File size: 19,031 Bytes
8a5da23
ee31436
 
 
 
f1a09e2
8a5da23
 
d49a57f
c755378
dac62d3
56bf4e8
ee31436
56bf4e8
 
010a64a
56bf4e8
39125ad
d062868
007425a
8a5da23
ad27ecb
8a5da23
 
caa4425
39125ad
4d9df48
 
 
24c1bd7
4d9df48
 
 
 
ee31436
39125ad
ee31436
4d9df48
7e4686f
4d9df48
06a62df
 
 
ad27ecb
d49a57f
4d9df48
010a64a
7e4686f
4d9df48
7e4686f
508b863
4d9df48
5c401da
8a5da23
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6162a3c
8a5da23
 
 
 
5c401da
8a5da23
 
5c401da
0537749
5c401da
 
 
 
 
8a5da23
 
5c401da
8a5da23
 
c755378
 
 
 
8a5da23
c755378
 
 
5c401da
 
56bf4e8
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ee31436
 
 
 
56bf4e8
 
ee31436
 
 
56bf4e8
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
508b863
56bf4e8
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ee31436
 
56bf4e8
ee31436
 
 
 
56bf4e8
ee31436
 
 
 
 
 
 
 
 
56bf4e8
 
ee31436
 
 
 
 
 
 
 
 
56bf4e8
 
ee31436
 
 
 
 
 
 
 
 
 
56bf4e8
 
ee31436
 
 
 
 
 
 
 
 
 
56bf4e8
ee31436
 
 
 
 
 
 
 
56bf4e8
ee31436
 
 
 
 
 
56bf4e8
ee31436
 
 
 
 
 
 
 
 
 
56bf4e8
ee31436
 
 
 
 
 
 
 
 
 
 
 
 
 
56bf4e8
 
ee31436
 
 
 
 
8067b48
bdf31b0
d062868
 
bdf31b0
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
508b863
bdf31b0
 
 
 
 
 
 
d062868
 
 
 
 
e157bd5
d062868
8067b48
d062868
f6be763
8067b48
 
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
import logging
import gradio as gr
import pandas as pd
from apscheduler.schedulers.background import BackgroundScheduler
from huggingface_hub import snapshot_download
import src.envs as envs
from main_backend import PENDING_STATUS, RUNNING_STATUS, FINISHED_STATUS, FAILED_STATUS
from src.backend import sort_queue
from src.envs import EVAL_REQUESTS_PATH, EVAL_RESULTS_PATH, QUEUE_REPO, RESULTS_REPO
import src.backend.manage_requests as manage_requests
import socket
import src.display.about as about
from src.display.css_html_js import custom_css
import src.display.utils as utils
import src.populate as populate
from src.populate import get_evaluation_queue_df, get_leaderboard_df
import src.submission.submit as submit
import os
import datetime
import spacy_transformers
import pprint
import src.backend.run_eval_suite as run_eval_suite

pp = pprint.PrettyPrinter(width=80)
TOKEN = os.environ.get("H4_TOKEN", None)
print("TOKEN", TOKEN)

def ui_snapshot_download(repo_id, local_dir, repo_type, tqdm_class, etag_timeout):
    try:
        print("local",local_dir)
        snapshot_download(repo_id=repo_id, local_dir=local_dir, repo_type=repo_type, tqdm_class=tqdm_class, etag_timeout=etag_timeout)
    except Exception as e:
        restart_space()

def restart_space():
    envs.API.restart_space(repo_id=envs.REPO_ID, token=TOKEN)

def init_space():
    #dataset_df = get_dataset_summary_table(file_path='blog/Hallucination-Leaderboard-Summary.csv')


    ui_snapshot_download(repo_id=QUEUE_REPO, local_dir=EVAL_REQUESTS_PATH, repo_type="dataset", tqdm_class=None, etag_timeout=30)
    ui_snapshot_download(repo_id=RESULTS_REPO, local_dir=EVAL_RESULTS_PATH, repo_type="dataset", tqdm_class=None, etag_timeout=30)

    original_df = get_leaderboard_df(EVAL_RESULTS_PATH, EVAL_REQUESTS_PATH, utils.COLS, utils.BENCHMARK_COLS)

    finished_eval_queue_df, running_eval_queue_df, pending_eval_queue_df = get_evaluation_queue_df(EVAL_REQUESTS_PATH, utils.EVAL_COLS)
    return original_df, finished_eval_queue_df, running_eval_queue_df, pending_eval_queue_df

