Update app.py

app.py

@@ -3,8 +3,8 @@ import logging
 import threading
 import boto3
 import torch
-from transformers import AutoTokenizer, AutoModelForCausalLM, GenerationConfig, StoppingCriteriaList
-from fastapi import FastAPI, HTTPException
+from transformers import AutoTokenizer, AutoModelForCausalLM, GenerationConfig, StoppingCriteriaList, AutoConfig
+from fastapi import FastAPI, HTTPException, Request
 from pydantic import BaseModel, field_validator
 from io import BytesIO
 from huggingface_hub import hf_hub_download
@@ -99,7 +99,7 @@ class S3ModelLoader:
     def download_model_from_huggingface(self, model_name):
         try:
             logging.info(f"Downloading model {model_name} from Hugging Face...")
-            model_dir = hf_hub_download(model_name, token=HUGGINGFACE_HUB_TOKEN)
+            model_dir = hf_hub_download(model_name, token=HUGGINGFACE_HUB_TOKEN, filename=model_name.split("/")[-1])
             self.s3_client.upload_file(model_dir, self.bucket_name, f"lilmeaty_garca/{model_name}")
             logging.info(f"Model {model_name} saved to S3 successfully.")
         except Exception as e: