app.py

import gradio as gr
from transformers import AutoModelForCausalLM, AutoTokenizer

# Load the model and tokenizer
from transformers import AutoModel, AutoTokenizer

from transformers.adapters import AutoAdapterModel
from transformers import AutoTokenizer

model_name = "unsloth/Meta-Llama-3.1-8B-Instruct"


tokenizer = AutoTokenizer.from_pretrained(model_name)

# Load the base model with adapters
model = AutoAdapterModel.from_pretrained(model_name)

model.load_adapter("Braszczynski/Llama-3.2-3B-Instruct-bnb-4bit-460steps")



def respond(
    message,
    history: list[tuple[str, str]],
    system_message,
    max_tokens,
    temperature,
    top_p,
):
    # Combine system message and chat history
    chat_history = f"{system_message}\n"
    for user_msg, bot_reply in history:
        chat_history += f"User: {user_msg}\nAssistant: {bot_reply}\n"
    chat_history += f"User: {message}\nAssistant:"

    # Tokenize the input
    inputs = tokenizer(chat_history, return_tensors="pt", truncation=True).to("cuda")

    # Generate response
    outputs = model.generate(
        inputs["input_ids"],
        max_new_tokens=max_tokens,
        temperature=temperature,
        top_p=top_p,
        pad_token_id=tokenizer.eos_token_id
    )

    # Decode and format the output
    response = tokenizer.decode(outputs[0], skip_special_tokens=True)
    response = response[len(chat_history):].strip()  # Remove input context from output
    return response

# Define the Gradio interface
demo = gr.ChatInterface(
    respond,
    additional_inputs=[
        gr.Textbox(value="You are a friendly assistant.", 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)"),
    ],
)

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