submit

Dockerfile

@@ -1,14 +1,22 @@
 FROM python:3.10-slim
-# Trigger rebuild 2025-03-14-16:57
+
+# 设置工作目录
 WORKDIR /code
 
+# 安装系统依赖
+RUN apt-get update && apt-get install -y --no-install-recommends libsndfile1 && \
+    rm -rf /var/lib/apt/lists/*
+
+# 复制依赖文件并安装 Python 依赖
 COPY requirements.txt .
-RUN apt-get update && apt-get install -y libsndfile1 && \
-    pip install --no-cache-dir -r requirements.txt
+RUN pip install --no-cache-dir -r requirements.txt
 
-# ✅ 把相关目录和文件都复制进去
+# 复制应用代码和模型
 COPY app.py .
 COPY ct2_model/ ct2_model/
-COPY whisper_processor/ whisper_processor/
 
-CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "7860"]
+# 暴露端口
+EXPOSE 7860
+
+# 启动 FastAPI 应用
+CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "7860"]

app.py

@@ -1,18 +1,18 @@
-
 from fastapi import FastAPI, UploadFile, File
 import soundfile as sf
-import ctranslate2
-from transformers import WhisperProcessor
+from faster_whisper import WhisperModel
 import os
+import tempfile
+
+# 设置 Hugging Face 缓存目录
 os.environ["HF_HOME"] = "/tmp/hf_cache"
 
 # 初始化 FastAPI 应用
 app = FastAPI()
 
-# 加载 Whisper 处理器和 CTranslate2 模型
+# 加载 Faster-Whisper 模型
 print("✅ THIS IS NEW APP.PY VERSION")
-processor = WhisperProcessor.from_pretrained("./whisper_processor")
-model = ctranslate2.Whisper("ct2_model", compute_type="int8", device="cpu")
+model = WhisperModel("ct2_model", compute_type="int8", device="cpu")
 
 @app.get("/")
 def root():
@@ -23,19 +23,17 @@ def root():
 
 @app.post("/transcribe")
 async def transcribe(file: UploadFile = File(...)):
-    # 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}
+    # 保存上传音频到临时文件
+    with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as temp_file:
+        temp_path = temp_file.name
+        temp_file.write(await file.read())
+
+    try:
+        # 使用 Faster-Whisper 进行推理
+        segments, info = model.transcribe(temp_path)
+        transcription = " ".join([segment.text for segment in segments])
+
+        return {"text": transcription}
+    finally:
+        # 删除临时文件
+        os.remove(temp_path)

requirements.txt

@@ -1,5 +1,4 @@
 fastapi
 uvicorn
-ctranslate2==3.20.0
-transformers
-soundfile
+soundfile
+faster-whisper