File size: 32,364 Bytes
54f3112 f5bfcdc 54f3112 e9e5051 f5bfcdc 54f3112 b7f2473 54f3112 43a97f8 54f3112 98e0fb3 54f3112 43a97f8 98e0fb3 43a97f8 54f3112 43a97f8 54f3112 43a97f8 54f3112 f5bfcdc 43a97f8 255a4d2 43a97f8 255a4d2 43a97f8 255a4d2 43a97f8 255a4d2 43a97f8 255a4d2 43a97f8 255a4d2 43a97f8 255a4d2 54f3112 43a97f8 54f3112 43a97f8 54f3112 43a97f8 255a4d2 43a97f8 54f3112 43a97f8 255a4d2 43a97f8 a7b9847 43a97f8 255a4d2 a7b9847 43a97f8 a7b9847 255a4d2 43a97f8 a7b9847 43a97f8 255a4d2 43a97f8 255a4d2 54f3112 cde8217 c4d90ef cde8217 20e4287 cde8217 467aefa 20e4287 c4d90ef 54f3112 c4d90ef 54f3112 20e4287 cde8217 54f3112 20e4287 cde8217 20e4287 cde8217 20e4287 04a0fb2 20e4287 04a0fb2 20e4287 04a0fb2 20e4287 43a97f8 20e4287 43a97f8 62dd8c2 20e4287 cde8217 20e4287 cde8217 43a97f8 20e4287 54f3112 20e4287 c4d90ef 20e4287 cde8217 43a97f8 04a0fb2 54f3112 04a0fb2 54f3112 04a0fb2 54f3112 04a0fb2 54f3112 43a97f8 20e4287 54f3112 43a97f8 62dd8c2 54f3112 cde8217 54f3112 cde8217 43a97f8 54f3112 c4d90ef 54f3112 c4d90ef 54f3112 c4d90ef 54f3112 43a97f8 f5bfcdc 43a97f8 54f3112 c4d90ef 54f3112 6379863 54f3112 43a97f8 54f3112 43a97f8 a7b9847 43a97f8 54f3112 43a97f8 54f3112 43a97f8 54f3112 43a97f8 cde8217 43a97f8 54f3112 43a97f8 54f3112 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 |
__all__ = ['block', 'make_clickable_model', 'make_clickable_user', 'get_submissions']
import gradio as gr
import pandas as pd
import json
import pdb
import tempfile
import re
from constants import *
from src.auto_leaderboard.model_metadata_type import ModelType
global data_component, filter_component
def validate_model_size(s):
pattern = r'^\d+B$|^-$'
if re.match(pattern, s):
return s
else:
return '-'
def upload_file(files):
file_paths = [file.name for file in files]
return file_paths
def prediction_analyse(prediction_content):
# pdb.set_trace()
predictions = prediction_content.split("\n")
# 读取 ground_truth JSON 文件
with open("./file/SEED-Bench-1.json", "r") as file:
ground_truth_data = json.load(file)["questions"]
# 将 ground_truth 数据转换为以 question_id 为键的字典
ground_truth = {item["question_id"]: item for item in ground_truth_data}
# 初始化结果统计字典
results = {i: {"correct": 0, "total": 0} for i in range(1, 13)}
# 遍历 predictions，计算每个 question_type_id 的正确预测数和总预测数
for prediction in predictions:
# pdb.set_trace()
prediction = prediction.strip()
if not prediction:
continue
try:
prediction = json.loads(prediction)
except json.JSONDecodeError:
print(f"Warning: Skipping invalid JSON data in line: {prediction}")
continue
question_id = prediction["question_id"]
if question_id not in ground_truth:
continue
gt_item = ground_truth[question_id]
question_type_id = gt_item["question_type_id"]
if prediction["prediction"] == gt_item["answer"]:
results[question_type_id]["correct"] += 1
results[question_type_id]["total"] += 1
return results
def prediction_analyse_v2(prediction_content):
# pdb.set_trace()
predictions = prediction_content.split("\n")
# 读取 ground_truth JSON 文件
with open("./file/SEED-Bench-2.json", "r") as file:
ground_truth_data = json.load(file)["questions"]
# 将 ground_truth 数据转换为以 question_id 为键的字典
ground_truth = {item["question_id"]: item for item in ground_truth_data}
# 初始化结果统计字典
results = {i: {"correct": 0, "total": 0} for i in range(1, 28)}
# 遍历 predictions，计算每个 question_type_id 的正确预测数和总预测数
for prediction in predictions:
