# https://www.gradio.app/playground import gradio as gr import pandas as pd from apscheduler.schedulers.background import BackgroundScheduler from huggingface_hub import snapshot_download from src.display.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, TOKEN, QUEUE_REPO, REPO_ID, RESULTS_REPO from src.populate import get_evaluation_queue_df, get_leaderboard_df from src.submission.submit import add_new_eval # blz ------------------------- import time app_start_time = time.time() """ import builtins # Global counter for file openings file_open_count = 0 def log_file_access(file_name): global file_open_count file_open_count += 1 print(f"File '{file_name}' opened for reading. Total count: {file_open_count}") # Patch built-in open original_open = builtins.open def custom_open(*args, **kwargs): file_name = args[0] if args else kwargs.get('file', 'Unknown') log_file_access(file_name) return original_open(*args, **kwargs) builtins.open = custom_open """ import json def print_first_json_content(directory): for root, dirs, files in os.walk(directory): for file in files: if file.endswith(".json"): json_file_path = os.path.join(root, file) try: with open(json_file_path, 'r') as json_file: data = json.load(json_file) print(f"Contents of {json_file_path}:") print(json.dumps(data, indent=4)) return True except Exception as e: print(f"Error reading {json_file_path}: {e}") return False print("No JSON file found.") return False import os def count_files_in_directory_tree(directory): file_count = 0 for root, dirs, files in os.walk(directory): file_count += len(files) return file_count from huggingface_hub import hf_hub_download, Repository def download_dataset(repo_id, local_dir): # Clone the repository repo_id_full = f"https://huggingface.co/datasets/{repo_id}" repo = Repository(local_dir=local_dir, clone_from=repo_id_full) files_cnt = count_files_in_directory_tree(local_dir) print(f"{local_dir}: {files_cnt}") # print_first_json_content(local_dir) # Alternatively, you can download specific files using hf_hub_download # file_path = hf_hub_download(repo_id, filename="your_file_name") # Usage #download_dataset("https://huggingface.co/datasets/open-llm-leaderboard/requests", "new/requests") #download_dataset("datasets/open-llm-leaderboard/results", "new/results") # blz end ------------------- def restart_space(): API.restart_space(repo_id=REPO_ID, token=TOKEN) try: print(EVAL_REQUESTS_PATH) download_dataset(repo_id=QUEUE_REPO, local_dir=EVAL_REQUESTS_PATH) #snapshot_download( # repo_id=QUEUE_REPO, local_dir=EVAL_REQUESTS_PATH, repo_type="dataset", tqdm_class=None, etag_timeout=30 #) except Exception: restart_space() try: print(EVAL_RESULTS_PATH) download_dataset(repo_id=RESULTS_REPO, local_dir=EVAL_RESULTS_PATH) #snapshot_download( # repo_id=RESULTS_REPO, local_dir=EVAL_RESULTS_PATH, repo_type="dataset", tqdm_class=None, etag_timeout=30 #) except Exception: restart_space() print("after downloading datasets") # BLZ raw_data, original_df = get_leaderboard_df(EVAL_RESULTS_PATH, EVAL_REQUESTS_PATH, COLS, BENCHMARK_COLS) leaderboard_df = original_df.copy() print("after get_leaderboard_df()") # BLZ ( finished_eval_queue_df, running_eval_queue_df, pending_eval_queue_df, ) = get_evaluation_queue_df(EVAL_REQUESTS_PATH, EVAL_COLS) print("after get_evaluation_queue_df()") # BLZ # Searching and filtering def update_table( hidden_df: pd.DataFrame, columns: list, type_query: list, precision_query: str, size_query: list, show_deleted: bool, query: str, recent7: bool, recent14: bool, recent21: bool ): filtered_df = filter_models(hidden_df, type_query, size_query, precision_query, show_deleted, recent7, recent14, recent21) filtered_df = filter_queries(query, filtered_df) df = select_columns(filtered_df, columns) return df def search_table(df: pd.DataFrame, query: str) -> pd.DataFrame: return df[(df[AutoEvalColumn.dummy.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] + [AutoEvalColumn.dummy.name] ] 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, recent7: bool, recent14: bool, recent21: 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] # blz if recent7: filtered_df = df[df[AutoEvalColumn.recent7.name] == True] if recent14: filtered_df = df[df[AutoEvalColumn.recent14.name] == True] if recent21: filtered_df = df[df[AutoEvalColumn.recent21.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("🏅 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(): 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(): deleted_models_visibility = gr.Checkbox( value=False, label="Show gated/private/deleted models", interactive=True ) with gr.Row(): filter_recent7 = gr.Checkbox( value=False, label="Recent (7 days)", interactive=True ) filter_recent14 = gr.Checkbox( value=False, label="Recent (14 days)", interactive=True ) filter_recent21 = gr.Checkbox( value=False, label="Recent (21 days)", 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.dummy.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, column_widths=["2%", "33%"] ) # 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, deleted_models_visibility, search_bar, filter_recent7, filter_recent14, filter_recent21, ], leaderboard_table, ) for selector in [shown_columns, filter_columns_type, filter_columns_precision, filter_columns_size, deleted_models_visibility, filter_recent7, filter_recent14, filter_recent21]: selector.change( update_table, [ hidden_leaderboard_table_for_search, shown_columns, filter_columns_type, filter_columns_precision, filter_columns_size, deleted_models_visibility, search_bar, filter_recent7, filter_recent14, filter_recent21, ], leaderboard_table, queue=True, ) 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, ) # blz -------------------- with gr.Row(): gr.Button("Download", link="/file=output/results.csv") # end blz ------------------- 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, ) # blz --- #import os #os.makedirs('output', exist_ok=True) #with open('output/results.csv', 'w') as f: # f.write('hello world') print(f"app start time {time.time() - app_start_time:.2f} seconds") # blz end ---- scheduler = BackgroundScheduler() scheduler.add_job(restart_space, "interval", seconds=1800) scheduler.start() demo.queue(default_concurrency_limit=40).launch(allowed_paths=["output/"]) # blz allowed paths print(f"app start after demo.queue {time.time() - app_start_time:.2f} seconds")