pm2
|
||||
```
|
||||
|
||||
### 9.2 方式 A：使用 ecosystem.config.js（推荐）
|
||||
|
||||
项目已内置 `ecosystem.config.js`：
|
||||
|
||||
```javascript
|
||||
module.exports = {
|
||||
apps: [{
|
||||
name: "trading_studio",
|
||||
script: "app.py",
|
||||
interpreter: "./venv/bin/python",
|
||||
cwd: __dirname,
|
||||
instances: 1,
|
||||
autorestart: true,
|
||||
max_memory_restart: "6G",
|
||||
env: {
|
||||
PYTHONUNBUFFERED: "1",
|
||||
CUDA_VISIBLE_DEVICES: "0",
|
||||
},
|
||||
}],
|
||||
};
|
||||
```
|
||||
|
||||
启动：
|
||||
|
||||
```bash
|
||||
cd ~/Trading_Studio
|
||||
mkdir -p logs
|
||||
|
||||
pm2 start ecosystem.config.js
|
||||
pm2 status
|
||||
pm2 logs trading_studio --lines 50
|
||||
```
|
||||
|
||||
### 9.3 方式 B：直接命令行
|
||||
|
||||
```bash
|
||||
cd ~/Trading_Studio
|
||||
|
||||
pm2 start app.py \
|
||||
--name "trading_studio" \
|
||||
--interpreter ./venv/bin/python
|
||||
|
||||
pm2 save
|
||||
```
|
||||
|
||||
### 9.4 开机自启
|
||||
|
||||
```bash
|
||||
pm2 startup
|
||||
# 按提示执行输出的 sudo 命令
|
||||
|
||||
pm2 save
|
||||
```
|
||||
|
||||
### 9.5 常用运维命令
|
||||
|
||||
```bash
|
||||
pm2 restart trading_studio # 重启（改代码后）
|
||||
pm2 stop trading_studio # 停止
|
||||
pm2 delete trading_studio # 移除
|
||||
pm2 monit # 实时监控 CPU/内存
|
||||
```
|
||||
|
||||
### 9.6 更新代码后重新部署
|
||||
|
||||
```bash
|
||||
cd ~/Trading_Studio
|
||||
git pull
|
||||
source venv/bin/activate
|
||||
pip install -r requirements.txt # 若有新依赖
|
||||
pm2 restart trading_studio
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 10. 迁移与故障排查
|
||||
|
||||
### 10.1 迁移到新机器
|
||||
|
||||
1. 复制 `speaker_emb.pt`（音色文件，在 `.gitignore` 中，需手动备份）
|
||||
2. 新机器按本文档完整部署
|
||||
3. 将 `speaker_emb.pt` 放回项目根目录
|
||||
4. `pm2 restart trading_studio`
|
||||
|
||||
### 10.2 CUDA / 显存问题
|
||||
|
||||
```bash
|
||||
# 查看显存占用
|
||||
nvidia-smi
|
||||
|
||||
# 若 OOM，确保无其他 GPU 进程
|
||||
fuser -v /dev/nvidia*
|
||||
```
|
||||
|
||||
Whisper 与 ChatTTS 不会同时常驻最大显存，但首次加载模型时峰值较高。建议：
|
||||
|
||||
- 锁定 120W 功耗墙
|
||||
- `max_memory_restart: "6G"` 已在 PM2 配置中设置
|
||||
|
||||
### 10.3 Whisper CUDA 报错
|
||||
|
||||
```
|
||||
错误: CUDA initialization failed / out of memory
|
||||
```
|
||||
|
||||
处理：
|
||||
|
||||
1. 重启 PM2 进程释放显存
|
||||
2. 确认 `compute_type="float16"`（已在 config.py 配置）
|
||||
3. 降级模型为 `base`（修改 `config.py` 中 `WHISPER_MODEL_SIZE`）
|
||||
|
||||
### 10.4 Ollama 超时
|
||||
|
||||
```
|
||||
连接 Ollama 超时（>60s）
|
||||
```
|
||||
|
||||
处理：
|
||||
|
||||
1. 确认 Ollama 节点模型已预加载：`ollama run huihui_ai/gemma-4-abliterated:e4b`
|
||||
2. 增大 `config.py` 中 `OLLAMA_TIMEOUT`
|
||||
3. 检查防火墙：`sudo ufw allow from 192.168.8.0/24 to any port 11434`（在 Ollama 节点）
|
||||
|
||||
### 10.5 ChatTTS 音色文件损坏
|
||||
|
||||
```bash
|
||||
rm speaker_emb.pt
|
||||
# 重新在 Web UI「音色锁定」上传参考人声
|
||||
```
|
||||
|
||||
### 10.6 端口 5683 被占用
|
||||
|
||||
```bash
|
||||
sudo lsof -i :5683
|
||||
# 或
|
||||
ss -tlnp | grep 5683
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 附录：防火墙（本机 Gradio）
|
||||
|
||||
若需局域网其他设备访问 Web UI：
|
||||
|
||||
```bash
|
||||
sudo ufw allow 5683/tcp
|
||||
sudo ufw reload
|
||||
```
|
||||
|
||||
访问地址：`http://<服务器局域网IP>:5683`
|
||||
|
||||
---
|
||||
|
||||
## 附录：config.py 关键常量速查
|
||||
|
||||
```python
|
||||
HOST = "0.0.0.0"
|
||||
PORT = 5683
|
||||
OLLAMA_URL = "http://192.168.8.64:11434/api/chat"
|
||||
MODEL_NAME = "huihui_ai/gemma-4-abliterated:e4b"
|
||||
WHISPER_MODEL_SIZE = "small"
|
||||
WHISPER_DEVICE = "cuda"
|
||||
WHISPER_COMPUTE_TYPE = "float16"
|
||||
SPEAKER_EMB_PATH = "speaker_emb.pt"
|
||||
TTS_SAMPLE_RATE = 24000
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
**部署完成后，请先在「音色锁定」模块完成首次音色提取，再进行日常复盘配音生产。**
|
||||
@@ -0,0 +1,197 @@
|
||||
# Trading Studio
|
||||
|
||||
**本地量化交易复盘 → B 站长视频配音生产流水线**
|
||||
|
||||
Trading Studio 是一套运行在 Ubuntu 物理服务器（RTX 3060 Ti）上的自动化配音系统，专为数字资产量化交易员设计。通过「盲录碎碎念 → 本地 GPU 识别 → 局域网大模型严厉润色 → 本地 GPU 声音克隆」的闭环，高效产出 B 站反思类长视频配音，辅助交易纪律的自我进化。
|
||||
|
||||
**Git 仓库：** https://git.bz121.com/dekun/Trading_Studio.git
|
||||
|
||||
---
|
||||
|
||||
## 系统定位
|
||||
|
||||
| 环节 | 技术栈 | 运行位置 |
|
||||
|------|--------|----------|
|
||||
| 碎碎念录音转写 | Faster-Whisper (CUDA float16) | 本地 3060 Ti |
|
||||
| 纪律审判式润色 | Gemma4 Abliterated @ Ollama | 局域网 `192.168.8.64` |
|
||||
| 固定音色配音 | ChatTTS (CUDA) | 本地 3060 Ti |
|
||||
| Web 中控 | Gradio | 端口 **5683** |
|
||||
|
||||
---
|
||||
|
||||
## 架构说明
|
||||
|
||||
```
|
||||
┌─────────────────────────────────────────────────────────────┐
|
||||
│ Gradio 中控 (app.py:5683) │
|
||||
├──────────────┬──────────────────────┬───────────────────────┤
|
||||
│ 音色锁定 │ 音频识别 │ 润色 + 合成 │
|
||||
│ tts_service │ whisper_service │ llm_service │
|
||||
│ │ │ tts_service │
|
||||
└──────┬───────┴──────────┬───────────┴──────────┬────────────┘
|
||||
│ │ │
|
||||
▼ ▼ ▼
|
||||
speaker_emb.pt Faster-Whisper Ollama HTTP
|
||||
(本地持久化) CUDA / small 192.168.8.64:11434
|
||||
gemma-4-abliterated
|
||||
```
|
||||
|
||||
### 模块解耦
|
||||
|
||||
| 文件 | 职责 |
|
||||
|------|------|
|
||||
| `config.py` | Ollama 地址、模型名、Prompt、路径等全局配置 |
|
||||
| `whisper_service.py` | Faster-Whisper CUDA 转写 |
|
||||
| `llm_service.py` | 远程 Ollama HTTP 非流式润色 |
|
||||
| `tts_service.py` | ChatTTS 音色提取与 wav 合成 |
|
||||
| `app.py` | Gradio 前端与流程编排 |
|
||||
|
||||
---
|
||||
|
||||
## 快速开始
|
||||
|
||||
> 完整环境配置请参阅 [DEPLOY.md](./DEPLOY.md)
|
||||
|
||||
```bash
|
||||
# 1. 克隆仓库
|
||||
git clone https://git.bz121.com/dekun/Trading_Studio.git
|
||||
cd Trading_Studio
|
||||
|
||||
# 2. 创建虚拟环境并安装依赖（详见 DEPLOY.md）
|
||||
python3 -m venv venv
|
||||
source venv/bin/activate
|
||||
pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121
|
||||
pip install -r requirements.txt
|
||||
|
||||
# 3. 启动中控
|
||||
python app.py
|
||||
# 浏览器访问: http://<服务器IP>:5683
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 使用流程
|
||||
|
||||
### 首次使用：锁定音色
|
||||
|
||||
1. 进入 **「音色锁定」** 标签页
|
||||
2. 上传 10-30 秒干净人声参考（你的碎碎念盲录样本）
|
||||
3. （可选）填写参考音频的精确转写，提升 zero-shot 还原度
|
||||
4. 点击 **锁定音色** → 生成 `speaker_emb.pt`
|
||||
|
||||
### 日常生产
|
||||
|
||||
**方式 A — 分步操作：**
|
||||
|
||||
1. **音频极速识别**：上传复盘录音 → Whisper 转写
|
||||
2. **Gemma4 纪律审判**：一键润色，生成逻辑清晰、语气严厉的反思稿
|
||||
3. **ChatTTS 合成**：输出 24kHz `.wav` 成品配音
|
||||
|
||||
**方式 B — 一键生产：**
|
||||
|
||||
上传录音后点击 **启动全流程**，系统自动串联三步。
|
||||
|
||||
---
|
||||
|
||||
## 核心配置（config.py）
|
||||
|
||||
| 配置项 | 默认值 |
|
||||
|--------|--------|
|
||||
| 中控端口 | `5683`（`0.0.0.0` 局域网可访问） |
|
||||
| Ollama 地址 | `http://192.168.8.64:11434` |
|
||||
| 模型名称 | `huihui_ai/gemma-4-abliterated:e4b` |
|
||||
| Whisper 模型 | `small` / CUDA / float16 |
|
||||
| 音色文件 | `speaker_emb.pt` |
|
||||
| 音频输出 | `outputs/` 目录 |
|
||||
|
||||
---
|
||||
|
||||
## PM2 守护运行
|
||||
|
||||
```bash
|
||||
# 方式 1：ecosystem 配置
|
||||
pm2 start ecosystem.config.js
|
||||
|
||||
# 方式 2：直接命令
|
||||
pm2 start app.py --name "trading_studio" --interpreter ./venv/bin/python
|
||||
|
||||
# 常用管理
|
||||
pm2 status
|
||||
pm2 logs trading_studio
|
||||
pm2 restart trading_studio
|
||||
pm2 save && pm2 startup # 开机自启
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## .gitignore 配置
|
||||
|
||||
提交 Git 时请确保忽略以下产物（已在 `.gitignore` 中预设）：
|
||||
|
||||
```gitignore
|
||||
venv/
|
||||
*.wav
|
||||
*.pt
|
||||
*.log
|
||||
uploads/
|
||||
outputs/
|
||||
```
|
||||
|
||||
**说明：**
|
||||
|
||||
- `venv/` — Python 虚拟环境，每台机器独立创建
|
||||
- `*.wav` — 录音与合成音频，体积大且含隐私
|
||||
- `*.pt` — ChatTTS 音色 Embedding 与模型权重
|
||||
- `*.log` — 运行日志
|
||||
|
||||
---
|
||||
|
||||
## 目录结构
|
||||
|
||||
```
|
||||
Trading_Studio/
|
||||
├── app.py # Gradio 主入口
|
||||
├── config.py # 全局配置
|
||||
├── whisper_service.py # Whisper CUDA 识别
|
||||
├── llm_service.py # Ollama 远程润色
|
||||
├── tts_service.py # ChatTTS 音色与合成
|
||||
├── ecosystem.config.js # PM2 守护配置
|
||||
├── requirements.txt # Python 依赖
|
||||
├── README.md # 本文件
|
||||
├── DEPLOY.md # 部署指南
|
||||
├── .gitignore
|
||||
├── speaker_emb.pt # 音色文件（运行时生成，不入库）
|
||||
├── uploads/ # 上传临时目录
|
||||
└── outputs/ # 合成 wav 输出
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 硬件要求
|
||||
|
||||
- **GPU：** NVIDIA RTX 3060 Ti（8GB 显存，建议锁定 120W 功耗墙）
|
||||
- **系统：** Ubuntu 22.04 / 24.04 LTS
|
||||
- **CUDA：** 12.1+（与 PyTorch cu121 匹配）
|
||||
- **局域网：** 可访问 `192.168.8.64:11434` 的 Ollama 节点
|
||||
|
||||
---
|
||||
|
||||
## 常见问题
|
||||
|
||||
**Q: Whisper 报 CUDA 错误？**
|
||||
A: 确认 `nvidia-smi` 正常，且未同时运行其他占显存任务。Whisper 使用 `float16` 已针对 8GB 优化。
|
||||
|
||||
**Q: Ollama 连接失败？**
|
||||
A: 在服务器上执行 `curl http://192.168.8.64:11434/api/tags` 验证连通性，确认模型已 `ollama pull`。
|
||||
|
||||
**Q: TTS 音色不稳定？**
|
||||
A: 重新锁定音色，填写参考音频精确转写，并保持 `temperature=0.3` 低随机性。
|
||||
|
||||
**Q: 合成音频为空或噪声？**
|
||||
A: 检查润色文本长度（过短可能导致异常），确认 `speaker_emb.pt` 存在且有效。
|
||||
|
||||
---
|
||||
|
||||
## License
|
||||
|
||||
Private — 仅供个人量化交易复盘使用。
|
||||
@@ -0,0 +1,378 @@
|
||||
"""
|
||||
Trading Studio — 自动化交易复盘视频配音系统
|
||||
Gradio Web 中控：音色锁定 → Whisper 识别 → Gemma4 润色 → ChatTTS 合成
|
||||
"""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import logging
|
||||
import shutil
|
||||
import sys
|
||||
import uuid
|
||||
from pathlib import Path
|
||||
|
||||
import gradio as gr
|
||||
|
||||
from config import (
|
||||
GIT_REPO_URL,
|
||||
HOST,
|
||||
MODEL_NAME,
|
||||
OLLAMA_URL,
|
||||
PORT,
|
||||
SPEAKER_EMB_PATH,
|
||||
UPLOAD_DIR,
|
||||
)
|
||||
from llm_service import check_ollama_health, polish_text
|
||||
from tts_service import generate_voice, save_fixed_speaker, speaker_is_ready
|
||||
from whisper_service import transcribe_audio
|
||||
|
||||
# ---------------------------------------------------------------------------
|
||||
# 日志
|
||||
# ---------------------------------------------------------------------------
|
||||
logging.basicConfig(
|
||||
level=logging.INFO,
|
||||
format="%(asctime)s [%(levelname)s] %(name)s: %(message)s",
|
||||
handlers=[
|
||||
logging.StreamHandler(sys.stdout),
|
||||
logging.FileHandler("trading_studio.log", encoding="utf-8"),
|
||||
],
|
||||
)
|
||||
logger = logging.getLogger("trading_studio")
|
||||
|
||||
# ---------------------------------------------------------------------------
|
||||
# 全局 UI 状态（Gradio State）
|
||||
# ---------------------------------------------------------------------------
|
||||
# raw_transcript / polished_script 在流水线中传递
|
||||
|
||||
|
||||
def _save_upload(upload_file) -> str | None:
|
||||
"""将 Gradio 上传文件复制到本地 uploads 目录，返回持久化路径。"""
|
||||
if upload_file is None:
|
||||
return None
|
||||
|
||||
src = Path(upload_file)
|
||||
if not src.exists():
|
||||
return None
|
||||
|
||||
dest = UPLOAD_DIR / f"{uuid.uuid4().hex}_{src.name}"
|
||||
shutil.copy2(src, dest)
|
||||
return str(dest)
|
||||
|
||||
|
||||
# ---------------------------------------------------------------------------
|
||||
# 模块 1：音色锁定
|
||||
# ---------------------------------------------------------------------------
|
||||
def ui_lock_speaker(audio_file, sample_transcript: str) -> str:
|
||||
"""【音色锁定】从参考人声提取并保存 Speaker Embedding。"""
|
||||
path = _save_upload(audio_file)
|
||||
if not path:
|
||||
return "请上传 10-30 秒干净参考人声（wav/mp3 均可）。"
|
||||
|
||||
ok, msg = save_fixed_speaker(path, sample_transcript or "")
|
||||
