app.py

import gradio as gr
import random
from datasets import load_dataset


# # Sample dataset with unique 10-digit IDs
# qa_dataset = {
#     "1234567890": {
#         "question": "What is the capital of France?",
#         "choices": ["A. Berlin", "B. Madrid", "C. Paris", "D. Lisbon"],
#         "answer": "C. Paris"
#     },
#     "0987654321": {
#         "question": "What is the largest planet in our solar system?",
#         "choices": ["A. Earth", "B. Jupiter", "C. Saturn", "D. Mars", "E. Venus"],
#         "answer": "B. Jupiter"
#     },
#     # Add more questions with unique IDs as needed
# }

truth_data = load_dataset("commonsense-index-dev/commonsense-candidates", "iter6-0520-error", split="train")

qa_dataset = {}
for item in truth_data:
    qa_dataset[item["id"]] = {
        "question": item["task"],
        "choices": item["choices"],
        "answer": item["answer"]
    }

def get_random_question():
    question_id = random.choice(list(qa_dataset.keys()))
    question_data = qa_dataset[question_id]
    return question_id, question_data["question"], question_data["choices"]

def get_question_by_id(question_id):
    if question_id in qa_dataset:
        question_data = qa_dataset[question_id]
        return question_id, question_data["question"], question_data["choices"]
    else:
        return None, "Invalid question ID", []

def check_answer(question_id, choice):
    correct_answer = qa_dataset[question_id]["answer"]
    return "Correct!" if choice == correct_answer else f"Incorrect. The correct answer is {correct_answer}."

def load_question(question_id=None):
    if question_id:
        question_id, question, choices = get_question_by_id(question_id)
    else:
        question_id, question, choices = get_random_question()
    
    question = f"## {question}"
    choices_markdown = "\n".join(choices)
    return question_id, question, choices_markdown, gr.update(visible=True), gr.update(value="", visible=True)

def show_buttons(choices_markdown):
    choices = choices_markdown.split("\n")
    visibility = [gr.update(visible=False)] * 10
    for i in range(len(choices)):
        visibility[i] = gr.update(visible=True, value=choices[i])
    return visibility

with gr.Blocks() as app:
    gr.Markdown("# Multiple Choice QA Dataset Viewer")

    question_id_input = gr.Textbox(label="Enter Question ID", placeholder="leave empty for random sampling")
    random_button = gr.Button("Retrieve or Random Sample")
    question_display = gr.Markdown(visible=True)
    choices_markdown = gr.Markdown(visible=False)
    choice_buttons = [gr.Button(visible=False) for _ in range(10)]
    result_display = gr.Markdown(visible=True)

    question_id = gr.State()

    question_id_input.submit(fn=load_question, inputs=question_id_input, outputs=[question_id, question_display, choices_markdown, result_display])
    random_button.click(fn=load_question, outputs=[question_id, question_display, choices_markdown, result_display])
    choices_markdown.change(fn=show_buttons, inputs=choices_markdown, outputs=choice_buttons)

    for i, button in enumerate(choice_buttons):
        button.click(fn=check_answer, inputs=[question_id, button], outputs=result_display)

app.launch()