# pdb.set_trace()
prediction = prediction.strip()
if not prediction:
continue
try:
prediction = json.loads(prediction)
except json.JSONDecodeError:
print(f"Warning: Skipping invalid JSON data in line: {prediction}")
continue
question_id = prediction["question_id"]
if question_id not in ground_truth:
continue
gt_item = ground_truth[question_id]
question_type_id = gt_item["question_type_id"]
if prediction["prediction"] == gt_item["answer"]:
results[question_type_id]["correct"] += 1
results[question_type_id]["total"] += 1
return results
def add_new_eval(
input_file,
model_name_textbox: str,
revision_name_textbox: str,
model_type: str,
model_link: str,
model_size: str,
benchmark_version: str,
LLM_type: str,
LLM_name_textbox: str,
Evaluation_dimension: str,
Evaluation_dimension_2: str,
Evaluation_method: str
):
if input_file is None:
return "Error! Empty file!"
else:
model_size = validate_model_size(model_size)
# v1 evaluation
if benchmark_version == 'v1':
content = input_file.decode("utf-8")
prediction = prediction_analyse(content)
csv_data = pd.read_csv(CSV_DIR)
csv_task_data = pd.read_csv(CSV_TASK_DIR)
Start_dimension, End_dimension = 1, 13
if Evaluation_dimension == 'Image':
End_dimension = 10
elif Evaluation_dimension == 'Video':
Start_dimension = 10
each_task_accuracy = {i: round(prediction[i]["correct"] / prediction[i]["total"] * 100, 1) if i >= Start_dimension and i < End_dimension else 0 for i in range(1, 13)}
# count for average image\video\all
total_correct_image = sum(prediction[i]["correct"] for i in range(1, 10))
total_correct_video = sum(prediction[i]["correct"] for i in range(10, 13))
total_image = sum(prediction[i]["total"] for i in range(1, 10))
total_video = sum(prediction[i]["total"] for i in range(10, 13))
if Evaluation_dimension != 'Video':
average_accuracy_image = round(total_correct_image / total_image * 100, 1)
average_task_accuracy_image = round(sum(each_task_accuracy[key] for key in range(1,10)) / 9, 1)
else:
average_accuracy_image = 0
average_task_accuracy_image = 0
if Evaluation_dimension != 'Image':
average_accuracy_video = round(total_correct_video / total_video * 100, 1)
average_task_accuracy_video = round(sum(each_task_accuracy[key] for key in range(10,13)) / 3, 1)
else:
average_accuracy_video = 0
average_task_accuracy_video = 0
if Evaluation_dimension == 'All':
overall_accuracy = round((total_correct_image + total_correct_video) / (total_image + total_video) * 100, 1)
overall_task_accuracy = round(sum(each_task_accuracy[key] for key in range(1,13)) / 12, 1)
else:
overall_accuracy = 0
overall_task_accuracy = 0
if LLM_type == 'Other':
LLM_name = LLM_name_textbox
else:
LLM_name = LLM_type
if revision_name_textbox == '':
col = csv_data.shape[0]
model_name = model_name_textbox
else:
model_name = revision_name_textbox
model_name_list = csv_data['Model']
name_list = [name.split(']')[0][1:] for name in model_name_list]
if revision_name_textbox not in name_list:
col = csv_data.shape[0]
else:
col = name_list.index(revision_name_textbox)
if model_link == '':
model_name = model_name # no url
else:
model_name = '[' + model_name + '](' + model_link + ')'
# add new data
new_data = [
model_type,
model_name,
LLM_name,
model_size,
Evaluation_method,
overall_accuracy,
average_accuracy_image,
average_accuracy_video,
