app.py

from transformers import AutoModelForCausalLM, AutoTokenizer
import gradio as gr

# Load pre-trained model and tokenizer
model_name = "mistralai/Mistral-7B-Instruct-v0.2"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name)

def generate_response(messages):
    """
    Generate response based on the given user messages.
    Parameters:
    - messages (list): A list of dictionaries containing user messages with roles.
    Returns:
    - response (str): The generated response.
    """
    # Apply chat template and encode messages
    encodeds = tokenizer.apply_chat_template(messages, return_tensors="pt")
    # Move inputs to device
    model_inputs = encodeds.to("cuda")  # Assuming CUDA device is available
    # Generate response
    generated_ids = model.generate(model_inputs, max_new_tokens=1000, do_sample=True)
    # Decode the generated response
    response = tokenizer.batch_decode(generated_ids)[0]
    return response

# Define Gradio interface components
input_chat = gr.Textbox(lines=5, label="Input Chat", placeholder="Enter chat messages...")
output_response = gr.Textbox(label="Generated Response", placeholder="Generated response will appear here...")

# Create Gradio interface
gr.Interface(generate_response, input_chat, output_response,
             title="Chat Response Generation",
             description="Generate responses based on user messages using Mistral AI model.",
             theme="default",
             allow_flagging="never").launch()