Update app.py

app.py

@@ -1,17 +1,15 @@
 import torch
 from transformers import AutoModelForCausalLM, AutoTokenizer
-import gradio as gr
 
+MODEL_ID = "rinna/bilingual-gpt-neox-4b-instruction-ppo"
 model = AutoModelForCausalLM.from_pretrained(
-    "rinna/bilingual-gpt-neox-4b-instruction-ppo",
-    use_auth_token="your_huggingface_token",
-    device_map="cpu"
+    MODEL_ID,
+    load_in_8bit=True,
+    device_map="auto"
 )
-
-MODEL_ID = "rinna/bilingual-gpt-neox-4b-instruction-ppo"
 tokenizer = AutoTokenizer.from_pretrained(MODEL_ID, use_fast=False)
 
-device = model.device
+import gradio as gr
 
 def generate_response(user_question,
              chat_history,
@@ -63,4 +61,17 @@ def generate_response(user_question,
     output = tokenizer.decode(tokens[0], skip_special_tokens=True)
     return output[len(prompt):]
 
+with gr.Blocks() as demo:
+    chat_history = gr.Chatbot()
+    user_message = gr.Textbox(label="Question:", placeholder="人工知能とは何ですか？")
+    clear = gr.ClearButton([user_message, chat_history])
+
+    def response(user_message, chat_history):
+        system_message = generate_response(user_message, chat_history)
+        chat_history.append((user_message, system_message))
+        return "", chat_history
+
+    user_message.submit(response, inputs=[user_message, chat_history], outputs=[user_message, chat_history])
 
+if __name__ == "__main__":
+    demo.launch()