leaderboard / app.py
ChrisWilson's picture
Update app.py
b8c0a50 verified
import gradio as gr
import pandas as pd
import matplotlib.pyplot as plt
from dataset_util import load_data, get_num_rows
import subnet_util
import datetime
import typing
import indexing_util
import weights
from io import BytesIO
FONT = """<link href="https://fonts.cdnfonts.com/css/intersect-c-brk" rel="stylesheet">"""
TITLE_FONT = """<link href="https://fonts.cdnfonts.com/css/promova" rel="stylesheet">"""
TITLE = """ <h1 align = "center" id = "space-title" class = "intersect"> D3 Subnet Leaderboard</h1> """
DESCRIPTION = """<marquee><h3 align= "center"> The D3 Subnet, standing for Decentralized Distributed Data Scraping subnet, plays a crucial role in the advancement of artificial intelligence by ensuring ample training data for all Bittensor AI networks. </h3></marquee>"""
IMAGE = """<a href="https://discord.com/channels/799672011265015819/1161764869280903240" target="_blank"><img src="https://github.com/gitphantomman/d3_subnet/blob/main/docs/d3.png?raw=true" alt="D3 Subnet" style="margin: auto; width: 20%; border: 0;" /></a>"""
last_refresh = None
demo = gr.Blocks(css="""
.intersect {font-family: 'Intersect C BRK', sans-serif; font-size:40px}
.promova {font-family: 'Promova', sans-serif; font-size:40px}
""")
twitter_text_dataset = load_data("bittensor-dataset/twitter-text-dataset")
twitter_text_num_rows = get_num_rows(twitter_text_dataset)
twitter_image_dataset = load_data("bittensor-dataset/twitter-image-dataset")
twitter_image_num_rows = get_num_rows(twitter_image_dataset)
tao_price = subnet_util.get_tao_price()
(subtensor, metagraph) = subnet_util.get_subtensor_and_metagraph()
last_refresh = datetime.datetime.now()
miners_data = subnet_util.get_subnet_data(subtensor, metagraph)
# url = "http://127.0.0.1:8001/"
# try:
# response = requests.get(url)
# response_body = response.json()
# twitter_text_num_rows = response_body['twitter_text_rows']
# twitter_image_num_rows = response_body['twitter_image_rows']
# except:
# twitter_text_num_rows = 0
# twitter_image_num_rows = 0
# bt.logging.error("Could not connect to the API")
daily_indexing_data = indexing_util.get_all(indexing_util.daily_indexing)
daily_df = pd.DataFrame(daily_indexing_data, columns=['Date', 'Value'])
daily_df['Date'] = pd.to_datetime(daily_df['Date'].str.decode('utf-8'))
daily_df['Value'] = daily_df['Value'].astype(int)
hotkey_indexing_data = indexing_util.get_all(indexing_util.hotkey_indexing)
hotkey_df = pd.DataFrame(hotkey_indexing_data, columns=['Hotkey', 'Value'])
hotkey_df['Hotkey'] = hotkey_df['Hotkey'].str.decode('utf-8')
hotkey_df['Value'] = hotkey_df['Value'].astype(int)
# Filter hotkey_df to only show value is over than 1000
hotkey_df = hotkey_df[hotkey_df['Value'] > 1000]
# Sort hotkey_df by Value
hotkey_df = hotkey_df.sort_values(by='Value', ascending=False)
hotkey_daily_indexing_data = indexing_util.get_all(indexing_util.hotkey_daily_indexing)
hotkey_daily_df = pd.DataFrame(hotkey_daily_indexing_data, columns=['Hotkey_Date', 'Value'])
hotkey_daily_df_= pd.DataFrame()
hotkey_daily_df_['Hotkey'] = hotkey_daily_df['Hotkey_Date'].str.decode('utf-8').str.split(' ').str[0]
hotkey_daily_df_['Date'] = hotkey_daily_df['Hotkey_Date'].str.decode('utf-8').str.split(' ').str[1]
hotkey_daily_df_['Value'] = hotkey_daily_df['Value'].astype(int)
dalily_df_max = daily_df['Value'].max()
hotkey_df_max = hotkey_df['Value'].max()
hotkey_daily_df_max = hotkey_daily_df_['Value'].max()
# print(hotkey_daily_df_)
weight_list, miners = weights.validator_info()
print(weight_list)
def leaderboard_data(
# show_stale: bool,
# scores: typing.Dict[int, typing.Dict[str, typing.Optional[float | str]]],
# competition_id: str,
):
value = [
[
c.hotkey[0:8] + "...",
c.uid,
c.url,
c.block,
]
for c in miners_data
# if c.incentive and c.url[0:8] == "https://"
]
return value
def validator_data():
value = [
[
c['uid'],
"{:,}".format(c['validator_stake']) + " τ",
c['vtrust'],
] + [
c ['miners_weight'].get(miner[0], 0)
