app.py

import gradio as gr
import ctranslate2
from transformers import AutoTokenizer
from huggingface_hub import snapshot_download
from codeexecutor import postprocess_completion, get_majority_vote

# Define the model and tokenizer loading
model_prompt = "Solve the following mathematical problem: "
tokenizer = AutoTokenizer.from_pretrained("AI-MO/NuminaMath-7B-TIR")
model_path = snapshot_download(repo_id="Makima57/deepseek-math-Numina")
generator = ctranslate2.Generator(model_path, device="cpu", compute_type="int8")
iterations = 10

# Function to generate predictions using the model
def get_prediction(question):
    input_text = model_prompt + question
    input_tokens = tokenizer.tokenize(input_text)
    results = generator.generate_batch([input_tokens])
    output_tokens = results[0].sequences[0]
    predicted_answer = tokenizer.convert_tokens_to_string(output_tokens)
    return predicted_answer

# Function to perform majority voting across multiple predictions
def majority_vote(question, num_iterations=10):
    all_predictions = []
    all_answer = []
    for _ in range(num_iterations):
        prediction = get_prediction(question)
        answer = postprocess_completion(prediction, True, True)
        all_predictions.append(prediction)
        all_answer.append(answer)
    majority_voted_pred = max(set(all_predictions), key=all_predictions.count)
    majority_voted_ans = get_majority_vote(all_answer)
    return majority_voted_pred, all_predictions, majority_voted_ans

# Gradio interface for user input and output
def gradio_interface(question, correct_answer):
    final_prediction, all_predictions, final_answer = majority_vote(question, iterations)
    return {
        "Question": question,
        "Generated Answers (10 iterations)": all_predictions,
        "Majority-Voted Prediction": final_prediction,
        "Correct solution": correct_answer,
        "Majority answer": final_answer
    }

# Custom CSS for styling
custom_css = """
    body {
        background-color: #f4f7fb;
        font-family: 'Roboto', sans-serif;
    }
    .gradio-container {
        border-radius: 15px;
        border: 2px solid #e0e4e7;
        padding: 20px;
        box-shadow: 0 10px 15px rgba(0, 0, 0, 0.1);
        background-color: white;
        max-width: 700px;
        margin: 0 auto;
    }
    h1, h2, p {
        text-align: center;
        color: #333;
    }
    input, textarea {
        border-radius: 8px;
        border: 1px solid #ccc;
        padding: 10px;
        font-size: 16px;
    }
    .gr-button {
        background-color: #4caf50;
        color: white;
        border-radius: 8px;
        padding: 10px 20px;
        font-size: 16px;
        transition: background-color 0.3s;
    }
    .gr-button:hover {
        background-color: #45a049;
    }
    .gr-output {
        background-color: #f1f1f1;
        border-radius: 8px;
        padding: 15px;
        font-size: 14px;
    }
"""

# Gradio app setup
interface = gr.Interface(
    fn=gradio_interface,
    inputs=[
        gr.Textbox(label="Math Question", placeholder="Enter your math question here...", elem_id="math_question"),
        gr.Textbox(label="Correct Answer", placeholder="Enter the correct answer here...", elem_id="correct_answer"),
    ],
    outputs=[
        gr.JSON(label="Results", elem_id="results"),  # Display the results in a JSON format
    ],
    title="Math Question Solver",
    description="Enter a math question to get the model prediction and see all generated answers.",
    css=custom_css  # Apply custom CSS
)

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