original_df, finished_eval_queue_df, running_eval_queue_df, pending_eval_queue_df = init_space()
leaderboard_df = original_df.copy()

def process_pending_evals():
    # if len(pending_eval_queue_df) == 0:
    #     print("No pending evaluations found.")
    #     return
    #
    # for _, eval_request in pending_eval_queue_df.iterrows():
    #     import re
    #     model_link = eval_request['model']
    #     match = re.search(r'>([^<]+)<', model_link)
    #     if match:
    #         eval_request['model'] = match.group(1)  # 赋值给 eval_request['model']
    #     else:
    #         eval_request['model'] = model_link  # 如果无法匹配,保留原始字符串
    #
    #     print(f"Evaluating model: {eval_request['model']}")
    #
    #     # 调用评估函数
    #     run_eval_suite.run_evaluation(
    #         eval_request=eval_request,
    #         local_dir=envs.EVAL_RESULTS_PATH_BACKEND,
    #         results_repo=envs.RESULTS_REPO,
    #         batch_size=1,
    #         device=envs.DEVICE,
    #         no_cache=True,
    #         need_check=False,  # 根据需要设定是否需要检查
    #         write_results=False  # 根据需要设定是否写入结果
    #     )
    #     print(f"Finished evaluation for model: {eval_request['model']}")
    #     # Update the status to FINISHED
    #     manage_requests.set_eval_request(
    #         api=envs.API,
    #         eval_request=eval_request,
    #         new_status="FINISHED",
    #         hf_repo=envs.QUEUE_REPO,
    #         local_dir=envs.EVAL_REQUESTS_PATH_BACKEND
    #     )
    current_pending_status = [PENDING_STATUS]
    print('_________________')
    manage_requests.check_completed_evals(
        api=envs.API,
        checked_status=RUNNING_STATUS,
        completed_status=FINISHED_STATUS,
        failed_status=FAILED_STATUS,
        hf_repo=envs.QUEUE_REPO,
        local_dir=envs.EVAL_REQUESTS_PATH_BACKEND,
        hf_repo_results=envs.RESULTS_REPO,
        local_dir_results=envs.EVAL_RESULTS_PATH_BACKEND
    )
    logging.info("Checked completed evals")
    eval_requests = manage_requests.get_eval_requests(
        job_status=current_pending_status,
        hf_repo=envs.QUEUE_REPO,
        local_dir=envs.EVAL_REQUESTS_PATH_BACKEND
    )
    logging.info("Got eval requests")
    eval_requests = sort_queue.sort_models_by_priority(api=envs.API, models=eval_requests)
    logging.info("Sorted eval requests")

    print(f"Found {len(eval_requests)} {','.join(current_pending_status)} eval requests")
    if len(eval_requests) == 0:
        print("No eval requests found. Exiting.")
        return

    for eval_request in eval_requests:
        pp.pprint(eval_request)
        run_eval_suite.run_evaluation(
            eval_request=eval_request,
            local_dir=envs.EVAL_RESULTS_PATH_BACKEND,
            results_repo=envs.RESULTS_REPO,
            batch_size=1,
            device=envs.DEVICE,
            no_cache=True,
            need_check= False,
            write_results= False
        )
        logging.info(f"Eval finished for model {eval_request.model}, now setting status to finished")

        # Update the status to FINISHED
        manage_requests.set_eval_request(
            api=envs.API,
            eval_request=eval_request,
            new_status=FINISHED_STATUS,
            hf_repo=envs.QUEUE_REPO,
            local_dir=envs.EVAL_REQUESTS_PATH_BACKEND
        )


# Searching and filtering
def update_table(
    hidden_df: pd.DataFrame,
    columns: list,
    type_query: list,
    precision_query: str,
    size_query: list,
    show_deleted: bool,
    query: str,
):
    filtered_df = filter_models(hidden_df, type_query, size_query, precision_query, show_deleted)
    filtered_df = filter_queries(query, filtered_df)
    df = select_columns(filtered_df, columns)
    return df


def search_table(df: pd.DataFrame, query: str) -> pd.DataFrame:
    return df[(df[utils.AutoEvalColumn.dummy.name].str.contains(query, case=False))]


def select_columns(df: pd.DataFrame, columns: list) -> pd.DataFrame:
    always_here_cols = [
        utils.AutoEvalColumn.model_type_symbol.name,
        utils.AutoEvalColumn.model.name,
    ]
    # We use COLS to maintain sorting
    filtered_df = df[
        always_here_cols + [c for c in utils.COLS if c in df.columns and c in columns] + [utils.AutoEvalColumn.dummy.name]
    ]
    return filtered_df


