app.py

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 (
    COLS,
    EVAL_COLS,
    EVAL_TYPES,
    NUMERIC_INTERVALS,
    TYPES,
    AutoEvalColumn,
    ModelType,
    fields,
    WeightType,
    Precision,
    generate_column_name
)
# from src.display.plot_curves import plot_curves
from src.envs import API, EVAL_REQUESTS_PATH, EVAL_RESULTS_PATH, QUEUE_REPO, REPO_ID, RESULTS_REPO, HF_TOKEN
from src.populate import get_evaluation_queue_df, get_leaderboard_df
from src.submission.submit import add_new_eval
from dotenv import load_dotenv


load_dotenv()


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

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=HF_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=HF_TOKEN
    )
except Exception:
    restart_space()


results, original_df = get_leaderboard_df(EVAL_RESULTS_PATH, COLS)
leaderboard_df = original_df.copy()
leaderboard_df.to_csv("leaderboard.csv", index=False)

(
    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,
    phenotypes: list,
    metrics: list,
    feature_sets: list,
    nb_shots: list,
    type_query: list,
    precision_query: str,
    size_query: list,
    show_deleted: bool,
    query: str,
):
    filtered_df = filter_models(hidden_df, type_query, size_query, precision_query, show_deleted, feature_sets, nb_shots)
    filtered_df = filter_queries(query, filtered_df)
    df = select_columns(filtered_df, columns, phenotypes, metrics)
    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, phenotypes: list, metrics:list) -> pd.DataFrame:
    always_here_cols = [
        AutoEvalColumn.model_type_symbol.name,
        AutoEvalColumn.model.name,
        AutoEvalColumn.feature_set.name,
        AutoEvalColumn.nb_shots.name,
    ]
    
    task_cols = []
    for phenotype in phenotypes:
        for metric in metrics:
            task_cols.append(generate_column_name(phenotype, metric))
    
    filtered_df = df[
        always_here_cols + [c for c in COLS if c in df.columns and c in columns] + sorted(task_cols)
    ]
    return filtered_df


def filter_queries(query: str, filtered_df: pd.DataFrame) -> pd.DataFrame:
    final_df = []
    if query != "":
        queries = [q.strip() for q in query.split(";")]
        for _q in queries:
            _q = _q.strip()
            if _q != "":
                temp_filtered_df = search_table(filtered_df, _q)
                if len(temp_filtered_df) > 0:
                    final_df.append(temp_filtered_df)
        if len(final_df) > 0:
            filtered_df = pd.concat(final_df)
            filtered_df = filtered_df.drop_duplicates(
                subset=[AutoEvalColumn.model.name, AutoEvalColumn.precision.name, AutoEvalColumn.revision.name]
            )

    return filtered_df


def filter_models(
    df: pd.DataFrame, type_query: list, size_query: list, precision_query: list, show_deleted: bool, feature_sets: list, nb_shots: list) -> pd.DataFrame:
    # Show all models
    if show_deleted:
        filtered_df = df
    else:  # Show only still on the hub models
        filtered_df = df[df[AutoEvalColumn.still_on_hub.name] == True]

    type_emoji = [t[0] for t in type_query]
    filtered_df = filtered_df.loc[df[AutoEvalColumn.model_type_symbol.name].isin(type_emoji)]
    filtered_df = filtered_df.loc[df[AutoEvalColumn.precision.name].isin(precision_query + ["None"])]
    if -1 not in nb_shots:
        filtered_df = filtered_df.loc[df[AutoEvalColumn.nb_shots.name].isin(nb_shots)]
    filtered_df = filtered_df.loc[df[AutoEvalColumn.feature_set.name].isin(feature_sets)]
    
    numeric_interval = pd.IntervalIndex(sorted([NUMERIC_INTERVALS[s] for s in size_query]))
    params_column = pd.to_numeric(df[AutoEvalColumn.params.name], errors="coerce")
    mask = params_column.apply(lambda x: any(numeric_interval.contains(x)))
    filtered_df = filtered_df.loc[mask]

    return filtered_df


def format_model_sample(sample):
    return f"{sample[0]}, {sample[1]}, {sample[2]}-shots"


def update_selected_models(selected_models, sample):
    sample_str = format_model_sample(sample)
    selected_models.append(sample_str)
    return selected_models


