Spaces:
AIR-Bench
/
Running on CPU Upgrade

File size: 21,912 Bytes
f766ce9
 
 
 
 
 
8a1daf9
f766ce9
36c5a0c
f766ce9
93fda91
 
 
 
 
 
 
 
 
f766ce9
93fda91
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
a0387d8
f766ce9
 
 
 
df659d0
e089630
 
 
 
 
 
 
 
f766ce9
93fda91
8a1daf9
 
973bd2a
8a1daf9
973bd2a
8a1daf9
f30cbcc
 
 
8a1daf9
4959887
ee1f0b7
 
af8395f
77ded94
2edd122
f30cbcc
ee1f0b7
 
 
77ded94
f766ce9
a0387d8
 
 
5808d8f
 
 
 
 
 
 
77ded94
 
5808d8f
77ded94
f30cbcc
 
 
 
 
 
 
77ded94
 
f30cbcc
77ded94
f766ce9
 
 
b80bda9
f766ce9
 
 
 
 
8ec7973
f766ce9
32ebf18
e8879cc
f766ce9
a0387d8
e8879cc
 
a0387d8
b80bda9
 
 
 
 
973bd2a
af8395f
a0387d8
77ded94
a0387d8
b80bda9
64a83a1
b80bda9
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
64a83a1
 
 
 
 
 
b80bda9
 
 
 
64a83a1
 
 
a0387d8
b80bda9
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
64a83a1
93fda91
64a83a1
 
 
 
 
 
 
 
93fda91
64a83a1
 
 
 
5808d8f
64a83a1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
f30cbcc
 
 
 
 
a0387d8
f30cbcc
 
a0387d8
 
f30cbcc
b80bda9
 
 
 
 
973bd2a
ee1f0b7
a0387d8
77ded94
a0387d8
b80bda9
64a83a1
b80bda9
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
64a83a1
 
 
 
 
 
b80bda9
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
f30cbcc
b80bda9
 
 
 
 
 
 
 
 
 
 
 
 
 
64a83a1
 
93fda91
64a83a1
 
 
 
 
 
 
 
 
93fda91
64a83a1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
8ec7973
36c5a0c
 
 
 
 
 
a30a228
 
f000c74
a30a228
f000c74
240d9ce
4cf5eb9
240d9ce
 
 
 
 
4cf5eb9
240d9ce
 
 
 
 
 
69a9e46
 
93fda91
 
69a9e46
 
4a6f9cd
d00fb74
2c777fc
36c5a0c
d00fb74
9400714
d00fb74
 
 
2c777fc
4a6f9cd
 
 
9f44d20
 
 
 
 
 
4a6f9cd
 
 
 
 
d00fb74
240d9ce
 
d00fb74
9400714
4a6f9cd
 
 
 
36c5a0c
8a1daf9
 
f766ce9
 
 
 
57ca843
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
import gradio as gr
from apscheduler.schedulers.background import BackgroundScheduler
from huggingface_hub import snapshot_download

from src.about import (
    INTRODUCTION_TEXT,
    BENCHMARKS_TEXT,
    TITLE,
    EVALUATION_QUEUE_TEXT
)
from src.benchmarks import (
    DOMAIN_COLS_QA,
    LANG_COLS_QA,
    DOMAIN_COLS_LONG_DOC,
    LANG_COLS_LONG_DOC,
    METRIC_LIST,
    DEFAULT_METRIC_QA,
    DEFAULT_METRIC_LONG_DOC
)
from src.display.css_html_js import custom_css
from src.display.utils import (
    COL_NAME_IS_ANONYMOUS,
    COL_NAME_REVISION,
    COL_NAME_TIMESTAMP,
    COL_NAME_RERANKING_MODEL,
    COL_NAME_RETRIEVAL_MODEL
)
from src.envs import (
    API,
    EVAL_RESULTS_PATH,
    REPO_ID,
    RESULTS_REPO,
    TOKEN,
    BM25_LINK,
    BENCHMARK_VERSION_LIST,
    LATEST_BENCHMARK_VERSION
)
from src.read_evals import (
    get_raw_eval_results,
    get_leaderboard_df
)
from src.utils import (
    update_metric,
    upload_file,
    get_default_cols,
    submit_results,
    reset_rank,
    remove_html
)
from src.display.gradio_formatting import (
    get_version_dropdown,
    get_search_bar,
    get_reranking_dropdown,
    get_metric_dropdown,
    get_domain_dropdown,
    get_language_dropdown,
    get_anonymous_checkbox,
    get_revision_and_ts_checkbox,
    get_leaderboard_table,
    get_noreranking_dropdown
)
from src.display.gradio_listener import set_listeners

