new prompting style

app.py

@@ -21,9 +21,9 @@ def generate_text(message, history):
     input_prompt = "Conversazione tra umano ed un assistente AI di nome cerbero-7b\n"
     for interaction in history:
         input_prompt += "[|Umano|] " + interaction[0] + "\n"
-        input_prompt += "[|AI|]" + interaction[1]
+        input_prompt += "[|Assistente|]" + interaction[1]
     
-    input_prompt += "[|Umano|] " + message + "\n[|AI|]"
+    input_prompt += "[|Umano|] " + message + "\n[|Assistente|]"
 
     print(input_prompt)
 
@@ -36,8 +36,7 @@ def generate_text(message, history):
         max_tokens=1024,
         stop=[
             "[|Umano|]",
-            "[|Human|]",
-            "[|AI|]",
+            "[|Assistente|]",
         ],
         stream=True,
     )
@@ -52,7 +51,7 @@ def generate_text(message, history):
 demo = gr.ChatInterface(
     generate_text,
     title="cerbero-7b running on CPU (quantized Q4_K)",
-    description="This is a quantized version of cerbero-7b running on CPU. It is less powerful than the original version, but it is much faster and it can even run on a Raspberry Pi 4.",
+    description="This is a quantized version of cerbero-7b running on CPU (very slow). It is less powerful than the original version, but it can even run on a Raspberry Pi 4.",
     examples=[
         "Dammi 3 idee di ricette che posso fare con i pistacchi",
         "Prepara un piano di esercizi da poter fare a casa",