return msg if ok else f"❌ {msg}"
|
||||
|
||||
|
||||
def ui_speaker_status() -> str:
|
||||
"""刷新音色状态。"""
|
||||
ok, msg = speaker_is_ready()
|
||||
return f"✅ {msg}" if ok else f"⚠️ {msg}"
|
||||
|
||||
|
||||
# ---------------------------------------------------------------------------
|
||||
# 模块 2：音频极速识别
|
||||
# ---------------------------------------------------------------------------
|
||||
def ui_transcribe(audio_file) -> tuple[str, str]:
|
||||
"""【Whisper 识别】返回 (转写文本, 状态日志)。"""
|
||||
path = _save_upload(audio_file)
|
||||
if not path:
|
||||
return "", "请上传待识别的碎碎念录音。"
|
||||
|
||||
ok, result = transcribe_audio(path)
|
||||
if ok:
|
||||
return result, f"✅ 识别完成，共 {len(result)} 字。"
|
||||
return "", f"❌ {result}"
|
||||
|
||||
|
||||
# ---------------------------------------------------------------------------
|
||||
# 模块 3：Gemma4 纪律审判
|
||||
# ---------------------------------------------------------------------------
|
||||
def ui_polish(raw_text: str) -> tuple[str, str]:
|
||||
"""【LLM 润色】对转写稿进行严厉自我反思式润色。"""
|
||||
if not raw_text or not raw_text.strip():
|
||||
return "", "请先完成语音识别或手动粘贴转写文本。"
|
||||
|
||||
ok, result = polish_text(raw_text)
|
||||
if ok:
|
||||
return result, f"✅ Gemma4 润色完成，共 {len(result)} 字。"
|
||||
return "", f"❌ {result}"
|
||||
|
||||
|
||||
def ui_check_ollama() -> str:
|
||||
"""检测远程 Ollama 节点状态。"""
|
||||
ok, msg = check_ollama_health()
|
||||
return f"✅ {msg}" if ok else f"❌ {msg}"
|
||||
|
||||
|
||||
# ---------------------------------------------------------------------------
|
||||
# 模块 4：ChatTTS 音频合成
|
||||
# ---------------------------------------------------------------------------
|
||||
def ui_synthesize(polished_text: str) -> tuple[str | None, str]:
|
||||
"""【TTS 合成】生成最终 wav 配音文件。"""
|
||||
if not polished_text or not polished_text.strip():
|
||||
return None, "请先完成 Gemma4 润色。"
|
||||
|
||||
ok, msg, wav_path = generate_voice(polished_text)
|
||||
if ok and wav_path:
|
||||
return wav_path, f"✅ {msg}"
|
||||
return None, f"❌ {msg}"
|
||||
|
||||
|
||||
# ---------------------------------------------------------------------------
|
||||
# 一键流水线
|
||||
# ---------------------------------------------------------------------------
|
||||
def ui_full_pipeline(
|
||||
audio_file,
|
||||
skip_polish: bool,
|
||||
manual_raw: str,
|
||||
) -> tuple[str, str, str | None, str]:
|
||||
"""
|
||||
串联执行：识别 → 润色（可跳过）→ 合成。
|
||||
返回 (raw, polished, wav_path, log)
|
||||
"""
|
||||
logs: list[str] = []
|
||||
|
||||
# Step 1: 识别
|
||||
if manual_raw and manual_raw.strip():
|
||||
raw = manual_raw.strip()
|
||||
logs.append(f"使用手动输入转写稿（{len(raw)} 字）。")
|
||||
else:
|
||||
path = _save_upload(audio_file)
|
||||
if not path:
|
||||
return "", "", None, "❌ 请上传录音或手动填写转写文本。"
|
||||
ok, result = transcribe_audio(path)
|
||||
if not ok:
|
||||
return "", "", None, f"❌ 识别失败: {result}"
|
||||
raw = result
|
||||
logs.append(f"✅ Whisper 识别完成（{len(raw)} 字）。")
|
||||
|
||||
# Step 2: 润色
|
||||
if skip_polish:
|
||||
polished = raw
|
||||
logs.append("已跳过 LLM 润色，直接使用原文。")
|
||||
else:
|
||||
ok, result = polish_text(raw)
|
||||
if not ok:
|
||||
return raw, "", None, f"❌ 润色失败: {result}\n" + "\n".join(logs)
|
||||
polished = result
|
||||
logs.append(f"✅ Gemma4 润色完成（{len(polished)} 字）。")
|
||||
|
||||
# Step 3: 合成
|
||||
ok, msg, wav_path = generate_voice(polished)
|
||||
if not ok:
|
||||
return raw, polished, None, f"❌ 合成失败: {msg}\n" + "\n".join(logs)
|
||||
|
||||
logs.append(f"✅ {msg}")
|
||||
return raw, polished, wav_path, "\n".join(logs)
|
||||
|
||||
|
||||
# ---------------------------------------------------------------------------
|
||||
# Gradio 界面
|
||||
# ---------------------------------------------------------------------------
|
||||
CUSTOM_CSS = """
|
||||
/* 硬核暗黑科技风 */
|
||||
.gradio-container {
|
||||
background: linear-gradient(160deg, #0a0a0f 0%, #12121a 40%, #0d0d12 100%) !important;
|
||||
color: #c8c8d0 !important;
|
||||
}
|
||||
.dark-panel {
|
||||
border: 1px solid #2a2a35;
|
||||
border-radius: 8px;
|
||||
padding: 16px;
|
||||
background: rgba(18, 18, 26, 0.85);
|
||||
margin-bottom: 12px;
|
||||
}
|
||||
h1, h2, h3 { color: #e8e8f0 !important; letter-spacing: 0.05em; }
|
||||
.status-bar {
|
||||
font-family: 'Consolas', 'Monaco', monospace;
|
||||
font-size: 0.85em;
|
||||
color: #7a7a90;
|
||||
}
|
||||
footer { visibility: hidden; }
|
||||
"""
|
||||
|
||||
|
||||
def build_app() -> gr.Blocks:
|
||||
"""构建 Gradio 主界面。"""
|
||||
theme = gr.themes.Monochrome(
|
||||
primary_hue="slate",
|
||||
secondary_hue="gray",
|
||||
neutral_hue="slate",
|
||||
font=[gr.themes.GoogleFont("Inter"), "system-ui", "sans-serif"],
|
||||
font_mono=[gr.themes.GoogleFont("JetBrains Mono"), "Consolas", "monospace"],
|
||||
).set(
|
||||
body_background_fill="#0a0a0f",
|
||||
body_background_fill_dark="#0a0a0f",
|
||||
block_background_fill="#12121a",
|
||||
block_background_fill_dark="#12121a",
|
||||
block_border_color="#2a2a35",
|
||||
block_label_text_color="#9090a0",
|
||||
input_background_fill="#1a1a24",
|
||||
button_primary_background_fill="#3a3a50",
|
||||
button_primary_background_fill_hover="#4a4a60",
|
||||
)
|
||||
|
||||
with gr.Blocks(
|
||||
title="Trading Studio | 交易复盘配音中控",
|
||||
theme=theme,
|
||||
css=CUSTOM_CSS,
|
||||
) as demo:
|
||||
gr.Markdown(
|
||||
f"""
|
||||
# ⚡ Trading Studio
|
||||
**本地量化交易复盘 → B 站配音生产流水线**
|
||||
|
||||
`Whisper(CUDA)` → `Gemma4 @ {OLLAMA_URL}` → `ChatTTS(CUDA)`
|
||||
|
||||
> 仓库: [{GIT_REPO_URL}]({GIT_REPO_URL})
|
||||
""",
|
||||
elem_classes=["dark-panel"],
|
||||
)
|
||||