def restart_space():
    API.restart_space(repo_id=REPO_ID)


try:
    snapshot_download(
        repo_id=RESULTS_REPO, local_dir=EVAL_RESULTS_PATH, repo_type="dataset", tqdm_class=None, etag_timeout=30,
        token=TOKEN
    )
except Exception as e:
    print(f'failed to download')
    restart_space()

raw_data = get_raw_eval_results(f"{EVAL_RESULTS_PATH}/{LATEST_BENCHMARK_VERSION}")

original_df_qa = get_leaderboard_df(
    raw_data, task='qa', metric=DEFAULT_METRIC_QA)
original_df_long_doc = get_leaderboard_df(
    raw_data, task='long-doc', metric=DEFAULT_METRIC_LONG_DOC)
print(f'raw data: {len(raw_data)}')
print(f'QA data loaded: {original_df_qa.shape}')
print(f'Long-Doc data loaded: {len(original_df_long_doc)}')

leaderboard_df_qa = original_df_qa.copy()
# leaderboard_df_qa = leaderboard_df_qa[has_no_nan_values(df, _benchmark_cols)]
shown_columns_qa, types_qa = get_default_cols(
    'qa', leaderboard_df_qa.columns, add_fix_cols=True)
leaderboard_df_qa = leaderboard_df_qa[~leaderboard_df_qa[COL_NAME_IS_ANONYMOUS]][shown_columns_qa]
leaderboard_df_qa.drop([COL_NAME_REVISION, COL_NAME_TIMESTAMP], axis=1, inplace=True)

leaderboard_df_long_doc = original_df_long_doc.copy()
shown_columns_long_doc, types_long_doc = get_default_cols(
    'long-doc', leaderboard_df_long_doc.columns, add_fix_cols=True)
leaderboard_df_long_doc = leaderboard_df_long_doc[~leaderboard_df_long_doc[COL_NAME_IS_ANONYMOUS]][shown_columns_long_doc]
leaderboard_df_long_doc.drop([COL_NAME_REVISION, COL_NAME_TIMESTAMP], axis=1, inplace=True)

# select reranking model
reranking_models = sorted(list(frozenset([eval_result.reranking_model for eval_result in raw_data])))


def update_metric_qa(
        metric: str,
        domains: list,
        langs: list,
        reranking_model: list,
        query: str,
        show_anonymous: bool,
        show_revision_and_timestamp,
):
    return update_metric(raw_data, 'qa', metric, domains, langs, reranking_model, query, show_anonymous, show_revision_and_timestamp)

def update_metric_long_doc(
        metric: str,
        domains: list,
        langs: list,
        reranking_model: list,
        query: str,
        show_anonymous: bool,
        show_revision_and_timestamp,
):
    return update_metric(raw_data, "long-doc", metric, domains, langs, reranking_model, query, show_anonymous, show_revision_and_timestamp)


demo = gr.Blocks(css=custom_css)

with demo:
    gr.HTML(TITLE)
    gr.Markdown(INTRODUCTION_TEXT, elem_classes="markdown-text")

    with gr.Tabs(elem_classes="tab-buttons") as tabs:
        with gr.TabItem("QA", elem_id="qa-benchmark-tab-table", id=0):
            with gr.Row():
                with gr.Column(min_width=320):
                    # select domain
                    with gr.Row():
                        selected_domains = get_domain_dropdown(DOMAIN_COLS_QA, DOMAIN_COLS_QA)
                    # select language
                    with gr.Row():
                        selected_langs = get_language_dropdown(LANG_COLS_QA, LANG_COLS_QA)

                with gr.Column():
                    with gr.Row():
                        selected_version = get_version_dropdown()
                    # select the metric
                    selected_metric = get_metric_dropdown(METRIC_LIST, DEFAULT_METRIC_QA)
                    with gr.Row():
                        show_anonymous = get_anonymous_checkbox()
                    with gr.Row():
                        show_revision_and_timestamp = get_revision_and_ts_checkbox()
            with gr.Tabs(elem_classes="tab-buttons") as sub_tabs:
                with gr.TabItem("Retrieval + Reranking", id=10):
                    with gr.Row():
                        # search retrieval models
                        with gr.Column():
                            search_bar = get_search_bar()
                        # select reranking models
                        with gr.Column():
                            selected_rerankings = get_reranking_dropdown(reranking_models)
                    leaderboard_table = get_leaderboard_table(leaderboard_df_qa, types_qa)
                    # Dummy leaderboard for handling the case when the user uses backspace key
                    hidden_leaderboard_table_for_search = get_leaderboard_table(original_df_qa, types_qa, visible=False)