each_task_accuracy[1],
each_task_accuracy[2],
each_task_accuracy[3],
each_task_accuracy[4],
each_task_accuracy[5],
each_task_accuracy[6],
each_task_accuracy[7],
each_task_accuracy[8],
each_task_accuracy[9],
each_task_accuracy[10],
each_task_accuracy[11],
each_task_accuracy[12],
]
csv_data.loc[col] = new_data
csv_data = csv_data.to_csv(CSV_DIR, index=False)
new_task_data = [
model_type,
model_name,
LLM_name,
model_size,
Evaluation_method,
overall_task_accuracy,
average_task_accuracy_image,
average_task_accuracy_video,
each_task_accuracy[1],
each_task_accuracy[2],
each_task_accuracy[3],
each_task_accuracy[4],
each_task_accuracy[5],
each_task_accuracy[6],
each_task_accuracy[7],
each_task_accuracy[8],
each_task_accuracy[9],
each_task_accuracy[10],
each_task_accuracy[11],
each_task_accuracy[12],
]
csv_task_data.loc[col] = new_data
csv_task_data = csv_task_data.to_csv(CSV_TASK_DIR, index=False)
# v2 evaluation
else:
content = input_file.decode("utf-8")
prediction = prediction_analyse_v2(content)
csv_data = pd.read_csv(CSV_V2_DIR)
csv_task_data = pd.read_csv(CSV_V2_TASK_DIR)
Start_dimension, End_dimension = 1, 28
if Evaluation_dimension_2 == 'Single':
End_dimension = 17
elif Evaluation_dimension_2 == 'L1':
End_dimension = 23
elif Evaluation_dimension_2 == 'L2':
End_dimension = 25
elif Evaluation_dimension_2 == 'L3':
End_dimension = 28
# pdb.set_trace()
each_task_accuracy = {i: round(prediction[i]["correct"] / prediction[i]["total"] * 100, 1) if i >= Start_dimension and i < End_dimension else 0 for i in range(1, 28)}
average_single = round(sum(prediction[i]["correct"] for i in range(1, 17)) / sum(prediction[i]["total"] for i in range(1, 17)) * 100, 1)
average_task_single = round(sum(each_task_accuracy[key] for key in range(1,17)) / 16, 1)
# Single
if Evaluation_dimension_2 == 'Single':
average_multi = 0
average_video = 0
average_p1 = 0
average_p2 = 0
average_p3 = 0
average_task_multi = 0
average_task_video = 0
average_task_p1 = 0
average_task_p2 = 0
average_task_p3 = 0
else:
average_multi = round(sum(prediction[i]["correct"] for i in range(17, 19)) / sum(prediction[i]["total"] for i in range(17, 19)) * 100, 1)
average_video = round(sum(prediction[i]["correct"] for i in range(19, 23)) / sum(prediction[i]["total"] for i in range(19, 23)) * 100, 1)
average_p1 = round(sum(prediction[i]["correct"] for i in range(1, 23)) / sum(prediction[i]["total"] for i in range(1, 23)) * 100, 1)
average_task_multi = round(sum(each_task_accuracy[key] for key in range(17,19)) / 2, 1)
average_task_video = round(sum(each_task_accuracy[key] for key in range(19,23)) / 4, 1)
average_task_p1 = round(sum(each_task_accuracy[key] for key in range(1,23)) / 22, 1)
# L2
if Evaluation_dimension_2 == 'L2':
average_p2 = round(sum(prediction[i]["correct"] for i in range(23, 25)) / sum(prediction[i]["total"] for i in range(23, 25)) * 100, 1)
average_task_p2 = round(sum(each_task_accuracy[key] for key in range(23,25)) / 2, 1)
average_p3 = 0
average_task_p3 = 0
# L3
elif Evaluation_dimension_2 == 'L3':
average_p2 = round(sum(prediction[i]["correct"] for i in range(23, 25)) / sum(prediction[i]["total"] for i in range(23, 25)) * 100, 1)
average_task_p2 = round(sum(each_task_accuracy[key] for key in range(23,25)) / 2, 1)
average_p3 = round(sum(prediction[i]["correct"] for i in range(25, 28)) / sum(prediction[i]["total"] for i in range(25, 28)) * 100, 1)
