Spaces:

debayan
/

ISM2023w

Runtime error

ISM2023w

File size: 8,838 Bytes

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'<div style="display: flex; justify-content: space-around;"><img src="{banner_url}" alt="Banner" style="width: 40vw; min-width: 300px; max-width: 600px;"> </div>'

TITLE = "<html> <head> <style> h1 {text-align: center;} </style> </head> <body> <h1> 🤗 Open Automatic Speech Recognition Leaderboard </b> </body> </html>"

INTRODUCTION_TEXT = "🏆 The ISM2023w Leaderboard ranks and evaluates models \
    on the Hugging Face Spaces. \
    \nWe report the Precision, Recall, Accuracy, Weighted [F1 Score](https://scikit-learn.org/stable/modules/generated/sklearn.metrics.f1_score.html)\n"

#CSV_TEXT = "<p>Please make sure that you upload the CSV in the correct format. Click the link to download a sample CSV!</p> <a href=\"https://huggingface.co/spaces/debayan/ism_2023w/raw/main/example.csv\" download=\"example.csv\">Download CSV</a>"
CSV_TEXT = "<p>Please use this link to upload your csvs and check leaderboards!</p> <a href=\"https://a37fcb50ccc3b4a060.gradio.live\" >https://a37fcb50ccc3b4a060.gradio.live</a>"



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"""
    <html>
    <head>
    <title>{title}</title>
    </head>
    <body>
    {content}
    </body>
    </html>
    """

    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)