File size: 22,197 Bytes
da92625 979b424 da92625 606c189 da96aa6 da92625 bfb3ae7 da92625 8d7f3c1 da92625 8d7f3c1 da92625 a6bb893 69f5df3 979b424 12e16dd 550d12e da92625 a4bf183 664fbb8 da92625 37b3751 606c189 3d0b0cf 37b3751 2642851 37b3751 da92625 606c189 428cea6 887dd92 da92625 d36c83d ee84fd2 d36c83d 3d0b0cf 9f4c149 d36c83d 5db0911 ee84fd2 f99d80b 606c189 9f4c149 606c189 37b3751 bed7fbb f99d80b 606c189 1062c17 606c189 9f4c149 606c189 bed7fbb 5f0ee8c 37b3751 e19ffb4 5f0ee8c 37b3751 e19ffb4 5f0ee8c 37b3751 bed7fbb f99d80b 606c189 9f4c149 606c189 9f4c149 606c189 9f4c149 606c189 9f4c149 606c189 9f4c149 606c189 5f0ee8c 37b3751 5f0ee8c 37b3751 5f0ee8c 606c189 3d0b0cf 606c189 2f5cc84 887dd92 606c189 3d0b0cf 606c189 8d7f3c1 5f0ee8c 8d7f3c1 5f0ee8c f99d80b 887dd92 4e841d6 8e21015 7fde741 3d0b0cf 7fde741 9ce778c 4e841d6 887dd92 da92625 2a7d8c1 174a614 da92625 174a614 da92625 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 |
import gradio as gr
from gradio_leaderboard import Leaderboard, ColumnFilter, SelectColumns, SearchColumns
import pandas as pd
from apscheduler.schedulers.background import BackgroundScheduler
from huggingface_hub import snapshot_download
from src.about import (
CITATION_BUTTON_LABEL,
CITATION_BUTTON_TEXT,
EVALUATION_QUEUE_TEXT,
INTRODUCTION_TEXT,
LLM_BENCHMARKS_TEXT,
TITLE,
SUB_TITLE,
EXTERNAL_LINKS,
COMING_SOON_TEXT
)
from src.display.css_html_js import custom_css
from src.display.utils import (
BENCHMARK_COLS,
COLS,
EVAL_COLS,
EVAL_TYPES,
AutoEvalColumn,
ModelType,
fields,
WeightType,
Precision
)
from src.envs import API, EVAL_REQUESTS_PATH, EVAL_RESULTS_PATH, QUEUE_REPO, REPO_ID, RESULTS_REPO, TOKEN
from src.populate import get_evaluation_queue_df, get_leaderboard_df, get_model_leaderboard_df
from src.submission.submit import add_new_eval
def restart_space():
API.restart_space(repo_id=REPO_ID)
### Space initialisation
try:
print(EVAL_REQUESTS_PATH)
snapshot_download(
repo_id=QUEUE_REPO, local_dir=EVAL_REQUESTS_PATH, repo_type="dataset", tqdm_class=None, etag_timeout=30, token=TOKEN
)
except Exception:
restart_space()
try:
print(EVAL_RESULTS_PATH)
snapshot_download(
repo_id=RESULTS_REPO, local_dir=EVAL_RESULTS_PATH, repo_type="dataset", tqdm_class=None, etag_timeout=30, token=TOKEN
)
except Exception:
restart_space()
LEADERBOARD_DF = get_leaderboard_df(EVAL_RESULTS_PATH, EVAL_REQUESTS_PATH, COLS, BENCHMARK_COLS)
(
finished_eval_queue_df,
running_eval_queue_df,
pending_eval_queue_df,
) = get_evaluation_queue_df(EVAL_REQUESTS_PATH, EVAL_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)],
select_columns=None,
# SelectColumns(
# default_selection=[c.name for c in fields(AutoEvalColumn) if c.displayed_by_default],
# cant_deselect=[c.name for c in fields(AutoEvalColumn) if c.never_hidden],
# label="Select Columns to Display:",
# ),
# search_columns=None,
# search_columns=[AutoEvalColumn.model.name, AutoEvalColumn.license.name],
search_columns=SearchColumns(primary_column=AutoEvalColumn.model.name, secondary_columns=[],