MODELS = [
    ["Model A", "Feature Set 1", 5],
    ["Model B", "Feature Set 2", 10],
    ["Model C", "Feature Set 3", 15]
]

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("🏅 LLM Benchmark", elem_id="llm-benchmark-tab-table", id=0):
            with gr.Row():
                with gr.Column():
                    with gr.Row():
                        search_bar = gr.Textbox(
                            placeholder=" 🔍 Search for your model (separate multiple queries with `;`) and press ENTER...",
                            show_label=False,
                            elem_id="search-bar",
                        )
                    with gr.Row():
                        with gr.Column(min_width=320):
                            shown_phenotypes = gr.CheckboxGroup(
                                choices=sorted(set([
                                    c.task.value.phenotype
                                    for c in fields(AutoEvalColumn)
                                    if not c.hidden and not c.never_hidden and c.is_task
                                ])),
                                label="Select phenotypes to show",
                                elem_id="phenotype-select",
                                interactive=True,
                            )
                            shown_metrics = gr.CheckboxGroup(
                                choices=sorted(set([
                                    c.task.value.metric.upper()
                                    for c in fields(AutoEvalColumn)
                                    if not c.hidden and not c.never_hidden and c.is_task
                                ])),
                                value=sorted(set([
                                    c.task.value.metric.upper()
                                    for c in fields(AutoEvalColumn)
                                    if not c.hidden and not c.never_hidden and c.is_task
                                ])),
                                label="Select metrics to show",
                                elem_id="metric-select",
                                interactive=True,
                            )
                    with gr.Row():
                        shown_columns = gr.CheckboxGroup(
                            choices=[
                                c.name
                                for c in fields(AutoEvalColumn)
                                if not c.hidden and not c.never_hidden and not c.is_task
                            ],
                            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,
                        )
                    with gr.Row():
                        deleted_models_visibility = gr.Checkbox(
                            value=True, label="Show gated/private/deleted models", interactive=True
                        )
                with gr.Column(min_width=320):
                    filter_features = gr.CheckboxGroup(
                        label="Features Set",
                        choices=[("Baseline (Age, Sex, BMI)", "baseline"), ("Expanded (Age, Sex, BMI, HDL, LDL, Total cholesterol, Triglycerides, Diastolic blood pressure, Smoking status, Snoring, Insomnia, Daytime napping, Sleep duration, Chronotype)", "expanded")],
                        value=["baseline"],
                        interactive=True,
                        elem_id="filter-feature-set",
                    )
                    filter_nb_shots = gr.CheckboxGroup(
                        label="Number of shots",
                        choices=[("Zero-Shot", 0), ("2-Shot", 2), ("4-Shot", 4), ("6-Shot", 6), ("All", -1)],
                        value=[0, 2, -1],
                        interactive=True,
                        elem_id="filter-nb-shots",
                    )
                    filter_columns_type = gr.CheckboxGroup(
                        label="Model types",
                        choices=[t.to_str() for t in ModelType],
                        value=[t.to_str() for t in ModelType],
                        interactive=True,
                        elem_id="filter-columns-type",
                    )
                    filter_columns_precision = gr.CheckboxGroup(
                        label="Precision",
                        choices=[i.value.name for i in Precision],
                        value=[i.value.name for i in Precision],
                        interactive=True,
                        elem_id="filter-columns-precision",
                    )
                    filter_columns_size = gr.CheckboxGroup(
                        label="Model sizes (in billions of parameters)",
                        choices=list(NUMERIC_INTERVALS.keys()),
                        value=list(NUMERIC_INTERVALS.keys()),
                        interactive=True,
                        elem_id="filter-columns-size",
                    )

            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,
            )
            search_bar.submit(
                update_table,
                [
                    hidden_leaderboard_table_for_search,
                    shown_columns,
                    shown_phenotypes,
                    shown_metrics,
                    filter_features,
                    filter_nb_shots,
                    filter_columns_type,
                    filter_columns_precision,
                    filter_columns_size,
                    deleted_models_visibility,
                    search_bar,
                ],
                leaderboard_table,
            )
            for selector in [shown_phenotypes, shown_metrics, shown_columns, filter_columns_type, filter_columns_precision, filter_columns_size, deleted_models_visibility, filter_nb_shots, filter_features]:
                selector.change(
                    update_table,
                    [
                        hidden_leaderboard_table_for_search,
                        shown_columns,
                        shown_phenotypes,
                        shown_metrics,
                        filter_features,
                        filter_nb_shots,
                        filter_columns_type,
                        filter_columns_precision,
                        filter_columns_size,
                        deleted_models_visibility,
                        search_bar,
                    ],
                    leaderboard_table,
                    queue=True,
                )
        
