File size: 8,178 Bytes
8f809e2
58c39e0
 
 
0607989
53fe897
8c47a22
53fe897
666860b
53fe897
cdcc48e
53fe897
cdcc48e
53fe897
cdcc48e
53fe897
cdcc48e
85a8a8b
0607989
 
 
7eb3c10
0607989
 
9e4233f
5b8d6d5
9e4233f
 
3573a39
 
85a8a8b
9e4233f
5b8d6d5
9e4233f
be473e6
136af2d
8092547
136af2d
96a1184
 
 
 
9e4233f
 
d2a76c0
9e4233f
 
 
8c47a22
 
 
 
 
 
 
3573a39
9e4233f
53fe897
 
3573a39
5559b52
d2a76c0
5559b52
7055d8b
5b8d6d5
 
d2a76c0
53fe897
5b8d6d5
53fe897
 
9e4233f
7055d8b
 
 
9e4233f
5311dba
666860b
3573a39
9e4233f
3573a39
 
 
be473e6
 
 
 
 
5b8d6d5
be473e6
 
3573a39
5b8d6d5
be473e6
 
3573a39
8f114e2
666860b
58c39e0
8092547
 
 
0607989
8092547
 
02cf07d
 
 
 
 
 
 
 
 
 
58c39e0
 
 
9e4233f
 
 
 
 
5b8d6d5
9e4233f
 
3573a39
8f809e2
53fe897
0607989
8f114e2
53fe897
 
 
3573a39
35be7f4
136af2d
 
8c47a22
 
 
8f114e2
8c47a22
5311dba
 
 
 
7f86019
35be7f4
 
666860b
35be7f4
 
7055d8b
8c47a22
 
666860b
 
cdcc48e
 
 
 
 
 
3573a39
 
8f809e2
3573a39
136af2d
3573a39
 
 
9e4233f
1c00552
 
 
 
 
 
 
 
 
 
3573a39
 
 
2694247
136af2d
3573a39
53fe897
 
5b8d6d5
 
 
 
 
 
53fe897
7055d8b
 
 
 
5311dba
7055d8b
 
53fe897
5b8d6d5
 
 
 
3573a39
5b8d6d5
3573a39
 
 
 
 
5b8d6d5
3573a39
 
7055d8b
5311dba
3573a39
 
5b8d6d5
7055d8b
3573a39
 
 
9e4233f
 
 
 
3573a39
9e4233f
8f809e2
3573a39
 
 
 
 
 
b56bfdc
 
 
 
 
 
 
 
 
3573a39
 
8f809e2
 
5b8d6d5
3573a39
8f809e2
8c47a22
cdcc48e
8c47a22
 
 
 
 
3573a39
 
 
8f809e2
5b8d6d5
8f809e2
8c47a22
cdcc48e
8c47a22
 
 
 
 
3573a39
 
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
import uuid

import gradio as gr

from io_utils import read_scanners, write_scanners
from text_classification_ui_helpers import (
    get_related_datasets_from_leaderboard,
    align_columns_and_show_prediction,
    get_dataset_splits,
    check_dataset,
    precheck_model_ds_enable_example_btn,
    try_submit,
    empty_column_mapping,
    write_column_mapping_to_config,
    enable_run_btn,
)

import logging
from wordings import (
  CONFIRM_MAPPING_DETAILS_MD, 
  INTRODUCTION_MD, 
  LOG_IN_TIPS, 
  CHECK_LOG_SECTION_RAW,
)

MAX_LABELS = 40
MAX_FEATURES = 20

EXAMPLE_MODEL_ID = "cardiffnlp/twitter-roberta-base-sentiment-latest"
CONFIG_PATH = "./config.yaml"
logger = logging.getLogger(__name__)

def get_demo():
    with gr.Row():
        gr.Markdown(INTRODUCTION_MD)
        uid_label = gr.Textbox(
            label="Evaluation ID:", value=uuid.uuid4, visible=False, interactive=False
        )
    
    with gr.Accordion(label="Login to Use This Space", open=True):
        gr.HTML(LOG_IN_TIPS)
        gr.LoginButton()
    with gr.Row():
        model_id_input = gr.Textbox(
            label="Hugging Face Model id",
            placeholder=EXAMPLE_MODEL_ID + " (press enter to confirm)",
        )

        with gr.Column():
            dataset_id_input = gr.Dropdown(
                choices=[],
                value="",
                allow_custom_value=True,
                label="Hugging Face Dataset id",
            )

    with gr.Row():
        dataset_config_input = gr.Dropdown(label="Dataset Config", visible=False, allow_custom_value=True)
        dataset_split_input = gr.Dropdown(label="Dataset Split", visible=False, allow_custom_value=True)

    with gr.Row():
        first_line_ds = gr.DataFrame(label="Dataset Preview", visible=False)
    with gr.Row():
        loading_dataset_info = gr.HTML(visible=True)
    with gr.Row():
        example_btn = gr.Button(
            "Validate Model & Dataset",
            visible=True,
            variant="primary",
            interactive=False,
        )
    with gr.Row():
        loading_validation = gr.HTML(visible=True)
    with gr.Row():
        validation_result = gr.HTML(visible=False)
    with gr.Row():
        example_input = gr.Textbox(label="Example Input", visible=False, interactive=False)
        example_prediction = gr.Label(label="Model Sample Prediction", visible=False)