placeholder="Search by the model name",
label="Searching"),
hide_columns=[c.name for c in fields(AutoEvalColumn) if c.hidden],
filter_columns=None,
# [
# ColumnFilter(AutoEvalColumn.model_type.name, type="checkboxgroup", label="Model types"),
# ColumnFilter(AutoEvalColumn.precision.name, type="checkboxgroup", label="Precision"),
# ColumnFilter(
# AutoEvalColumn.params.name,
# type="slider",
# min=0.01,
# max=150,
# label="Select the number of parameters (B)",
# ),
# ColumnFilter(
# AutoEvalColumn.still_on_hub.name, type="boolean", label="Deleted/incomplete", default=True
# ),
# ],
# bool_checkboxgroup_label="Hide models",
interactive=False,
)
# model_result_path = "./src/results/models_2024-10-07-14:50:12.666068.jsonl"
# model_result_path = "./src/results/models_2024-10-08-03:10:26.811832.jsonl"
# model_result_path = "./src/results/models_2024-10-08-03:25:44.801310.jsonl"
model_result_path = "./src/results/models_2024-10-08-17:39:21.001582.jsonl"
# model_leaderboard_df = get_model_leaderboard_df(model_result_path)
def overall_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)],
select_columns=None,
search_columns=SearchColumns(primary_column=AutoEvalColumn.model.name, secondary_columns=[],
placeholder="Search by the model name",
label="Searching"),
hide_columns=[c.name for c in fields(AutoEvalColumn) if c.hidden],
filter_columns=None,
interactive=False,
)
demo = gr.Blocks(css=custom_css)
with demo:
gr.HTML(TITLE)
gr.HTML(SUB_TITLE)
gr.HTML(EXTERNAL_LINKS)
gr.Markdown(INTRODUCTION_TEXT, elem_classes="markdown-text")
with gr.Tabs(elem_classes="tab-buttons") as tabs:
with gr.TabItem("๐
Overview", elem_id="llm-benchmark-tab-table", id=0):
DESCRIPTION_TEXT = """
Total #models: 52 (Last updated: 2024-10-08)
This page provids a comprehensive overview of model ranks across various dimensions. Models are sorted based on their averaged rank across all dimensions.
(Some missing values are due to the slow or problemtic model responses, and we will update the leaderboard once we have the complete results.)
"""
gr.Markdown(DESCRIPTION_TEXT, elem_classes="markdown-text")
leaderboard = overall_leaderboard(
get_model_leaderboard_df(
model_result_path,
benchmark_cols=[
# AutoEvalColumn.rank_overall.name,
AutoEvalColumn.model.name,
AutoEvalColumn.rank_overall.name,
AutoEvalColumn.rank_math_algebra.name,
AutoEvalColumn.rank_math_geometry.name,
AutoEvalColumn.rank_math_probability.name,
AutoEvalColumn.rank_reason_logical.name,
AutoEvalColumn.rank_reason_social.name,
],
rank_col=[],
)
)
with gr.TabItem("๐ฏ Overall", elem_id="llm-benchmark-tab-table", id=1):
DESCRIPTION_TEXT = """
Overall dimension measures the comprehensive performance of LLMs across diverse tasks.
We start with diverse questions from the widely-used [MT-Bench](https://arxiv.org/abs/2306.05685),
coving a wide range of domains, including writing, roleplay, extraction, reasoning, math, coding, knowledge I (STEM), and knowledge II (humanities/social science).
