ubowang commited on
Commit
003d870
β€’
1 Parent(s): 06e0b4f

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +40 -325
app.py CHANGED
@@ -1,345 +1,60 @@
1
- import subprocess
2
- import gradio as gr
3
- import pandas as pd
4
- from apscheduler.schedulers.background import BackgroundScheduler
5
- from huggingface_hub import snapshot_download
6
 
7
- from src.about import (
8
- CITATION_BUTTON_LABEL,
9
- CITATION_BUTTON_TEXT,
10
- EVALUATION_QUEUE_TEXT,
11
- INTRODUCTION_TEXT,
12
- LLM_BENCHMARKS_TEXT,
13
- TITLE,
14
- )
15
- from src.display.css_html_js import custom_css
16
- from src.display.utils import (
17
- BENCHMARK_COLS,
18
- COLS,
19
- EVAL_COLS,
20
- EVAL_TYPES,
21
- NUMERIC_INTERVALS,
22
- TYPES,
23
- AutoEvalColumn,
24
- ModelType,
25
- fields,
26
- WeightType,
27
- Precision
28
- )
29
- from src.envs import API, EVAL_REQUESTS_PATH, EVAL_RESULTS_PATH, QUEUE_REPO, REPO_ID, RESULTS_REPO, TOKEN
30
- from src.populate import get_evaluation_queue_df, get_leaderboard_df
31
- from src.submission.submit import add_new_eval
32
 
 
33
 
34
- def restart_space():
35
- API.restart_space(repo_id=REPO_ID)
36
-
37
- try:
38
- print(EVAL_REQUESTS_PATH)
39
- snapshot_download(
40
- repo_id=QUEUE_REPO, local_dir=EVAL_REQUESTS_PATH, repo_type="dataset", tqdm_class=None, etag_timeout=30, token=TOKEN
41
- )
42
- except Exception:
43
- restart_space()
44
- try:
45
- print(EVAL_RESULTS_PATH)
46
- snapshot_download(
47
- repo_id=RESULTS_REPO, local_dir=EVAL_RESULTS_PATH, repo_type="dataset", tqdm_class=None, etag_timeout=30, token=TOKEN
48
  )
49
- except Exception:
50
- restart_space()
51
-
52
-
53
- raw_data, original_df = get_leaderboard_df(EVAL_RESULTS_PATH, EVAL_REQUESTS_PATH, COLS, BENCHMARK_COLS)
54
- leaderboard_df = original_df.copy()
55
-
56
- (
57
- finished_eval_queue_df,
58
- running_eval_queue_df,
59
- pending_eval_queue_df,
60
- ) = get_evaluation_queue_df(EVAL_REQUESTS_PATH, EVAL_COLS)
61
-
62
-
63
- # Searching and filtering
64
- def update_table(
65
- hidden_df: pd.DataFrame,
66
- columns: list,
67
- type_query: list,
68
- precision_query: str,
69
- size_query: list,
70
- show_deleted: bool,
71
- query: str,
72
- ):
73
- filtered_df = filter_models(hidden_df, type_query, size_query, precision_query, show_deleted)
74
- filtered_df = filter_queries(query, filtered_df)
75
- df = select_columns(filtered_df, columns)
76
- return df
77
-
78
-
79
- def search_table(df: pd.DataFrame, query: str) -> pd.DataFrame:
80
- return df[(df[AutoEvalColumn.model.name].str.contains(query, case=False))]
81
-
82
-
83
- def select_columns(df: pd.DataFrame, columns: list) -> pd.DataFrame:
84
- always_here_cols = [
85
- AutoEvalColumn.model_type_symbol.name,
86
- AutoEvalColumn.model.name,
87
- ]
88
- # We use COLS to maintain sorting
89
- filtered_df = df[
90
- always_here_cols + [c for c in COLS if c in df.columns and c in columns]
91
- ]
92
- return filtered_df
93
-
94
-
95
- def filter_queries(query: str, filtered_df: pd.DataFrame) -> pd.DataFrame:
96
- final_df = []
97
- if query != "":
98
- queries = [q.strip() for q in query.split(";")]
99
- for _q in queries:
100
- _q = _q.strip()
101
- if _q != "":
102
- temp_filtered_df = search_table(filtered_df, _q)
103
- if len(temp_filtered_df) > 0:
104
- final_df.append(temp_filtered_df)
105
- if len(final_df) > 0:
106
- filtered_df = pd.concat(final_df)
107
- filtered_df = filtered_df.drop_duplicates(
108
- subset=[AutoEvalColumn.model.name, AutoEvalColumn.precision.name, AutoEvalColumn.revision.name]
109
- )
110
-
111
- return filtered_df
112
-
113
-
114
- def filter_models(
115
- df: pd.DataFrame, type_query: list, size_query: list, precision_query: list, show_deleted: bool
116
- ) -> pd.DataFrame:
117
- # Show all models
118
- if show_deleted:
119
- filtered_df = df
120
- else: # Show only still on the hub models
121
- filtered_df = df[df[AutoEvalColumn.still_on_hub.name] == True]
122
-
123
- type_emoji = [t[0] for t in type_query]
124
- filtered_df = filtered_df.loc[df[AutoEvalColumn.model_type_symbol.name].isin(type_emoji)]
125
- filtered_df = filtered_df.loc[df[AutoEvalColumn.precision.name].isin(precision_query + ["None"])]
126
-
127
- numeric_interval = pd.IntervalIndex(sorted([NUMERIC_INTERVALS[s] for s in size_query]))
128
- params_column = pd.to_numeric(df[AutoEvalColumn.params.name], errors="coerce")
129
- mask = params_column.apply(lambda x: any(numeric_interval.contains(x)))
130
- filtered_df = filtered_df.loc[mask]
131
-
132
- return filtered_df
133
-
134
-
135
- demo = gr.Blocks(css=custom_css)
136
- with demo:
137
- gr.HTML(TITLE)
138
- gr.Markdown(INTRODUCTION_TEXT, elem_classes="markdown-text")
139
-
140
  with gr.Tabs(elem_classes="tab-buttons") as tabs:
141
- with gr.TabItem("πŸ… LLM Benchmark", elem_id="llm-benchmark-tab-table", id=0):
 