    with gr.Row():
        with gr.Accordion(
            label="Label and Feature Mapping", visible=False, open=False
        ) as column_mapping_accordion:
            with gr.Row():
                gr.Markdown(CONFIRM_MAPPING_DETAILS_MD)
            column_mappings = []
            with gr.Row():
                with gr.Column():
                    gr.Markdown("# Label Mapping")
                    for _ in range(MAX_LABELS):
                        column_mappings.append(gr.Dropdown(visible=False))
                with gr.Column():
                    gr.Markdown("# Feature Mapping")
                    for _ in range(MAX_LABELS, MAX_LABELS + MAX_FEATURES):
                        column_mappings.append(gr.Dropdown(visible=False))

    with gr.Accordion(label="Scanner Advanced Config (optional)", open=False):
        scanners = gr.CheckboxGroup(visible=True)

        @gr.on(triggers=[uid_label.change], inputs=[uid_label], outputs=[scanners])
        def get_scanners(uid):
            selected = read_scanners(uid)
            # we remove data_leakage from the default scanners
            # Reason: data_leakage barely raises any issues and takes too many requests
            # when using inference API, causing rate limit error
            scan_config = [
                "ethical_bias", 
                "text_perturbation", 
                "robustness",
                "performance",
                "underconfidence",
                "overconfidence",
                "spurious_correlation",
                "data_leakage",
                ]
            return gr.update(
                choices=scan_config, value=selected, label="Scan Settings", visible=True
            )

    with gr.Row():
        run_btn = gr.Button(
            "Get Evaluation Result",
            variant="primary",
            interactive=False,
            size="lg",
        )

    with gr.Row():
        logs = gr.Textbox(
            value=CHECK_LOG_SECTION_RAW,
            label="Log",
            visible=False,
            every=0.5,
        )

    
    scanners.change(write_scanners, inputs=[scanners, uid_label])

    gr.on(
        triggers=[model_id_input.change],
        fn=get_related_datasets_from_leaderboard,
        inputs=[model_id_input, dataset_id_input],
        outputs=[dataset_id_input],
    ).then(
        fn=check_dataset,
        inputs=[dataset_id_input],
        outputs=[dataset_config_input, dataset_split_input, loading_dataset_info],
    )
    
    gr.on(
        triggers=[dataset_id_input.input, dataset_id_input.select],
        fn=check_dataset,
        inputs=[dataset_id_input],
        outputs=[dataset_config_input, dataset_split_input, loading_dataset_info]
    )

    dataset_config_input.change(fn=get_dataset_splits, inputs=[dataset_id_input, dataset_config_input], outputs=[dataset_split_input])

    gr.on(
        triggers=[model_id_input.change, dataset_id_input.change, dataset_config_input.change],
        fn=empty_column_mapping,
        inputs=[uid_label]
    )

    gr.on(
        triggers=[label.change for label in column_mappings],
        fn=write_column_mapping_to_config,
        inputs=[
            uid_label,
            *column_mappings,
        ],
    )

    # label.change sometimes does not pass the changed value
    gr.on(
        triggers=[label.input for label in column_mappings],
        fn=write_column_mapping_to_config,
        inputs=[
            uid_label,
            *column_mappings,
        ],
    )

    gr.on(
        triggers=[
            model_id_input.change,
            model_id_input.input,
            dataset_id_input.change,
            dataset_config_input.change,
            dataset_split_input.change,
        ],
        fn=precheck_model_ds_enable_example_btn,
        inputs=[
            model_id_input,
            dataset_id_input,
            dataset_config_input,
            dataset_split_input,
        ],
        outputs=[
            example_btn, 
            first_line_ds,
            validation_result,
            example_input,
            example_prediction,
            column_mapping_accordion,],
    )

    gr.on(
        triggers=[
            example_btn.click,
        ],
        fn=align_columns_and_show_prediction,
        inputs=[
            model_id_input,
            dataset_id_input,
            dataset_config_input,
            dataset_split_input,
            uid_label,
        ],
        outputs=[
            validation_result,
            example_input,
            example_prediction,
            column_mapping_accordion,
            run_btn,
            loading_validation,
            *column_mappings,
        ],
    )

    gr.on(
        triggers=[
            run_btn.click,
        ],
        fn=try_submit,
        inputs=[
            model_id_input,
            dataset_id_input,
            dataset_config_input,
            dataset_split_input,
            uid_label,
        ],
        outputs=[
            run_btn, 
            logs, 
            uid_label,            
            validation_result,
            example_input,
            example_prediction,
            column_mapping_accordion,
          ],
    )

    gr.on(
        triggers=[
            scanners.input,
        ],
        fn=enable_run_btn,
        inputs=[
            uid_label,
            model_id_input, 
            dataset_id_input, 
            dataset_config_input, 
            dataset_split_input
        ],
        outputs=[run_btn],
    )

    gr.on(
        triggers=[label.input for label in column_mappings],
        fn=enable_run_btn,
        inputs=[
            uid_label,
            model_id_input, 
            dataset_id_input, 
            dataset_config_input, 
            dataset_split_input
        ],  # FIXME
        outputs=[run_btn],
    )