"""
gr.Markdown(DESCRIPTION_TEXT, elem_classes="markdown-text")
leaderboard = overall_leaderboard(
get_model_leaderboard_df(
model_result_path,
benchmark_cols=[
AutoEvalColumn.rank_overall.name,
AutoEvalColumn.model.name,
AutoEvalColumn.score_overall.name,
AutoEvalColumn.sd_overall.name,
AutoEvalColumn.license.name,
AutoEvalColumn.organization.name,
AutoEvalColumn.knowledge_cutoff.name,
],
rank_col=[AutoEvalColumn.rank_overall.name],
))
with gr.TabItem("๐ข Math", elem_id="math-tab-table", id=2):
DESCRIPTION_TEXT="""
Algebra, Geometry, and Probability are the current three main math domains in the leaderboard.
To mitigate the potential impact of data contimination, we have carefully selected the datasets from various sources.
We prioritize **recent math datasets** and focus on **college and beyond level** math questions.
The current datasets include
[MATH](https://arxiv.org/abs/2103.03874),
[MATH-500](https://github.com/openai/prm800k/tree/main/prm800k/math_splits),
[Omni](https://omni-math.github.io/),
[MathQA](https://arxiv.org/abs/1905.13319),
[MathBench](https://arxiv.org/abs/2405.12209),
[SciBench](https://arxiv.org/abs/2307.10635), and more!
We plan to include more math domains, such as calculus, number theory, and more in the future.
"""
gr.Markdown(DESCRIPTION_TEXT, elem_classes="markdown-text")
# leaderboard = init_leaderboard(LEADERBOARD_DF)
with gr.TabItem("๐งฎ Algebra", elem_id="algebra_subtab", id=0, elem_classes="subtab"):
leaderboard = overall_leaderboard(
get_model_leaderboard_df(
model_result_path,
benchmark_cols=[
AutoEvalColumn.rank_math_algebra.name,
AutoEvalColumn.model.name,
AutoEvalColumn.score_math_algebra.name,
AutoEvalColumn.sd_math_algebra.name,
AutoEvalColumn.license.name,
AutoEvalColumn.organization.name,
AutoEvalColumn.knowledge_cutoff.name,
],
rank_col=[AutoEvalColumn.rank_math_algebra.name],
)
)
with gr.TabItem("๐ Geometry", elem_id="geometry_subtab", id=1, elem_classes="subtab"):
leaderboard = overall_leaderboard(
get_model_leaderboard_df(
model_result_path,
benchmark_cols=[
AutoEvalColumn.rank_math_geometry.name,
AutoEvalColumn.model.name,
AutoEvalColumn.score_math_geometry.name,
AutoEvalColumn.sd_math_geometry.name,
AutoEvalColumn.license.name,
AutoEvalColumn.organization.name,
AutoEvalColumn.knowledge_cutoff.name,
],
rank_col=[AutoEvalColumn.rank_math_geometry.name],
)
)
with gr.TabItem("๐ Probability", elem_id="prob_subtab", id=2, elem_classes="subtab"):
leaderboard = overall_leaderboard(
get_model_leaderboard_df(
model_result_path,
benchmark_cols=[
AutoEvalColumn.rank_math_probability.name,
AutoEvalColumn.model.name,
AutoEvalColumn.score_math_probability.name,
AutoEvalColumn.sd_math_probability.name,
AutoEvalColumn.license.name,
AutoEvalColumn.organization.name,
AutoEvalColumn.knowledge_cutoff.name,
],
rank_col=[AutoEvalColumn.rank_math_probability.name],
)
)
with gr.TabItem("๐ง Reasoning", elem_id="reasonong-tab-table", id=3):
DESCRIPTION_TEXT = """
Reasoning is a broad domain for evaluating LLMs, but traditional tasks like commonsense reasoning have become less effective in differentiating modern LLMs.
We now present two challenging types of reasoning: logical reasoning and social reasoning, both of which present more meaningful and sophisticated ways to assess LLM performance.
