Spaces:
Running
Running
File size: 7,756 Bytes
1ddbee0 0fb3ba6 1ddbee0 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 |
import os
import json
import gradio as gr
import pandas as pd
import numpy as np
from collections import defaultdict
LENGTHS = ["dataset_total_score", "4k", "8k", "16k", "32k", "64k", "128k"]
datasets_params = json.load(open("datasets_config.json", "r"))
TASKS = datasets_params.keys()
def make_default_md():
leaderboard_md = f"""
🏅 LIBRA LeaderBoard
| [GitHub](https://github.com/ai-forever/LIBRA) | [Datasets](https://huggingface.co/datasets/ai-forever/LIBRA) |
"""
return leaderboard_md
def make_model_desc_md():
with open("docs/description.md", "r") as f:
description = f.read()
return description
def make_overall_table_by_tasks(files):
results = defaultdict(list)
result_dct = {}
for file in files:
if not file.endswith("json"): continue
path = "results/" + file
data = json.load(open(path))
model_name = file.split('/')[-1].split(".json")[0]
result_dct[model_name] = {}
for dataset in data.keys():
if dataset == "total_score":
result_dct[model_name][dataset] = round(data[dataset] * 100, 1)
continue
result_dct[model_name][dataset] = round(data[dataset]["dataset_total_score"] * 100, 1)
for file in files:
if not file.endswith("json"): continue
model_name = file.split('/')[-1].split(".json")[0]
results['Model'].append(model_name)
for key in result_dct[model_name].keys():
if key == "total_score":
results["Total Score"].append(result_dct[model_name][key])
else:
results[datasets_params[key]["name"]].append(result_dct[model_name][key])
table = pd.DataFrame(results).sort_values(['Total Score'], ascending=False)
cols = table.columns.tolist()
cols = [cols[0]] + [cols[22]] + cols[1:22]
return table[cols]
def make_overall_table_by_lengths(files):
results = defaultdict(list)
result_dct = {}
for file in files:
if not file.endswith("json"): continue
path = "results/" + file
data = json.load(open(path))
model_name = file.split('/')[-1].split(".json")[0]
result_dct[model_name] = {}
for dataset in data.keys():
if dataset == "total_score":
result_dct[model_name][dataset] = data[dataset]
continue
for length in data[dataset].keys():
if length == "dataset_total_score": continue
if length not in result_dct[model_name]:
result_dct[model_name][length] = []
result_dct[model_name][length].append(data[dataset][length])
for model_name in result_dct.keys():
for length in result_dct[model_name].keys():
result_dct[model_name][length] = round(np.mean(result_dct[model_name][length]) * 100, 1)
for file in files:
if not file.endswith("json"): continue
model_name = file.split('/')[-1].split(".json")[0]
results['Model'].append(model_name)
for key in result_dct[model_name].keys():
if key == "total_score":
results["Total Score"].append(result_dct[model_name][key])
else:
results[key].append(result_dct[model_name][key])
table = pd.DataFrame(results).sort_values(['Total Score'], ascending=False)
cols = table.columns.tolist()
cols = [cols[0]] + [cols[7]] + cols[1:7]
return table[cols]
def load_model(files, tab_name):
results = defaultdict(list)
for file in files:
if not file.endswith("json"): continue
model_name = file.split('/')[-1].split(".json")[0]
results['Model'].append(model_name)
result = json.load(open("results/" + file, "r"))
for length in LENGTHS:
if length in result[tab_name].keys():
if length == "dataset_total_score":
results["Dataset Total Score"].append(round(result[tab_name][length] * 100, 1))
continue
results[length].append(round(result[tab_name][length] * 100, 1))
else:
results[length].append("-")
return pd.DataFrame(results).sort_values(['Dataset Total Score'], ascending=False)
def build_leaderboard_tab(files):
default_md = make_default_md()
md_1 = gr.Markdown(default_md, elem_id="leaderboard_markdown")
with gr.Tabs() as tabs:
with gr.Tab("Results by Lengths", id=0):
df = make_overall_table_by_lengths(files)
gr.Dataframe(
headers=[
"Model",
] + LENGTHS,
datatype=[
"markdown",
"str",
"str",
"str",
"str",
"str",
"str",
"str",
],
value=df,
elem_id="arena_leaderboard_dataframe",
height=700,
wrap=True,
)
with gr.Tab("Results by Tasks", id=1):
df = make_overall_table_by_tasks(files)
gr.Dataframe(
headers=[
"Model",
] + LENGTHS,
datatype=[
"markdown",
"str",
"str",
"str",
"str",
"str",
"str",
"str",
"str",
"str",
"str",
"str",
"str",
"str",
"str",
"str",
"str",
"str",
"str",
"str",
"str",
"str",
"str"
],
value=df,
elem_id="arena_leaderboard_dataframe",
height=700,
wrap=False,
)
for tab_id, tab_name in enumerate(TASKS):
df = load_model(files, tab_name)
with gr.Tab(datasets_params[tab_name]["name"], id=tab_id+2):
gr.Dataframe(
headers=[
"Model",
] + LENGTHS,
datatype=[
"markdown",
"str",
"str",
"str",
"str",
"str",
"str",
"str",
],
value=df,
elem_id="arena_leaderboard_dataframe",
height=700,
wrap=True,
)
with gr.Tab("Description", id=tab_id + 3):
desc_md = make_model_desc_md()
gr.Markdown(desc_md, elem_id="leaderboard_markdown")
return [md_1]
def build_demo(files):
text_size = gr.themes.sizes.text_lg
with gr.Blocks(title="LIBRA leaderboard",
theme=gr.themes.Base(text_size=text_size)) as demo:
build_leaderboard_tab(files)
return demo
if __name__ == "__main__":
files = os.listdir("results")
demo = build_demo(files)
demo.launch(share=False)
|