|
||||
with gr.Row():
|
||||
ollama_status = gr.Textbox(
|
||||
label="Ollama 节点",
|
||||
value=f"模型: {MODEL_NAME}",
|
||||
interactive=False,
|
||||
scale=3,
|
||||
elem_classes=["status-bar"],
|
||||
)
|
||||
speaker_status = gr.Textbox(
|
||||
label="音色状态",
|
||||
value="检测中...",
|
||||
interactive=False,
|
||||
scale=2,
|
||||
elem_classes=["status-bar"],
|
||||
)
|
||||
refresh_btn = gr.Button("🔄 刷新状态", scale=1)
|
||||
|
||||
refresh_btn.click(
|
||||
fn=lambda: (ui_check_ollama(), ui_speaker_status()),
|
||||
outputs=[ollama_status, speaker_status],
|
||||
)
|
||||
|
||||
with gr.Tabs():
|
||||
# ---- Tab 1: 音色锁定 ----
|
||||
with gr.Tab("🎙️ 音色锁定"):
|
||||
gr.Markdown(
|
||||
"上传 **10-30 秒** 干净人声样本，系统将提取 Speaker Embedding "
|
||||
f"并保存至 `{SPEAKER_EMB_PATH.name}`，后续合成 100% 还原音色。"
|
||||
)
|
||||
with gr.Row():
|
||||
spk_audio = gr.Audio(
|
||||
label="参考人声（碎碎念盲录样本）",
|
||||
type="filepath",
|
||||
sources=["upload", "microphone"],
|
||||
)
|
||||
spk_transcript = gr.Textbox(
|
||||
label="参考音频精确转写（可选，提升还原度）",
|
||||
placeholder="尽量与参考音频内容完全一致...",
|
||||
lines=6,
|
||||
)
|
||||
lock_btn = gr.Button("🔒 锁定音色", variant="primary")
|
||||
lock_log = gr.Textbox(label="锁定结果", lines=4, interactive=False)
|
||||
lock_btn.click(ui_lock_speaker, [spk_audio, spk_transcript], lock_log)
|
||||
|
||||
# ---- Tab 2: 分步操作 ----
|
||||
with gr.Tab("🔧 分步流水线"):
|
||||
with gr.Row():
|
||||
with gr.Column(scale=1):
|
||||
gr.Markdown("### Step 1 · 音频极速识别")
|
||||
rec_audio = gr.Audio(
|
||||
label="交易复盘碎碎念录音",
|
||||
type="filepath",
|
||||
sources=["upload", "microphone"],
|
||||
)
|
||||
transcribe_btn = gr.Button("⚡ Faster-Whisper 识别", variant="primary")
|
||||
transcribe_log = gr.Textbox(label="识别日志", lines=2, interactive=False)
|
||||
|
||||
with gr.Column(scale=1):
|
||||
gr.Markdown("### Step 2 · Gemma4 纪律审判")
|
||||
raw_text = gr.Textbox(
|
||||
label="转写原文（可编辑）",
|
||||
lines=10,
|
||||
placeholder="识别结果将显示在此，也可手动粘贴...",
|
||||
)
|
||||
polish_btn = gr.Button("⚖️ 远程 Gemma4 严厉润色", variant="primary")
|
||||
polish_log = gr.Textbox(label="润色日志", lines=2, interactive=False)
|
||||
|
||||
with gr.Column(scale=1):
|
||||
gr.Markdown("### Step 3 · ChatTTS 配音合成")
|
||||
polished_text = gr.Textbox(
|
||||
label="润色配音稿（可编辑）",
|
||||
lines=10,
|
||||
placeholder="润色结果将显示在此...",
|
||||
)
|
||||
synth_btn = gr.Button("🔊 合成配音 WAV", variant="primary")
|
||||
synth_log = gr.Textbox(label="合成日志", lines=2, interactive=False)
|
||||
output_audio = gr.Audio(label="成品配音", type="filepath")
|
||||
|
||||
transcribe_btn.click(ui_transcribe, rec_audio, [raw_text, transcribe_log])
|
||||
polish_btn.click(ui_polish, raw_text, [polished_text, polish_log])
|
||||
synth_btn.click(ui_synthesize, polished_text, [output_audio, synth_log])
|
||||
|
||||
# ---- Tab 3: 一键生产 ----
|
||||
with gr.Tab("🚀 一键生产"):
|
||||
gr.Markdown(
|
||||
"上传碎碎念录音，系统自动完成 **识别 → 润色 → 合成** 全流程。"
|
||||
)
|
||||
with gr.Row():
|
||||
pipe_audio = gr.Audio(
|
||||
label="复盘录音",
|
||||
type="filepath",
|
||||
sources=["upload", "microphone"],
|
||||
)
|
||||
pipe_manual = gr.Textbox(
|
||||
label="或手动输入转写（跳过识别）",
|
||||
lines=4,
|
||||
placeholder="若已有转写文本，可直接粘贴，留空则走 Whisper 识别",
|
||||
)
|
||||
skip_polish_cb = gr.Checkbox(
|
||||
label="跳过 Gemma4 润色（仅测试 TTS）",
|
||||
value=False,
|
||||
)
|
||||
pipeline_btn = gr.Button("▶ 启动全流程", variant="primary", size="lg")
|
||||
pipeline_log = gr.Textbox(label="流水线日志", lines=6, interactive=False)
|
||||
with gr.Row():
|
||||
pipe_raw = gr.Textbox(label="转写原文", lines=6)
|
||||
pipe_polished = gr.Textbox(label="润色稿", lines=6)
|
||||
pipe_output = gr.Audio(label="成品配音", type="filepath")
|
||||
|
||||
pipeline_btn.click(
|
||||
ui_full_pipeline,
|
||||
[pipe_audio, skip_polish_cb, pipe_manual],
|
||||
[pipe_raw, pipe_polished, pipe_output, pipeline_log],
|
||||
)
|
||||
|
||||
demo.load(
|
||||
fn=lambda: (ui_check_ollama(), ui_speaker_status()),
|
||||
outputs=[ollama_status, speaker_status],
|
||||
)
|
||||
|
||||
return demo
|
||||
|
||||
|
||||
def main() -> None:
|
||||
"""主入口：启动 Gradio 服务。"""
|
||||
logger.info("Trading Studio 启动中... HOST=%s PORT=%s", HOST, PORT)
|
||||
app = build_app()
|
||||
app.launch(
|
||||
server_name=HOST,
|
||||
server_port=PORT,
|
||||
share=False,
|
||||
show_error=True,
|
||||
allowed_paths=[str(Path(__file__).resolve().parent / "outputs")],
|
||||
)
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
main()
|
||||
@@ -0,0 +1,82 @@
|
||||
"""
|
||||
Trading Studio 全局配置模块
|
||||
统一存放局域网节点、模型名称、固定 Prompt 及本地路径。
|
||||
"""
|
||||
|
||||
from pathlib import Path
|
||||
|
||||
# ---------------------------------------------------------------------------
|
||||
# 网络与服务
|
||||
# ---------------------------------------------------------------------------
|
||||
# 远程 Ollama 节点（局域网大模型审查润色）
|
||||
OLLAMA_HOST = "192.168.8.64"
|
||||
OLLAMA_PORT = 11434
|
||||
OLLAMA_URL = f"http://{OLLAMA_HOST}:{OLLAMA_PORT}/api/chat"
|
||||
|
||||
# 指定无限制版 Gemma4 模型
|
||||
MODEL_NAME = "huihui_ai/gemma-4-abliterated:e4b"
|
||||
|
||||
# Gradio 中控固定端口（硬性死规则）
|
||||
HOST = "0.0.0.0"
|
||||
PORT = 5683
|
||||
|
||||
# HTTP 请求超时（秒）
|
||||
OLLAMA_TIMEOUT = 60
|
||||
|
||||
# ---------------------------------------------------------------------------
|
||||
# LLM 系统 Prompt
|
||||
# ---------------------------------------------------------------------------
|
||||
SYSTEM_PROMPT = (
|
||||
"你是一个冷静、极其严格的数字资产量化交易员。"
|
||||
"请把下面这段口语化、包含结巴和逻辑混乱的交易复盘录音转写，"
|
||||
"润色成一段逻辑清晰、行文通顺的 B 站长视频反思配音稿。"
|
||||
"语气要内向、克制、严谨。"
|
||||
"如果原视频中有由于心态不好、违背交易纪律（如手贱乱开仓、提前平仓）"
|
||||
