import os import logging import gradio as gr import pandas as pd from apscheduler.schedulers.background import BackgroundScheduler from huggingface_hub import snapshot_download from gradio_space_ci import enable_space_ci from src.display.about import ( CITATION_BUTTON_LABEL, CITATION_BUTTON_TEXT, EVALUATION_QUEUE_TEXT, FAQ_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, Precision, WeightType, fields, ) from src.envs import ( API, DYNAMIC_INFO_FILE_PATH, DYNAMIC_INFO_PATH, DYNAMIC_INFO_REPO, EVAL_REQUESTS_PATH, EVAL_RESULTS_PATH, H4_TOKEN, IS_PUBLIC, QUEUE_REPO, REPO_ID, RESULTS_REPO, ) from src.populate import get_evaluation_queue_df, get_leaderboard_df from src.scripts.update_all_request_files import update_dynamic_files from src.submission.submit import add_new_eval from src.tools.collections import update_collections from src.tools.plots import create_metric_plot_obj, create_plot_df, create_scores_df # Start ephemeral Spaces on PRs (see config in README.md) enable_space_ci() def restart_space(): API.restart_space(repo_id=REPO_ID, token=H4_TOKEN) def download_dataset(repo_id, local_dir, repo_type="dataset", max_attempts=3): """Attempt to download dataset with retries.""" attempt = 0 while attempt < max_attempts: try: print(f"Downloading {repo_id} to {local_dir}") snapshot_download( repo_id=repo_id, local_dir=local_dir, repo_type=repo_type, tqdm_class=None, etag_timeout=30, max_workers=8, ) return except Exception as e: logging.error(f"Error downloading {repo_id}: {e}") attempt += 1 if attempt == max_attempts: restart_space() def init_space(full_init: bool = True): """Initializes the application space, loading only necessary data.""" if full_init: # These downloads only occur on full initialization download_dataset(QUEUE_REPO, EVAL_REQUESTS_PATH) download_dataset(DYNAMIC_INFO_REPO, DYNAMIC_INFO_PATH) download_dataset(RESULTS_REPO, EVAL_RESULTS_PATH) # Always retrieve the leaderboard DataFrame raw_data, original_df = get_leaderboard_df( results_path=EVAL_RESULTS_PATH, requests_path=EVAL_REQUESTS_PATH, dynamic_path=DYNAMIC_INFO_FILE_PATH, cols=COLS, benchmark_cols=BENCHMARK_COLS, ) if full_init: # Collection update only happens on full initialization update_collections(original_df) leaderboard_df = original_df.copy() # Evaluation queue DataFrame retrieval is independent of initialization detail level eval_queue_dfs = get_evaluation_queue_df(EVAL_REQUESTS_PATH, EVAL_COLS) return leaderboard_df, raw_data, original_df, eval_queue_dfs # Convert the environment variable "LEADERBOARD_FULL_INIT" to a boolean value, defaulting to True if the variable is not set. # This controls whether a full initialization should be performed. do_full_init = os.getenv("LEADERBOARD_FULL_INIT", "True") == "True" # Calls the init_space function with the `full_init` parameter determined by the `do_full_init` variable. # This initializes various DataFrames used throughout the application, with the level of initialization detail controlled by the `do_full_init` flag. leaderboard_df, raw_data, original_df, eval_queue_dfs = init_space(full_init=do_full_init) finished_eval_queue_df, running_eval_queue_df, pending_eval_queue_df = eval_queue_dfs # Data processing for plots now only on demand in the respective Gradio tab def load_and_create_plots(): plot_df = create_plot_df(create_scores_df(raw_data)) return plot_df # Searching and filtering def update_table( hidden_df: pd.DataFrame, columns: list, type_query: list, precision_query: str, size_query: list, hide_models: list, query: str, ): filtered_df = filter_models( df=hidden_df, type_query=type_query, size_query=size_query, precision_query=precision_query, hide_models=hide_models, ) filtered_df = filter_queries(query, filtered_df) df = select_columns(filtered_df, columns) return df def load_query(request: gr.Request): # triggered only once at startup => read query parameter if it exists query = request.query_params.get("query") or "" return ( query, query, ) # return one for the "search_bar", one for a hidden component that triggers a reload only if value has changed def search_model(df: pd.DataFrame, query: str) -> pd.DataFrame: return df[(df[AutoEvalColumn.fullname.name].str.contains(query, case=False, na=False))] def search_license(df: pd.DataFrame, query: str) -> pd.DataFrame: return df[df[AutoEvalColumn.license.name].str.contains(query, case=False, na=False)] def select_columns(df: pd.DataFrame, columns: list) -> pd.DataFrame: always_here_cols = [c.name for c in fields(AutoEvalColumn) if c.never_hidden] dummy_col = [AutoEvalColumn.fullname.name] filtered_df = df[always_here_cols + [c for c in COLS if c in df.columns and c in columns] + dummy_col] return filtered_df def filter_queries(query: str, df: pd.DataFrame): tmp_result_df = [] # Empty query return the same df if query == "": return df # all_queries = [q.strip() for q in query.split(";")] # license_queries = [] all_queries = [q.strip() for q in query.split(";") if q.strip() != ""] model_queries = [q for q in all_queries if not q.startswith("licence")] license_queries_raw = [q for q in all_queries if q.startswith("license")] license_queries = [ q.replace("license:", "").strip() for q in license_queries_raw if q.replace("license:", "").strip() != "" ] # Handling model name search for