app.py

import gradio as gr
import spaces
from transformers import pipeline
import torch

# Global variable to store the pipeline
pipe = None

@spaces.GPU
def initialize_model():
    global pipe
    if pipe is None:
        pipe = pipeline(
            "text-generation", 
            model="apexion-ai/Orion-V1-4B",
            torch_dtype=torch.float16,
            device_map="auto"
        )
    return pipe

@spaces.GPU
def generate_response(message, history, max_length=512, temperature=0.7, top_p=0.9):
    """Generate response using the Orion model"""
    
    # Initialize model inside the GPU-decorated function
    model_pipe = initialize_model()
    
    # Format the conversation history
    messages = []
    
    # Add conversation history
    for user_msg, assistant_msg in history:
        messages.append({"role": "user", "content": user_msg})
        if assistant_msg:
            messages.append({"role": "assistant", "content": assistant_msg})
    
    # Add current message
    messages.append({"role": "user", "content": message})
    
    # Generate response
    try:
        response = model_pipe(
            messages,
            max_length=max_length,
            temperature=temperature,
            top_p=top_p,
            do_sample=True,
            pad_token_id=model_pipe.tokenizer.eos_token_id
        )
        
        # Extract the generated text
        generated_text = response[0]['generated_text']
        
        # Get the last assistant message
        if isinstance(generated_text, list):
            assistant_response = generated_text[-1]['content']
        else:
            # Fallback parsing if needed
            assistant_response = str(generated_text).split("assistant")[-1].strip()
        
        return assistant_response
        
    except Exception as e:
        return f"Error generating response: {str(e)}"

# Create the Gradio interface
def create_interface():
    with gr.Blocks(title="Orion-V1-4B Chat", theme=gr.themes.Soft()) as demo:
        gr.Markdown("""
        # 🚀 Orion-V1-4B Chat
        
        Chat with the Orion-V1-4B model by Apexion AI. This is a 4B parameter language model optimized for conversation.
        
        **Model:** `apexion-ai/Orion-V1-4B`
        """)
        
        chatbot = gr.Chatbot(
            height=400,
            placeholder="Start chatting with Orion-V1-4B...",
            label="Chat"
        )
        
        msg = gr.Textbox(
            placeholder="Type your message here...",
            label="Message",
            lines=2
        )
        
        with gr.Row():
            submit_btn = gr.Button("Send", variant="primary")
            clear_btn = gr.Button("Clear Chat", variant="secondary")
        
        with gr.Accordion("Advanced Settings", open=False):
            max_length = gr.Slider(
                minimum=50,
                maximum=2048,
                value=512,
                step=50,
                label="Max Length"
            )
            temperature = gr.Slider(
                minimum=0.1,
                maximum=2.0,
                value=0.7,
                step=0.1,
                label="Temperature"
            )
            top_p = gr.Slider(
                minimum=0.1,
                maximum=1.0,
                value=0.9,
                step=0.1,
                label="Top P"
            )
        
        # Event handlers
        def user_message(message, history):
            return "", history + [[message, None]]
        
        def bot_response(history, max_len, temp, top_p):
            if history:
                user_message = history[-1][0]
                bot_message = generate_response(
                    user_message, 
                    history[:-1], 
                    max_len, 
                    temp, 
                    top_p
                )
                history[-1][1] = bot_message
            return history
        
        # Wire up the events
        msg.submit(user_message, [msg, chatbot], [msg, chatbot]).then(
            bot_response, [chatbot, max_length, temperature, top_p], chatbot
        )
        
        submit_btn.click(user_message, [msg, chatbot], [msg, chatbot]).then(
            bot_response, [chatbot, max_length, temperature, top_p], chatbot
        )
        
        clear_btn.click(lambda: None, None, chatbot, queue=False)
        
        gr.Markdown("""
        ---
        
        ### About Orion-V1-4B
        
        Orion-V1-4B is a 4 billion parameter language model developed by Apexion AI. 
        It's designed for efficient text generation and conversation.
        
        **Features:**
        - 4B parameters for efficient inference
        - Optimized for conversational AI
        - Supports various text generation tasks
        
        This Space uses ZeroGPU for efficient GPU allocation.
        """)
    
    return demo

# Launch the app
if __name__ == "__main__":
    demo = create_interface()
    demo.launch()