        """
        with gr.TabItem("📈 ROC/PR Curves", elem_id="llm-benchmark-tab-table", id=2):
            with gr.Row():
                with gr.Column():
                    shown_phenotypes_curve = gr.CheckboxGroup(
                        choices=sorted(set([
                            c.task.value.phenotype
                            for c in fields(AutoEvalColumn)
                            if not c.hidden and not c.never_hidden and c.is_task
                        ])),
                        label="Select phenotypes",
                        elem_id="phenotype-select-curve",
                        interactive=True,
                    )
                with gr.Column():
                    selected_models = gr.Dropdown(
                        choices=[format_model_sample(sample) for sample in MODELS],
                        label="Selected models",
                        elem_id="selected-models",
                        interactive=True,
                        multiselect=True,
                    )

        with gr.TabItem("📝 About", elem_id="llm-benchmark-tab-table", id=3):
            gr.Markdown(LLM_BENCHMARKS_TEXT, elem_classes="markdown-text")
        """
        
        with gr.TabItem("🚀 Submit here! ", elem_id="llm-benchmark-tab-table", id=4):
            with gr.Column():
                with gr.Row():
                    gr.Markdown(EVALUATION_QUEUE_TEXT, elem_classes="markdown-text")

                with gr.Column():
                    with gr.Accordion(
                        f"✅ Finished Evaluations ({len(finished_eval_queue_df)})",
                        open=False,
                    ):
                        with gr.Row():
                            finished_eval_table = gr.components.Dataframe(
                                value=finished_eval_queue_df,
                                headers=EVAL_COLS,
                                datatype=EVAL_TYPES,
                                row_count=5,
                            )
                    with gr.Accordion(
                        f"🔄 Running Evaluation Queue ({len(running_eval_queue_df)})",
                        open=False,
                    ):
                        with gr.Row():
                            running_eval_table = gr.components.Dataframe(
                                value=running_eval_queue_df,
                                headers=EVAL_COLS,
                                datatype=EVAL_TYPES,
                                row_count=5,
                            )

                    with gr.Accordion(
                        f"⏳ Pending Evaluation Queue ({len(pending_eval_queue_df)})",
                        open=False,
                    ):
                        with gr.Row():
                            pending_eval_table = gr.components.Dataframe(
                                value=pending_eval_queue_df,
                                headers=EVAL_COLS,
                                datatype=EVAL_TYPES,
                                row_count=5,
                            )
            with gr.Row():
                gr.Markdown("# ✉️✨ Submit your model here!", elem_classes="markdown-text")

            with gr.Row():
                with gr.Column():
                    model_name_textbox = gr.Textbox(label="Model name")
                    revision_name_textbox = gr.Textbox(label="Revision commit", placeholder="main")
                    model_type = gr.Dropdown(
                        choices=[t.to_str(" : ") for t in ModelType if t != ModelType.Unknown],
                        label="Model type",
                        multiselect=False,
                        value=None,
                        interactive=True,
                    )

                with gr.Column():
                    precision = gr.Dropdown(
                        choices=[i.value.name for i in Precision if i != Precision.Unknown],
                        label="Precision",
                        multiselect=False,
                        value="float16",
                        interactive=True,
                    )
                    weight_type = gr.Dropdown(
                        choices=[i.value.name for i in WeightType],
                        label="Weights type",
                        multiselect=False,
                        value="Original",
                        interactive=True,
                    )
                    base_model_name_textbox = gr.Textbox(label="Base model (for delta or adapter weights)")

            submit_button = gr.Button("Submit Eval")
            submission_result = gr.Markdown()
            submit_button.click(
                add_new_eval,
                [
                    model_name_textbox,
                    base_model_name_textbox,
                    revision_name_textbox,
                    precision,
                    weight_type,
                    model_type,
                ],
                submission_result,
            )

    with gr.Row():
        with gr.Accordion("📙 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()