query in model_queries: tmp_df = search_model(df, query) if len(tmp_df) > 0: tmp_result_df.append(tmp_df) if not tmp_result_df and not license_queries: # Nothing is found, no license_queries -> return empty df return pd.DataFrame(columns=df.columns) if tmp_result_df: df = pd.concat(tmp_result_df) df = df.drop_duplicates( subset=[AutoEvalColumn.model.name, AutoEvalColumn.precision.name, AutoEvalColumn.revision.name] ) if not license_queries: return df # Handling license search tmp_result_df = [] for query in license_queries: tmp_df = search_license(df, query) if len(tmp_df) > 0: tmp_result_df.append(tmp_df) if not tmp_result_df: # Nothing is found, return empty df return pd.DataFrame(columns=df.columns) df = pd.concat(tmp_result_df) df = df.drop_duplicates( subset=[AutoEvalColumn.model.name, AutoEvalColumn.precision.name, AutoEvalColumn.revision.name] ) return df def filter_models( df: pd.DataFrame, type_query: list, size_query: list, precision_query: list, hide_models: list ) -> pd.DataFrame: # Show all models if "Private or deleted" in hide_models: filtered_df = df[df[AutoEvalColumn.still_on_hub.name] == True] else: filtered_df = df if "Contains a merge/moerge" in hide_models: filtered_df = filtered_df[filtered_df[AutoEvalColumn.merged.name] == False] if "MoE" in hide_models: filtered_df = filtered_df[filtered_df[AutoEvalColumn.moe.name] == False] if "Flagged" in hide_models: filtered_df = filtered_df[filtered_df[AutoEvalColumn.flagged.name] == False] 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"])] 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 leaderboard_df = filter_models( df=leaderboard_df, type_query=[t.to_str(" : ") for t in ModelType], size_query=list(NUMERIC_INTERVALS.keys()), precision_query=[i.value.name for i in Precision], hide_models=["Private or deleted", "Contains a merge/moerge", "Flagged"], # Deleted, merges, flagged, MoEs ) 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 models or licenses (e.g., 'model_name; license: MIT') and press ENTER...", show_label=False, elem_id="search-bar", ) 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.dummy ], 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(): hide_models = gr.CheckboxGroup( label="Hide models", choices=["Private or deleted", "Contains a merge/moerge", "Flagged", "MoE"], value=["Private or deleted", "Contains a merge/moerge", "Flagged"], interactive=True, ) with gr.Column(min_width=320): # with gr.Box(elem_id="box-filter"): 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 + [AutoEvalColumn.fullname.name] ], 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, filter_columns_type, filter_columns_precision, filter_columns_size, hide_models, search_bar, ], leaderboard_table, ) # Define a hidden component that will trigger a reload only if a query parameter has been set hidden_search_bar = gr.Textbox(value="", visible=False) hidden_search_bar.change( update_table, [ hidden_leaderboard_table_for_search, shown_columns, filter_columns_type, filter_columns_precision, filter_columns_size, hide_models, search_bar, ], leaderboard_table, ) # Check query parameter once at startup and update search bar + hidden component demo.load(load_query, inputs=[], outputs=[search_bar, hidden_search_bar]) for selector in [ shown_columns, filter_columns_type, filter_columns_precision, filter_columns_size, hide_models, ]: selector.change( update_table, [ hidden_leaderboard_table_for_search, shown_columns, filter_columns_type, filter_columns_precision, filter_columns_size, hide_models, search_bar, ], leaderboard_table, queue=True, ) with gr.TabItem("📈 Metrics through time", elem_id="llm-benchmark-tab-table", id=2): with gr.Row(): with gr.Column(): plot_df = load_and_create_plots() chart = create_metric_plot_obj( plot_df, [AutoEvalColumn.average.name], title="Average of Top Scores and Human Baseline Over Time (from last update)", ) gr.Plot(value=chart, min_width=500) with gr.Column(): plot_df = load_and_create_plots() chart = create_metric_plot_obj( plot_df, BENCHMARK_COLS, title="Top Scores and Human Baseline Over Time (from last update)", ) gr.Plot(value=chart, min_width=500) with gr.TabItem("📝 About", elem_id="llm-benchmark-tab-table", id=3): gr.Markdown(LLM_BENCHMARKS_TEXT, elem_classes="markdown-text") with gr.TabItem("❗FAQ", elem_id="llm-benchmark-tab-table", id=4): gr.Markdown(FAQ_TEXT, elem_classes="markdown-text") with gr.TabItem("🚀 Submit ", elem_id="llm-benchmark-tab-table", id=5): with gr.Column(): with gr.Row(): gr.Markdown(EVALUATION_QUEUE_TEXT, elem_classes="markdown-text") 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") private = gr.Checkbox(False, label="Private", visible=not IS_PUBLIC) model_type = gr.Dropdown( choices=[t.to_str(" : ") for t in ModelType if t != ModelType.Unknown], label="Model type", multiselect=False, value=ModelType.FT.to_str(" : "), 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)") 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, ) 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, private, 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", hours=3) # restarted every 3h scheduler.add_job(update_dynamic_files, "interval", hours=2) # launched every 2 hour scheduler.start() demo.queue(default_concurrency_limit=40).launch()