average_task_p3 = round(sum(each_task_accuracy[key] for key in range(25,28)) / 3, 1)
# L1
else:
average_p2 = 0
average_task_p2 = 0
average_p3 = 0
average_task_p3 = 0
if LLM_type == 'Other':
LLM_name = LLM_name_textbox
else:
LLM_name = LLM_type
if revision_name_textbox == '':
col = csv_data.shape[0]
model_name = model_name_textbox
else:
model_name = revision_name_textbox
model_name_list = csv_data['Model']
name_list = [name.split(']')[0][1:] for name in model_name_list]
if revision_name_textbox not in name_list:
col = csv_data.shape[0]
else:
col = name_list.index(revision_name_textbox)
if model_link == '':
model_name = model_name # no url
else:
model_name = '[' + model_name + '](' + model_link + ')'
# add new data
new_data = [
model_name,
LLM_name,
model_size,
Evaluation_method,
average_single,
average_multi,
average_video,
average_p1,
average_p2,
average_p3,
each_task_accuracy[1],
each_task_accuracy[2],
each_task_accuracy[3],
each_task_accuracy[4],
each_task_accuracy[5],
each_task_accuracy[6],
each_task_accuracy[7],
each_task_accuracy[8],
each_task_accuracy[9],
each_task_accuracy[10],
each_task_accuracy[11],
each_task_accuracy[12],
each_task_accuracy[13],
each_task_accuracy[14],
each_task_accuracy[15],
each_task_accuracy[16],
each_task_accuracy[17],
each_task_accuracy[18],
each_task_accuracy[19],
each_task_accuracy[20],
each_task_accuracy[21],
each_task_accuracy[22],
each_task_accuracy[23],
each_task_accuracy[24],
each_task_accuracy[25],
each_task_accuracy[26],
each_task_accuracy[27]
]
csv_data.loc[col] = new_data
csv_data = csv_data.to_csv(CSV_V2_DIR, index=False)
new_task_data = [
model_name,
LLM_name,
model_size,
Evaluation_method,
average_task_single,
average_task_multi,
average_task_video,
average_task_p1,
average_task_p2,
average_task_p3,
each_task_accuracy[1],
each_task_accuracy[2],
each_task_accuracy[3],
each_task_accuracy[4],
each_task_accuracy[5],
each_task_accuracy[6],
each_task_accuracy[7],
each_task_accuracy[8],
each_task_accuracy[9],
each_task_accuracy[10],
each_task_accuracy[11],
each_task_accuracy[12],
each_task_accuracy[13],
each_task_accuracy[14],
each_task_accuracy[15],
each_task_accuracy[16],
each_task_accuracy[17],
each_task_accuracy[18],
each_task_accuracy[19],
each_task_accuracy[20],
each_task_accuracy[21],
each_task_accuracy[22],
each_task_accuracy[23],
each_task_accuracy[24],
each_task_accuracy[25],
each_task_accuracy[26],
each_task_accuracy[27]
]
csv_task_data.loc[col] = new_task_data
csv_task_data = csv_task_data.to_csv(CSV_V2_TASK_DIR, index=False)
return 0
def get_baseline_df(average_type):
if average_type == 'All Average':
df = pd.read_csv(CSV_DIR)
else:
df = pd.read_csv(CSV_TASK_DIR)
df = df.sort_values(by="Avg. All", ascending=False)
present_columns = MODEL_INFO + checkbox_group.value
df = df[present_columns]
return df
def get_baseline_v2_df(average_type):
# pdb.set_trace()
if average_type == 'All Average':
df = pd.read_csv(CSV_V2_DIR)
else:
df = pd.read_csv(CSV_V2_TASK_DIR)
df = df.sort_values(by="Avg. Single", ascending=False)
present_columns = MODEL_INFO_V2 + checkbox_group_v2.value
# pdb.set_trace()
df = df[present_columns]
return df
def get_all_df():
df = pd.read_csv(CSV_DIR)
df = df.sort_values(by="Avg. All", ascending=False)
return df
def get_all_v2_df():
