File size: 5,006 Bytes
193db9d 973519b 55d797c 193db9d 55d797c 0bab47c 193db9d 55d797c 3b39b49 9756440 55d797c 193db9d 633b045 973519b 55d797c 973519b 55d797c 973519b 193db9d 0bab47c 193db9d 973519b 55d797c 193db9d 55d797c 193db9d 55d797c 22e8b31 55d797c 22e8b31 55d797c 193db9d 22e8b31 55d797c 193db9d 22e8b31 55d797c 193db9d 0bab47c 193db9d 22e8b31 55d797c 193db9d 0bab47c 193db9d 22e8b31 55d797c 9756440 193db9d 22e8b31 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 |
import datasets
import gradio as gr
from apscheduler.schedulers.background import BackgroundScheduler
from huggingface_hub import snapshot_download
from loguru import logger
import populate
from about import LEADERBOARD_INTRODUCTION_TEXT, LEADERBOARD_TITLE
from app_configs import DEFAULT_SELECTIONS, THEME
from components.quizbowl.bonus import BonusInterface
from components.quizbowl.tossup import TossupInterface
from display.css_html_js import fonts_header, js_head, leaderboard_css
from display.custom_css import css_bonus, css_pipeline, css_tossup
from display.guide import BUILDING_MARKDOWN, GUIDE_MARKDOWN, QUICKSTART_MARKDOWN
from display.utils import AutoEvalColumn, fields
# Constants
from envs import (
API,
EVAL_REQUESTS_PATH,
EVAL_RESULTS_PATH,
LEADERBOARD_REFRESH_INTERVAL,
PLAYGROUND_DATASET_NAMES,
QUEUE_REPO,
REPO_ID,
RESULTS_REPO,
SERVER_REFRESH_INTERVAL,
)
from workflows import factory
from workflows.configs import AVAILABLE_MODELS
def restart_space():
API.restart_space(repo_id=REPO_ID)
def download_dataset_snapshot(repo_id, local_dir):
try:
logger.info(f"Downloading dataset snapshot from {repo_id} to {local_dir}")
snapshot_download(
repo_id=repo_id,
local_dir=local_dir,
repo_type="dataset",
tqdm_class=None,
)
except Exception as e:
logger.error(f"Error downloading dataset snapshot from {repo_id} to {local_dir}: {e}. Restarting space.")
restart_space()
download_dataset_snapshot(QUEUE_REPO, EVAL_REQUESTS_PATH)
def fetch_leaderboard_df():
logger.info("Leaderboard fetched...")
download_dataset_snapshot(RESULTS_REPO, EVAL_RESULTS_PATH)
return populate.get_leaderboard_df(EVAL_RESULTS_PATH)
def load_dataset(mode: str):
if mode == "tossup":
ds = datasets.load_dataset(PLAYGROUND_DATASET_NAMES["tossup"], split="eval")
ds = ds.filter(lambda x: x["qid"].split("-")[2] == "1" and int(x["qid"].split("-")[3]) <= 10)
elif mode == "bonus":
ds = datasets.load_dataset(PLAYGROUND_DATASET_NAMES["bonus"], split="eval")
ds = ds.filter(lambda x: x["qid"].split("-")[2] == "1" and int(x["qid"].split("-")[3]) <= 10)
else:
raise ValueError(f"Invalid mode: {mode}")
return ds
def get_default_tab_id(request: gr.Request):
logger.info(f"Request: {request}")
tab_key_value = request.query_params.get("tab", "tossup")
return gr.update(selected=tab_key_value)
if __name__ == "__main__":
scheduler = BackgroundScheduler()
scheduler.add_job(restart_space, "interval", seconds=SERVER_REFRESH_INTERVAL)
scheduler.start()
css = css_pipeline + css_tossup + css_bonus + leaderboard_css
head = fonts_header + js_head
tossup_ds = load_dataset("tossup")
bonus_ds = load_dataset("bonus")
with gr.Blocks(
css=css,
head=head,
theme=THEME,
title="Quizbowl Bot",
) as demo:
with gr.Row():
gr.Markdown("## Welcome to Quizbowl Bot! This is a tool for creating and testing quizbowl agents.")
with gr.Tabs() as gtab:
with gr.Tab("🛎️ Tossup Agents", id="tossup"):
defaults = DEFAULT_SELECTIONS["tossup"] | {
"init_workflow": factory.create_simple_qb_tossup_workflow(),
}
tossup_interface = TossupInterface(demo, tossup_ds, AVAILABLE_MODELS, defaults)
with gr.Tab("🙋🏻♂️ Bonus Round Agents", id="bonus"):
defaults = DEFAULT_SELECTIONS["bonus"] | {
"init_workflow": factory.create_simple_qb_bonus_workflow(),
}
bonus_interface = BonusInterface(demo, bonus_ds, AVAILABLE_MODELS, defaults)
with gr.Tab("🏅 Leaderboard", elem_id="llm-benchmark-tab-table", id="leaderboard"):
leaderboard_timer = gr.Timer(LEADERBOARD_REFRESH_INTERVAL)
gr.Markdown("<a id='leaderboard' href='#leaderboard'>QANTA Leaderboard</a>")
gr.Markdown(LEADERBOARD_INTRODUCTION_TEXT)
refresh_btn = gr.Button("🔄 Refresh")
leaderboard_table = gr.Dataframe(
value=fetch_leaderboard_df,
every=leaderboard_timer,
headers=[c.name for c in fields(AutoEvalColumn)],
datatype=[c.type for c in fields(AutoEvalColumn)],
elem_id="leaderboard-table",
interactive=False,
visible=True,
)
refresh_btn.click(fn=fetch_leaderboard_df, inputs=[], outputs=leaderboard_table)
with gr.Tab("❓ Help", id="help"):
with gr.Row():
with gr.Column():
gr.Markdown(QUICKSTART_MARKDOWN)
with gr.Column():
gr.Markdown(BUILDING_MARKDOWN)
demo.queue(default_concurrency_limit=40).launch()
|