Spaces:
Running
Running
File size: 14,462 Bytes
105dc1a 72e365d b9504aa 72e365d 105dc1a 6d9a5f4 72e365d 6d9a5f4 105dc1a 4914e39 105dc1a 6d9a5f4 105dc1a b9504aa 105dc1a 6d9a5f4 105dc1a 6d9a5f4 105dc1a 6d9a5f4 105dc1a 6d9a5f4 105dc1a 6d9a5f4 105dc1a b9504aa 105dc1a 4914e39 105dc1a 72e365d 105dc1a 4914e39 105dc1a 4914e39 105dc1a c67332f 105dc1a | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 | 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 task_keys_star_first():
"""LIBRA Mini datasets (display name ends with ' *') first, then others; order within each group follows datasets_config."""
keys = list(datasets_params.keys())
star = [k for k in keys if datasets_params[k]["name"].endswith(" *")]
rest = [k for k in keys if k not in star]
return star + rest
TASK_TAB_ORDER = task_keys_star_first()
LIBRA_MINI_TASK_KEYS = frozenset(
k for k in datasets_params if datasets_params[k]["name"].endswith(" *")
)
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)
front = ["Model", "Total Score"]
rest = [datasets_params[t]["name"] for t in TASK_TAB_ORDER if datasets_params[t]["name"] in table.columns]
rest += [c for c in table.columns if c not in front and c not in rest]
return table[front + rest]
def make_overall_table_by_tasks_mini(files):
"""Per-task scores for LIBRA Mini datasets only; Total Score = mean over mini tasks."""
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] = {}
mini_raw = []
for dataset in data.keys():
if dataset == "total_score":
continue
if dataset not in LIBRA_MINI_TASK_KEYS:
continue
result_dct[model_name][dataset] = round(data[dataset]["dataset_total_score"] * 100, 1)
mini_raw.append(data[dataset]["dataset_total_score"])
if mini_raw:
result_dct[model_name]["total_score"] = round(float(np.mean(mini_raw)) * 100, 1)
else:
result_dct[model_name]["total_score"] = float("nan")
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)
front = ["Model", "Total Score"]
mini_tab_order = [t for t in TASK_TAB_ORDER if t in LIBRA_MINI_TASK_KEYS]
rest = [datasets_params[t]["name"] for t in mini_tab_order if datasets_params[t]["name"] in table.columns]
rest += [c for c in table.columns if c not in front and c not in rest]
return table[front + rest]
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)
front = ["Model", "Total Score"]
length_cols = [c for c in LENGTHS if c in table.columns]
return table[front + length_cols]
def make_overall_table_by_lengths_mini(files):
"""Aggregate by context length using only LIBRA Mini datasets (names ending with ' *')."""
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] = {}
mini_scores = []
for dataset in data.keys():
if dataset == "total_score":
continue
if dataset not in LIBRA_MINI_TASK_KEYS:
continue
block = data[dataset]
if "dataset_total_score" in block:
mini_scores.append(block["dataset_total_score"])
for length in block.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(block[length])
if mini_scores:
result_dct[model_name]["total_score"] = float(np.mean(mini_scores))
else:
result_dct[model_name]["total_score"] = float("nan")
for model_name in result_dct.keys():
for length in list(result_dct[model_name].keys()):
if length == "total_score":
continue
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(round(result_dct[model_name][key] * 100, 1))
else:
results[key].append(result_dct[model_name][key])
table = pd.DataFrame(results).sort_values(["Total Score"], ascending=False)
front = ["Model", "Total Score"]
length_cols = [c for c in LENGTHS if c in table.columns]
return table[front + length_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"))
task_block = result.get(tab_name)
if not task_block:
results["Dataset Total Score"].append("-")
for length in LENGTHS:
if length == "dataset_total_score":
continue
results[length].append("-")
continue
for length in LENGTHS:
if length in task_block.keys():
if length == "dataset_total_score":
results["Dataset Total Score"].append(round(task_block[length] * 100, 1))
continue
results[length].append(round(task_block[length] * 100, 1))
else:
results[length].append("-")
df = pd.DataFrame(results)
df["_sort_key"] = pd.to_numeric(df["Dataset Total Score"], errors="coerce").fillna(-1)
return df.sort_values("_sort_key", ascending=False).drop(columns=["_sort_key"])
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 length (LIBRA-mini)", id=0):
df = make_overall_table_by_lengths_mini(files)
gr.Dataframe(
headers=[
"Model",
] + LENGTHS,
datatype=[
"markdown",
"str",
"str",
"str",
"str",
"str",
"str",
"str",
],
value=df,
elem_id="arena_leaderboard_dataframe",
max_height=700,
wrap=True,
)
with gr.Tab("Results by length (LIBRA)", id=1):
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",
max_height=700,
wrap=True,
)
with gr.Tab("Results by task (LIBRA-mini)", id=2):
df = make_overall_table_by_tasks_mini(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",
max_height=700,
wrap=False,
)
with gr.Tab("Results by task (LIBRA)", id=3):
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",
max_height=700,
wrap=False,
)
for tab_id, tab_name in enumerate(TASK_TAB_ORDER):
df = load_model(files, tab_name)
with gr.Tab(datasets_params[tab_name]["name"], id=tab_id + 4):
gr.Dataframe(
headers=[
"Model",
] + LENGTHS,
datatype=[
"markdown",
"str",
"str",
"str",
"str",
"str",
"str",
"str",
],
value=df,
elem_id="arena_leaderboard_dataframe",
max_height=700,
wrap=True,
)
with gr.Tab("Description", id=len(TASK_TAB_ORDER) + 4):
desc_md = make_model_desc_md()
gr.Markdown(desc_md, elem_id="leaderboard_markdown")
return [md_1]
def build_demo(files):
with gr.Blocks(title="LIBRA leaderboard") as demo:
build_leaderboard_tab(files)
return demo
if __name__ == "__main__":
files = os.listdir("results")
demo = build_demo(files)
text_size = gr.themes.sizes.text_lg
demo.launch(theme=gr.themes.Base(text_size=text_size), share=False)
|