for miner in miners
]
for c in weight_list
]
return value
with demo:
gr.HTML(FONT)
gr.HTML(TITLE_FONT)
gr.HTML(TITLE)
gr.HTML(IMAGE)
gr.HTML(DESCRIPTION)
with gr.Tabs():
with gr.Accordion("Dataset Stats"):
with gr.Row():
with gr.Column(scale=1):
gr.HTML(f"<h2 align = 'center' style = 'font-size: 25px' >Current Size of Text Dataset: <span style = 'font-size: 30px; color: green;'>{twitter_text_num_rows}</span></h2>")
with gr.Column(scale=1):
gr.HTML(f"<h2 align = 'center' style = 'font-size: 25px' >Current Size of Image Dataset: <span style = 'font-size: 30px; color: green;'>{twitter_image_num_rows}</span></h2>")
gr.HTML("<br/>")
with gr.Accordion("Subnet Stats"):
gr.HTML(f"""<h2 align = 'center' class="promova" style = 'font-size: 35px;' > Miner Stats</h2>""")
gr.LinePlot(
daily_df,
x="Date",
y="Value",
title="Daliy scraped data amount",
# color="Date",
# tooltip=["Date", "Value"],
y_lim=[0, dalily_df_max * 1.5],
x_title="Date",
y_title="Amount of data scraped",
height=500,
width=1000,
scale=5,
# color="Value",
color_legend_position="top",
# elem_classes="daily_scraped_data",
)
gr.BarPlot(
hotkey_df,
x="Hotkey",
y="Value",
title="Scraped data amount of each Miner",
# color="Date",
# tooltip=["Hotkey", "Value"],
y_lim=[0, hotkey_df_max * 1.5],
x_title="Date",
y_title="Amount of data scraped",
height=500,
width=1000,
scale=5,
# color="Value",
x_label_angle=-30,
color_legend_position="top",
# elem_classes="daily_scraped_data",
)
gr.ScatterPlot(
hotkey_daily_df_,
x="Date",
y="Value",
title="Daily scraped data amount of each Miner",
# color="Date",
# tooltip=["Hotkey"],
y_lim=[0, hotkey_daily_df_max * 1.5],
x_title="Date",
y_title="Amount of data scraped",
height=500,
width=1000,
scale=5,
# color="Hotkey",
x_label_angle=-30,
color_legend_position="top",
# elem_classes="daily_scraped_data",
)
gr.HTML("<br/>")
with gr.Tab(label="Miners Data"):
class_denominator = sum(
miners_data[i].incentive #TODO: emssion to incentive
for i in range(0, min(10, len(miners_data)))
if miners_data[i].incentive
)
if class_denominator == 0:
class_values = {
f"(uid={miners_data[i].uid}, hotkey={miners_data[i].hotkey[0:8]}) - {miners_data[i].url} · ${round(miners_data[i].emission * tao_price, 2):,}{round(miners_data[i].emission, 2):,})": miners_data[i].incentive + 1 / (class_denominator + 1)
for i in range(0, min(10, len(miners_data)))
}
else:
class_values = {
f"(uid={miners_data[i].uid}, hotkey={miners_data[i].hotkey[0:8]}) - {miners_data[i].url} · ${round(miners_data[i].emission * tao_price, 2):,}{round(miners_data[i].emission, 2):,})": miners_data[i].incentive / class_denominator
for i in range(0, min(10, len(miners_data)))
# if miners_data[i].incentive
}
gr.Label(
label="Top 10 Miners",
value=class_values,
num_top_classes=10,
)
# miner_table = gr.components.Dataframe(
# value=miners_data
# )
gr.HTML("<br/>")
with gr.Accordion("Miner stats"):
gr.HTML(
f"""<h3>{last_refresh.strftime("refreshed at %H:%M on %Y-%m-%d")}</h3>"""
)
# with gr.Tabs():
# for entry in miners_data:
# name = f"uid={entry.uid} : commit={entry.commit[0:8]} : url={entry.url}"
# with gr.Tab(name):
# gr.Chatbot()
leaderboard_table = gr.components.Dataframe(
value=leaderboard_data(),
headers = [
"Hotkey",
"UID",
"Url",
"Block",
],
datatype=[
"markdown",
"number",
"markdown",
"number",
],
elem_id="leaderboard_table",
interactive=False,
visible=True,
)
gr.HTML("<br/>")
with gr.Accordion("Validator stats"):
gr.HTML(
f"""<h3>{last_refresh.strftime("refreshed at %H:%M on %Y-%m-%d")}</h3>"""
)
# with gr.Tabs():
# for entry in miners_data:
# name = f"uid={entry.uid} : commit={entry.commit[0:8]} : url={entry.url}"
# with gr.Tab(name):
# gr.Chatbot()
leaderboard_table = gr.components.Dataframe(
value=validator_data(),
headers = [
"UID",
"Stake",
"V-Trust"
] + [
f"Miner-{miner[0]} (Incentive: {miner[1]})"
for miner in miners
],
datatype=[
"number",
"number",
"number"
] + [
"number"
for miner in miners
],
elem_id="leaderboard_table",
interactive=False,
visible=True,
)
demo.launch()