"导致少赚或亏损的部分，请用冷酷、严厉的语气狠狠地自我吐槽、反思该点。"
|
||||
"去掉所有无意义的口头禅，字数不做删减。"
|
||||
)
|
||||
|
||||
# ---------------------------------------------------------------------------
|
||||
# Faster-Whisper 配置
|
||||
# ---------------------------------------------------------------------------
|
||||
WHISPER_MODEL_SIZE = "small"
|
||||
WHISPER_DEVICE = "cuda"
|
||||
WHISPER_COMPUTE_TYPE = "float16"
|
||||
WHISPER_LANGUAGE = "zh"
|
||||
|
||||
# ---------------------------------------------------------------------------
|
||||
# ChatTTS 配置
|
||||
# ---------------------------------------------------------------------------
|
||||
# 项目根目录
|
||||
BASE_DIR = Path(__file__).resolve().parent
|
||||
|
||||
# 固定音色 Embedding 存储路径
|
||||
SPEAKER_EMB_PATH = BASE_DIR / "speaker_emb.pt"
|
||||
|
||||
# 合成音频输出目录
|
||||
OUTPUT_DIR = BASE_DIR / "outputs"
|
||||
OUTPUT_DIR.mkdir(parents=True, exist_ok=True)
|
||||
|
||||
# ChatTTS 采样率（Hz）
|
||||
TTS_SAMPLE_RATE = 24000
|
||||
|
||||
# 音色样本时长建议（秒）
|
||||
SPEAKER_SAMPLE_MIN_SEC = 10
|
||||
SPEAKER_SAMPLE_MAX_SEC = 30
|
||||
|
||||
# TTS 推理默认参数（低 temperature 有助于音色稳定）
|
||||
TTS_TEMPERATURE = 0.3
|
||||
TTS_TOP_P = 0.7
|
||||
TTS_TOP_K = 20
|
||||
TTS_SPEED_PROMPT = "[speed_5]"
|
||||
|
||||
# ---------------------------------------------------------------------------
|
||||
# 上传临时文件目录
|
||||
# ---------------------------------------------------------------------------
|
||||
UPLOAD_DIR = BASE_DIR / "uploads"
|
||||
UPLOAD_DIR.mkdir(parents=True, exist_ok=True)
|
||||
|
||||
# ---------------------------------------------------------------------------
|
||||
# Git 仓库（文档引用）
|
||||
# ---------------------------------------------------------------------------
|
||||
GIT_REPO_URL = "https://git.bz121.com/dekun/Trading_Studio.git"
|
||||
@@ -0,0 +1,26 @@
|
||||
/**
|
||||
* PM2 进程守护配置
|
||||
* 用法: pm2 start ecosystem.config.js
|
||||
*/
|
||||
module.exports = {
|
||||
apps: [
|
||||
{
|
||||
name: "trading_studio",
|
||||
script: "app.py",
|
||||
interpreter: "./venv/bin/python",
|
||||
cwd: __dirname,
|
||||
instances: 1,
|
||||
autorestart: true,
|
||||
watch: false,
|
||||
max_memory_restart: "6G",
|
||||
env: {
|
||||
PYTHONUNBUFFERED: "1",
|
||||
CUDA_VISIBLE_DEVICES: "0",
|
||||
},
|
||||
error_file: "./logs/pm2-error.log",
|
||||
out_file: "./logs/pm2-out.log",
|
||||
log_date_format: "YYYY-MM-DD HH:mm:ss",
|
||||
merge_logs: true,
|
||||
},
|
||||
],
|
||||
};
|
||||
+162
@@ -0,0 +1,162 @@
|
||||
"""
|
||||
远程 Ollama LLM 润色服务
|
||||
通过局域网 HTTP 请求 Gemma4 模型，对交易复盘转写稿进行纪律审判式润色。
|
||||
"""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import logging
|
||||
from typing import Tuple
|
||||
|
||||
import requests
|
||||
|
||||
from config import MODEL_NAME, OLLAMA_TIMEOUT, OLLAMA_URL, SYSTEM_PROMPT
|
||||
|
||||
logger = logging.getLogger(__name__)
|
||||
|
||||
|
||||
def _build_payload(raw_text: str) -> dict:
|
||||
"""构造 Ollama /api/chat 非流式请求体。"""
|
||||
return {
|
||||
"model": MODEL_NAME,
|
||||
"messages": [
|
||||
{"role": "system", "content": SYSTEM_PROMPT},
|
||||
{
|
||||
"role": "user",
|
||||
"content": (
|
||||
"以下是我的交易复盘录音转写原文，请严格按系统要求润色：\n\n"
|
||||
f"{raw_text}"
|
||||
),
|
||||
},
|
||||
],
|
||||
"stream": False,
|
||||
"options": {
|
||||
"temperature": 0.7,
|
||||
"num_predict": 4096,
|
||||
},
|
||||
}
|
||||
|
||||
|
||||
def _extract_content(response_json: dict) -> str:
|
||||
"""从 Ollama 响应 JSON 中提取 assistant 文本。"""
|
||||
# /api/chat 标准格式
|
||||
message = response_json.get("message")
|
||||
if isinstance(message, dict):
|
||||
content = message.get("content", "").strip()
|
||||
if content:
|
||||
return content
|
||||
|
||||
# 兼容 /api/generate 格式（部分旧版或代理）
|
||||
if "response" in response_json:
|
||||
content = str(response_json["response"]).strip()
|
||||
if content:
|
||||
return content
|
||||
|
||||
raise ValueError(f"无法从 Ollama 响应中解析文本内容: {response_json}")
|
||||
|
||||
|
||||
def polish_text(raw_text: str) -> Tuple[bool, str]:
|
||||
"""
|
||||
调用远程 Ollama 对原始转写文本进行润色。
|
||||
|
||||
Args:
|
||||
raw_text: Whisper 转写得到的原始口语文本
|
||||
|
||||
Returns:
|
||||
(success, polished_text_or_error_message)
|
||||
"""
|
||||
if not raw_text or not raw_text.strip():
|
||||
return False, "润色输入为空，请先完成语音识别。"
|
||||
|
||||
payload = _build_payload(raw_text.strip())
|
||||
|
||||
try:
|
||||
logger.info("正在请求 Ollama: %s, model=%s", OLLAMA_URL, MODEL_NAME)
|
||||
response = requests.post(
|
||||
OLLAMA_URL,
|
||||
json=payload,
|
||||
timeout=OLLAMA_TIMEOUT,
|
||||
)
|
||||
response.raise_for_status()
|
||||
|
||||
data = response.json()
|
||||
polished = _extract_content(data)
|
||||
|
||||
if not polished:
|
||||
return False, "Ollama 返回内容为空，请检查模型是否正常加载。"
|
||||
|
||||
logger.info("润色完成，输出字数: %d", len(polished))
|
||||
return True, polished
|
||||
|
||||
except requests.exceptions.ConnectTimeout:
|
||||
err = (
|
||||
f"连接 Ollama 超时（>{OLLAMA_TIMEOUT}s）。"
|
||||
f"请确认 {OLLAMA_URL} 可达且 Ollama 服务已启动。"
|
||||
)
|
||||
logger.error(err)
|
||||
return False, err
|
||||
|
||||
except requests.exceptions.ReadTimeout:
|
||||
err = (
|
||||
f"Ollama 响应超时（>{OLLAMA_TIMEOUT}s）。"
|
||||
"模型可能正在加载或生成长度过长，请稍后重试。"
|
||||
)
|
||||
logger.error(err)
|
||||
return False, err
|
||||
|
||||
except requests.exceptions.ConnectionError as exc:
|
||||
err = (
|
||||
f"无法连接到 Ollama 节点 ({OLLAMA_URL})。"
|
||||
"请检查局域网连通性、防火墙及 Ollama 是否监听 0.0.0.0:11434。\n"
|
||||
f"详情: {exc}"
|
||||
)
|
||||
logger.error(err)
|
||||
return False, err
|
||||
|
||||
except requests.exceptions.HTTPError as exc:
|
||||
status = exc.response.status_code if exc.response is not None else "?"
