app.py

import subprocess
import gradio as gr
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 (
    BENCHMARK_COLS,
    COLS,
    EVAL_COLS,
    EVAL_TYPES,
    NUMERIC_INTERVALS,
    TYPES,
    AutoEvalColumn,
    ModelType,
    fields,
    WeightType,
    Precision
)
from src.envs import API, EVAL_REQUESTS_PATH, EVAL_RESULTS_PATH, DATA_REPO, REPO_ID, TOKEN
from src.populate import get_evaluation_queue_df, get_leaderboard_df
from src.submission.submit import add_new_eval


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

try:
    print(EVAL_REQUESTS_PATH)
    snapshot_download(
        repo_id=DATA_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=DATA_REPO, local_dir=EVAL_RESULTS_PATH, repo_type="dataset", tqdm_class=None, etag_timeout=30, token=TOKEN
    )
except Exception:
    restart_space()


raw_data, original_df = get_leaderboard_df(EVAL_RESULTS_PATH, EVAL_REQUESTS_PATH, COLS, BENCHMARK_COLS)
leaderboard_df = original_df.copy()

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


# Searching and filtering
def update_table(
    hidden_df: pd.DataFrame,
    columns: list,
):
    df = select_columns(hidden_df, columns)
    return df


def search_table(df: pd.DataFrame, query: str) -> pd.DataFrame:
    return df[(df[AutoEvalColumn.model.name].str.contains(query, case=False))]


def select_columns(df: pd.DataFrame, columns: list) -> pd.DataFrame:
    always_here_cols = [
        AutoEvalColumn.model_type_symbol.name,
        AutoEvalColumn.model.name,
    ]
    # We use COLS to maintain sorting
    filtered_df = df[
        always_here_cols + [c for c in COLS if c in df.columns and c in columns] 
    ]
    return filtered_df

demo = gr.Blocks(css=custom_css)
with demo:
    gr.HTML(TITLE)
    gr.Markdown(INTRODUCTION_TEXT, elem_classes="markdown-text")

    with gr.Tabs(elem_classes="tab-buttons") as tabs:
        with gr.TabItem("Leaderboard", elem_id="llm-benchmark-tab-table", id=0):
            with gr.Row():
                shown_columns = gr.CheckboxGroup(
                    choices=[
                        c.name
                        for c in fields(AutoEvalColumn)
                        if not c.hidden and not c.never_hidden
                    ],
                    value=[
                        c.name
                        for c in fields(AutoEvalColumn)
                        if c.displayed_by_default and not c.hidden and not c.never_hidden
                    ],
                    label="Select columns to show",
                    elem_id="column-select",
                    interactive=True,
                )

            leaderboard_table = gr.components.Dataframe(
                value=leaderboard_df[
                    [c.name for c in fields(AutoEvalColumn) if c.never_hidden]
                    + shown_columns.value
                ],
                headers=[c.name for c in fields(AutoEvalColumn) if c.never_hidden] + shown_columns.value,
                datatype=TYPES,
                elem_id="leaderboard-table",
                interactive=False,
                visible=True,
            )

            # Dummy leaderboard for handling the case when the user uses backspace key
            hidden_leaderboard_table_for_search = gr.components.Dataframe(
                value=original_df[COLS],
                headers=COLS,
                datatype=TYPES,
                visible=False,
            )
            for selector in [shown_columns]:
                selector.change(
                    update_table,
                    [
                        hidden_leaderboard_table_for_search,
                        shown_columns,
                    ],
                    leaderboard_table,
                    queue=True,
                )

        with gr.TabItem("Submit Algorithm", elem_id="llm-benchmark-tab-table", id=3):
            with gr.Column():
                with gr.Row():
                    gr.Markdown(EVALUATION_QUEUE_TEXT, elem_classes="markdown-text")

            with gr.Row():
                gr.Markdown("# Submission Form\nSubmitted files will be stored and made public.", elem_classes="markdown-text")

            with gr.Row():
                with gr.Column():
                    submission_file = gr.File(label="Evaluation result (JSON file generated by `run_benchmark.py`, one algorithm on all challenges)")

                with gr.Column():
                    algo_name = gr.Textbox(label="Algorithm display name")
                    algo_info = gr.Textbox(label="Comments & extra information")
                    algo_link = gr.Textbox(label="One external link (e.g. GitHub repo, paper, project page)")
                    submitter_email = gr.Textbox(label="Email address for contact (will be kept confidential)")

            submit_button = gr.Button("Submit Algorithm")
            submission_result = gr.Markdown()
            submit_button.click(
                add_new_eval,
                [
                    submission_file,
                    algo_name,
                    algo_info,
                    algo_link,
                    submitter_email,
                ],
                submission_result,
            )

    with gr.Row():
        with gr.Accordion("About & Citation", open=False):
            citation_button = gr.Textbox(
                value=CITATION_BUTTON_TEXT,
                label=CITATION_BUTTON_LABEL,
                lines=20,
                elem_id="citation-button",
                show_copy_button=True,
            )

scheduler = BackgroundScheduler()
scheduler.add_job(restart_space, "interval", seconds=1800)
scheduler.start()
demo.queue(default_concurrency_limit=40).launch()