Update app.py

app.py

@@ -13,25 +13,37 @@ MAX_INPUT_TOKEN_LENGTH = int(os.getenv("MAX_INPUT_TOKEN_LENGTH", "4096"))
 
 DESCRIPTION = """\
 # L-MChat
-This Space demonstrates [L-MChat](https://huggingface.co/collections/Artples/l-mchat-663265a8351231c428318a8f) by L-AI. <br> To select the Model that you want to use please go to the Adavanced Inputs, the Quality-Model (L-MChat-7b) is activated by default.
+
+This Space demonstrates **L-MChat**, a pair of chat-optimized language models:
+
+- **Fast-Model**: `Artples/L-MChat-Small`
+- **Quality-Model**: `Artples/L-MChat-7b`
+
+By default the **Quality-Model** is selected. You can switch to the Fast-Model if you want
+lower latency at the cost of quality.
+
+Use the *System prompt* field to steer the assistant’s behavior (for example:
+“Act as a helpful programming tutor”). The sliders allow you to configure the
+generation parameters.
 """
 
 if not torch.cuda.is_available():
-    DESCRIPTION += "\n<p>Running on CPU! This demo does not work on CPU.</p>"
+    DESCRIPTION += "\n\n<p>Running on CPU! This demo does not work on CPU.</p>"
 
 model_dict = {
     "Fast-Model": "Artples/L-MChat-Small",
-    "Quality-Model": "Artples/L-MChat-7b"
+    "Quality-Model": "Artples/L-MChat-7b",
 }
 
+
 @spaces.GPU(enable_queue=True, duration=90)
 def generate(
     message: str,
     chat_history: list[tuple[str, str]],
     system_prompt: str,
     model_choice: str,
-    max_new_tokens: int = 1024,
-    temperature: float = 0.1,
+    max_new_tokens: int = DEFAULT_MAX_NEW_TOKENS,
+    temperature: float = 0.6,
     top_p: float = 0.9,
     top_k: int = 50,
     repetition_penalty: float = 1.2,
@@ -41,45 +53,68 @@ def generate(
     tokenizer = AutoTokenizer.from_pretrained(model_id)
     tokenizer.use_default_system_prompt = False
 
-    conversation = []
+    conversation: list[dict] = []
     if system_prompt:
         conversation.append({"role": "system", "content": system_prompt})
+
     for user, assistant in chat_history:
-        conversation.extend([{"role": "user", "content": user}, {"role": "assistant", "content": assistant}])
+        conversation.append({"role": "user", "content": user})
+        if assistant is not None:
+            conversation.append({"role": "assistant", "content": assistant})
+
     conversation.append({"role": "user", "content": message})
 
-    input_ids = tokenizer.apply_chat_template(conversation, return_tensors="pt", add_generation_prompt=True)
+    input_ids = tokenizer.apply_chat_template(
+        conversation,
+        return_tensors="pt",
+        add_generation_prompt=True,
+    )
+
     if input_ids.shape[1] > MAX_INPUT_TOKEN_LENGTH:
         input_ids = input_ids[:, -MAX_INPUT_TOKEN_LENGTH:]
-        gr.Warning(f"Trimmed input from conversation as it was longer than {MAX_INPUT_TOKEN_LENGTH} tokens.")
+        gr.Warning(
+            f"Trimmed input from conversation as it was longer than "
+            f"{MAX_INPUT_TOKEN_LENGTH} tokens."
+        )
+
     input_ids = input_ids.to(model.device)
 
-    streamer = TextIteratorStreamer(tokenizer, timeout=10.0, skip_prompt=True, skip_special_tokens=True)
+    streamer = TextIteratorStreamer(
+        tokenizer,
+        timeout=10.0,
+        skip_prompt=True,
+        skip_special_tokens=True,
+    )
+
     generate_kwargs = dict(
-        {"input_ids": input_ids},
+        input_ids=input_ids,
         streamer=streamer,
-        max_new_tokens=max_new_tokens,
+        max_new_tokens=min(max_new_tokens, MAX_MAX_NEW_TOKENS),
         do_sample=True,
+        temperature=temperature,
         top_p=top_p,
         top_k=top_k,
-        temperature=temperature,
-        num_beams=1,
         repetition_penalty=repetition_penalty,
     )
-    t = Thread(target=model.generate, kwargs=generate_kwargs)
-    t.start()
 
-    outputs = []
+    thread = Thread(target=model.generate, kwargs=generate_kwargs)
+    thread.start()
+
+    outputs: list[str] = []
     for text in streamer:
         outputs.append(text)
         yield "".join(outputs)
 
+
 chat_interface = gr.ChatInterface(
-    theme='ehristoforu/RE_Theme',
     fn=generate,
     additional_inputs=[
         gr.Textbox(label="System prompt", lines=6),
-        gr.Radio(["Fast-Model", "Quality-Model"], label="Model", value="Quality-Model"),
+        gr.Radio(
+            ["Fast-Model", "Quality-Model"],
+            label="Model",
+            value="Quality-Model",
+        ),
         gr.Slider(
             label="Max new tokens",
             minimum=1,
@@ -126,7 +161,7 @@ chat_interface = gr.ChatInterface(
     ],
 )
 
-with gr.Blocks(css="style.css") as demo:
+with gr.Blocks(css="styles.css") as demo:
     gr.Markdown(DESCRIPTION)
     chat_interface.render()