app.py

import gradio as gr
from transformers import AutoTokenizer, AutoModelForCausalLM
from peft import PeftModel

# Load the base model and tokenizer
base_model_name = "google/gemma-2-2b-it" 
tokenizer = AutoTokenizer.from_pretrained(base_model_name)
model = AutoModelForCausalLM.from_pretrained(base_model_name)

# Load the adapter configuration
adapter_name = "hemhemoh/Gemma-2-2b-it-wazobia-bot"
model = PeftModel.from_pretrained(model, adapter_name)
from transformers import (AutoModelForCausalLM, AutoTokenizer)


def respond(
    message,
    history: list[tuple[str, str]],
    system_message,
    max_tokens,
    temperature,
    top_p,
):
    # Construct messages using the system prompt and conversation history
    messages = [{"role": "system", "content": system_message}]

    for val in history:
        if val[0]:
            messages.append({"role": "user", "content": val[0]})
        if val[1]:
            messages.append({"role": "assistant", "content": val[1]})

    # Add the latest user message
    messages.append({"role": "user", "content": message})

    # Convert the conversation into the appropriate input format
    prompt = tokenizer.apply_chat_template(
        messages, tokenize=False, add_generation_prompt=True
    )
    inputs = tokenizer(
        prompt, return_tensors="pt", padding=True, truncation=True
    ).to("cuda")  # Adjust device as necessary

    # Generate response using your local model
    outputs = model.generate(
        **inputs,
        max_length=max_tokens,  # Use max_tokens slider value
        num_return_sequences=1,
        top_k=50,
        top_p=top_p,  # Use top_p slider value
        temperature=temperature,  # Use temperature slider value
        no_repeat_ngram_size=3,
    )

    # Decode and clean up the output
    text = tokenizer.decode(outputs[0], skip_special_tokens=True)
    response = text.split("model")[1].strip()  # Adjust if "model" split is unnecessary
    yield response


# Gradio ChatInterface with additional inputs
demo = gr.ChatInterface(
    respond,
    additional_inputs=[
        gr.Textbox(
            value="You are a highly skilled and empathetic mental health therapist fluent in English, Yoruba, Igbo, and Hausa. Respond to each user's concerns in the language they use to ensure comfort and understanding.",
            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()