Spaces:

megatrump
/

test-FunAudioLLM

Paused

App Files Files Community

megatrump commited on Mar 6

Commit

9587328

1 Parent(s): f9e54c0

Add Gradio Pages

Browse files

Files changed (4) hide show

api.py +70 -27
awake.py +81 -0
requirements.txt +4 -3
start.sh +2 -2

api.py CHANGED Viewed

@@ -14,6 +14,7 @@ from funasr import AutoModel
 from dotenv import load_dotenv
 import os
 import time
 # 加载环境变量
 load_dotenv()
@@ -286,40 +287,82 @@ async def transcribe_audio(
         }
-@app.get("/", response_class=HTMLResponse)
-async def root():
-    html_content = """
     <!DOCTYPE html>
     <html>
         <head>
-            <title>SenseVoice API</title>
-            <style>
-                body { font-family: Arial, sans-serif; max-width: 800px; margin: 40px auto; padding: 0 20px; line-height: 1.6; }
-                h1 { color: #2c3e50; }
-                .api-info { background: #f8f9fa; padding: 20px; border-radius: 5px; margin: 20px 0; }
-                .api-link { display: inline-block; background: #3498db; color: white; padding: 10px 20px; text-decoration: none; border-radius: 5px; margin-top: 20px; }
-                .api-link:hover { background: #2980b9; }
-            </style>
         </head>
         <body>
-            <h1>欢迎使用 SenseVoice API</h1>
-            <div class="api-info">
-                <h2>服务信息</h2>
-                <p>版本：1.0.0</p>
-                <p>描述：多语言语音识别服务，支持中文、英语、粤语、日语、韩语等多种语言的语音转写。</p>
-                <h2>主要功能</h2>
-                <ul>
-                    <li>支持多种音频格式：MP3、WAV、FLAC、OGG、M4A</li>
-                    <li>自动语言检测</li>
-                    <li>情感和事件识别</li>
-                    <li>高性能语音识别引擎</li>
-                </ul>
-            </div>
-            <a href="/docs" class="api-link">查看API文档</a>
         </body>
     </html>
-    """
-    return html_content
 if __name__ == "__main__":
     import uvicorn

 from dotenv import load_dotenv
 import os
 import time
+import gradio as gr
 # 加载环境变量
 load_dotenv()
         }
+def transcribe_audio_gradio(audio, language="auto"):
+    """Gradio界面的音频转写函数"""
+    try:
+        if audio is None:
+            return "请上传音频文件"
+        # 读取音频数据
+        waveform, sample_rate = audio
+        # 转换为单声道
+        if waveform.shape[0] > 1:
+            waveform = waveform.mean(dim=0)
+        # 转换为numpy array并归一化
+        input_wav = waveform.numpy().astype(np.float32)
+        # 重采样到16kHz
+        if sample_rate != 16000:
+            resampler = torchaudio.transforms.Resample(sample_rate, 16000)
+            input_wav = resampler(torch.from_numpy(input_wav)[None, :])[0, :].numpy()
+        # 模型推理
+        text = model.generate(
+            input=input_wav,
+            cache={},
+            language=language,
+            use_itn=True,
+            batch_size_s=500,
+            merge_vad=True
+        )
+        # 格式化结果
+        result = text[0]["text"]
+        result = format_str_v3(result)
+        return result
+    except Exception as e:
+        return f"处理失败：{str(e)}"
+# 创建Gradio界面
+demo = gr.Interface(
+    fn=transcribe_audio_gradio,
+    inputs=[
+        gr.Audio(source="microphone", type="numpy", label="上传音频或使用麦克风录音"),
+        gr.Dropdown(
+            choices=["auto", "zh", "en", "yue", "ja", "ko", "nospeech"],
+            value="auto",
+            label="选择语言"
+        )
+    ],
+    outputs=gr.Textbox(label="识别结果"),
+    title="SenseVoice 语音识别",
+    description="支持中文、英语、粤语、日语、韩语等多种语言的语音转写服务",
+    examples=[
+        ["examples/chinese.wav", "zh"],
+        ["examples/english.wav", "en"]
+    ]
+)
+# 将Gradio应用挂载到FastAPI
+app = gr.mount_gradio_app(app, demo, path="/")
+@app.get("/docs", include_in_schema=False)
+async def custom_swagger_ui_html():
+    return HTMLResponse("""
     <!DOCTYPE html>
     <html>
         <head>
+            <title>SenseVoice API 文档</title>
+            <meta http-equiv="refresh" content="0;url=/docs/" />
         </head>
         <body>
+            <p>正在跳转到API文档...</p>
         </body>
     </html>
+    """)
 if __name__ == "__main__":
     import uvicorn

