app.py

import gradio as gr
from unsloth import FastLanguageModel
from transformers import TextStreamer
import torch

# Load the model and tokenizer locally
max_seq_length = 2048
dtype = None
model_name_or_path = "michailroussos/model_llama_8d"

# Load model and tokenizer using unsloth
model, tokenizer = FastLanguageModel.from_pretrained(
    model_name=model_name_or_path,
    max_seq_length=max_seq_length,
    dtype=dtype,
    load_in_4bit=True,
)
FastLanguageModel.for_inference(model)  # Enable optimized inference

# Define the response function
def respond(message, history, system_message, max_tokens, temperature, top_p):
    # Combine system message and conversation history
    messages = [{"role": "system", "content": system_message}]
    for user_msg, assistant_msg in history:
        if user_msg:
            messages.append({"role": "user", "content": user_msg})
        if assistant_msg:
            messages.append({"role": "assistant", "content": assistant_msg})
    messages.append({"role": "user", "content": message})
    
    # Tokenize inputs
    inputs = tokenizer.apply_chat_template(
        messages,
        tokenize=True,
        add_generation_prompt=True,
        return_tensors="pt",
    ).to("cuda" if torch.cuda.is_available() else "cpu")

    # Use TextStreamer to process and yield outputs incrementally
    class GradioStreamer(TextStreamer):
        def __init__(self, tokenizer, *args, **kwargs):
            super().__init__(tokenizer, *args, **kwargs)
            self.generated_text = ""

        def on_token(self, token_id):
            token = self.tokenizer.decode(token_id, skip_special_tokens=True)
            self.generated_text += token
            yield self.generated_text

    # Initialize Gradio-compatible streamer
    streamer = GradioStreamer(tokenizer, skip_prompt=True)
    
    # Generate response with streaming
    _ = model.generate(
        input_ids=inputs,
        max_new_tokens=max_tokens,
        use_cache=True,
        temperature=temperature,
        top_p=top_p,
        streamer=streamer,
    )



# Define the Gradio interface
demo = gr.ChatInterface(
    respond,
    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)"),
    ],
    type="messages"
)

if __name__ == "__main__":
    demo.launch(share=True)