For logical reasoning, we leverage datasets from sources such as
[BIG-Bench Hard (BBH)](https://arxiv.org/abs/2210.09261),
[FOLIO](https://arxiv.org/abs/2209.00840),
[LogiQA2.0](https://github.com/csitfun/LogiQA2.0),
[PrOntoQA](https://arxiv.org/abs/2210.01240),
[ReClor](https://arxiv.org/abs/2002.04326),
These cover a range of tasks including deductive reasoning, object counting and tracking, pattern recognition,
temporal reasoning, first-order logic reaosning, etc.
For social reasoning, we collect datasets from
[MMToM-QA (Text-only)](https://arxiv.org/abs/2401.08743),
[BigToM](https://arxiv.org/abs/2306.15448),
[Adv-CSFB](https://arxiv.org/abs/2305.14763),
[SocialIQA](https://arxiv.org/abs/1904.09728),
[NormBank](https://arxiv.org/abs/2305.17008), covering challenging social reasoning tasks,
such as social commonsense reasoning, social normative reasoning, Theory of Mind (ToM) reasoning, etc.
"""
gr.Markdown(DESCRIPTION_TEXT, elem_classes="markdown-text")
with gr.TabItem("๐งฉ Logical", elem_id="logical_subtab", id=0, elem_classes="subtab"):
leaderboard = overall_leaderboard(
get_model_leaderboard_df(
model_result_path,
benchmark_cols=[
AutoEvalColumn.rank_reason_logical.name,
AutoEvalColumn.model.name,
AutoEvalColumn.score_reason_logical.name,
AutoEvalColumn.sd_reason_logical.name,
AutoEvalColumn.license.name,
AutoEvalColumn.organization.name,
AutoEvalColumn.knowledge_cutoff.name,
],
rank_col=[AutoEvalColumn.rank_reason_logical.name],
)
)
with gr.TabItem("๐ฃ๏ธ Social", elem_id="social_subtab", id=1, elem_classes="subtab"):
leaderboard = overall_leaderboard(
get_model_leaderboard_df(
model_result_path,
benchmark_cols=[
AutoEvalColumn.rank_reason_social.name,
AutoEvalColumn.model.name,
AutoEvalColumn.score_reason_social.name,
AutoEvalColumn.sd_reason_social.name,
AutoEvalColumn.license.name,
AutoEvalColumn.organization.name,
AutoEvalColumn.knowledge_cutoff.name,
],
rank_col=[AutoEvalColumn.rank_reason_social.name],
)
)
with gr.TabItem("๐ฌ Science", elem_id="science-table", id=4):
CURRENT_TEXT = """
# Coming soon!
We are working on adding more tasks on scientific domains to the leaderboard. The forthcoming ones are biology, chemistry, and physics.
We have diversely and aggressively collected recent science datasets, including but not limited to
[GPQA](https://arxiv.org/abs/2311.12022),
[JEEBench](https://aclanthology.org/2023.emnlp-main.468/),
[MMLU-Pro](https://arxiv.org/abs/2406.01574),
[OlympiadBench](https://arxiv.org/abs/2402.14008),
[SciBench](https://arxiv.org/abs/2307.10635),
[SciEval](https://arxiv.org/abs/2308.13149).
"""
gr.Markdown(CURRENT_TEXT, elem_classes="markdown-text")
with gr.TabItem("</> Coding", elem_id="coding-table", id=5):
CURRENT_TEXT = """
# Coming soon!
We are working on adding more tasks in coding domains to the leaderboard.
The forthcoming ones focus on Python, Java, and C++, with plans to expand to more languages.
We collect a variety of recent coding datasets, including
[HumanEval](https://huggingface.co/datasets/openai/openai_humaneval),
[MBPP](https://huggingface.co/datasets/google-research-datasets/mbpp),
[HumanEvalFix](https://huggingface.co/datasets/bigcode/humanevalpack),
[newly crawled LeetCode data](https://leetcode.com/problemset/),
filtered code-related queries from [Arena-Hard-Auto](https://github.com/lmarena/arena-hard-auto) and more!
Our efforts also include synthesizing new code-related queries to ensure diversity!
