import gradio as gr import pandas as pd import json import os from utils_display import AutoEvalColumn, fields, make_clickable_model, styled_error, styled_message from datetime import datetime, timezone from pathlib import Path from request_api import make_get_request, make_post_request from io import BytesIO, StringIO LEADERBOARD_PATH = "/home/Bhattacharya/ism_leaderboard/files/leaderboard" # Directory where request by models are stored DIR_OUTPUT_REQUESTS = Path("requested_models") EVAL_REQUESTS_PATH = Path("eval_requests") ########################## # Text definitions # ########################## banner_url = "https://huggingface.co/spaces/debayan/ism_2023w/resolve/main/logo_leaderboard.png" BANNER = f'

' TITLE = "

a37fcb50ccc3b4a060.gradio.live" METRICS_TAB_TEXT = """ Here you will find details about the multi-class classification metrics and datasets reported in our leaderboard. ## Metrics 🎯 Word Error Rate (WER) and Real-Time Factor (RTF) are popular metrics for evaluating the accuracy of speech recognition models by estimating how accurate the predictions from the models are and how fast they are returned. We explain them each below. """ LAST_UPDATED = "Sep 28th 2023" def create_html_page(title, content): html_page = f""" {title} {content} """ return html_page def fetch_leaderboard(phase=1): """ Fetch the leaderboard from the local disk """ params = {"phase": phase} leaderboard_df = make_get_request("fetch-leaderboard", params) csv_data = leaderboard_df.content.decode('utf-8') print(csv_data) # create dataframe from csv data leaderboard_df = pd.read_csv(StringIO(csv_data)) leaderboard_df = leaderboard_df.dropna() return leaderboard_df def process_csv_text(temp_file): if isinstance(temp_file, str): df = pd.read_csv(StringIO(temp_file)) else: try: df = pd.read_csv(temp_file.name) except: return None return df def app(team_name, password, csv_file): # convert csv_file from bytes to file # Check if the team is registered # make get request to get team name from password returned_team_name_json = make_get_request("get-team-name", {"password": password}).json() # handle error by checking if the returned json has an error key if "error" in returned_team_name_json: return create_html_page("Error", "Invalid team name or password.") else: returned_team_name = returned_team_name_json["team_name"] if team_name != returned_team_name: return create_html_page("Error", "Invalid team name or password.") else: print("team name and password are correct") # convert csv_file from bytes to file # read csv file as dataframe #df = pd.read_csv(csv_file.name,encoding='utf-8') # convert tempfile to bytes df = process_csv_text(csv_file) if df is None: return create_html_page("Error", "Invalid CSV file format.") csv_data = df.to_csv(index=False) # upload csv file to the server upload_csv_response = make_post_request("upload", {"password": password}, data = csv_data) # handle error by checking if the returned json has an error key if "error" in upload_csv_response.json(): return create_html_page("Error", upload_csv_response.json()["error"] ) else: return create_html_page("Success", upload_csv_response.json()["message"]) def formatter(x): x = round(x, 2) return x def format_leaderboard(df,sort_by="F1 Score"): # Formats the columns for col in df.columns: if col == "Team Name": # do nothing continue elif col == "Team Members": # do nothing continue else: #if col type if numeric then format it else leave it as it is if df[col].dtype == "float64" or df[col].dtype == "int64": df[col] = df[col].apply(formatter) # For numerical values df.sort_values(by=sort_by, inplace=True, ascending=False) return df leaderboard_df_1 = format_leaderboard(fetch_leaderboard(phase=1)) leaderboard_df_2 = format_leaderboard(fetch_leaderboard(phase=2)) COLS = [c.name for c in fields(AutoEvalColumn)] TYPES = [c.type for c in fields(AutoEvalColumn)] with gr.Blocks() as demo: leaderboard_table_1 = None leaderboard_table_2 = None def button_clicked_phase_1(inputs): leaderboard_df_1 = format_leaderboard(fetch_leaderboard(phase=1)) leaderboard_table_1 = gr.components.Dataframe( value=leaderboard_df_1, datatype=TYPES, max_rows=None, elem_id="leaderboard-table-1", interactive=False, visible=True, ) return leaderboard_table_1 def button_clicked_phase_2(): leaderboard_df_2 = format_leaderboard(fetch_leaderboard(phase=2)) leaderboard_table_2 = gr.components.Dataframe( value=leaderboard_df_2, datatype=TYPES, max_rows=None, elem_id="leaderboard-table-2", interactive=False, visible=True, ) return leaderboard_table_2 #outputs.update(leaderboard_df_2) #inputs.value = leaderboard_df_1 gr.HTML(BANNER, elem_id="banner") gr.Markdown(INTRODUCTION_TEXT, elem_classes="markdown-text") gr.HTML(CSV_TEXT, elem_classes="markdown-text") with gr.Tabs(elem_classes="tab-buttons") as tabs: with gr.TabItem("🏅 Leaderboard Phase 1", elem_id="od-benchmark-tab-table-1", id=0): leaderboard_table_1 = gr.components.Dataframe( value=leaderboard_df_1, datatype=TYPES, max_rows=None, elem_id="leaderboard-table-1", interactive=False, visible=True, ) button = gr.Button("Refresh Leaderboard",interactive=True) button.click(fn=button_clicked_phase_1,outputs=[leaderboard_table_1]) with gr.TabItem("🏅 Leaderboard Phase 2", elem_id="od-benchmark-tab-table-2", id=1): leaderboard_table_2 = gr.components.Dataframe( value=leaderboard_df_2, datatype=TYPES, max_rows=None, elem_id="leaderboard-table-2", interactive=False, visible=True, ) button = gr.Button("Refresh Leaderboard",interactive=True) button.click(fn=button_clicked_phase_2,outputs=[leaderboard_table_2]) with gr.TabItem("📈 Upload CSV", elem_id="od-benchmark-tab-table", id=2): gr.Interface( fn=app, title='CSV Upload and Leaderboard', description='Upload a CSV file and see your team\'s rank on the leaderboard.', layout='vertical', theme='compact', inputs=[ gr.Textbox(label='Team Name'), gr.Textbox(label='Password'), gr.File(file_count= "single", file_types =[".txt",".csv"] ,label='CSV File') # only allow csv files ], outputs=gr.outputs.HTML() ) gr.Markdown(f"Last updated on {LAST_UPDATED}", elem_classes="markdown-text") demo.launch(debug=True,share=True)