Upload 4 files

app.py

@@ -0,0 +1,94 @@
+from funasr import AutoModel
+from funasr.utils.postprocess_utils import rich_transcription_postprocess
+from modelscope import snapshot_download
+
+import io
+import os
+import tempfile
+import json
+from typing import Optional
+
+import torch
+
+from fastapi import FastAPI, File, Form, UploadFile, HTTPException
+from fastapi.responses import StreamingResponse, Response
+
+from config import model_config
+
+device = "cuda:0" if torch.cuda.is_available() else "cpu"
+model_dir = snapshot_download(model_config['model_dir'])
+
+class SynthesizeResponse(Response):
+    media_type = 'text/plain'
+
+app = FastAPI()
+
+@app.post('/asr', response_class=SynthesizeResponse)
+async def generate(
+    file: UploadFile = File(...),
+    vad_model: str = Form("fsmn-vad"),
+    vad_kwargs: str = Form('{"max_single_segment_time": 30000}'),
+    ncpu: int = Form(4),
+    batch_size: int = Form(1),
+    language: str = Form("auto"),
+    use_itn: bool = Form(True),
+    batch_size_s: int = Form(60),
+    merge_vad: bool = Form(True),
+    merge_length_s: int = Form(15),
+    batch_size_threshold_s: int = Form(50),
+    hotword: Optional[str] = Form(" "),
+    spk_model: str = Form("cam++"),
+    ban_emo_unk: bool = Form(False),
+) -> StreamingResponse:
+    try:
+        # 将字符串转换为字典
+        vad_kwargs = json.loads(vad_kwargs)
+
+        # 创建临时文件并保存上传的音频文件
+        with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as temp_file:
+            temp_file_path = temp_file.name
+            input_wav_bytes = await file.read()
+            temp_file.write(input_wav_bytes)
+
+        try:
+            # 初始化模型
+            model = AutoModel(
+                model=model_dir,
+                trust_remote_code=False,
+                remote_code="./model.py",
+                vad_model=vad_model,
+                vad_kwargs=vad_kwargs,
+                ncpu=ncpu,
+                batch_size=batch_size,
+                hub="ms",
+                device=device,
+            )
+
+            # 生成结果
+            res = model.generate(
+                input=temp_file_path,  # 使用临时文件路径作为输入
+                cache={},
+                language=language,
+                use_itn=use_itn,
+                batch_size_s=batch_size_s,
+                merge_vad=merge_vad,
+                merge_length_s=merge_length_s,
+                batch_size_threshold_s=batch_size_threshold_s,
+                hotword=hotword,
+                spk_model=spk_model,
+                ban_emo_unk=ban_emo_unk
+            )
+
+            # 处理结果
+            text = rich_transcription_postprocess(res[0]["text"])
+            
+            # 返回结果
+            return StreamingResponse(io.BytesIO(text.encode('utf-8')), media_type="text/plain")
+        
+        finally:
+            # 确保在处理完毕后删除临时文件
+            if os.path.exists(temp_file_path):
+                os.remove(temp_file_path)
+    
+    except Exception as e:
+        raise HTTPException(status_code=500, detail=str(e))

config.py

@@ -0,0 +1,3 @@
+model_config = {
+	'model_dir': 'iic/SenseVoiceSmall'
+}

requirements.txt

@@ -0,0 +1,14 @@
+--extra-index-url https://download.pytorch.org/whl/cpu
+torch
+torchaudio
+funasr
+modelscope
+huggingface
+huggingface_hub
+uvicorn
+fastapi
+python-dotenv
+numpy
+gradio
+rotary_embedding_torch
+

run.py

@@ -0,0 +1,11 @@
+import uvicorn
+import os
+
+from app import app
+from dotenv import load_dotenv
+
+load_dotenv()
+port = int(os.getenv('PORT', 3151))
+
+if __name__ == '__main__':
+        uvicorn.run(app, host='0.0.0.0', port=port)