Update app.py

app.py

@@ -2,7 +2,7 @@ from huggingface_hub import InferenceClient
 
 import gradio as gr
 import random
-client=InferenceClient("HuggingFaceH4/zephyr-7b-beta")
+client=InferenceClient("Qwen/Qwen2.5-72B-Instruct")
 
 def respond(message, history):
     #responses = ["Yes", "No"]
@@ -17,7 +17,8 @@ def respond(message, history):
     if history:
         messages.extend(history)
     
-    response = client.chat_completion(messages, max_tokens=100)
+    response = client.chat_completion(messages, max_tokens=100, temperature=1.3, top_p=0.2)
+    # temperature and top_p control randomness
     
     return response["choices"][0]["message"]["content"].strip()