app.py

import gradio as gr
import json
import os
import random
import datasets #load_dataset, save_to_disk, load_from_disk

HF_TOKEN = os.getenv('HF_DIPROMATS2024_T2_LEADERBOARD-TOKEN')

# Use the Hugging Face dataset
DATASET_NAME = "NLP-UNED/dipromats2024-t2_leaderboard-data"
SPLIT = 'results'

try:
    dataset = datasets.load_dataset(DATASET_NAME, token=HF_TOKEN)

except Exception as e:
    print(f"Error loading dataset: {e}")
    dataset = datasets.Dataset.from_dict({"email": [], "team_name": [], "run_id": [], "description": [], "lenient_f1": [], "strict_f1": [], "average_f1": []})
    dataset = datasets.DatasetDict({SPLIT: dataset})

# Función para convertir el dataset en tabla
def data_to_table():
    global dataset
    table_data = []
    for item in dataset[SPLIT]:
        table_data.append([item.get("team_name", ""), item.get("run_id", ""),
                           item.get("lenient_f1", ""), item.get("strict_f1", ""), item.get("average_f1", "")])
    return table_data


# Función para subir los resultados al leaderboard
def update_leaderboard(email, team_input, run_id, description, lenient_f1, strict_f1, average_f1):
    global datataset
    new_data = dataset[SPLIT].add_item({
        "email": email,
        "team_name": team_input,
        "run_id": run_id,
        "description": description,
        "lenient_f1": lenient_f1,
        "strict_f1": strict_f1,
        "average_f1": average_f1
    })
    dataset[SPLIT] = new_data
    new_data.push_to_hub(DATASET_NAME, split=SPLIT, token=HF_TOKEN)
    return data_to_table(), gr.Tabs(selected=0), gr.Button(visible=True), gr.Button(visible=False), "", "", "", "", None, None, None, None


# Función para evaluar los resultados
def evaluate_results(file_path):
    lenient_f1 = random.random()
    strict_f1 = random.random()
    average_f1 = (lenient_f1 + strict_f1) / 2
    return lenient_f1, strict_f1, average_f1


# Función para procesar el archivo de resultados
def process_file(file_path, team_input, run_id, description, email):
    warn = False
    if not file_path:
        gr.Warning("File cannot be blank")
        warn=True
    if not team_input:
        gr.Warning("Team name cannot be blank")
        warn=True
    if not run_id:
        gr.Warning("Run ID cannot be blank")
        warn=True
    if not file_path:
        gr.Warning("File cannot be blank")
        warn=True
    if not description:
        gr.Warning("Description cannot be blank")
        warn=True
    if not email:
        gr.Warning("Email cannot be blank")
        warn=True

    if warn:
        return gr.Button(visible=True), gr.Button(visible=False), None, None, None

    lenient_f1, strict_f1, average_f1 = evaluate_results(file_path)

    return gr.Button(visible=False), gr.Button(visible=True), lenient_f1, strict_f1, average_f1


# Main

with gr.Blocks() as leaderboard:

    gr.Markdown(
        """
        # Dipromats 2024 Task 2 Leaderboard
        # Automatic Detection of Narratives from Diplomats of Major Powers
        This is...
        You can...
        """)
    with gr.Tabs() as tabs:

        # Tab Leaderboard
        with gr.TabItem("Leaderboard", id=0):
            gr.Markdown(
                """
                #
                # Leaderboard
                """)
            leaderboard_table = gr.Dataframe(headers=["Team", "Run ID", "Lenient F1", "Strict F1", "Average F1"],
                        value=data_to_table(),
                        interactive=False)

        # Tab Evaluate
        with gr.TabItem("Evaluate your results", id=1):
            gr.Markdown(
                """
                # Upload your results and get evaluated
                Then you can decide to submit your results to the leaderboard or not.
                Make sure that you upload a file with the json format described in...
                """)

            # Submission Form
            with gr.Row():
                with gr.Column():
                    with gr.Row():
                        team_input = gr.Textbox(label="Team Name")
                        run_id = gr.Textbox(label="Run ID")
                    email_input = gr.Textbox(label="Email (only for submission verification, it won't be shown)")
                description_input = gr.Textbox(label="System description", lines=6)
            file_input = gr.File(label="Upload a JSON file", file_types=[".json"], type="filepath", file_count="single")
            evaluate_button = gr.Button("Evaluate")

            # System results table
            with gr.Row(visible=True):
                lenient_f1 = gr.Number(label="Lenient F1", interactive=False)
                strict_f1 = gr.Number(label="Strict F1", interactive=False)
                average_f1 = gr.Number(label="Average F1", interactive=False)

            # Submit to leaderboard
            submit_button = gr.Button("Submit to leaderboard", visible=False)

    evaluate_button.click(process_file,
                        inputs=[file_input, team_input, run_id, description_input, email_input],
                        outputs=[evaluate_button,submit_button,lenient_f1, strict_f1, average_f1])
    submit_button.click(update_leaderboard,
                        inputs=[email_input, team_input, run_id, description_input, lenient_f1, strict_f1, average_f1],
                        outputs=[leaderboard_table, tabs, evaluate_button, submit_button, team_input, run_id, description_input, email_input, file_input,lenient_f1, strict_f1, average_f1])
leaderboard.launch()