bluebench / app.py
jbnayahu's picture
.
bc7425f unverified
import io
import gradio as gr
from gradio_leaderboard import Leaderboard
from apscheduler.schedulers.background import BackgroundScheduler
from src.about import (
CITATION_BUTTON_LABEL,
CITATION_BUTTON_TEXT,
INTRODUCTION_TEXT,
LLM_BENCHMARKS_TEXT,
TITLE,
TITLE_IMAGE,
)
from src.display.css_html_js import custom_css
from src.display.utils import (
BENCHMARK_COLS,
COLS,
AutoEvalColumn,
fields,
)
from src.envs import API, EVAL_RESULTS_PATH, REPO_ID
from src.populate import get_leaderboard_df
def restart_space():
API.restart_space(repo_id=REPO_ID)
LEADERBOARD_DF = get_leaderboard_df(EVAL_RESULTS_PATH, COLS, BENCHMARK_COLS)
def init_leaderboard(dataframe):
if dataframe is None or dataframe.empty:
raise ValueError("Leaderboard DataFrame is empty or None.")
return Leaderboard(
value=dataframe,
datatype=[c.type for c in fields(AutoEvalColumn)],
search_columns=[AutoEvalColumn.model.name],
interactive=False
)
def download_csv():
with io.StringIO() as buffer:
LEADERBOARD_DF.to_csv(buffer, index=False)
return buffer.getvalue()
gui = gr.Blocks(css=custom_css)
with gui:
gr.HTML(TITLE_IMAGE)
gr.HTML(TITLE)
gr.Markdown(INTRODUCTION_TEXT, elem_classes="markdown-text")
with gr.Tabs(elem_classes="tab-buttons") as tabs:
with gr.TabItem("πŸ… LLM Benchmark", elem_id="llm-benchmark-tab-table", id=0):
leaderboard = init_leaderboard(LEADERBOARD_DF.style.highlight_max(color = 'lightgreen', axis=0).data)
with gr.TabItem("πŸ“ About", elem_id="llm-benchmark-tab-table", id=2):
gr.Markdown(LLM_BENCHMARKS_TEXT, elem_classes="markdown-text")
with gr.Row():
with gr.Accordion("πŸ’Ύ Download as CSV", open=False):
download_button = gr.Button("Click to Generate CSV File")
csv_output = gr.File(label="Generated File")
def generate_csv_file():
csv_content = download_csv()
csv_output.visible=True
with open("bluebench.csv", "w", encoding="utf-8") as f:
f.write(csv_content)
return "bluebench.csv"
download_button.click(fn=generate_csv_file, outputs=csv_output)
with gr.Row():
with gr.Accordion("πŸ“™ 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()
gui.queue(default_concurrency_limit=40).launch()