                    set_listeners(
                        "qa",
                        leaderboard_table,
                        hidden_leaderboard_table_for_search,
                        search_bar,
                        selected_domains,
                        selected_langs,
                        selected_rerankings,
                        show_anonymous,
                        show_revision_and_timestamp,
                    )

                    # set metric listener
                    selected_metric.change(
                        update_metric_qa,
                        [
                            selected_metric,
                            selected_domains,
                            selected_langs,
                            selected_rerankings,
                            search_bar,
                            show_anonymous,
                            show_revision_and_timestamp,
                        ],
                        leaderboard_table,
                        queue=True
                    )
                with gr.TabItem("Retrieval Only", id=11):
                    with gr.Row():
                        with gr.Column(scale=1):
                            search_bar_retriever = get_search_bar()
                        with gr.Column(scale=1):
                            selected_noreranker = get_noreranking_dropdown()
                    lb_df_retriever = leaderboard_df_qa[leaderboard_df_qa[COL_NAME_RERANKING_MODEL] == "NoReranker"]
                    lb_df_retriever = reset_rank(lb_df_retriever)
                    lb_table_retriever = get_leaderboard_table(lb_df_retriever, types_qa)
                    # Dummy leaderboard for handling the case when the user uses backspace key
                    hidden_lb_df_retriever = original_df_qa[original_df_qa[COL_NAME_RERANKING_MODEL] == "NoReranker"]
                    hidden_lb_df_retriever = reset_rank(hidden_lb_df_retriever)
                    hidden_lb_table_retriever = get_leaderboard_table(hidden_lb_df_retriever, types_qa, visible=False)

                    set_listeners(
                        "qa",
                        lb_table_retriever,
                        hidden_lb_table_retriever,
                        search_bar_retriever,
                        selected_domains,
                        selected_langs,
                        selected_noreranker,
                        show_anonymous,
                        show_revision_and_timestamp,
                    )

                    # set metric listener
                    selected_metric.change(
                        update_metric_qa,
                        [
                            selected_metric,
                            selected_domains,
                            selected_langs,
                            selected_noreranker,
                            search_bar_retriever,
                            show_anonymous,
                            show_revision_and_timestamp,
                        ],
                        lb_table_retriever,
                        queue=True
                    )
                with gr.TabItem("Reranking Only", id=12):
                    lb_df_reranker = leaderboard_df_qa[leaderboard_df_qa[COL_NAME_RETRIEVAL_MODEL] == BM25_LINK]
                    lb_df_reranker = reset_rank(lb_df_reranker)
                    reranking_models_reranker = lb_df_reranker[COL_NAME_RERANKING_MODEL].apply(remove_html).unique().tolist()
                    with gr.Row():
                        with gr.Column(scale=1):
                            selected_rerankings_reranker = get_reranking_dropdown(reranking_models_reranker)
                        with gr.Column(scale=1):
                            search_bar_reranker = gr.Textbox(show_label=False, visible=False)
                    lb_table_reranker = get_leaderboard_table(lb_df_reranker, types_qa)
                    hidden_lb_df_reranker = original_df_qa[original_df_qa[COL_NAME_RETRIEVAL_MODEL] == BM25_LINK]
                    hidden_lb_df_reranker = reset_rank(hidden_lb_df_reranker)
                    hidden_lb_table_reranker = get_leaderboard_table(
                        hidden_lb_df_reranker, types_qa, visible=False
                    )