|
||||
body = exc.response.text[:500] if exc.response is not None else ""
|
||||
err = (
|
||||
f"Ollama HTTP 错误 ({status})。"
|
||||
f"请确认模型 `{MODEL_NAME}` 已通过 ollama pull 下载。\n"
|
||||
f"响应片段: {body}"
|
||||
)
|
||||
logger.error(err)
|
||||
return False, err
|
||||
|
||||
except ValueError as exc:
|
||||
logger.error("Ollama 响应解析失败: %s", exc)
|
||||
return False, str(exc)
|
||||
|
||||
except requests.exceptions.RequestException as exc:
|
||||
err = f"Ollama 请求异常: {exc}"
|
||||
logger.exception(err)
|
||||
return False, err
|
||||
|
||||
except Exception as exc:
|
||||
err = f"润色过程发生未知错误: {exc}"
|
||||
logger.exception(err)
|
||||
return False, err
|
||||
|
||||
|
||||
def check_ollama_health() -> Tuple[bool, str]:
|
||||
"""
|
||||
快速检测 Ollama 节点是否在线（不触发完整推理）。
|
||||
|
||||
Returns:
|
||||
(online, message)
|
||||
"""
|
||||
base_url = OLLAMA_URL.rsplit("/api/", 1)[0]
|
||||
try:
|
||||
resp = requests.get(f"{base_url}/api/tags", timeout=10)
|
||||
resp.raise_for_status()
|
||||
tags = resp.json().get("models", [])
|
||||
model_names = [m.get("name", "") for m in tags]
|
||||
if any(MODEL_NAME.split(":")[0] in name for name in model_names):
|
||||
return True, f"Ollama 在线，已检测到模型: {MODEL_NAME}"
|
||||
return True, (
|
||||
f"Ollama 在线，但未找到模型 {MODEL_NAME}，"
|
||||
f"请执行: ollama pull {MODEL_NAME}"
|
||||
)
|
||||
except Exception as exc:
|
||||
return False, f"Ollama 不可达: {exc}"
|
||||
@@ -0,0 +1,23 @@
|
||||
# Trading Studio 依赖清单
|
||||
# CUDA 版 PyTorch 请按 DEPLOY.md 单独安装（cu121），此处不重复指定
|
||||
|
||||
# Web 中控
|
||||
gradio>=4.44.0
|
||||
|
||||
# 语音识别（CUDA 加速）
|
||||
faster-whisper>=1.0.0
|
||||
|
||||
# 远程 LLM 通信
|
||||
requests>=2.31.0
|
||||
|
||||
# 语音合成
|
||||
ChatTTS @ git+https://github.com/2noise/ChatTTS.git
|
||||
torchaudio>=2.1.0
|
||||
scipy>=1.11.0
|
||||
numpy>=1.24.0
|
||||
librosa>=0.10.0
|
||||
|
||||
# 音频处理辅助
|
||||
soundfile>=0.12.0
|
||||
|
||||
# PM2 通过 Node.js 全局安装，不在 pip 范围内
|
||||
+305
@@ -0,0 +1,305 @@
|
||||
"""
|
||||
ChatTTS 本地语音合成服务
|
||||
支持从参考人声提取 Speaker Embedding 并固定音色合成配音。
|
||||
"""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import logging
|
||||
import traceback
|
||||
import uuid
|
||||
from datetime import datetime
|
||||
from pathlib import Path
|
||||
from typing import Any, Dict, Optional, Tuple
|
||||
|
||||
import numpy as np
|
||||
import torch
|
||||
from scipy.io import wavfile
|
||||
|
||||
from config import (
|
||||
OUTPUT_DIR,
|
||||
SPEAKER_EMB_PATH,
|
||||
SPEAKER_SAMPLE_MAX_SEC,
|
||||
SPEAKER_SAMPLE_MIN_SEC,
|
||||
TTS_SAMPLE_RATE,
|
||||
TTS_SPEED_PROMPT,
|
||||
TTS_TEMPERATURE,
|
||||
TTS_TOP_K,
|
||||
TTS_TOP_P,
|
||||
)
|
||||
|
||||
logger = logging.getLogger(__name__)
|
||||
|
||||
# 全局 ChatTTS 实例
|
||||
_chat = None
|
||||
_chat_error: Optional[str] = None
|
||||
|
||||
|
||||
def _load_audio_for_chattts(audio_path: str, sample_rate: int = TTS_SAMPLE_RATE) -> np.ndarray:
|
||||
"""
|
||||
加载音频并重采样到 ChatTTS 所需采样率。
|
||||
优先使用 ChatTTS 自带工具，回退到 librosa。
|
||||
"""
|
||||
try:
|
||||
from ChatTTS.utils import load_audio
|
||||
|
||||
return load_audio(audio_path, sample_rate)
|
||||
except ImportError:
|
||||
pass
|
||||
|
||||
try:
|
||||
from tools.audio import load_audio
|
||||
|
||||
return load_audio(audio_path, sample_rate)
|
||||
except ImportError:
|
||||
pass
|
||||
|
||||
import librosa
|
||||
|
||||
audio, _ = librosa.load(audio_path, sr=sample_rate, mono=True)
|
||||
return audio
|
||||
|
||||
|
||||
def _get_audio_duration_sec(audio: np.ndarray, sample_rate: int) -> float:
|
||||
"""计算音频时长（秒）。"""
|
||||
if audio is None or len(audio) == 0:
|
||||
return 0.0
|
||||
return len(audio) / float(sample_rate)
|
||||
|
||||
|
||||
def get_chattts_instance():
|
||||
"""
|
||||
获取或初始化 ChatTTS 模型。
|
||||
启用 GPU 加速，compile=False 以兼容 3060 Ti 8GB 显存。
|
||||
"""
|
||||
global _chat, _chat_error
|
||||
|
||||
if _chat is not None:
|
||||
return _chat, None
|
||||
|
||||
if _chat_error is not None:
|
||||
return None, _chat_error
|
||||
|
||||
try:
|
||||
import ChatTTS
|
||||
|
||||
logger.info("正在加载 ChatTTS 模型...")