df = pd.read_csv(CSV_V2_DIR)
df = df.sort_values(by="Avg. P1", ascending=False)
return df
def switch_version(version):
return f"当前版本: {version}"
block = gr.Blocks()
with block:
gr.Markdown(
LEADERBORAD_INTRODUCTION
)
with gr.Tabs(elem_classes="tab-buttons") as tabs:
with gr.TabItem("🏅 SEED Benchmark v2", elem_id="seed-benchmark-tab-table", id=0):
with gr.Row():
with gr.Accordion("Citation", open=False):
citation_button = gr.Textbox(
value=CITATION_BUTTON_TEXT,
label=CITATION_BUTTON_LABEL,
elem_id="citation-button",
).style(show_copy_button=True)
gr.Markdown(
TABLE_INTRODUCTION
)
# selection for column part:
checkbox_group_v2 = gr.CheckboxGroup(
choices=TASK_V2_INFO,
value=AVG_V2_INFO,
label="Evaluation Dimension",
interactive=True,
)
with gr.Row():
# selection for model size part:
model_size_v2 = gr.CheckboxGroup(
choices=MODEL_SIZE,
value=MODEL_SIZE,
label="Model Size",
interactive=True,
)
# selection for model size part:
evaluation_method_v2 = gr.CheckboxGroup(
choices=EVALUATION_METHOD,
value=EVALUATION_METHOD,
label="Evaluation Method",
interactive=True,
)
average_type_v2 = gr.Radio(AVERAGE_TYPE, label="Performance Average Type", value="All Average")
baseline_v2_value = get_baseline_v2_df(average_type_v2.value)
baseline_v2_header = MODEL_INFO_V2 + checkbox_group_v2.value
baseline_v2_datatype = ['markdown'] * 3 + ['number'] * len(checkbox_group_v2.value)
# pdb.set_trace()
# 创建数据帧组件
data_component_v2 = gr.components.Dataframe(
value=baseline_v2_value,
headers=baseline_v2_header,
type="pandas",
datatype=baseline_v2_datatype,
interactive=False,
visible=True,
)
def on_filter_model_size_method_v2_change(selected_model_size, selected_evaluation_method, selected_columns):
updated_data = get_all_v2_df()
# model_size & evaluation_method:
# 自定义过滤函数
def custom_filter(row, model_size_filters, evaluation_method_filters):
model_size = row['Model Size']
evaluation_method = row['Evaluation Method']
if model_size == '-':
size_filter = '-' in model_size_filters
elif 'B' in model_size:
size = float(model_size.replace('B', ''))
size_filter = ('>=10B' in model_size_filters and size >= 10) or ('<10B' in model_size_filters and size < 10)
else:
size_filter = False
method_filter = evaluation_method in evaluation_method_filters
return size_filter and method_filter
# 使用自定义过滤函数过滤数据
mask = updated_data.apply(custom_filter, axis=1, model_size_filters=selected_model_size, evaluation_method_filters=selected_evaluation_method)
updated_data = updated_data[mask]
# columns:
selected_columns = [item for item in TASK_V2_INFO if item in selected_columns]
present_columns = MODEL_INFO_V2 + selected_columns
updated_data = updated_data[present_columns]
updated_data = updated_data.sort_values(by=selected_columns[0], ascending=False)
updated_headers = present_columns
update_datatype = [DATA_TITILE_V2_TYPE[COLUMN_V2_NAMES.index(x)] for x in updated_headers]
filter_component = gr.components.Dataframe(
value=updated_data,
headers=updated_headers,
type="pandas",
datatype=update_datatype,
interactive=False,
visible=True,
)
# pdb.set_trace()
return filter_component.value
def on_average_type_v2_change(average_type_v2):
return get_baseline_v2_df(average_type_v2)
average_type_v2.change(fn=on_average_type_v2_change, inputs=[average_type_v2], outputs=data_component_v2)
