This repository has been archived on 2025-08-20. You can view files and clone it, but cannot push or open issues or pull requests.
Files
enginex-bi_series-asr-backup/bi_v100-sensevoice/server_sensevoice.py

42 lines
1.1 KiB
Python

from fastapi import FastAPI, Request
from contextlib import asynccontextmanager
from funasr import AutoModel
from funasr.utils.postprocess_utils import rich_transcription_postprocess
import os
import torch
model_dir = os.getenv("MODEL_DIR", "/model/iic/SenseVoiceSmall")
model = None
@asynccontextmanager
async def lifespan(app: FastAPI):
global model
if model is None:
device = "cuda" if torch.cuda.is_available() else "cpu"
model = AutoModel(model=model_dir, disable_update=True, device=device)
yield
pass
app = FastAPI(lifespan=lifespan)
@app.post("/recognition")
async def asr(request: Request, language: str = "auto"):
audio_data = await request.body()
res = model.generate(input=audio_data, use_itn=True, ban_emo_unk=True)
text = rich_transcription_postprocess(res[0]["text"])
return {
"RecognitionStatus": "Success",
"DisplayText": text
}
@app.get("/health")
@app.get("/ready")
def ready():
return {"status": "ok"}
if __name__ == "__main__":
import uvicorn
uvicorn.run(app, host="0.0.0.0", port=80)