|
||||
chat = ChatTTS.Chat()
|
||||
|
||||
# 兼容不同版本 API：load_models（旧版）/ load（新版）
|
||||
if hasattr(chat, "load_models"):
|
||||
chat.load_models(compile=False)
|
||||
elif hasattr(chat, "load"):
|
||||
chat.load(compile=False)
|
||||
else:
|
||||
raise RuntimeError("当前 ChatTTS 版本缺少 load / load_models 方法。")
|
||||
|
||||
_chat = chat
|
||||
logger.info("ChatTTS 模型加载成功。")
|
||||
return _chat, None
|
||||
|
||||
except ImportError as exc:
|
||||
_chat_error = (
|
||||
"未安装 ChatTTS，请参考 DEPLOY.md 安装。\n"
|
||||
f"原始错误: {exc}"
|
||||
)
|
||||
logger.exception("ChatTTS 导入失败")
|
||||
return None, _chat_error
|
||||
|
||||
except Exception as exc:
|
||||
_chat_error = f"ChatTTS 模型加载失败: {exc}\n{traceback.format_exc()}"
|
||||
logger.exception("ChatTTS 初始化异常")
|
||||
return None, _chat_error
|
||||
|
||||
|
||||
def _encode_spk_emb(chat, tensor_or_str: Any) -> str:
|
||||
"""将 Speaker Embedding 编码为 ChatTTS 可用的字符串格式。"""
|
||||
if isinstance(tensor_or_str, str):
|
||||
return tensor_or_str
|
||||
|
||||
if hasattr(chat, "_encode_spk_emb"):
|
||||
return chat._encode_spk_emb(tensor_or_str)
|
||||
|
||||
# 兜底：直接转字符串（部分版本可接受 tensor）
|
||||
return tensor_or_str
|
||||
|
||||
|
||||
def save_fixed_speaker(
|
||||
audio_sample_path: str,
|
||||
sample_transcript: str = "",
|
||||
) -> Tuple[bool, str]:
|
||||
"""
|
||||
从 10-30 秒干净人声中提取 Speaker Embedding 并序列化保存。
|
||||
|
||||
Args:
|
||||
audio_sample_path: 参考人声 wav/mp3 等路径
|
||||
sample_transcript: 参考音频的精确转写（可选，有助于 zero-shot 音色还原）
|
||||
|
||||
Returns:
|
||||
(success, message)
|
||||
"""
|
||||
if not audio_sample_path:
|
||||
return False, "未提供音色参考音频。"
|
||||
|
||||
chat, init_err = get_chattts_instance()
|
||||
if chat is None:
|
||||
return False, init_err or "ChatTTS 不可用。"
|
||||
|
||||
try:
|
||||
audio = _load_audio_for_chattts(audio_sample_path, TTS_SAMPLE_RATE)
|
||||
duration = _get_audio_duration_sec(audio, TTS_SAMPLE_RATE)
|
||||
|
||||
if duration < SPEAKER_SAMPLE_MIN_SEC:
|
||||
return False, (
|
||||
f"参考音频过短（{duration:.1f}s），建议 {SPEAKER_SAMPLE_MIN_SEC}-"
|
||||
f"{SPEAKER_SAMPLE_MAX_SEC} 秒干净人声。"
|
||||
)
|
||||
if duration > SPEAKER_SAMPLE_MAX_SEC + 5:
|
||||
logger.warning("参考音频超过建议时长 %.1fs，将截取前 %ds", duration, SPEAKER_SAMPLE_MAX_SEC)
|
||||
max_samples = SPEAKER_SAMPLE_MAX_SEC * TTS_SAMPLE_RATE
|
||||
audio = audio[:max_samples]
|
||||
|
||||
# 从参考音频提取音色特征
|
||||
spk_smp = chat.sample_audio_speaker(audio)
|
||||
|
||||
# 同时保存编码后的 spk_emb 字符串，便于 infer 时直接使用
|
||||
spk_emb = _encode_spk_emb(chat, spk_smp)
|
||||
|
||||
payload: Dict[str, Any] = {
|
||||
"spk_emb": spk_emb,
|
||||
"spk_smp": spk_smp,
|
||||
"txt_smp": sample_transcript.strip(),
|
||||
"created_at": datetime.now().isoformat(),
|
||||
"source_audio": str(audio_sample_path),
|
||||
}
|
||||
|
||||
torch.save(payload, SPEAKER_EMB_PATH)
|
||||
|
||||
msg = (
|
||||
f"音色已锁定并保存至 {SPEAKER_EMB_PATH}\n"
|
||||
f"参考时长: {duration:.1f}s"
|
||||
)
|
||||
if not sample_transcript.strip():
|
||||
msg += "\n提示：填写参考音频精确转写可进一步提升音色还原度。"
|
||||
|
||||
logger.info("Speaker Embedding 保存成功: %s", SPEAKER_EMB_PATH)
|
||||
return True, msg
|
||||
|
||||
except Exception as exc:
|
||||
err = f"音色提取失败: {exc}\n{traceback.format_exc()}"
|
||||
logger.exception("save_fixed_speaker 失败")
|
||||
return False, err
|
||||
|
||||
|
||||
def _load_speaker_payload() -> Tuple[Optional[Dict[str, Any]], Optional[str]]:
|
||||
"""加载本地 speaker_emb.pt。"""
|
||||
if not SPEAKER_EMB_PATH.exists():
|
||||
return None, (
|
||||
f"未找到固定音色文件 `{SPEAKER_EMB_PATH.name}`。"
|
||||
"请先在【音色锁定】模块上传 10-30 秒参考人声。"
|
||||
)
|
||||
|
||||
try:
|
||||
payload = torch.load(SPEAKER_EMB_PATH, map_location="cpu", weights_only=False)
|
||||
|
||||
# 兼容旧版仅保存 tensor 的文件
|
||||
if isinstance(payload, torch.Tensor):
|
||||
chat, err = get_chattts_instance()
|
||||
if chat is None:
|
||||
return None, err
|
||||
return {
|
||||
"spk_emb": _encode_spk_emb(chat, payload),
|
||||
"spk_smp": None,
|
||||
"txt_smp": "",
|
||||
}, None
|
||||
|
||||
if not isinstance(payload, dict):
|
||||
return None, "speaker_emb.pt 格式无效，请重新锁定音色。"
|
||||
|
||||
return payload, None
|
||||
|
||||
except Exception as exc:
|
||||
return None, f"读取 speaker_emb.pt 失败: {exc}"
|
||||
|
||||
|
||||
def speaker_is_ready() -> Tuple[bool, str]:
|
||||
"""检查固定音色是否已配置。"""
|
||||
payload, err = _load_speaker_payload()
|
||||
if payload is None:
|
||||
return False, err or "音色未配置。"
|
||||
return True, f"已加载固定音色: {SPEAKER_EMB_PATH}"
|
||||
|
||||
|
||||
def generate_voice(refined_text: str) -> Tuple[bool, str, Optional[str]]:
|
||||
"""
|
||||
使用 ChatTTS 将润色后的文稿合成为 wav 配音。
|
||||
|
||||
Args:
|
||||
refined_text: LLM 润色后的配音稿
|
||||
|
||||
Returns:
|
||||
(success, message, output_wav_path_or_none)
|
||||
"""
|
||||
if not refined_text or not refined_text.strip():
|
||||
return False, "合成文本为空，请先完成润色。", None
|
||||
|
||||
chat, init_err = get_chattts_instance()
|
||||
if chat is None:
|
||||
return False, init_err or "ChatTTS 不可用。", None
|
||||
|
||||
payload, spk_err = _load_speaker_payload()
|
||||
if payload is None:
|
||||
return False, spk_err or "请先锁定音色。", None
|
||||
|
||||
try:
|
||||
import ChatTTS
|
||||
|
||||
spk_emb = payload.get("spk_emb")
|
||||
spk_smp = payload.get("spk_smp")
|
||||
txt_smp = payload.get("txt_smp", "")
|
||||
|
||||
params_infer_code = ChatTTS.Chat.InferCodeParams(
|
||||
prompt=TTS_SPEED_PROMPT,
|
||||
spk_emb=spk_emb,
|