def filter_queries(query: str, filtered_df: pd.DataFrame) -> pd.DataFrame:
    final_df = []
    if query != "":
        queries = [q.strip() for q in query.split(";")]
        for _q in queries:
            _q = _q.strip()
            if _q != "":
                temp_filtered_df = search_table(filtered_df, _q)
                if len(temp_filtered_df) > 0:
                    final_df.append(temp_filtered_df)
        if len(final_df) > 0:
            filtered_df = pd.concat(final_df)
            filtered_df = filtered_df.drop_duplicates(
                subset=[utils.AutoEvalColumn.model.name, utils.AutoEvalColumn.precision.name, utils.AutoEvalColumn.revision.name]
            )

    return filtered_df


def filter_models(
    df: pd.DataFrame, type_query: list, size_query: list, precision_query: list, show_deleted: bool
) -> pd.DataFrame:
    # Show all models
    # if show_deleted:
    #   filtered_df = df
    # else:  # Show only still on the hub models
        # filtered_df = df[df[utils.AutoEvalColumn.still_on_hub.name]]

    filtered_df = df
    
    type_emoji = [t[0] for t in type_query]
    filtered_df = filtered_df.loc[df[utils.AutoEvalColumn.model_type_symbol.name].isin(type_emoji)]
    filtered_df = filtered_df.loc[df[utils.AutoEvalColumn.precision.name].isin(precision_query + ["None"])]

    numeric_interval = pd.IntervalIndex(sorted([utils.NUMERIC_INTERVALS[s] for s in size_query]))
    params_column = pd.to_numeric(df[utils.AutoEvalColumn.params.name], errors="coerce")
    mask = params_column.apply(lambda x: any(numeric_interval.contains(x)))
    filtered_df = filtered_df.loc[mask]

    return filtered_df


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

    with gr.Tabs(elem_classes="tab-buttons") as tabs:
        with gr.TabItem("🏅 LLM Benchmark", elem_id="llm-benchmark-tab-table", id=0):
            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",
                        )
                    with gr.Row():
                        shown_columns = gr.CheckboxGroup(
                            choices=[
                                c.name
                                for c in utils.fields(utils.AutoEvalColumn)
                                if not c.hidden and not c.never_hidden and not c.dummy
                            ],
                            value=[
                                c.name
                                for c in utils.fields(utils.AutoEvalColumn)
                                if c.displayed_by_default and not c.hidden and not c.never_hidden
                            ],
                            label="Select columns to show",
                            elem_id="column-select",
                            interactive=True,
                        )
                    with gr.Row():
                        deleted_models_visibility = gr.Checkbox(
                            value=False, label="Show gated/private/deleted models", interactive=True
                        )
                with gr.Column(min_width=320):
                    #with gr.Box(elem_id="box-filter"):
                    filter_columns_type = gr.CheckboxGroup(
                        label="Model types",
                        choices=[t.to_str() for t in utils.ModelType],
                        value=[t.to_str() for t in utils.ModelType],
                        interactive=True,
                        elem_id="filter-columns-type",
                    )
                    filter_columns_precision = gr.CheckboxGroup(
                        label="Precision",
                        choices=[i.value.name for i in utils.Precision],
                        value=[i.value.name for i in utils.Precision],
                        interactive=True,
                        elem_id="filter-columns-precision",
                    )
                    filter_columns_size = gr.CheckboxGroup(
                        label="Model sizes (in billions of parameters)",
                        choices=list(utils.NUMERIC_INTERVALS.keys()),
                        value=list(utils.NUMERIC_INTERVALS.keys()),
                        interactive=True,
                        elem_id="filter-columns-size",
                    )

            leaderboard_table = gr.components.Dataframe(
                value=leaderboard_df[
                    [c.name for c in utils.fields(utils.AutoEvalColumn) if c.never_hidden]
                    + shown_columns.value
                    + [utils.AutoEvalColumn.dummy.name]
                ],
                headers=[c.name for c in utils.fields(utils.AutoEvalColumn) if c.never_hidden] + shown_columns.value,
                datatype=utils.TYPES,
                elem_id="leaderboard-table",
                interactive=False,
                visible=True,
                column_widths=["2%", "33%"]
            )