142
  with gr.Row():
143
- with gr.Column():
144
- with gr.Row():
145
- search_bar = gr.Textbox(
146
- placeholder=" πŸ” Search for your model (separate multiple queries with `;`) and press ENTER...",
147
- show_label=False,
148
- elem_id="search-bar",
149
- )
150
- with gr.Row():
151
- shown_columns = gr.CheckboxGroup(
152
- choices=[
153
- c.name
154
- for c in fields(AutoEvalColumn)
155
- if not c.hidden and not c.never_hidden
156
- ],
157
- value=[
158
- c.name
159
- for c in fields(AutoEvalColumn)
160
- if c.displayed_by_default and not c.hidden and not c.never_hidden
161
- ],
162
- label="Select columns to show",
163
- elem_id="column-select",
164
- interactive=True,
165
- )
166
- with gr.Row():
167
- deleted_models_visibility = gr.Checkbox(
168
- value=False, label="Show gated/private/deleted models", interactive=True
169
- )
170
- with gr.Column(min_width=320):
171
- #with gr.Box(elem_id="box-filter"):
172
- filter_columns_type = gr.CheckboxGroup(
173
- label="Model types",
174
- choices=[t.to_str() for t in ModelType],
175
- value=[t.to_str() for t in ModelType],
176
- interactive=True,
177
- elem_id="filter-columns-type",
178
- )
179
- filter_columns_precision = gr.CheckboxGroup(
180
- label="Precision",
181
- choices=[i.value.name for i in Precision],
182
- value=[i.value.name for i in Precision],
183
- interactive=True,
184
- elem_id="filter-columns-precision",
185
- )
186
- filter_columns_size = gr.CheckboxGroup(
187
- label="Model sizes (in billions of parameters)",
188
- choices=list(NUMERIC_INTERVALS.keys()),
189
- value=list(NUMERIC_INTERVALS.keys()),
190
- interactive=True,
191
- elem_id="filter-columns-size",
192
  )
 
 
 
