|
import requests |
|
from bs4 import BeautifulSoup |
|
import pandas as pd |
|
import gradio as gr |
|
import io |
|
|
|
BASE_URL = "https://scale.com/leaderboard" |
|
|
|
LEADERBOARDS = { |
|
"Coding": "/coding", |
|
"Adversarial Robustness": "/adversarial_robustness", |
|
"Instruction Following": "/instruction_following", |
|
"Math": "/math" |
|
} |
|
|
|
def scrape_leaderboard(leaderboard): |
|
url = BASE_URL + LEADERBOARDS[leaderboard] |
|
|
|
response = requests.get(url) |
|
soup = BeautifulSoup(response.content, 'html.parser') |
|
|
|
leaderboard_div = soup.find('div', class_='flex flex-col gap-4 sticky top-20') |
|
|
|
if not leaderboard_div: |
|
raise ValueError("Leaderboard div not found. The page structure might have changed.") |
|
|
|
table = leaderboard_div.find('table', class_='w-full caption-bottom text-sm') |
|
|
|
if not table: |
|
raise ValueError("Leaderboard table not found within the div.") |
|
|
|
data = [] |
|
for row in table.find('tbody').find_all('tr'): |
|
cols = row.find_all('td') |
|
rank = cols[0].find('div', class_='flex').text.strip().split()[0] |
|
model = cols[0].find('a').text.strip() |
|
score = cols[1].text.strip() |
|
confidence = cols[2].text.strip() |
|
data.append([rank, model, score, confidence]) |
|
|
|
df = pd.DataFrame(data, columns=['Rank', 'Model', 'Score', '95% Confidence']) |
|
return df |
|
|
|
def update_leaderboard(leaderboard): |
|
try: |
|
df = scrape_leaderboard(leaderboard) |
|
return df, create_interactive_table(df) |
|
except Exception as e: |
|
return None, f"An error occurred: {str(e)}" |
|
|
|
def create_interactive_table(df): |
|
html = f""" |
|
<script src="https://cdn.jsdelivr.net/npm/ag-grid-community/dist/ag-grid-community.min.js"></script> |
|
<div id="myGrid" style="height: 500px; width: 100%;" class="ag-theme-alpine"></div> |
|
<script> |
|
var gridOptions = {{ |
|
columnDefs: [ |
|
{{field: "Rank", sortable: true, filter: true}}, |
|
{{field: "Model", sortable: true, filter: true}}, |
|
{{field: "Score", sortable: true, filter: true}}, |
|
{{field: "95% Confidence", sortable: true, filter: true}} |
|
], |
|
rowData: {df.to_dict(orient='records')}, |
|
defaultColDef: {{ |
|
flex: 1, |
|
minWidth: 100, |
|
resizable: true, |
|
}}, |
|
domLayout: 'autoHeight' |
|
}}; |
|
|
|
document.addEventListener('DOMContentLoaded', function() {{ |
|
var gridDiv = document.querySelector('#myGrid'); |
|
new agGrid.Grid(gridDiv, gridOptions); |
|
}}); |
|
</script> |
|
""" |
|
return html |
|
|
|
def export_to_excel(df): |
|
if df is not None: |
|
output = io.BytesIO() |
|
with pd.ExcelWriter(output, engine='openpyxl') as writer: |
|
df.to_excel(writer, index=False, sheet_name='Leaderboard') |
|
output.seek(0) |
|
return output |
|
return None |
|
|
|
|
|
with gr.Blocks() as iface: |
|
gr.Markdown("# Scale AI Leaderboard Viewer") |
|
with gr.Row(): |
|
dropdown = gr.Dropdown(choices=list(LEADERBOARDS.keys()), label="Select Leaderboard", value="Coding") |
|
export_button = gr.Button("Export to Excel") |
|
|
|
table_output = gr.HTML() |
|
df_state = gr.State() |
|
|
|
def on_load(): |
|
df, html = update_leaderboard("Coding") |
|
return df, html |
|
|
|
dropdown.change(update_leaderboard, inputs=[dropdown], outputs=[df_state, table_output]) |
|
export_button.click(export_to_excel, inputs=[df_state], outputs=[gr.File(label="Download Excel")]) |
|
|
|
iface.load(on_load, outputs=[df_state, table_output]) |
|
|
|
|
|
iface.launch() |
