app.py

import gradio as gr
from gradio_leaderboard import Leaderboard, ColumnFilter, SelectColumns
import pandas as pd
from apscheduler.schedulers.background import BackgroundScheduler
from huggingface_hub import snapshot_download

from src.about import (
    CITATION_BUTTON_LABEL,
    CITATION_BUTTON_TEXT,
    EVALUATION_QUEUE_TEXT,
    INTRODUCTION_TEXT,
    LLM_BENCHMARKS_TEXT,
    TITLE,
)
from src.display.css_html_js import custom_css
from src.display.utils import (
    COLS,
    ST_BENCHMARK_COLS,
    AGENTIC_BENCHMARK_COLS,
    EVAL_COLS,
    EVAL_TYPES,
    AutoEvalColumn,
    ModelType,
    fields,
    WeightType,
    Precision
)
from src.envs import API, EVAL_REQUESTS_PATH, EVAL_RESULTS_PATH, QUEUE_REPO, REPO_ID, RESULTS_REPO, TOKEN
from src.populate import get_evaluation_queue_df, get_leaderboard_df, TASK_NAME_INVERSE_MAP
from src.submission.submit import add_new_eval


def restart_space():
    API.restart_space(repo_id=REPO_ID)

### Space initialisation
try:
    print(EVAL_REQUESTS_PATH)
    snapshot_download(
        repo_id=QUEUE_REPO, local_dir=EVAL_REQUESTS_PATH, repo_type="dataset", tqdm_class=None, etag_timeout=30, token=TOKEN
    )
except Exception:
    restart_space()
try:
    print(EVAL_RESULTS_PATH)
    snapshot_download(
        repo_id=RESULTS_REPO, local_dir=EVAL_RESULTS_PATH, repo_type="dataset", tqdm_class=None, etag_timeout=30, token=TOKEN
    )
except Exception:
    restart_space()


ST_LEADERBOARD_DF = get_leaderboard_df(EVAL_RESULTS_PATH, EVAL_REQUESTS_PATH, COLS, ST_BENCHMARK_COLS)
AGENTIC_LEADERBOARD_DF = get_leaderboard_df(EVAL_RESULTS_PATH, EVAL_REQUESTS_PATH, COLS, AGENTIC_BENCHMARK_COLS)

(
    finished_eval_queue_df,
    running_eval_queue_df,
    pending_eval_queue_df,
) = get_evaluation_queue_df(EVAL_REQUESTS_PATH, EVAL_COLS)

def init_leaderboard(dataframe, benchmark_type):
    if dataframe is None or dataframe.empty:
        raise ValueError("Leaderboard DataFrame is empty or None.")
    
    AutoEvalColumnSubset = [c for c in fields(AutoEvalColumn) if ((c.name=="Model") or (TASK_NAME_INVERSE_MAP.get(c.name, dict()).get("type", "")==benchmark_type))]

    return gr.components.Dataframe(
        value=dataframe,
        datatype=[c.type for c in AutoEvalColumnSubset],
        column_widths=["150px" if c.name != "Model" else "250px" for c in AutoEvalColumnSubset],
        wrap=False,
    )


demo = gr.Blocks(css=custom_css, theme=gr.themes.Default(primary_hue=gr.themes.colors.pink))
with demo:
    gr.HTML(f"""
    <div id="page-header" style="text-align:center; margin-bottom:1rem;">
        <div id="left-container">
            <img src="/gradio_api/file=src/assets/logo-icon.png">
        </div>
        <div id="centre-container">
            <h1 style="margin-bottom: 0.25rem;">{TITLE}</h1>
            <p style="color:#eb088a; margin:0; font-size:1.2rem;">Performance Insights &amp; Comparison</p>
        </div>
        <div id="right-container"></div>
    </div>
    """)
    # gr.HTML(TITLE)
    with gr.Group(elem_classes="intro-block"):
        gr.Markdown(INTRODUCTION_TEXT, elem_classes="markdown-text")
    # gr.Markdown(INTRODUCTION_TEXT, elem_classes="markdown-text")

    with gr.Tabs(elem_classes=["leaderboard-table", "tab-buttons"]) as tabs:
        with gr.TabItem("Base Benchmark", elem_id="llm-benchmark-tab-table", id=0):
            leaderboard = init_leaderboard(ST_LEADERBOARD_DF, "base")

        with gr.TabItem("Agentic Benchmark", elem_id="llm-benchmark-tab-table", id=1):
            leaderboard = init_leaderboard(AGENTIC_LEADERBOARD_DF, "agentic")

        with gr.TabItem("About", elem_id="llm-benchmark-tab-table", id=2):
            gr.Markdown(LLM_BENCHMARKS_TEXT, elem_classes="markdown-text")


scheduler = BackgroundScheduler()
scheduler.add_job(restart_space, "interval", seconds=1800)
scheduler.start()
demo.queue(default_concurrency_limit=40).launch(allowed_paths=["src/assets/logo-icon.png"])