                    set_listeners(
                        "qa",
                        lb_table_reranker,
                        hidden_lb_table_reranker,
                        search_bar_reranker,
                        selected_domains,
                        selected_langs,
                        selected_rerankings_reranker,
                        show_anonymous,
                        show_revision_and_timestamp,
                    )
                    # set metric listener
                    selected_metric.change(
                        update_metric_qa,
                        [
                            selected_metric,
                            selected_domains,
                            selected_langs,
                            selected_rerankings_reranker,
                            search_bar_reranker,
                            show_anonymous,
                            show_revision_and_timestamp,
                        ],
                        lb_table_reranker,
                        queue=True
                    )
        with gr.TabItem("Long Doc", elem_id="long-doc-benchmark-tab-table", id=1):
            with gr.Row():
                with gr.Column(min_width=320):
                    # select domain
                    with gr.Row():
                        selected_domains = get_domain_dropdown(DOMAIN_COLS_LONG_DOC, DOMAIN_COLS_LONG_DOC)
                    # select language
                    with gr.Row():
                        selected_langs = get_language_dropdown(
                            LANG_COLS_LONG_DOC, LANG_COLS_LONG_DOC
                        )
                with gr.Column():
                    with gr.Row():
                        selected_version = get_version_dropdown()
                    # select the metric
                    with gr.Row():
                        selected_metric = get_metric_dropdown(METRIC_LIST, DEFAULT_METRIC_LONG_DOC)
                    with gr.Row():
                        show_anonymous = get_anonymous_checkbox()
                    with gr.Row():
                        show_revision_and_timestamp = get_revision_and_ts_checkbox()
            with gr.Tabs(elem_classes="tab-buttons") as sub_tabs:
                with gr.TabItem("Retrieval + Reranking", id=20):
                    with gr.Row():
                        with gr.Column():
                            search_bar = get_search_bar()
                        # select reranking model
                        with gr.Column():
                            selected_rerankings = get_reranking_dropdown(reranking_models)

                    lb_table = get_leaderboard_table(
                        leaderboard_df_long_doc, types_long_doc
                    )

                    # Dummy leaderboard for handling the case when the user uses backspace key
                    hidden_lb_table_for_search = get_leaderboard_table(
                        original_df_long_doc, types_long_doc, visible=False
                    )

                    set_listeners(
                        "long-doc",
                        lb_table,
                        hidden_lb_table_for_search,
                        search_bar,
                        selected_domains,
                        selected_langs,
                        selected_rerankings,
                        show_anonymous,
                        show_revision_and_timestamp,
                    )

                    # set metric listener
                    selected_metric.change(
                        update_metric_long_doc,
                        [
                            selected_metric,
                            selected_domains,
                            selected_langs,
                            selected_rerankings,
                            search_bar,
                            show_anonymous,
                            show_revision_and_timestamp
                        ],
                        lb_table,
                        queue=True
                    )
                with gr.TabItem("Retrieval Only", id=21):
                    with gr.Row():
                        with gr.Column(scale=1):
                            search_bar_retriever = get_search_bar()
                        with gr.Column(scale=1):
                            selected_noreranker = get_noreranking_dropdown()
                    lb_df_retriever_long_doc = leaderboard_df_long_doc[
                        leaderboard_df_long_doc[COL_NAME_RERANKING_MODEL] == "NoReranker"
                    ]
                    lb_df_retriever_long_doc = reset_rank(lb_df_retriever_long_doc)
                    hidden_lb_db_retriever_long_doc = original_df_long_doc[
                        original_df_long_doc[COL_NAME_RERANKING_MODEL] == "NoReranker"
                    ]
                    hidden_lb_db_retriever_long_doc = reset_rank(hidden_lb_db_retriever_long_doc)
                    lb_table_retriever_long_doc = get_leaderboard_table(
                        lb_df_retriever_long_doc, types_long_doc)
                    hidden_lb_table_retriever_long_doc = get_leaderboard_table(
                        hidden_lb_db_retriever_long_doc, types_long_doc, visible=False
                    )

                    set_listeners(
                        "long-doc",
                        lb_table_retriever_long_doc,
                        hidden_lb_table_retriever_long_doc,
                        search_bar_retriever,
                        selected_domains,
                        selected_langs,
                        selected_noreranker,
                        show_anonymous,
                        show_revision_and_timestamp,
                    )