            # Dummy leaderboard for handling the case when the user uses backspace key
            hidden_leaderboard_table_for_search = gr.components.Dataframe(
                value=original_df[utils.COLS],
                headers=utils.COLS,
                datatype=utils.TYPES,
                visible=False,
            )
            search_bar.submit(
                update_table,
                [
                    hidden_leaderboard_table_for_search,
                    shown_columns,
                    filter_columns_type,
                    filter_columns_precision,
                    filter_columns_size,
                    deleted_models_visibility,
                    search_bar,
                ],
                leaderboard_table,
            )
            for selector in [shown_columns, filter_columns_type, filter_columns_precision, filter_columns_size, deleted_models_visibility]:
                selector.change(
                    update_table,
                    [
                        hidden_leaderboard_table_for_search,
                        shown_columns,
                        filter_columns_type,
                        filter_columns_precision,
                        filter_columns_size,
                        deleted_models_visibility,
                        search_bar,
                    ],
                    leaderboard_table,
                    queue=True,
                )

        with gr.TabItem("📝 About", elem_id="llm-benchmark-tab-table", id=2):
            gr.Markdown(about.LLM_BENCHMARKS_TEXT, elem_classes="markdown-text")

        with gr.TabItem("🚀 Submit here! ", elem_id="llm-benchmark-tab-table", id=3):
            with gr.Column():
                with gr.Row():
                    gr.Markdown(about.EVALUATION_QUEUE_TEXT, elem_classes="markdown-text")

                with gr.Column():
                    with gr.Accordion(
                        f"✅ Finished Evaluations ({len(finished_eval_queue_df)})",
                        open=False,
                    ):
                        with gr.Row():
                            finished_eval_table = gr.components.Dataframe(
                                value=finished_eval_queue_df,
                                headers=utils.EVAL_COLS,
                                datatype=utils.EVAL_TYPES,
                                row_count=5,
                            )
                    with gr.Accordion(
                        f"🔄 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=utils.EVAL_COLS,
                                datatype=utils.EVAL_TYPES,
                                row_count=5,
                            )

                    with gr.Accordion(
                        f"⏳ 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=utils.EVAL_COLS,
                                datatype=utils.EVAL_TYPES,
                                row_count=5,
                            )
            with gr.Row():
                gr.Markdown("# ✉️✨ 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 utils.ModelType if t != utils.ModelType.Unknown],
                        label="Model type",
                        multiselect=False,
                        value=None,
                        interactive=True,
                    )

                with gr.Column():
                    precision = gr.Dropdown(
                        choices=[i.value.name for i in utils.Precision if i != utils.Precision.Unknown],
                        label="Precision",
                        multiselect=False,
                        value="float16",
                        interactive=True,
                    )
                    weight_type = gr.Dropdown(
                        choices=[i.value.name for i in utils.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(
                submit.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("📙 Citation", open=False):
            citation_button = gr.Textbox(
                value=about.CITATION_BUTTON_TEXT,
                label=about.CITATION_BUTTON_LABEL,
                lines=20,
                elem_id="citation-button",
                show_copy_button=True,
            )


# 在初始化完成后调用
# original_df, finished_eval_queue_df, running_eval_queue_df, pending_eval_queue_df = init_space()
# process_pending_evals()

# try:
#     print(envs.EVAL_REQUESTS_PATH)
#     snapshot_download(
#         repo_id=envs.QUEUE_REPO, local_dir=envs.EVAL_REQUESTS_PATH, repo_type="dataset", tqdm_class=None, etag_timeout=30
#     )
# except Exception:
#     restart_space()
# try:
#     print(envs.EVAL_RESULTS_PATH)
#     snapshot_download(
#         repo_id=envs.RESULTS_REPO, local_dir=envs.EVAL_RESULTS_PATH, repo_type="dataset", tqdm_class=None, etag_timeout=30
#     )
# except Exception:
#     restart_space()

# raw_data, original_df = populate.get_leaderboard_df(envs.RESULTS_REPO, envs.QUEUE_REPO, utils.COLS, utils.BENCHMARK_COLS)




(
    finished_eval_queue_df,
    running_eval_queue_df,
    pending_eval_queue_df,
) = populate.get_evaluation_queue_df(envs.EVAL_REQUESTS_PATH, utils.EVAL_COLS)



def background_init_and_process():
    global original_df, finished_eval_queue_df, running_eval_queue_df, pending_eval_queue_df
    original_df, finished_eval_queue_df, running_eval_queue_df, pending_eval_queue_df = init_space()
    #process_pending_evals()

scheduler = BackgroundScheduler()
scheduler.add_job(background_init_and_process, 'date', run_date=datetime.datetime.now())  # 立即执行
scheduler.add_job(restart_space, "interval", seconds=36000)
scheduler.start()
demo.queue(default_concurrency_limit=40).launch()