import os 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 try: from commutil import dbg except: dbg = print 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, TYPES, AutoEvalColumn, ModelType, fields, WeightType, Precision from src.envs import API, EVAL_REQUESTS_PATH, EVAL_RESULTS_PATH, QUEUE_REPO, REPO_ID, RESULTS_REPO, TOKEN, EVAL_REQUESTS_PATH_BACKEND, EVAL_RESULTS_PATH_BACKEND 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) def uncheck_all(): return [], [], [], [], [], [], [] # Searching and filtering def update_table( hidden_df: pd.DataFrame, # en_shown_columns_TAD: list, # en_shown_columns_ACM: list, # en_shown_columns_KUT: list, # en_shown_columns_NLP: list, # en_shown_columns_FQA: list, # en_shown_columns_CDP: list, # shown_columns2: list, shown_columns_cyberkut: list, shown_columns_cybernlp: list, shown_columns_cyberdsa: list, shown_columns: list, filter_columns_type: list, filter_columns_precision: list, filter_columns_size: list, show_deleted: bool, search_bar: str, ): # dbg(shown_columns) # return hidden_df # Combine all column selections # selected_columns = shown_columns # selected_columns = (en_shown_columns_TAD + en_shown_columns_ACM + en_shown_columns_KUT + en_shown_columns_NLP + en_shown_columns_FQA + en_shown_columns_CDP + shown_columns2) selected_columns = shown_columns_cyberkut + shown_columns_cybernlp + shown_columns_cyberdsa + shown_columns # Filter models based on queries filtered_df = filter_models(hidden_df, filter_columns_type, filter_columns_size, filter_columns_precision, show_deleted) filtered_df = filter_queries(search_bar, filtered_df) # print(filtered_df) df = select_columns(filtered_df, selected_columns) # dbg(df) return 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) -> pd.DataFrame: # return df # Show all models if show_deleted: filtered_df = df else: filtered_df = df[df[AutoEvalColumn.still_on_hub.name] == True] if "All" not in type_query: if "?" in type_query: filtered_df = filtered_df.loc[~df[AutoEvalColumn.model_type_symbol.name].isin([t for t in ModelType if t != "?"])] else: type_emoji = [t[0] for t in type_query] filtered_df = filtered_df.loc[df[AutoEvalColumn.model_type_symbol.name].isin(type_emoji)] if "All" not in precision_query: if "?" in precision_query: filtered_df = filtered_df.loc[df[AutoEvalColumn.precision.name].isna()] else: filtered_df = filtered_df.loc[df[AutoEvalColumn.precision.name].isin(precision_query + ["None"])] if "All" not in size_query: filtered_df = df # if "?" in size_query: # filtered_df = filtered_df.loc[df[AutoEvalColumn.params.name].isna()] # else: # 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 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: # return df always_here_cols = [ AutoEvalColumn.model_type_symbol.name, AutoEvalColumn.model.name, ] # Ensure no duplicates and add the new average columns unique_columns = set(always_here_cols + columns) # We use COLS to maintain sorting filtered_df = df[[c for c in COLS if c in df.columns and c in unique_columns]] # Debugging print to see if the new columns are included # print(f"Columns included in DataFrame: {filtered_df.columns.tolist()}") # dbg(filtered_df) return filtered_df ## Space initialisation if not os.getenv("ENVIRONMENT"): 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() # LEADERBOARD_DF = get_leaderboard_df(EVAL_RESULTS_PATH, EVAL_REQUESTS_PATH, COLS, BENCHMARK_COLS) LEADERBOARD_DF = get_leaderboard_df(EVAL_RESULTS_PATH, EVAL_REQUESTS_PATH, COLS, BENCHMARK_COLS) # LEADERBOARD_DF.rename(columns={ # '4、malicious_URL_acc': 'mali', # }) LEADERBOARD_DF_CP = LEADERBOARD_DF.copy() # dbg(LEADERBOARD_DF) ( finished_eval_queue_df, running_eval_queue_df, pending_eval_queue_df, ) = get_evaluation_queue_df(EVAL_REQUESTS_PATH, EVAL_COLS) def init_leaderboard(): pass # if dataframe is None or dataframe.empty: # raise ValueError("Leaderboard DataFrame is empty or None.") # print(dataframe) # return Leaderboard( # value=dataframe, # datatype=[c.type for c in fields(AutoEvalColumn)], # select_columns=SelectColumns( # default_selection=[c.name for c in fields(AutoEvalColumn) if c.displayed_by_default], # cant_deselect=[c.name for c in fields(AutoEvalColumn) if c.never_hidden], # label="Select Columns to Display:", # ), # search_columns=[AutoEvalColumn.model.name, AutoEvalColumn.license.name], # hide_columns=[c.name for c in fields(AutoEvalColumn) if c.hidden], # filter_columns=[ # ColumnFilter(AutoEvalColumn.model_type.name, type="checkboxgroup", label="Model types"), # ColumnFilter(AutoEvalColumn.precision.name, type="checkboxgroup", label="Precision"), # ColumnFilter( # AutoEvalColumn.params.name, # type="slider", # min=0.01, # max=150, # label="Select the number of parameters (B)", # ), # # ColumnFilter( # # AutoEvalColumn.still_on_hub.name, type="boolean", label="Deleted/incomplete", default=True # # ), # ], # bool_checkboxgroup_label="Hide models", # interactive=False, # ) with gr.Column(): # with gr.Row(): # with gr.Column(): 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.Column(min_width=320): # with gr.Box(elem_id="box-filter"): with gr.Row(): filter_columns_type = gr.CheckboxGroup( label="Model types", choices=["All"] + [t.to_str() for t in ModelType], value=["All"], interactive=True, elem_id="filter-columns-type", ) filter_columns_precision = gr.CheckboxGroup( label="Precision", choices=["All"] + [i.value.name for i in Precision], value=["All"], interactive=True, elem_id="filter-columns-precision", ) # with gr.Row(): # deleted_models_visibility = gr.Checkbox( # value=True, label="Show gated/private/deleted models", interactive=True # ) # with gr.Row(): # with gr.Accordion("Select columns to show"): # with gr.Tab("Other"): with gr.Accordion("Select columns to show"): choice_list = ["Overall", "KUT", "NLP", "DSA", "CLS", "GEN", "REA"] shown_columns = gr.CheckboxGroup( choices=[c.name for c in fields(AutoEvalColumn) if c.name not in ["Model", "T"] and any(x in c.name for x in choice_list)], value=[c.name for c in fields(AutoEvalColumn) if c.name not in ["Model", "T"] and any(x in c.name for x in choice_list)], # choices=[c.name for c in fields(AutoEvalColumn) if c.never_hidden] + [c.name for c in fields(AutoEvalColumn) if c.displayed_by_default and not c.never_hidden], # value=[c.name for c in fields(AutoEvalColumn) if c.never_hidden] + [c.name for c in fields(AutoEvalColumn) if c.displayed_by_default and not c.never_hidden], label="Select Columns to Display:", interactive=True, ) # with gr.Tab("CyberTAD"): # datasets = ["4", "6"] # en_shown_columns_TAD = gr.CheckboxGroup( # choices=[c.name for c in fields(AutoEvalColumn) if c.name not in ["Model", "T"] and any(x in c.name for x in datasets)], # value=[c.name for c in fields(AutoEvalColumn) if c.displayed_by_default and c.name not in ["Model", "T"] and any(x in c.name for x in datasets)], # interactive=True # ) # with gr.Tab("CyberACM"): # datasets = ["1", "2"] # en_shown_columns_ACM = gr.CheckboxGroup( # choices=[c.name for c in fields(AutoEvalColumn) if c.name not in ["Model", "T"] and any(x in c.name for x in datasets)], # value=[c.name for c in fields(AutoEvalColumn) if c.displayed_by_default and c.name not in ["Model", "T"] and any(x in c.name for x in datasets)], # interactive=True # ) # with gr.Tab("CyberKUT"): # datasets = ["8"] # en_shown_columns_KUT = gr.CheckboxGroup( # choices=[c.name for c in fields(AutoEvalColumn) if c.name not in ["Model", "T"] and any(x in c.name for x in datasets)], # value=[c.name for c in fields(AutoEvalColumn) if c.displayed_by_default and c.name not in ["Model", "T"] and any(x in c.name for x in datasets)], # interactive=True # ) # with gr.Tab("CyberNLP"): # datasets = ["9", "3"] # en_shown_columns_NLP = gr.CheckboxGroup( # choices=[c.name for c in fields(AutoEvalColumn) if c.name not in ["Model", "T"] and any(x in c.name for x in datasets)], # value=[c.name for c in fields(AutoEvalColumn) if c.displayed_by_default and c.name not in ["Model", "T"] and any(x in c.name for x in datasets)], # interactive=True # ) # with gr.Tab("CyberFQA"): # datasets = ["11", "13"] # en_shown_columns_FQA = gr.CheckboxGroup( # choices=[c.name for c in fields(AutoEvalColumn) if c.name not in ["Model", "T"] and any(x in c.name for x in datasets)], # value=[c.name for c in fields(AutoEvalColumn) if c.displayed_by_default and c.name not in ["Model", "T"] and any(x in c.name for x in datasets)], # interactive=True # ) # with gr.Tab("CyberCDP"): # datasets = ["10"] # en_shown_columns_CDP = gr.CheckboxGroup( # choices=[c.name for c in fields(AutoEvalColumn) if c.name not in ["Model", "T"] and any(x in c.name for x in datasets)], # value=[c.name for c in fields(AutoEvalColumn) if c.displayed_by_default and c.name not in ["Model", "T"] and any(x in c.name for x in datasets)], # interactive=True # ) # dbg(fields(AutoEvalColumn)) # dbg([c.name for c in fields(AutoEvalColumn)]) # dbg([c.name for c in fields(AutoEvalColumn) if any(x in c.name for x in ["Averge"])]) with gr.Tab("CyberKUT"): datasets = ["NetQA", "Embed", "Metric", "CodeQA"] shown_columns_cyberkut = gr.CheckboxGroup(choices=[c.name for c in fields(AutoEvalColumn) if c.name not in ["Model", "T"] and any(x in c.name for x in datasets)], value=[], interactive=True) with gr.Tab("CyberNLP"): datasets = ["Corpus", "CDTier", "NER", "HackerNews"] shown_columns_cybernlp = gr.CheckboxGroup(choices=[c.name for c in fields(AutoEvalColumn) if c.name not in ["Model", "T"] and any(x in c.name for x in datasets)], value=[], interactive=True) with gr.Tab("CyberDSA"): datasets = ["MaliURLs", "CSIC2010", "BETH", "MITRE"] shown_columns_cyberdsa = gr.CheckboxGroup(choices=[c.name for c in fields(AutoEvalColumn) if c.name not in ["Model", "T"] and any(x in c.name for x in datasets)], value=[], interactive=True) # with gr.Accordion("[zh] Select columns to show"): # shown_columns2 = gr.CheckboxGroup( # choices=[c.name for c in fields(AutoEvalColumn) if c.name not in ["Model", "T"]], # value=[c.name for c in fields(AutoEvalColumn) if c.displayed_by_default and c.name not in ["Model", "T"]], # # choices=[c.name for c in fields(AutoEvalColumn) if c.never_hidden] + [c.name for c in fields(AutoEvalColumn) if c.displayed_by_default and not c.never_hidden], # # value=[c.name for c in fields(AutoEvalColumn) if c.never_hidden] + [c.name for c in fields(AutoEvalColumn) if c.displayed_by_default and not c.never_hidden], # label="Select