Update app.py

app.py

@@ -1,6 +1,7 @@
 import gradio as gr
 import torch
 from transformers import AutoModel, AutoTokenizer
+import deepspeed
 
 # Model name
 model_name = "OpenGVLab/InternVideo2_5_Chat_8B"
@@ -13,7 +14,8 @@ model = AutoModel.from_pretrained(
     model_name, 
     trust_remote_code=True,
     torch_dtype=torch.float16,  # Use float16 for lower memory usage
-    device_map="auto"           # Automatically place model on available GPU
+    device_map="auto", # Automatically place model on available GPU
+    deepspeed={"stage": 3}
 )
 
 # Define inference function