Update app.py

app.py

@@ -9,9 +9,9 @@ import os
 
 import subprocess
 subprocess.run('pip install flash-attn==2.7.4.post1 --no-build-isolation', env={'FLASH_ATTENTION_SKIP_CUDA_BUILD': "TRUE"}, shell=True)
-os.system("pip install transformers")
+os.system("pip install transformers==4.51.1")
 os.system("pip install numpy==1.26.4")
-os.system("pip install vllm")
+os.system("pip install vllm==0.8.2")
 import sys
 import threading
 import re
@@ -36,7 +36,7 @@ def parse_args():
     parser.add_argument(
         "--backend",
         type=str,
-        default="vllm",
+        default="transformers",
         choices=["transformers", "vllm"],
         help="Backend to use for inference",
     )