Lazy-import numpy/onnxruntime/transformers: fix slow startup (Starting 고착)

app.py

@@ -12,18 +12,19 @@ import time
 from pathlib import Path
 from typing import Optional
 
-import numpy as np
-import onnxruntime as ort
 from fastapi import FastAPI, HTTPException
 from fastapi.middleware.cors import CORSMiddleware
 from fastapi.responses import FileResponse, StreamingResponse
 from fastapi.staticfiles import StaticFiles
-from huggingface_hub import snapshot_download
 from pydantic import BaseModel
-from transformers import AutoTokenizer
 
 MODEL_REPO = os.getenv("MODEL_REPO", "onnx-community/EXAONE-3.5-2.4B-Instruct")
-# EXAONE-3.5-2.4B 구조 (다른 모델 쓰려면 수정 필요)
+
+# 무거운 패키지는 generator.load() 안에서 지연 import
+# (uvicorn이 먼저 뜨고 나서 백그라운드 스레드가 로딩)
+np = None
+ort = None
+
 NUM_LAYERS = 30
 NUM_KV_HEADS = 8
 HEAD_DIM = 80
@@ -37,9 +38,6 @@ app.add_middleware(
 )
 
 
-# ============================================================
-# 모델 (gen.py의 Generator와 동일 로직)
-# ============================================================
 class Generator:
     def __init__(self):
         self.tokenizer = None
@@ -50,6 +48,14 @@ class Generator:
         self.status = "initializing"
 
     def load(self):
+        import numpy as _np
+        import onnxruntime as _ort
+        from huggingface_hub import snapshot_download
+        from transformers import AutoTokenizer
+        global np, ort
+        np = _np
+        ort = _ort
+
         try:
             self.status = "downloading model"
             self.model_dir = Path(snapshot_download(