Spaces:
AIR-Bench
/
Running on CPU Upgrade

File size: 13,656 Bytes
f766ce9
 
 
 
 
 
8a1daf9
f766ce9
36c5a0c
f766ce9
 
8a1daf9
 
a0c2cea
4a6f9cd
f30cbcc
7ff98ba
f766ce9
 
 
 
 
e22a0ca
 
 
 
 
 
 
f766ce9
1a2dba5
8a1daf9
 
 
 
1a2dba5
8a1daf9
f30cbcc
 
 
8a1daf9
2edd122
 
 
f30cbcc
1a2dba5
2edd122
f766ce9
5808d8f
 
 
 
 
 
 
 
8a1daf9
f30cbcc
 
 
 
 
 
 
 
1a2dba5
f766ce9
 
 
 
 
 
 
 
8ec7973
f766ce9
 
8ec7973
f766ce9
 
 
 
 
 
8ec7973
 
 
 
 
 
 
 
 
e8879cc
f766ce9
e8879cc
 
 
 
 
f766ce9
 
e8879cc
 
2edd122
e8879cc
 
 
 
2edd122
e8879cc
 
8ec7973
8a1daf9
f766ce9
e8879cc
 
 
 
 
 
f766ce9
 
 
8a1daf9
7ff98ba
f766ce9
 
 
 
 
 
f8b3d0f
8a1daf9
7ff98ba
f8b3d0f
 
 
 
5808d8f
 
61eca2d
 
 
 
 
 
 
 
 
 
 
5808d8f
 
61eca2d
 
 
f8b3d0f
 
 
 
 
 
 
 
 
 
 
 
f766ce9
5808d8f
 
 
 
 
 
 
 
 
 
 
 
 
 
f30cbcc
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2edd122
f30cbcc
 
 
 
2edd122
f30cbcc
 
 
8a1daf9
f30cbcc
 
 
 
 
 
 
 
 
 
 
7ff98ba
f30cbcc
 
 
 
 
 
 
 
7ff98ba
f30cbcc
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
8ec7973
36c5a0c
 
 
 
 
 
a30a228
 
 
2edd122
 
 
 
a30a228
9f44d20
a30a228
 
4a6f9cd
d00fb74
2c777fc
36c5a0c
d00fb74
 
 
 
 
2c777fc
4a6f9cd
 
 
9f44d20
 
 
 
 
 
4a6f9cd
 
 
 
 
d00fb74
 
 
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
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.display.css_html_js import custom_css
from src.leaderboard.read_evals import get_raw_eval_results, get_leaderboard_df

from src.envs import API, EVAL_RESULTS_PATH, REPO_ID, RESULTS_REPO, TOKEN
from utils import update_table, update_metric, update_table_long_doc, upload_file, get_default_cols, submit_results
from src.benchmarks import DOMAIN_COLS_QA, LANG_COLS_QA, DOMAIN_COLS_LONG_DOC, LANG_COLS_LONG_DOC, metric_list
from src.display.utils import TYPES_QA, TYPES_LONG_DOC


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:
    restart_space()

raw_data = get_raw_eval_results(f"{EVAL_RESULTS_PATH}/AIR-Bench_24.04")

original_df_qa = get_leaderboard_df(
    raw_data, task='qa', metric='ndcg_at_3')
original_df_long_doc = get_leaderboard_df(
    raw_data, task='long-doc', metric='ndcg_at_3')
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()
shown_columns_qa = get_default_cols('qa', leaderboard_df_qa.columns, add_fix_cols=True)
leaderboard_df_qa = leaderboard_df_qa[shown_columns_qa]

leaderboard_df_long_doc = original_df_long_doc.copy()
shown_columns_long_doc = get_default_cols('long-doc', leaderboard_df_long_doc.columns, add_fix_cols=True)
leaderboard_df_long_doc = leaderboard_df_long_doc[shown_columns_long_doc]


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

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


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():
                    # search bar for model name
                    with gr.Row():
                        search_bar = gr.Textbox(
                            placeholder=" 🔍 Search for your model (separate multiple queries with `;`) and press ENTER...",
                            show_label=False,
                            elem_id="search-bar",
                        )
                    # select the metric
                    selected_metric = gr.Dropdown(
                        choices=metric_list,
                        value=metric_list[1],
                        label="Select the metric",
                        interactive=True,
                        elem_id="metric-select",
                    )
                with gr.Column(min_width=320):
                    # select domain
                    with gr.Row():
                        selected_domains = gr.CheckboxGroup(
                            choices=DOMAIN_COLS_QA,
                            value=DOMAIN_COLS_QA,
                            label="Select the domains",
                            elem_id="domain-column-select",
                            interactive=True,
                        )
                    # select language
                    with gr.Row():
                        selected_langs = gr.Dropdown(
                            choices=LANG_COLS_QA,
                            value=LANG_COLS_QA,
                            label="Select the languages",
                            elem_id="language-column-select",
                            multiselect=True,
                            interactive=True
                        )
                    # select reranking model
                    reranking_models = list(frozenset([eval_result.reranking_model for eval_result in raw_data]))
                    with gr.Row():
                        selected_rerankings = gr.CheckboxGroup(
                            choices=reranking_models,
                            value=reranking_models,
                            label="Select the reranking models",
                            elem_id="reranking-select",
                            interactive=True
                        )

            leaderboard_table = gr.components.Dataframe(
                value=leaderboard_df_qa,
                datatype=TYPES_QA,
                elem_id="leaderboard-table",
                interactive=False,
                visible=True,
            )