model_size_v2.change(fn=on_filter_model_size_method_v2_change, inputs=[model_size_v2, evaluation_method_v2, checkbox_group_v2], outputs=data_component_v2)
evaluation_method_v2.change(fn=on_filter_model_size_method_v2_change, inputs=[model_size_v2, evaluation_method_v2, checkbox_group_v2], outputs=data_component_v2)
checkbox_group_v2.change(fn=on_filter_model_size_method_v2_change, inputs=[model_size_v2, evaluation_method_v2, checkbox_group_v2], outputs=data_component_v2)
# table seed-bench-v1
with gr.TabItem("🏅 SEED Benchmark v1", elem_id="seed-benchmark-tab-table", id=1):
with gr.Row():
with gr.Accordion("Citation", open=False):
citation_button = gr.Textbox(
value=CITATION_BUTTON_TEXT,
label=CITATION_BUTTON_LABEL,
elem_id="citation-button",
).style(show_copy_button=True)
gr.Markdown(
TABLE_INTRODUCTION
)
# selection for column part:
checkbox_group = gr.CheckboxGroup(
choices=TASK_INFO,
value=AVG_INFO,
label="Evaluation Dimension",
interactive=True,
)
with gr.Row():
# selection for model size part:
model_size = gr.CheckboxGroup(
choices=MODEL_SIZE,
value=MODEL_SIZE,
label="Model Size",
interactive=True,
)
# selection for model size part:
evaluation_method = gr.CheckboxGroup(
choices=EVALUATION_METHOD,
value=EVALUATION_METHOD,
label="Evaluation Method",
interactive=True,
)
average_type = gr.Radio(AVERAGE_TYPE, label="Performance Average Type", value="All Average")
baseline_value = get_baseline_df(average_type.value)
baseline_header = MODEL_INFO + checkbox_group.value
baseline_datatype = ['markdown'] * 4 + ['number'] * len(checkbox_group.value)
# pdb.set_trace()
# 创建数据帧组件
data_component = gr.components.Dataframe(
value=baseline_value,
headers=baseline_header,
type="pandas",
datatype=baseline_datatype,
interactive=False,
visible=True,
)
def on_filter_model_size_method_change(selected_model_size, selected_evaluation_method, selected_columns):
updated_data = get_all_df()
# model_size & evaluation_method:
# 自定义过滤函数
def custom_filter(row, model_size_filters, evaluation_method_filters):
model_size = row['Model Size']
evaluation_method = row['Evaluation Method']
if model_size == '-':
size_filter = '-' in model_size_filters
elif 'B' in model_size:
size = float(model_size.replace('B', ''))
size_filter = ('>=10B' in model_size_filters and size >= 10) or ('<10B' in model_size_filters and size < 10)
else:
size_filter = False
method_filter = evaluation_method in evaluation_method_filters
return size_filter and method_filter
# 使用自定义过滤函数过滤数据
mask = updated_data.apply(custom_filter, axis=1, model_size_filters=selected_model_size, evaluation_method_filters=selected_evaluation_method)
updated_data = updated_data[mask]
# columns:
selected_columns = [item for item in TASK_INFO if item in selected_columns]
present_columns = MODEL_INFO + selected_columns
updated_data = updated_data[present_columns]
updated_data = updated_data.sort_values(by=selected_columns[0], ascending=False)
updated_headers = present_columns
update_datatype = [DATA_TITILE_TYPE[COLUMN_NAMES.index(x)] for x in updated_headers]
filter_component = gr.components.Dataframe(
value=updated_data,
headers=updated_headers,
type="pandas",
datatype=update_datatype,
interactive=False,
visible=True,
)
# pdb.set_trace()
return filter_component.value
def on_average_type_change(average_type):
return get_baseline_df(average_type)