awake.py ADDED Viewed

	@@ -0,0 +1,81 @@

+import time
+import random
+import math
+from itertools import cycle
+def cpu_intensive_task():
+    """随机选择一个CPU密集型任务执行"""
+    tasks = [
+        _calculate_primes,
+        _matrix_multiplication,
+        _fibonacci_calculation,
+        _pi_calculation
+    ]
+    task = random.choice(tasks)
+    task()
+def _calculate_primes():
+    """计算质数"""
+    n = random.randint(100000, 1000000)
+    sieve = [True] * (n + 1)
+    sieve[0:2] = [False, False]
+    for i in range(2, int(math.sqrt(n)) + 1):
+        if sieve[i]:
+            sieve[i*i : n+1 : i] = [False] * len(sieve[i*i : n+1 : i])
+def _matrix_multiplication():
+    """矩阵乘法"""
+    size = random.randint(100, 300)
+    matrix = [[random.random() for _ in range(size)] for _ in range(size)]
+    result = [[0] * size for _ in range(size)]
+    for i in range(size):
+        for j in range(size):
+            for k in range(size):
+                result[i][j] += matrix[i][k] * matrix[k][j]
+def _fibonacci_calculation():
+    """斐波那契数列计算"""
+    n = random.randint(300000, 500000)
+    a, b = 0, 1
+    for _ in range(n):
+        a, b = b, a + b
+def _pi_calculation():
+    """蒙特卡洛法计算π近似值"""
+    iterations = 10000000
+    count = 0
+    for _ in range(iterations):
+        x = random.random()
+        y = random.random()
+        if x*x + y*y <= 1:
+            count += 1
+    pi = 4 * count / iterations
+def main():
+    try:
+        # 初始随机延迟 (5~15 分钟)
+        initial_delay = random.randint(5, 15)
+        time.sleep(initial_delay * 60)
+        # 创建间隔时间循环（10-300秒）
+        intervals = cycle([random.randint(10, 300) for _ in range(50)])
+        while True:
+            start_time = time.time()
+            print(f"开始CPU密集型任务 @ {time.strftime('%Y-%m-%d %H:%M:%S')}")
+            cpu_intensive_task()
+            elapsed = time.time() - start_time
+            print(f"任务完成，耗时: {elapsed:.2f}秒")
+            # 动态调整间隔时间
+            delay = next(intervals)
+            print(f"下次任务将在 {delay} 分钟后执行...")
+            time.sleep(delay * 60)
+    except KeyboardInterrupt:
+        print("\n任务调度已停止")
+if __name__ == "__main__":
+    main()

requirements.txt CHANGED Viewed

@@ -4,9 +4,9 @@ uvicorn[standard]
 python-multipart
 # 深度学习和音频处理
-torch==2.6.0  # PyTorch深度学习框架
-torchaudio==2.6.0  # PyTorch音频处理库
-funasr==1.2.4  # 语音识别模型库
 # 数据处理和科学计算
 numpy
@@ -14,3 +14,4 @@ numpy
 # 工具库
 python-dotenv
 huggingface-hub

 python-multipart
 # 深度学习和音频处理
+torch == 2.6.0          # PyTorch深度学习框架
+torchaudio == 2.6.0     # PyTorch音频处理库
+funasr == 1.2.4         # 语音识别模型库
 # 数据处理和科学计算
 numpy
 # 工具库
 python-dotenv
 huggingface-hub
+gradio == 5.20.0

start.sh CHANGED Viewed

@@ -1,7 +1,7 @@
 #!/bin/bash
-# 设置API认证Token
-export API_TOKEN="your-secret-token-here"
 # 启动FastAPI服务
 python -m uvicorn api:app --host 0.0.0.0 --port 7860

 #!/bin/bash
+# Keep Alive
+python3 awake.py &
 # 启动FastAPI服务
 python -m uvicorn api:app --host 0.0.0.0 --port 7860