Update app.py

app.py

@@ -1,6 +1,6 @@
 import torch
-import gradio as gr
 from transformers import AutoModelForCausalLM, AutoTokenizer
+import gradio as gr
 
 model_id = "cody82/unitrip"
 
@@ -11,28 +11,37 @@ model.to(device)
 
 system_message = "Ты — умный помощник по Университету Иннополис."
 
-def respond(user_input, history=None):
+def respond(user_message, history):
     if history is None:
         history = []
-
-    prompt = f"{system_message}\nUser: {user_input}\nAssistant:"
+    # Формируем полный контекст (если нужно)
+    prompt = system_message + "\n"
+    for user_text, bot_text in history:
+        prompt += f"User: {user_text}\nAssistant: {bot_text}\n"
+    prompt += f"User: {user_message}\nAssistant:"
 
     inputs = tokenizer(prompt, return_tensors="pt").to(device)
 
     with torch.no_grad():
         outputs = model.generate(
             **inputs,
-            max_new_tokens=128,
-            do_sample=False,
+            max_new_tokens=150,
             pad_token_id=tokenizer.eos_token_id,
             eos_token_id=tokenizer.eos_token_id,
-            use_cache=True,
+            do_sample=False,
         )
 
-    generated = outputs[0][inputs["input_ids"].shape[1]:]
-    answer = tokenizer.decode(generated, skip_special_tokens=True)
-    history.append((user_input, answer.strip()))
+    generated_text = tokenizer.decode(outputs[0][inputs["input_ids"].shape[1]:], skip_special_tokens=True).strip()
+
+    history.append((user_message, generated_text))
     return history, history
 
-iface = gr.ChatInterface(fn=respond, title="Innopolis Q&A")
-iface.launch()
+with gr.Blocks() as demo:
+    chatbot = gr.Chatbot()
+    message = gr.Textbox(placeholder="Введите вопрос...")
+    state = gr.State([])  # История сообщений
+
+    message.submit(respond, inputs=[message, state], outputs=[chatbot, state])
+    message.submit(lambda: "", None, message)  # Очистить поле ввода после отправки
+
+demo.launch(share=True)