||||
spk_smp=spk_smp if spk_smp else None,
|
||||
txt_smp=txt_smp if txt_smp else None,
|
||||
temperature=TTS_TEMPERATURE,
|
||||
top_P=TTS_TOP_P,
|
||||
top_K=TTS_TOP_K,
|
||||
)
|
||||
|
||||
# 内向克制语气：降低 oral 强度
|
||||
params_refine_text = ChatTTS.Chat.RefineTextParams(
|
||||
prompt="[oral_2][laugh_0][break_4]",
|
||||
)
|
||||
|
||||
wavs = chat.infer(
|
||||
refined_text.strip(),
|
||||
skip_refine_text=False,
|
||||
params_refine_text=params_refine_text,
|
||||
params_infer_code=params_infer_code,
|
||||
)
|
||||
|
||||
if not wavs or len(wavs) == 0:
|
||||
return False, "ChatTTS 未生成有效音频。", None
|
||||
|
||||
wav_array = np.asarray(wavs[0], dtype=np.float32)
|
||||
|
||||
# 归一化并转 int16
|
||||
peak = np.max(np.abs(wav_array)) or 1.0
|
||||
wav_int16 = (wav_array / peak * 32767).astype(np.int16)
|
||||
|
||||
timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
|
||||
filename = f"voiceover_{timestamp}_{uuid.uuid4().hex[:6]}.wav"
|
||||
output_path = OUTPUT_DIR / filename
|
||||
|
||||
wavfile.write(str(output_path), TTS_SAMPLE_RATE, wav_int16)
|
||||
|
||||
msg = f"配音合成成功: {output_path}"
|
||||
logger.info(msg)
|
||||
return True, msg, str(output_path)
|
||||
|
||||
except Exception as exc:
|
||||
err = f"语音合成失败: {exc}\n{traceback.format_exc()}"
|
||||
logger.exception("generate_voice 失败")
|
||||
return False, err, None
|
||||
@@ -0,0 +1,155 @@
|
||||
"""
|
||||
Faster-Whisper CUDA 语音识别服务
|
||||
封装本地 GPU 加速的音频转写逻辑，适配 RTX 3060 Ti 8GB 显存。
|
||||
"""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import logging
|
||||
import traceback
|
||||
from typing import Optional, Tuple
|
||||
|
||||
from config import (
|
||||
WHISPER_COMPUTE_TYPE,
|
||||
WHISPER_DEVICE,
|
||||
WHISPER_LANGUAGE,
|
||||
WHISPER_MODEL_SIZE,
|
||||
)
|
||||
|
||||
logger = logging.getLogger(__name__)
|
||||
|
||||
# 全局懒加载模型实例，避免 Gradio 重复初始化占用显存
|
||||
_model = None
|
||||
_model_error: Optional[str] = None
|
||||
|
||||
|
||||
def _is_cuda_error(exc: BaseException) -> bool:
|
||||
"""判断异常是否与 CUDA/GPU 相关。"""
|
||||
msg = str(exc).lower()
|
||||
cuda_keywords = (
|
||||
"cuda",
|
||||
"cudnn",
|
||||
"cublas",
|
||||
"gpu",
|
||||
"out of memory",
|
||||
"no kernel image",
|
||||
"device-side assert",
|
||||
)
|
||||
return any(k in msg for k in cuda_keywords)
|
||||
|
||||
|
||||
def get_whisper_model():
|
||||
"""
|
||||
获取或初始化 Faster-Whisper 模型。
|
||||
强制 device=cuda, compute_type=float16。
|
||||
"""
|
||||
global _model, _model_error
|
||||
|
||||
if _model is not None:
|
||||
return _model, None
|
||||
|
||||
if _model_error is not None:
|
||||
return None, _model_error
|
||||
|
||||
try:
|
||||
from faster_whisper import WhisperModel
|
||||
|
||||
logger.info(
|
||||
"正在加载 Whisper 模型: size=%s, device=%s, compute_type=%s",
|
||||
WHISPER_MODEL_SIZE,
|
||||
WHISPER_DEVICE,
|
||||
WHISPER_COMPUTE_TYPE,
|
||||
)
|
||||
_model = WhisperModel(
|
||||
WHISPER_MODEL_SIZE,
|
||||
device=WHISPER_DEVICE,
|
||||
compute_type=WHISPER_COMPUTE_TYPE,
|
||||
)
|
||||
logger.info("Whisper 模型加载成功。")
|
||||
return _model, None
|
||||
|
||||
except ImportError as exc:
|
||||
_model_error = (
|
||||
"未安装 faster-whisper，请执行: pip install faster-whisper\n"
|
||||
f"原始错误: {exc}"
|
||||
)
|
||||
logger.exception("faster-whisper 导入失败")
|
||||
return None, _model_error
|
||||
|
||||
except Exception as exc:
|
||||
if _is_cuda_error(exc):
|
||||
_model_error = (
|
||||
"CUDA 初始化失败，请检查 NVIDIA 驱动、CUDA 运行时及 cuDNN 是否正确安装。\n"
|
||||
f"错误详情: {exc}\n"
|
||||
f"{traceback.format_exc()}"
|
||||
)
|
||||
else:
|
||||
_model_error = f"Whisper 模型加载失败: {exc}\n{traceback.format_exc()}"
|
||||
logger.exception("Whisper 模型加载异常")
|
||||
return None, _model_error
|
||||
|
||||
|
||||
def transcribe_audio(audio_path: str) -> Tuple[bool, str]:
|
||||
"""
|
||||
将音频文件转写为中文文本。
|
||||
|
||||
Args:
|
||||
audio_path: 本地音频文件绝对或相对路径
|
||||
|
||||
Returns:
|
||||
(success, text_or_error_message)
|
||||
"""
|
||||
if not audio_path:
|
||||
return False, "未提供音频文件路径。"
|
||||
|
||||
model, init_error = get_whisper_model()
|
||||
if model is None:
|
||||
return False, init_error or "Whisper 模型不可用。"
|
||||
|
||||
try:
|
||||
segments, info = model.transcribe(
|
||||
audio_path,
|
||||
language=WHISPER_LANGUAGE,
|
||||
beam_size=5,
|
||||
vad_filter=True,
|
||||
)
|
||||
|
||||
text_parts = []
|
||||
for segment in segments:
|
||||
text_parts.append(segment.text.strip())
|
||||
|
||||
result_text = "".join(text_parts).strip()
|
||||
|
||||
if not result_text:
|
||||
return False, (
|
||||
"识别结果为空，请检查音频是否有效、音量是否足够，"
|
||||
f"或尝试更换格式。检测到语言: {getattr(info, 'language', 'unknown')}"
|
||||
)
|
||||
|
||||
logger.info(
|
||||
"转写完成: 语言=%s, 概率=%.2f, 字数=%d",
|
||||
getattr(info, "language", "?"),
|
||||
getattr(info, "language_probability", 0.0),
|
||||
len(result_text),
|
||||
)
|
||||
return True, result_text
|
||||
|
||||
except Exception as exc:
|
||||
if _is_cuda_error(exc):
|
||||
err = (
|
||||
"CUDA 推理异常：显存可能不足或 GPU 状态异常。"
|
||||
"建议关闭其他占用显存的进程后重试。\n"
|
||||
f"错误详情: {exc}"
|
||||
)
|
||||
else:
|
||||
err = f"音频转写失败: {exc}\n{traceback.format_exc()}"
|
||||
|
||||
logger.exception("transcribe_audio 失败")
|
||||
return False, err
|
||||
|
||||
|
||||
def reset_whisper_model() -> None:
|
||||
"""释放模型引用（用于调试或显存回收）。"""
|
||||
global _model, _model_error
|
||||
_model = None
|
||||
_model_error = None
|
||||
Reference in New Issue
Block a user