Update app.py

app.py

@@ -1,6 +1,7 @@
 import os
 import shutil
 import tempfile
+import torch
 import spaces
 import gradio as gr
 from fastapi import FastAPI, UploadFile, File, Form, HTTPException
@@ -10,17 +11,18 @@ from transformers import pipeline
 # 1. 声明加载的模型
 MODEL_NAME = "openai/whisper-small"
 
-# 2. 全局初始化 Pipeline
+# 2. 初始化 Pipeline
+device = "cuda" if torch.cuda.is_available() else "cpu"
 pipe = pipeline(
     "automatic-speech-recognition",
     model=MODEL_NAME,
     chunk_length_s=30,
-    device="cuda"
+    device=device
 )
 
-# 3. 核心计算函数
+# 3. 【关键点】在最外层定义带 @spaces.GPU 的核心推理函数
 @spaces.GPU
-def transcribe_core(audio_path: str, target_language: str = None, is_translate: bool = False):
+def run_whisper_inference(audio_path: str, target_language: str = None, is_translate: bool = False):
     generate_kwargs = {}
     if target_language:
         generate_kwargs["language"] = target_language
@@ -32,12 +34,13 @@ def transcribe_core(audio_path: str, target_language: str = None, is_translate:
     result = pipe(audio_path, generate_kwargs=generate_kwargs)
     return result["text"]
 
+
 # --- Gradio 界面 ---
 def gradio_predict(audio_path):
     if audio_path is None:
         return "请先上传音频或录音！"
     try:
-        return transcribe_core(audio_path)
+        return run_whisper_inference(audio_path)
     except Exception as e:
         return f"错误: {str(e)}"
 
@@ -49,22 +52,21 @@ demo = gr.Interface(
     description="【完美兼容 OpenAI 规范】"
 )
 
-# ===========================================================================
-# 关键修改部分：正确初始化 FastAPI 应用
-# ===========================================================================
+
+# --- FastAPI 接口配置 ---
 app = FastAPI()
 
-# 辅助函数：处理 OpenAI 请求
-async def process_openai_audio_request(file, response_format, language, is_translate):
-    suffix = os.path.splitext(file.filename)[1] or ".mp3"
+async def process_openai_audio_request(file: UploadFile, response_format: str, language: str, is_translate: bool):
+    suffix = os.path.splitext(file.filename)[1] or ".wav"
     with tempfile.NamedTemporaryFile(delete=False, suffix=suffix) as temp_file:
         shutil.copyfileobj(file.file, temp_file)
         temp_path = temp_file.name
 
     try:
-        text = transcribe_core(temp_path, target_language=language, is_translate=is_translate)
+        # 调用最外层带 @spaces.GPU 装饰的函数
+        text = run_whisper_inference(temp_path, target_language=language, is_translate=is_translate)
     except Exception as e:
-        raise HTTPException(status_code=500, detail=f"OpenAI Audio API failed: {str(e)}")
+        raise HTTPException(status_code=500, detail=f"Inference failed: {str(e)}")
     finally:
         if os.path.exists(temp_path):
             os.remove(temp_path)
@@ -75,7 +77,6 @@ async def process_openai_audio_request(file, response_format, language, is_trans
     return JSONResponse(content={"text": text})
 
 
-# 4. 完美兼容接口一：语音转录 (Transcriptions)
 @app.post("/v1/audio/transcriptions")
 async def transcribe_api(
     file: UploadFile = File(...),
@@ -93,7 +94,6 @@ async def transcribe_api(
     )
 
 
-# 5. 完美兼容接口二：语音翻译 (Translations)
 @app.post("/v1/audio/translations")
 async def translate_api(
     file: UploadFile = File(...),
@@ -109,5 +109,5 @@ async def translate_api(
         is_translate=True
     )
 
-# 6. 将 Gradio 挂载到 FastAPI 应用的根路径（必须放在所有 FastAPI 路由之后！）
+# 挂载 Gradio 页面到 FastAPI 根路径
 app = gr.mount_gradio_app(app, demo, path="/")