File size: 7,410 Bytes
8f809e2
58c39e0
 
 
53fe897
 
 
 
 
 
 
 
 
 
58c39e0
9e4233f
5b8d6d5
9e4233f
 
3573a39
 
 
 
9e4233f
5b8d6d5
9e4233f
be473e6
136af2d
8092547
136af2d
9e4233f
 
 
 
 
 
 
 
 
 
3573a39
9e4233f
53fe897
 
3573a39
5559b52
 
 
 
5b8d6d5
 
053c12c
53fe897
5b8d6d5
53fe897
 
5b8d6d5
9e4233f
5704515
9e4233f
3573a39
 
9e4233f
3573a39
 
 
be473e6
 
 
 
 
5b8d6d5
be473e6
 
3573a39
5b8d6d5
be473e6
 
3573a39
 
9e4233f
5b8d6d5
58c39e0
 
 
 
 
3573a39
58c39e0
 
 
 
8092547
 
 
 
 
 
 
58c39e0
 
 
9e4233f
 
 
 
 
5b8d6d5
9e4233f
 
3573a39
8f809e2
53fe897
 
 
 
 
 
3573a39
136af2d
53fe897
58c39e0
5559b52
136af2d
 
 
53fe897
136af2d
5559b52
53fe897
 
 
 
 
5559b52
136af2d
 
 
 
1c00552
58c39e0
5b8d6d5
58c39e0
 
 
1c00552
58c39e0
 
 
 
 
3573a39
 
8f809e2
3573a39
136af2d
3573a39
 
 
9e4233f
1c00552
 
 
 
 
 
 
 
 
 
3573a39
 
 
136af2d
3573a39
53fe897
 
5b8d6d5
 
 
 
 
 
53fe897
5559b52
53fe897
5b8d6d5
 
 
 
3573a39
5b8d6d5
3573a39
 
 
 
 
5b8d6d5
3573a39
 
 
 
 
5b8d6d5
5559b52
3573a39
 
 
9e4233f
 
 
 
3573a39
9e4233f
8f809e2
3573a39
 
 
 
8f809e2
5b8d6d5
 
3573a39
 
60f78e5
3573a39
 
8f809e2
3573a39
 
8f809e2
 
5b8d6d5
 
 
 
3573a39
8f809e2
 
3573a39
 
 
8f809e2
5b8d6d5
8f809e2
53fe897
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
import uuid

import gradio as gr

from io_utils import get_logs_file, read_scanners, write_scanners
from text_classification_ui_helpers import (
    align_columns_and_show_prediction,
    check_dataset,
    deselect_run_inference,
    precheck_model_ds_enable_example_btn,
    select_run_mode,
    try_submit,
    write_column_mapping_to_config,
)
from wordings import CONFIRM_MAPPING_DETAILS_MD, INTRODUCTION_MD

MAX_LABELS = 40
MAX_FEATURES = 20

EXAMPLE_MODEL_ID = "cardiffnlp/twitter-roberta-base-sentiment-latest"
EXAMPLE_DATA_ID = "tweet_eval"
CONFIG_PATH = "./config.yaml"


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.Row():
        model_id_input = gr.Textbox(
            label="Hugging Face model id",
            placeholder=EXAMPLE_MODEL_ID + " (press enter to confirm)",
        )

        dataset_id_input = gr.Textbox(
            label="Hugging Face Dataset id",
            placeholder=EXAMPLE_DATA_ID + " (press enter to confirm)",
        )

    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_status = gr.HTML(visible=True)
    with gr.Row():
        example_btn = gr.Button(
            "Validate model & dataset",
            visible=True,
            variant="primary",
            interactive=False,
        )

    with gr.Row():
        example_input = gr.HTML(visible=False)
    with gr.Row():
        example_prediction = gr.Label(label="Model Prediction Sample", 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="Model Wrap Advance Config (optional)", open=False):
        run_local = gr.Checkbox(value=True, label="Run in this Space")
        run_inference = gr.Checkbox(value=False, label="Run with Inference API")
        inference_token = gr.Textbox(
            value="",
            label="HF Token for Inference API",
            visible=False,
            interactive=True,
        )

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

        @gr.on(triggers=[uid_label.change], inputs=[uid_label], outputs=[scanners])
        def get_scanners(uid):
            selected = read_scanners(uid)
            # currently 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 = selected + ["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=get_logs_file,
            label="Giskard Bot Evaluation Log:",
            visible=False,
            every=0.5,
        )

    dataset_id_input.change(
        check_dataset,
        inputs=[dataset_id_input],
        outputs=[dataset_config_input, dataset_split_input, first_line_ds, loading_status],
    )

    dataset_config_input.change(
        check_dataset,
        inputs=[dataset_id_input, dataset_config_input],
        outputs=[dataset_config_input, dataset_split_input, first_line_ds, loading_status],
    )

    dataset_split_input.change(
        check_dataset,
        inputs=[dataset_id_input, dataset_config_input, dataset_split_input],
        outputs=[dataset_config_input, dataset_split_input, first_line_ds, loading_status],
    )

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

    run_inference.change(
        select_run_mode,
        inputs=[run_inference],
        outputs=[inference_token, run_local],
    )

    run_local.change(
        deselect_run_inference,
        inputs=[run_local],
        outputs=[inference_token, run_inference],
    )

    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,
            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, loading_status],
    )

    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=[
            example_input,
            example_prediction,
            column_mapping_accordion,
            run_btn,
            loading_status,
            *column_mappings,
        ],
    )

    gr.on(
        triggers=[
            run_btn.click,
        ],
        fn=try_submit,
        inputs=[
            model_id_input,
            dataset_id_input,
            dataset_config_input,
            dataset_split_input,
            run_local,
            run_inference,
            inference_token,
            uid_label,
        ],
        outputs=[run_btn, logs, uid_label],
    )

    def enable_run_btn():
        return gr.update(interactive=True)

    gr.on(
        triggers=[
            run_inference.input,
            run_local.input,
            inference_token.input,
            scanners.input,
        ],
        fn=enable_run_btn,
        inputs=None,
        outputs=[run_btn],
    )

    gr.on(
        triggers=[label.input for label in column_mappings],
        fn=enable_run_btn,
        inputs=None,  # FIXME
        outputs=[run_btn],
    )