import os from fastapi import FastAPI from pydantic import BaseModel from transformers import AutoTokenizer, AutoModelForCausalLM # 🔐 캐시 경로 바꿔서 PermissionError 해결 os.environ["HF_HOME"] = "/data" os.environ["TRANSFORMERS_CACHE"] = "/data" app = FastAPI() tokenizer = AutoTokenizer.from_pretrained("skt/kogpt2-base-v2") model = AutoModelForCausalLM.from_pretrained("skt/kogpt2-base-v2") class InputText(BaseModel): data: list @app.post("/run/predict") async def generate(input_text: InputText): prompt = input_text.data[0] input_ids = tokenizer.encode(prompt, return_tensors="pt") output = model.generate(input_ids, max_new_tokens=50, pad_token_id=tokenizer.eos_token_id) result = tokenizer.decode(output[0], skip_special_tokens=True) reply = result.split("챗봇:")[-1].strip() return {"data": [reply]}