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, nc_tasks, nr_tasks, lp_tasks, ) from src.display.css_html_js import custom_css from src.display.utils import ( BENCHMARK_COLS, #COLS, COLS_NC, COLS_NR, COLS_LP, EVAL_COLS, EVAL_TYPES, NUMERIC_INTERVALS, TYPES, AutoEvalColumn_NodeClassification, AutoEvalColumn_NodeRegression, AutoEvalColumn_LinkPrediction, #AutoEvalColumn, ModelType, TASK_LIST, OFFICIAL, HONOR, 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 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=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() restart_go = 1 # Searching and filtering def update_table( hidden_df: pd.DataFrame, columns: list, query: str, ): #filtered_df = filter_models(hidden_df, size_query, show_deleted) filtered_df = filter_queries(query, hidden_df) print(columns) df = select_columns(filtered_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 = [ "Model" ] # We use COLS to maintain sorting #print(df) #print(df.columns) #print([c for c in df.columns if c in columns]) filtered_df = df[ always_here_cols + [c for c in df.columns if c in columns] ] #print(filtered_df) 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] ) return filtered_df def filter_models( df: pd.DataFrame, size_query: list, show_deleted: bool ) -> 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"])] 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 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("🏅 Entity Classification Leaderboard", elem_id="llm-benchmark-tab-table", id=0): global COLS COLS = COLS_NC AutoEvalColumn = AutoEvalColumn_NodeClassification original_df = get_leaderboard_df(EVAL_REQUESTS_PATH, "Node Classification") leaderboard_df = original_df.copy() 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(): 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, ) #print(leaderboard_df) #print(shown_columns.value) 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, search_bar, ], leaderboard_table, ) for selector in [shown_columns]: selector.change( update_table, [ hidden_leaderboard_table_for_search, shown_columns, search_bar, ], leaderboard_table, queue=True, ) gr.Markdown("Evaluation metric: AUROC ⬆️") with gr.TabItem("🏅 Entity Regression Leaderboard", elem_id="llm-benchmark-tab-table", id=1): COLS = COLS_NR AutoEvalColumn = AutoEvalColumn_NodeRegression original_df = get_leaderboard_df(EVAL_REQUESTS_PATH, "Node Regression") leaderboard_df = original_df.copy() 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(): 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, ) #print(leaderboard_df) #print(shown_columns) 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, search_bar, ], leaderboard_table, ) for selector in [shown_columns]: selector.change( update_table, [ hidden_leaderboard_table_for_search, shown_columns, search_bar, ], leaderboard_table, queue=True, ) gr.Markdown("Evaluation metric: MAE ⬇️") with gr.TabItem("🏅 Recommendation Leaderboard", elem_id="llm-benchmark-tab-table", id=2): COLS = COLS_LP AutoEvalColumn = AutoEvalColumn_LinkPrediction original_df = get_leaderboard_df(EVAL_REQUESTS_PATH, "Link Prediction") leaderboard_df = original_df.copy() 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(): 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, ) #print(leaderboard_df) #print(shown_columns) 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, search_bar, ], leaderboard_table, ) for selector in [shown_columns]: selector.change( update_table, [ hidden_leaderboard_table_for_search, shown_columns, search_bar, ], leaderboard_table, queue=True, ) gr.Markdown("Evaluation metric: MAP ⬆️") with gr.TabItem("🚀 Submit here! ", 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("# ✉️✨ Submit your model here!", elem_classes="markdown-text") with gr.Row(): with gr.Column(): author_name_textbox = gr.Textbox(label="Your name") email_textbox = gr.Textbox(label="Your email") relbench_version_textbox = gr.Textbox(label="RelBench version") model_name_textbox = gr.Textbox(label="Model name") ''' dataset_name_textbox = gr.Dropdown( choices=[t.value.name for t in TASK_LIST], label="Task name (e.g. rel-amazon-user-churn)", multiselect=False, value=None, interactive=True, ) ''' official_or_not = gr.Dropdown( choices=[i.value.name for i in OFFICIAL], label="Is it an official submission?", multiselect=False, value=None, interactive=True, ) paper_url_textbox = gr.Textbox(label="Paper URL Link") github_url_textbox = gr.Textbox(label="GitHub URL Link") #parameters_textbox = gr.Textbox(label="Number of parameters") task_track = gr.Dropdown( choices=['Entity Classification', 'Entity Regression', 'Recommendation'], label="Choose the task track", multiselect=False, value=None, interactive=True, ) honor_code = gr.Dropdown( choices=[i.value.name for i in HONOR], label="Do you agree to the honor code?", multiselect=False, value=None, interactive=True, ) with gr.Column(): test_performance = gr.Textbox(lines = 16, label="Test set performance, use {task: [mean,std]} format e.g. {'rel-amazon/user-churn': [0.352,0.023], 'rel-amazon/user-ltv': [0.304,0.022], ...}") valid_performance = gr.Textbox(lines = 16, label="Validation set performance, use {task: [mean,std]} format e.g. {'rel-amazon/user-churn': [0.352,0.023], 'rel-amazon/user-ltv': [0.304,0.022], ...}") submit_button = gr.Button("Submit Eval") submission_result = gr.Markdown() submit_button.click( add_new_eval, [ author_name_textbox, email_textbox, relbench_version_textbox, model_name_textbox, official_or_not, test_performance, valid_performance, paper_url_textbox, github_url_textbox, #parameters_textbox, honor_code, task_track ], submission_result, ) scheduler = BackgroundScheduler() scheduler.add_job(restart_space, "interval", seconds=1800) scheduler.start() demo.queue(default_concurrency_limit=40).launch()