Update main.py

from transformers import AutoTokenizer, AutoModelForCausalLM, BitsAndBytesConfig

main.py

@@ -2,6 +2,7 @@ from fastapi import FastAPI
 from pydantic import BaseModel
 from huggingface_hub import InferenceClient
 import uvicorn
+from transformers import AutoTokenizer, AutoModelForCausalLM, BitsAndBytesConfig
 
 
 app = FastAPI()