average_type.change(fn=on_average_type_change, inputs=[average_type], outputs=data_component)
model_size.change(fn=on_filter_model_size_method_change, inputs=[model_size, evaluation_method, checkbox_group], outputs=data_component)
evaluation_method.change(fn=on_filter_model_size_method_change, inputs=[model_size, evaluation_method, checkbox_group], outputs=data_component)
checkbox_group.change(fn=on_filter_model_size_method_change, inputs=[model_size, evaluation_method, checkbox_group], outputs=data_component)
# table 2
with gr.TabItem("📝 About", elem_id="seed-benchmark-tab-table", id=2):
gr.Markdown(LEADERBORAD_INFO, elem_classes="markdown-text")
# table 3
with gr.TabItem("🚀 Submit here! ", elem_id="seed-benchmark-tab-table", id=3):
gr.Markdown(LEADERBORAD_INTRODUCTION, elem_classes="markdown-text")
with gr.Row():
gr.Markdown(SUBMIT_INTRODUCTION, elem_classes="markdown-text")
with gr.Row():
gr.Markdown("# ✉️✨ Submit your model evaluation json file here!", elem_classes="markdown-text")
with gr.Row():
with gr.Column():
model_name_textbox = gr.Textbox(
label="Model name", placeholder="LLaMA-7B"
)
revision_name_textbox = gr.Textbox(
label="Revision Model Name", placeholder="LLaMA-7B"
)
model_type = gr.Dropdown(
choices=[
"LLM",
"ImageLLM",
"VideoLLM",
"Other",
],
label="Model type",
multiselect=False,
value="ImageLLM",
interactive=True,
)
model_link = gr.Textbox(
label="Model Link", placeholder="https://huggingface.co/decapoda-research/llama-7b-hf"
)
model_size = gr.Textbox(
label="Model size", placeholder="7B(Input content format must be 'number+B' or '-', default is '-')"
)
benchmark_version= gr.Dropdown(
choices=["v1", "v2"],
label="Benchmark version",
multiselect=False,
value="v1",
interactive=True,
)
with gr.Column():
LLM_type = gr.Dropdown(
choices=["Vicuna-7B", "Flan-T5-XL", "LLaMA-7B", "Other"],
label="LLM type",
multiselect=False,
value="LLaMA-7B",
interactive=True,
)
LLM_name_textbox = gr.Textbox(
label="LLM model (Required for Other)",
placeholder="LLaMA-13B",
value="LLaMA-13B",
)
Evaluation_dimension = gr.Dropdown(
choices=["All", "Image", "Video"],
label="Evaluation dimension for SEED-Bench 1(for evaluate SEED-Bench 1)",
multiselect=False,
value="All",
interactive=True,
)
Evaluation_dimension_2 = gr.Dropdown(
choices=["Single", "L1", "L2", "L3"],
label="Evaluation dimension for SEED-Bench 2(for evaluate SEED-Bench 2)",
multiselect=False,
value="L2",
interactive=True,
)
Evaluation_method = gr.Dropdown(
choices=EVALUATION_METHOD,
label="Evaluation method",
multiselect=False,
value=EVALUATION_METHOD[0],
interactive=True,
)
with gr.Column():
input_file = gr.inputs.File(label = "Click to Upload a json File", file_count="single", type='binary')
submit_button = gr.Button("Submit Eval")
submission_result = gr.Markdown()
submit_button.click(
add_new_eval,
inputs = [
input_file,
model_name_textbox,
revision_name_textbox,
model_type,
model_link,
model_size,
benchmark_version,
LLM_type,
LLM_name_textbox,
Evaluation_dimension,
Evaluation_dimension_2,
Evaluation_method
],
)
def refresh_data():
value1 = get_baseline_df(average_type)
value2 = get_baseline_v2_df(average_type_v2)
return value1, value2
with gr.Row():
data_run = gr.Button("Refresh")
data_run.click(
refresh_data, outputs=[data_component, data_component_v2]
)
# block.load(get_baseline_df, outputs=data_title)
block.launch() |