ct2

Dockerfile

@@ -3,8 +3,9 @@ FROM python:3.10-slim
 WORKDIR /code
 
 COPY requirements.txt .
-RUN pip install --no-cache-dir -r requirements.txt
+RUN apt-get update && apt-get install -y libsndfile1 && \
+    pip install --no-cache-dir -r requirements.txt
 
 COPY app.py .
 
-CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "7860"]
+CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "7860"]

app.py

@@ -1,39 +1,38 @@
-from faster_whisper import WhisperModel
-from fastapi import FastAPI, UploadFile, File
-import uvicorn
 import os
-os.environ["HF_HOME"] = "/tmp/hf_cache"  
+from fastapi import FastAPI, UploadFile, File
+import soundfile as sf
+import ctranslate2
+from transformers import WhisperProcessor
 
+# 初始化 FastAPI 应用
 app = FastAPI()
 
-# ✅ 加载模型（small模型 + CPU）
-model = WhisperModel("./ct2_model", device="cpu", compute_type="int8", local_files_only=True)
+# 加载 Whisper 处理器和 CTranslate2 模型
+processor = WhisperProcessor.from_pretrained("openai/whisper-small")
+model = ctranslate2.Whisper("ct2_model", compute_type="int8", device="cpu")
 
 @app.get("/")
 def root():
-    return {"message": "Whisper API is running."}
-
-
+    return {
+        "message": "CTranslate2 Whisper API is running.",
+        "usage": "POST /transcribe with an audio file (.wav, .mp3, etc.)"
+    }
 
 @app.post("/transcribe")
 async def transcribe(file: UploadFile = File(...)):
-    try:
-        temp_path = f"/tmp/{file.filename}"
-        with open(temp_path, "wb") as f:
-            f.write(await file.read())
-
-        segments, info = model.transcribe(temp_path, beam_size=1)
-
-        results = [
-            {"start": s.start, "end": s.end, "text": s.text}
-            for s in segments
-        ]
-
-        return {"language": info.language, "segments": results}
-
-    except Exception as e:
-        return {"error": str(e)}
-
-
-if __name__ == "__main__":
-    uvicorn.run(app, host="0.0.0.0", port=7860)
+    # 保存上传音频
+    temp_path = f"/tmp/{file.filename}"
+    with open(temp_path, "wb") as f:
+        f.write(await file.read())
+
+    # 加载音频并提取特征
+    audio_input, sample_rate = sf.read(temp_path)
+    inputs = processor(audio_input, sampling_rate=sample_rate, return_tensors="np")
+    features = inputs.input_features[0]
+
+    # 运行 CTranslate2 推理
+    results = model.generate(features)
+    tokens = results[0]["tokens"]
+    text = processor.decode(tokens, skip_special_tokens=True)
+
+    return {"text": text}

requirements.txt

@@ -1,3 +1,5 @@
-faster-whisper==0.9.0
 fastapi
-uvicorn
+uvicorn
+ctranslate2
+transformers
+soundfile