"""
gr.Markdown(CURRENT_TEXT, elem_classes="markdown-text")
with gr.TabItem("๐ About", elem_id="llm-benchmark-tab-table", id=6):
ABOUT_TEXT = """
# About Us
[Decentralized Arena](https://de-arena.maitrix.org/) is an open-source project that automates and scales the evaluation of large language models (LLMs) across various fine-grained dimensions,
developed by reseachers from UCSD, CMU, MBZUAI, [Maitrix.org](https://maitrix.org/) and [LLM360](https://www.llm360.ai/).
Stay tuned for more updates and new features!
## Team members
Yanbin Yin, [Zhen Wang](https://zhenwang9102.github.io/), [Kun Zhou](https://lancelot39.github.io/), Xiangdong Zhang,
[Shibo Hao](https://ber666.github.io/), [Yi Gu](https://www.yigu.page/), Jieyuan Liu, Somanshu Singla, [Tianyang Liu](https://leolty.github.io/),
[Eric P. Xing](https://www.cs.cmu.edu/~epxing/), [Zhengzhong Liu](https://hunterhector.github.io/), [Haojian Jin](https://www.haojianj.in/),
[Zhiting Hu](https://zhiting.ucsd.edu/)
## Contact Us
- Follow us on X, [Maitrix.org](https://twitter.com/MaitrixOrg) and [LLM360](https://twitter.com/llm360)
"""
gr.Markdown(ABOUT_TEXT, elem_classes="markdown-text")
'''
with gr.TabItem("๐ Submit here! ", elem_id="llm-benchmark-tab-table", id=3):
with gr.Column():
with gr.Row():
gr.Markdown(EVALUATION_QUEUE_TEXT, elem_classes="markdown-text")
with gr.Column():
with gr.Accordion(
f"โ
Finished Evaluations ({len(finished_eval_queue_df)})",
open=False,
):
with gr.Row():
finished_eval_table = gr.components.Dataframe(
value=finished_eval_queue_df,
headers=EVAL_COLS,
datatype=EVAL_TYPES,
row_count=5,
)
with gr.Accordion(
f"๐ Running Evaluation Queue ({len(running_eval_queue_df)})",
open=False,
):
with gr.Row():
running_eval_table = gr.components.Dataframe(
value=running_eval_queue_df,
headers=EVAL_COLS,
datatype=EVAL_TYPES,
row_count=5,
)
with gr.Accordion(
f"โณ Pending Evaluation Queue ({len(pending_eval_queue_df)})",
open=False,
):
with gr.Row():
pending_eval_table = gr.components.Dataframe(
value=pending_eval_queue_df,
headers=EVAL_COLS,
datatype=EVAL_TYPES,
row_count=5,
)
with gr.Row():
gr.Markdown("# โ๏ธโจ Submit your model here!", elem_classes="markdown-text")
with gr.Row():
with gr.Column():
model_name_textbox = gr.Textbox(label="Model name")
revision_name_textbox = gr.Textbox(label="Revision commit", placeholder="main")
model_type = gr.Dropdown(
choices=[t.to_str(" : ") for t in ModelType if t != ModelType.Unknown],
label="Model type",
multiselect=False,
value=None,
interactive=True,
)
with gr.Column():
precision = gr.Dropdown(
choices=[i.value.name for i in Precision if i != Precision.Unknown],
label="Precision",
multiselect=False,
value="float16",
interactive=True,
)
weight_type = gr.Dropdown(
choices=[i.value.name for i in WeightType],
label="Weights type",
multiselect=False,
value="Original",
interactive=True,
)
base_model_name_textbox = gr.Textbox(label="Base model (for delta or adapter weights)")
submit_button = gr.Button("Submit Eval")
submission_result = gr.Markdown()
submit_button.click(
add_new_eval,
[
model_name_textbox,
base_model_name_textbox,
revision_name_textbox,
precision,
weight_type,
model_type,
],
submission_result,
)
'''
with gr.Row():
with gr.Accordion("๐ 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()
demo.queue(default_concurrency_limit=40).launch() |