Update app.py

app.py

@@ -16,13 +16,24 @@ try:
     tools = mcp_client.get_tools()
 
     # model = InferenceClientModel()
-    model = TransformersModel(
-    model_id="Qwen/Qwen2.5-Coder-32B-Instruct",
-    device="cuda",
-    max_new_tokens=5000,
+    # model = TransformersModel(
+    # model_id="Qwen/Qwen2.5-Coder-32B-Instruct",
+    # device="cuda",
+    # max_new_tokens=5000,
+    # )
+    model_id = "unsloth/Llama-3.2-1B"
+
+    model = AutoModelForCausalLM.from_pretrained(
+      model_id,
+      torch_dtype=torch.bfloat16,
+      device_map="auto"
     )
+
+
     agent = CodeAgent(tools=[*tools], model=model)
 
+
+
     # Define Gradio ChatInterface
     demo = gr.ChatInterface(
         fn=lambda message, history: str(agent.run(message)),