Update app.py

app.py

@@ -1,70 +1,75 @@
+import spaces
 import gradio as gr
-from huggingface_hub import InferenceClient
-
-
-def respond(
-    message,
-    history: list[dict[str, str]],
-    system_message,
-    max_tokens,
-    temperature,
-    top_p,
-    hf_token: gr.OAuthToken,
-):
-    """
-    For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
-    """
-    client = InferenceClient(token=hf_token.token, model="openai/gpt-oss-20b")
-
-    messages = [{"role": "system", "content": system_message}]
-
-    messages.extend(history)
+import torch
+from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
+from threading import Thread
 
+@spaces.GPU
+def predict(message, history):
+    torch.set_default_device("cuda")
+    
+    # Load model and tokenizer
+    model_id = "kurakurai/Luth-LFM2-1.2B"
+    tokenizer = AutoTokenizer.from_pretrained(model_id)
+    model = AutoModelForCausalLM.from_pretrained(
+        model_id,
+        device_map="auto",
+        torch_dtype=torch.bfloat16,
+        trust_remote_code=True,
+        load_in_4bit=True,  # Keeping 4-bit quantization for efficiency
+        # attn_implementation="flash_attention_2"  # Uncomment on compatible GPU
+    )
+    
+    # Format conversation history for chat template
+    messages = [{"role": "user" if i % 2 == 0 else "assistant", "content": msg} 
+                for conv in history for i, msg in enumerate(conv) if msg]
     messages.append({"role": "user", "content": message})
-
-    response = ""
-
-    for message in client.chat_completion(
+    
+    # Apply chat template
+    input_ids = tokenizer.apply_chat_template(
         messages,
-        max_tokens=max_tokens,
-        stream=True,
-        temperature=temperature,
-        top_p=top_p,
-    ):
-        choices = message.choices
-        token = ""
-        if len(choices) and choices[0].delta.content:
-            token = choices[0].delta.content
+        add_generation_prompt=True,
+        return_tensors="pt",
+        tokenize=True
+    ).to('cuda')
+    
+    # Setup streamer for real-time output
+    streamer = TextIteratorStreamer(tokenizer, timeout=10., skip_prompt=True, skip_special_tokens=True)
+    
+    # Generation parameters
+    generate_kwargs = dict(
+        input_ids=input_ids,
+        streamer=streamer,
+        max_new_tokens=256,
+        do_sample=True,
+        temperature=0.3,
+        min_p=0.15,
+        repetition_penalty=1.05,
+        pad_token_id=tokenizer.eos_token_id
+    )
+    
+    # Start generation in separate thread
+    t = Thread(target=model.generate, kwargs=generate_kwargs)
+    t.start()
+    
+    # Stream tokens
+    partial_message = ""
+    for new_token in streamer:
+        partial_message += new_token
+        yield partial_message
 
-        response += token
-        yield response
-
-
-"""
-For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
-"""
-chatbot = gr.ChatInterface(
-    respond,
-    type="messages",
-    additional_inputs=[
-        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
-        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
-        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
-        gr.Slider(
-            minimum=0.1,
-            maximum=1.0,
-            value=0.95,
-            step=0.05,
-            label="Top-p (nucleus sampling)",
-        ),
+# Setup Gradio interface
+gr.ChatInterface(
+    predict,
+    description="""
+    <center><h2>Kurakura AI Luth-LFM2-1.2B Chat</h2></center>
+    
+    Chat with [Luth-LFM2-1.2B](https://huggingface.co/kurakurai/Luth-LFM2-1.2B), a French-tuned version of LFM2-1.2B.
+    """,
+    examples=[
+    "Peux-tu résoudre l'équation 3x - 7 = 11 pour x ?",
+    "Explique la photosynthèse en termes simples.",
+    "Écris un petit poème sur l'intelligence artificielle."
     ],
-)
-
-with gr.Blocks() as demo:
-    with gr.Sidebar():
-        gr.LoginButton()
-    chatbot.render()
-
-
-if __name__ == "__main__":
-    demo.launch()
+    theme=gr.themes.Soft(primary_hue="blue"),
+).launch()