File size: 5,923 Bytes
9e4233f
8f809e2
3573a39
 
 
 
 
 
 
8f809e2
3573a39
 
 
 
 
 
 
9e4233f
 
 
 
3573a39
 
 
 
9e4233f
8f809e2
136af2d
9e4233f
be473e6
136af2d
 
 
9e4233f
 
 
 
 
 
 
 
 
 
3573a39
9e4233f
3573a39
 
 
9e4233f
3573a39
9e4233f
3573a39
 
9e4233f
3573a39
 
 
be473e6
 
 
 
 
 
 
3573a39
be473e6
 
3573a39
 
9e4233f
136af2d
9e4233f
3573a39
 
136af2d
8f809e2
 
 
3573a39
 
 
 
9e4233f
 
 
 
 
 
 
 
3573a39
8f809e2
 
 
3573a39
136af2d
 
 
 
 
 
 
 
 
 
 
 
 
 
3573a39
 
8f809e2
3573a39
 
 
 
136af2d
3573a39
 
 
9e4233f
3025552
 
 
 
 
 
 
 
 
 
 
 
 
3573a39
 
 
136af2d
3573a39
 
 
8f809e2
3573a39
 
 
 
 
 
 
 
 
 
 
 
 
9e4233f
 
 
 
3573a39
9e4233f
8f809e2
3573a39
 
 
 
8f809e2
3573a39
 
 
 
 
8f809e2
3573a39
 
8f809e2
 
3573a39
 
 
 
 
 
 
8f809e2
 
3573a39
 
 
8f809e2
 
 
 
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
import gradio as gr
import uuid
from io_utils import (
    read_scanners,
    write_scanners,
    read_inference_type,
    write_inference_type,
    get_logs_file,
)
from wordings import INTRODUCTION_MD, CONFIRM_MAPPING_DETAILS_MD
from text_classification_ui_helpers import (
    try_submit,
    check_dataset_and_get_config,
    check_dataset_and_get_split,
    check_model_and_show_prediction,
    write_column_mapping_to_config,
)

MAX_LABELS = 20
MAX_FEATURES = 20

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


def get_demo(demo):
    uid = uuid.uuid4()
    with gr.Row():
        gr.Markdown(INTRODUCTION_MD)
        uid_label = gr.Textbox(
            label="Evaluation ID:", value=uid, 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)
        dataset_split_input = gr.Dropdown(label="Dataset Split", visible=False)

    with gr.Row():
        example_input = gr.Markdown("Example Input", 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():
                    for _ in range(MAX_LABELS):
                        column_mappings.append(gr.Dropdown(visible=False))
                with gr.Column():
                    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")
        use_inference = read_inference_type(uid) == "hf_inference_api"
        run_inference = gr.Checkbox(value=use_inference, label="Run with Inference API")

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

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

    with gr.Row():
        logs = gr.Textbox(label="Giskard Bot Evaluation Log:", visible=False)
        demo.load(get_logs_file, uid_label, logs, every=0.5)

    dataset_id_input.change(
        check_dataset_and_get_config, inputs=[dataset_id_input, uid_label], outputs=[dataset_config_input]
    )

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

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

    run_inference.change(write_inference_type, inputs=[run_inference, uid_label])

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

    gr.on(
        triggers=[
            model_id_input.change,
            dataset_id_input.change,
            dataset_config_input.change,
            dataset_split_input.change,
        ],
        fn=check_model_and_show_prediction,
        inputs=[
            model_id_input,
            dataset_id_input,
            dataset_config_input,
            dataset_split_input,
        ],
        outputs=[
            example_input,
            example_prediction,
            column_mapping_accordion,
            *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,
            uid_label,
        ],
        outputs=[run_btn, logs],
    )

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

    gr.on(
        triggers=[
            model_id_input.change,
            dataset_config_input.change,
            dataset_split_input.change,
            run_inference.change,
            run_local.change,
            scanners.change,
        ],
        fn=enable_run_btn,
        inputs=None,
        outputs=[run_btn],
    )

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