193
 
194
- leaderboard_table = gr.components.Dataframe(
195
- value=leaderboard_df[
196
- [c.name for c in fields(AutoEvalColumn) if c.never_hidden]
197
- + shown_columns.value
198
- ],
199
- headers=[c.name for c in fields(AutoEvalColumn) if c.never_hidden] + shown_columns.value,
200
- datatype=TYPES,
201
- elem_id="leaderboard-table",
202
  interactive=False,
203
  visible=True,
204
  )
 
 
205
 
206
- # Dummy leaderboard for handling the case when the user uses backspace key
207
- hidden_leaderboard_table_for_search = gr.components.Dataframe(
208
- value=original_df[COLS],
209
- headers=COLS,
210
- datatype=TYPES,
211
- visible=False,
212
- )
213
- search_bar.submit(
214
- update_table,
215
- [
216
- hidden_leaderboard_table_for_search,
217
- shown_columns,
218
- filter_columns_type,
219
- filter_columns_precision,
220
- filter_columns_size,
221
- deleted_models_visibility,
222
- search_bar,
223
- ],
224
- leaderboard_table,
225
- )
226
- for selector in [shown_columns, filter_columns_type, filter_columns_precision, filter_columns_size, deleted_models_visibility]:
227
- selector.change(
228
- update_table,
229
- [
230
- hidden_leaderboard_table_for_search,
231
- shown_columns,
232
- filter_columns_type,
233
- filter_columns_precision,
234
- filter_columns_size,
235
- deleted_models_visibility,
236
- search_bar,
237
- ],
238
- leaderboard_table,
239
- queue=True,
240
- )
241
-
242
- with gr.TabItem("πŸ“ About", elem_id="llm-benchmark-tab-table", id=2):
243
- gr.Markdown(LLM_BENCHMARKS_TEXT, elem_classes="markdown-text")
244
 
245
- with gr.TabItem("πŸš€ Submit here! ", elem_id="llm-benchmark-tab-table", id=3):
246
- with gr.Column():
247
- with gr.Row():
248
- gr.Markdown(EVALUATION_QUEUE_TEXT, elem_classes="markdown-text")
249
-
250
- with gr.Column():
251
- with gr.Accordion(
252
- f"βœ… Finished Evaluations ({len(finished_eval_queue_df)})",
253
- open=False,
254
- ):
255
- with gr.Row():
256
- finished_eval_table = gr.components.Dataframe(
257
- value=finished_eval_queue_df,
258
- headers=EVAL_COLS,
259
- datatype=EVAL_TYPES,
260
- row_count=5,
261
- )
262
- with gr.Accordion(
263
- f"πŸ”„ Running Evaluation Queue ({len(running_eval_queue_df)})",
264
- open=False,
265
- ):
266
- with gr.Row():
267
- running_eval_table = gr.components.Dataframe(
268
- value=running_eval_queue_df,
269
- headers=EVAL_COLS,
270
- datatype=EVAL_TYPES,
271
- row_count=5,
272
- )
273
-
274
- with gr.Accordion(
275
- f"⏳ Pending Evaluation Queue ({len(pending_eval_queue_df)})",
276
- open=False,
277
- ):
278
- with gr.Row():
279
- pending_eval_table = gr.components.Dataframe(
280
- value=pending_eval_queue_df,
281
- headers=EVAL_COLS,
282
- datatype=EVAL_TYPES,
283
- row_count=5,
284
- )
285
  with gr.Row():
286
- gr.Markdown("# βœ‰οΈβœ¨ Submit your model here!", elem_classes="markdown-text")
287
 
288
  with gr.Row():
289
- with gr.Column():
290
- model_name_textbox = gr.Textbox(label="Model name")
291
- revision_name_textbox = gr.Textbox(label="Revision commit", placeholder="main")
292
- model_type = gr.Dropdown(
293
- choices=[t.to_str(" : ") for t in ModelType if t != ModelType.Unknown],
294
- label="Model type",
295
- multiselect=False,
296
- value=None,
297
- interactive=True,
298
- )
299
 
