improvements

app.py

@@ -16,13 +16,12 @@ subprocess.run(
 
 
 DESCRIPTION = """\
-# Gemma 2 9B IT
+# Phi 3.5 mini ITA 💬 🇮🇹
 
-Gemma 2 is Google's latest iteration of open LLMs.
-This is a demo of [`google/gemma-2-9b-it`](https://huggingface.co/google/gemma-2-9b-it), fine-tuned for instruction following.
-For more details, please check [our post](https://huggingface.co/blog/gemma2).
+Fine-tuned version of Microsoft/Phi-3.5-mini-instruct to improve the performance on the Italian language.
+Small (3.82 B parameters) but capable model, with 128k context length.
 
-👉 Looking for a larger and more powerful version? Try the 27B version in [HuggingChat](https://huggingface.co/chat/models/google/gemma-2-27b-it).
+For more details, check out the [model card](https://huggingface.co/anakin87/Phi-3.5-mini-ITA).
 """
 
 MAX_MAX_NEW_TOKENS = 2048
@@ -47,13 +46,14 @@ model.eval()
 def generate(
     message: str,
     chat_history: list[tuple[str, str]],
+    system_message: str = "",
     max_new_tokens: int = 1024,
     temperature: float = 0.001,
     top_p: float = 1.0,
     top_k: int = 50,
     repetition_penalty: float = 1.0,
 ) -> Iterator[str]:
-    conversation = []
+    conversation = [{"role": "system", "content": system_message}]
     for user, assistant in chat_history:
         conversation.extend(
             [
@@ -93,6 +93,11 @@ def generate(
 chat_interface = gr.ChatInterface(
     fn=generate,
     additional_inputs=[
+        gr.Textbox(
+            value="",
+            label="System message",
+            render=False,
+        ),        
         gr.Slider(
             label="Max new tokens",
             minimum=1,
@@ -138,7 +143,6 @@ chat_interface = gr.ChatInterface(
         ["Fammi un elenco puntato dei pro e contro di vivere in Italia. Massimo 2 pro e 2 contro."],
         ["Inventa una breve storia con animali sul valore dell'amicizia."],
         ["Scrivi un articolo di 100 parole sui 'Benefici dell'open-source nella ricerca sull'intelligenza artificiale'"],
-        ["Hello there! How are you doing?"],
         ["Can you explain briefly to me what is the Python programming language?"],
         ["How many hours does it take a man to eat a Helicopter?"],
         ["Write a 100-word article on 'Benefits of Open-Source in AI research'"],
@@ -146,7 +150,7 @@ chat_interface = gr.ChatInterface(
     cache_examples=False,
 )
 
-with gr.Blocks(css="style.css", fill_height=True) as demo:
+with gr.Blocks(css="style.css", fill_height=True, theme="soft") as demo:
     gr.Markdown(DESCRIPTION)
     gr.DuplicateButton(value="Duplicate Space for private use", elem_id="duplicate-button")
     chat_interface.render()

style.css

@@ -0,0 +1,11 @@
+h1 {
+  text-align: center;
+  display: block;
+}
+
+#duplicate-button {
+  margin: auto;
+  color: #fff;
+  background: #1565c0;
+  border-radius: 100vh;
+}