                    selected_metric.change(
                        update_metric_long_doc,
                        [
                            selected_metric,
                            selected_domains,
                            selected_langs,
                            selected_noreranker,
                            search_bar_retriever,
                            show_anonymous,
                            show_revision_and_timestamp,
                        ],
                        lb_table_retriever_long_doc,
                        queue=True
                    )
                with gr.TabItem("Reranking Only", id=22):
                    lb_df_reranker_ldoc = leaderboard_df_long_doc[
                        leaderboard_df_long_doc[COL_NAME_RETRIEVAL_MODEL] == BM25_LINK
                        ]
                    lb_df_reranker_ldoc = reset_rank(lb_df_reranker_ldoc)
                    reranking_models_reranker_ldoc = lb_df_reranker_ldoc[COL_NAME_RERANKING_MODEL].apply(remove_html).unique().tolist()
                    with gr.Row():
                        with gr.Column(scale=1):
                            selected_rerankings_reranker_ldoc = get_reranking_dropdown(reranking_models_reranker_ldoc)
                        with gr.Column(scale=1):
                            search_bar_reranker_ldoc = gr.Textbox(show_label=False, visible=False)
                    lb_table_reranker_ldoc = get_leaderboard_table(lb_df_reranker_ldoc, types_long_doc)
                    hidden_lb_df_reranker_ldoc = original_df_long_doc[original_df_long_doc[COL_NAME_RETRIEVAL_MODEL] == BM25_LINK]
                    hidden_lb_df_reranker_ldoc = reset_rank(hidden_lb_df_reranker_ldoc)
                    hidden_lb_table_reranker_ldoc = get_leaderboard_table(
                        hidden_lb_df_reranker_ldoc, types_long_doc, visible=False
                    )

                    set_listeners(
                        "long-doc",
                        lb_table_reranker_ldoc,
                        hidden_lb_table_reranker_ldoc,
                        search_bar_reranker_ldoc,
                        selected_domains,
                        selected_langs,
                        selected_rerankings_reranker_ldoc,
                        show_anonymous,
                        show_revision_and_timestamp,
                    )
                    selected_metric.change(
                        update_metric_long_doc,
                        [
                            selected_metric,
                            selected_domains,
                            selected_langs,
                            selected_rerankings_reranker_ldoc,
                            search_bar_reranker_ldoc,
                            show_anonymous,
                            show_revision_and_timestamp,
                        ],
                        lb_table_reranker_ldoc,
                        queue=True
                    )

        with gr.TabItem("🚀Submit here!", elem_id="submit-tab-table", id=2):
            with gr.Column():
                with gr.Row():
                    gr.Markdown(EVALUATION_QUEUE_TEXT, elem_classes="markdown-text")
                with gr.Row():
                    gr.Markdown("## ✉️Submit your model here!", elem_classes="markdown-text")
                with gr.Row():
                    with gr.Column():
                        model_name = gr.Textbox(label="Retrieval Method name")
                    with gr.Column():
                        model_url = gr.Textbox(label="Retrieval Method URL")
                with gr.Row():
                    with gr.Column():
                        reranking_model_name = gr.Textbox(
                            label="Reranking Model name",
                            info="Optional",
                            value="NoReranker"
                        )
                    with gr.Column():
                        reranking_model_url = gr.Textbox(
                            label="Reranking Model URL",
                            info="Optional",
                            value=""
                        )
                with gr.Row():
                    with gr.Column():
                        benchmark_version = gr.Dropdown(
                            BENCHMARK_VERSION_LIST,
                            value=LATEST_BENCHMARK_VERSION,
                            interactive=True,
                            label="AIR-Bench Version")
                with gr.Row():
                    upload_button = gr.UploadButton("Click to upload search results", file_count="single")
                with gr.Row():
                    file_output = gr.File()
                with gr.Row():
                    is_anonymous = gr.Checkbox(
                        label="Nope. I want to submit anonymously 🥷",
                        value=False,
                        info="Do you want to shown on the leaderboard by default?")
                with gr.Row():
                    submit_button = gr.Button("Submit")
                with gr.Row():
                    submission_result = gr.Markdown()
                upload_button.upload(
                    upload_file,
                    [
                        upload_button,
                    ],
                    file_output)
                submit_button.click(
                    submit_results,
                    [
                        file_output,
                        model_name,
                        model_url,
                        reranking_model_name,
                        reranking_model_url,
                        benchmark_version,
                        is_anonymous
                    ],
                    submission_result,
                    show_progress="hidden"
                )

        with gr.TabItem("📝 About", elem_id="llm-benchmark-tab-table", id=3):
            gr.Markdown(BENCHMARKS_TEXT, elem_classes="markdown-text")

scheduler = BackgroundScheduler()
scheduler.add_job(restart_space, "interval", seconds=1800)
scheduler.start()
demo.queue(default_concurrency_limit=40).launch()