            # Dummy leaderboard for handling the case when the user uses backspace key
            hidden_leaderboard_table_for_search = gr.components.Dataframe(
                value=leaderboard_df_qa,
                datatype=TYPES_QA,
                # headers=COLS,
                # datatype=TYPES,
                visible=False,
            )

            # Set search_bar listener
            search_bar.submit(
                update_table,
                [
                    hidden_leaderboard_table_for_search,
                    selected_domains,
                    selected_langs,
                    selected_rerankings,
                    search_bar,
                ],
                leaderboard_table,
            )

            # Set column-wise listener
            for selector in [
                selected_domains, selected_langs, selected_rerankings
            ]:
                selector.change(
                    update_table,
                    [
                        hidden_leaderboard_table_for_search,
                        selected_domains,
                        selected_langs,
                        selected_rerankings,
                        search_bar,
                    ],
                    leaderboard_table,
                    queue=True,
                )

            # set metric listener
            selected_metric.change(
                update_metric_qa,
                [
                    selected_metric,
                    selected_domains,
                    selected_langs,
                    selected_rerankings,
                    search_bar,
                ],
                leaderboard_table,
                queue=True
            )

        with gr.TabItem("Long Doc", elem_id="long-doc-benchmark-tab-table", id=1):
            with gr.Row():
                with gr.Column():
                    with gr.Row():
                        search_bar = gr.Textbox(
                            placeholder=" 🔍 Search for your model (separate multiple queries with `;`) and press ENTER...",
                            show_label=False,
                            elem_id="search-bar-long-doc",
                        )
                        # select the metric
                    selected_metric = gr.Dropdown(
                        choices=metric_list,
                        value=metric_list[1],
                        label="Select the metric",
                        interactive=True,
                        elem_id="metric-select-long-doc",
                    )
                with gr.Column(min_width=320):
                    # select domain
                    with gr.Row():
                        selected_domains = gr.CheckboxGroup(
                            choices=DOMAIN_COLS_LONG_DOC,
                            value=DOMAIN_COLS_LONG_DOC,
                            label="Select the domains",
                            elem_id="domain-column-select-long-doc",
                            interactive=True,
                        )
                    # select language
                    with gr.Row():
                        selected_langs = gr.Dropdown(
                            choices=LANG_COLS_LONG_DOC,
                            value=LANG_COLS_LONG_DOC,
                            label="Select the languages",
                            elem_id="language-column-select-long-doc",
                            multiselect=True,
                            interactive=True
                        )
                    # select reranking model
                    reranking_models = list(frozenset([eval_result.reranking_model for eval_result in raw_data]))
                    with gr.Row():
                        selected_rerankings = gr.CheckboxGroup(
                            choices=reranking_models,
                            value=reranking_models,
                            label="Select the reranking models",
                            elem_id="reranking-select-long-doc",
                            interactive=True
                        )

            leaderboard_table_long_doc = gr.components.Dataframe(
                value=leaderboard_df_long_doc,
                datatype=TYPES_LONG_DOC,
                elem_id="leaderboard-table-long-doc",
                interactive=False,
                visible=True,
            )

            # Dummy leaderboard for handling the case when the user uses backspace key
            hidden_leaderboard_table_for_search = gr.components.Dataframe(
                value=leaderboard_df_long_doc,
                datatype=TYPES_LONG_DOC,
                visible=False,
            )

            # Set search_bar listener
            search_bar.submit(
                update_table_long_doc,
                [
                    hidden_leaderboard_table_for_search,
                    selected_domains,
                    selected_langs,
                    selected_rerankings,
                    search_bar,
                ],
                leaderboard_table_long_doc,
            )

            # Set column-wise listener
            for selector in [
                selected_domains, selected_langs, selected_rerankings
            ]:
                selector.change(
                    update_table_long_doc,
                    [
                        hidden_leaderboard_table_for_search,
                        selected_domains,
                        selected_langs,
                        selected_rerankings,
                        search_bar,
                    ],
                    leaderboard_table_long_doc,
                    queue=True,
                )

            # set metric listener
            selected_metric.change(
                update_metric_long_doc,
                [
                    selected_metric,
                    selected_domains,
                    selected_langs,
                    selected_rerankings,
                    search_bar,
                ],
                leaderboard_table_long_doc,
                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():
                        benchmark_version = gr.Dropdown(
                            ["AIR-Bench_24.04",],
                            value="AIR-Bench_24.04",
                            interactive=True,
                            label="AIR-Bench Version")
                    with gr.Column():
                        model_name = gr.Textbox(label="Model name")
                    with gr.Column():
                        model_url = gr.Textbox(label="Model URL")
                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():
                    submit_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,
                        benchmark_version,
                        submit_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()