300
- with gr.Column():
301
- precision = gr.Dropdown(
302
- choices=[i.value.name for i in Precision if i != Precision.Unknown],
303
- label="Precision",
304
- multiselect=False,
305
- value="float16",
306
- interactive=True,
307
- )
308
- weight_type = gr.Dropdown(
309
- choices=[i.value.name for i in WeightType],
310
- label="Weights type",
311
- multiselect=False,
312
- value="Original",
313
- interactive=True,
314
- )
315
- base_model_name_textbox = gr.Textbox(label="Base model (for delta or adapter weights)")
316
 
317
- submit_button = gr.Button("Submit Eval")
318
- submission_result = gr.Markdown()
319
- submit_button.click(
320
- add_new_eval,
321
- [
322
- model_name_textbox,
323
- base_model_name_textbox,
324
- revision_name_textbox,
325
- precision,
326
- weight_type,
327
- model_type,
328
- ],
329
- submission_result,
330
- )
331
 
332
- with gr.Row():
333
- with gr.Accordion("πŸ“™ Citation", open=False):
334
- citation_button = gr.Textbox(
335
- value=CITATION_BUTTON_TEXT,
336
- label=CITATION_BUTTON_LABEL,
337
- lines=20,
338
- elem_id="citation-button",
339
- show_copy_button=True,
340
- )
341
 
342
- scheduler = BackgroundScheduler()
343
- scheduler.add_job(restart_space, "interval", seconds=1800)
344
- scheduler.start()
345
- demo.queue(default_concurrency_limit=40).launch()
 
1
+ from utils import *
 
 
 
 
2
 
3
+ global data_component
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4
 
5
+ block = gr.Blocks()
6
 
7
+ with block:
8
+ gr.Markdown(
9
+ LEADERBORAD_INTRODUCTION
 
 
 
 
 
 
 
 
 
 
 
10
  )
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
11
  with gr.Tabs(elem_classes="tab-buttons") as tabs:
12
+ # Table 1
13
+ with gr.TabItem("πŸ“Š MMLU-Pro", elem_id="qa-tab-table1", id=1):
14
  with gr.Row():
15
+ with gr.Accordion("Citation", open=False):
16
+ citation_button = gr.Textbox(
17
+ value=CITATION_BUTTON_TEXT,
18
+ label=CITATION_BUTTON_LABEL,
19
+ elem_id="citation-button",
20
+ lines=20,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
21
  )
22
+ gr.Markdown(
23
+ TABLE_INTRODUCTION
24
+ )
25
 
26
+ data_component = gr.components.Dataframe(
27
+ value=get_df(),
28
+ headers=COLUMN_NAMES,
29
+ type="pandas",
30
+ datatype=DATA_TITILE_TYPE,
 
 
 
31
  interactive=False,
32
  visible=True,
33
  )
34
+ refresh_button = gr.Button("Refresh")
35
+ refresh_button.click(fn=refresh_data, outputs=data_component)
36
 
37
+ # table 2
38
+ with gr.TabItem("πŸ“ About", elem_id="qa-tab-table2", id=2):
39
+ gr.Markdown(LEADERBORAD_INFO, elem_classes="markdown-text")
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
40
 
41
+ # table 3
42
+ with gr.TabItem("πŸš€ Submit here! ", elem_id="submit-tab", id=3):
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
43
  with gr.Row():
44
+ gr.Markdown(SUBMIT_INTRODUCTION, elem_classes="markdown-text")
45
 
46
  with gr.Row():
47
+ gr.Markdown("# βœ‰οΈβœ¨ Submit your json file here!", elem_classes="markdown-text")
 
 
 
 
 
 
 
 
 
48
 
49
+ with gr.Column():
50
+ input_file = gr.components.File(label="Click to Upload a json File", file_count="single", type='binary')
51
+ submit_button = gr.Button("Submit Results")
 
 
 
 
 
 
 
 
 
 
 
 
 
52
 
53
+ submission_result = gr.Markdown()
54
+ submit_button.click(
55
+ add_new_eval,
56
+ inputs=[input_file],
57
+ )
 
 
 
 
 
 
 
 
 
58
 
59
+ block.launch()
 
 
 
 
 
 
 
 
60