try to avoid tokenising inputs multiple times

app.py

@@ -7,19 +7,18 @@ import torch
 import gradio as gr
 
 
-def chat(message, history, bot_input_ids):
+def chat(message, history):
     history = history or []
-    bot_input_ids = bot_input_ids or []
     new_user_input_ids = tokenizer.encode(message+tokenizer.eos_token, return_tensors='pt')
     # append the new user input tokens to the chat history
+    bot_input_ids = [item[2], item[3] for item in history]
     bot_input_ids = torch.cat([chat_history_ids, new_user_input_ids], dim=-1) if chat_history_ids is not None else new_user_input_ids
-
     # generated a response while limiting the total chat history to 1000 tokens, 
     chat_history_ids = model.generate(bot_input_ids, max_length=5000, pad_token_id=tokenizer.eos_token_id)
     print("The text is ", [text])
     # pretty print last ouput tokens from bot
     reponse =  tokenizer.decode(chat_history_ids[:, bot_input_ids.shape[-1]:][0], skip_special_tokens=True)
-    history.append((message, response))
+    history.append((message, response, new_user_input_ids, chat_history_ids))
     return history, bot_input_ids, feedback(message)
 
 
@@ -38,8 +37,8 @@ def feedback(text):
 
 iface = gr.Interface(
     chat,
-    ["text", "state", "state"],
-    ["chatbot", "state", "state", "text"],
+    ["text", "state"],
+    ["chatbot", "state", "text"],
     allow_screenshot=False,
     allow_flagging="never",
 )