Columns to Display:", # interactive=True, # ) # shown_columns = [en_shown_columns_TAD, en_shown_columns_ACM, en_shown_columns_KUT, en_shown_columns_NLP, en_shown_columns_FQA, en_shown_columns_CDP, shown_columns2] shown_columns = [shown_columns_cyberkut, shown_columns_cybernlp, shown_columns_cyberdsa, shown_columns] # dbg(shown_columns, "before") # with gr.Row(): uncheck_all_button = gr.Button("Uncheck All") uncheck_all_button.click( uncheck_all, inputs=[], outputs=[ *shown_columns, ], ) # with gr.Row(): deleted_models_visibility = gr.Checkbox(value=True, label="Show gated/private/deleted models", interactive=True, visible=False) filter_columns_size = gr.CheckboxGroup( label="Model sizes (in billions of parameters)", # choices=["All"] + list(NUMERIC_INTERVALS.keys()) + ["?"], choices=["All"] + ["?"], value=["All"], interactive=True, elem_id="filter-columns-size", visible=False, ) # # TODO 添加新列的顺序 # leaderboard_table = gr.Dataframe( # value=LEADERBOARD_DF_CP[[c.name for c in fields(AutoEvalColumn) if c.never_hidden] + ["Averge KUT⬆️", "Averge NLP⬆️", "Averge DSA⬆️"] + [c.name for c in fields(AutoEvalColumn) if c.displayed_by_default and not c.never_hidden]], # headers=[[c.name for c in fields(AutoEvalColumn) if c.never_hidden] + ["Averge KUT⬆️", "Averge NLP⬆️", "Averge DSA⬆️"] + [c.name for c in fields(AutoEvalColumn) if c.displayed_by_default and not c.never_hidden]], # datatype=TYPES, # elem_id="leaderboard-table", # interactive=False, # visible=True, # ) # dbg([(c.name, c.never_hidden) for c in fields(AutoEvalColumn)]) # dbg([c.name for c in fields(AutoEvalColumn) if c.never_hidden] + [c.name for c in fields(AutoEvalColumn) if c.displayed_by_default and not c.never_hidden]) leaderboard_table = gr.Dataframe( value=LEADERBOARD_DF_CP[[c.name for c in fields(AutoEvalColumn) if c.never_hidden] + [c.name for c in fields(AutoEvalColumn) if c.displayed_by_default and not c.never_hidden]], headers=[[c.name for c in fields(AutoEvalColumn) if c.never_hidden] + [c.name for c in fields(AutoEvalColumn) if c.displayed_by_default and not c.never_hidden]], datatype=TYPES, elem_id="leaderboard-table", interactive=False, visible=True, # height=800, # min_width=120, ) # dbg(len(LEADERBOARD_DF)) # print(f"[DEBUG]---------- COLS :", COLS) # print(f"[DEBUG]---------- TYPES :", TYPES) # Dummy leaderboard for handling the case when the user uses backspace key hidden_leaderboard_table_for_search = gr.Dataframe( value=LEADERBOARD_DF[COLS], headers=COLS, datatype=TYPES, visible=False, ) search_bar.submit( update_table, inputs=[ hidden_leaderboard_table_for_search, *shown_columns, filter_columns_type, filter_columns_precision, filter_columns_size, deleted_models_visibility, search_bar, ], outputs=leaderboard_table, queue=True, ) for selector in [*shown_columns, filter_columns_type, filter_columns_precision, filter_columns_size, deleted_models_visibility, filter_columns_size]: selector.change( update_table, inputs=[ hidden_leaderboard_table_for_search, *shown_columns, filter_columns_type, filter_columns_precision, filter_columns_size, deleted_models_visibility, search_bar, ], outputs=leaderboard_table, queue=True, ) 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): init_leaderboard() with gr.TabItem("📝 About", elem_id="llm-benchmark-tab-table", id=2): gr.Markdown(LLM_BENCHMARKS_TEXT, elem_classes="markdown-text") 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.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()