File size: 9,134 Bytes
73d9a01
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import gradio as gr
from datasets import disable_caching, load_dataset
from transformer_ranker import TransformerRanker, prepare_popular_models
import traceback

from utils import (
    DISABLED_BUTTON_VARIANT, ENABLED_BUTTON_VARIANT, CSS, HEADLINE, FOOTER,
    EmbeddingProgressTracker, check_dataset_exists, check_dataset_is_loaded,
    compute_ratio, ensure_one_lm_selected, get_dataset_info
)

disable_caching()

THEME = "pseudolab/huggingface-korea-theme"
DEFAULT_SAMPLES = 1000
MAX_SAMPLES = 5000
LANGUAGE_MODELS = prepare_popular_models('base') + prepare_popular_models('large')

# Add a tiny model for demonstration on CPU
LANGUAGE_MODELS = ['prajjwal1/bert-tiny'] + list(dict.fromkeys(LANGUAGE_MODELS))
LANGUAGE_MODELS.insert(LANGUAGE_MODELS.index("bert-base-cased") + 1, "bert-base-uncased")

# Preselect some small models
DEFAULT_MODELS = [
    "prajjwal1/bert-tiny", "google/electra-small-discriminator", 
    "distilbert-base-cased", "sentence-transformers/all-MiniLM-L12-v2"
]


with gr.Blocks(css=CSS, theme=THEME) as demo:

    ########## STEP 1: Load the Dataset ##########

    gr.Markdown(HEADLINE)

    gr.Markdown("## Step 1: Load a Dataset")
    with gr.Group():
        dataset = gr.State(None)

        dataset_name = gr.Textbox(
            label="Enter the name of your dataset",
            placeholder="Examples: trec, ag_news, sst2, conll2003, leondz/wnut_17",
            max_lines=1,
        )
        select_dataset_button = gr.Button(
            value="Load dataset", interactive=False, variant=DISABLED_BUTTON_VARIANT
        )

        # Activate the "Load dataset" button if dataset was found
        dataset_name.change(
            check_dataset_exists, inputs=dataset_name, outputs=select_dataset_button
        )

    gr.Markdown(
        "*The number of samples that can be used in this demo is limited to save resources. "
        "To run an estimate on the full dataset, check out the "
        "[library](https://github.com/flairNLP/transformer-ranker).*"
    )

    ########## Step 1.1 Dataset preprocessing ##########

    with gr.Accordion("Dataset settings", open=False) as dataset_config:
        with gr.Row() as dataset_details:
            dataset_name_label = gr.Label("", label="Dataset Name")
            num_samples = gr.State(0)
            num_samples_label = gr.Label("", label="Number of Samples")
            num_samples.change(
                lambda x: str(x), inputs=[num_samples], outputs=[num_samples_label]
            )

        with gr.Row():
            text_column = gr.Dropdown("", label="Text Column")
            text_pair_column = gr.Dropdown("", label="Text Pair Column")

        with gr.Row():
            label_column = gr.Dropdown("", label="Label Column")
            task_category = gr.Dropdown("", label="Task Type")

        with gr.Group():
            downsample_ratio = gr.State(0.0)
            num_samples_to_use = gr.Slider(
                20, MAX_SAMPLES, label="Samples to use", value=DEFAULT_SAMPLES, step=1
            )
            downsample_ratio_label = gr.Label("", label="Ratio of dataset to use")
            downsample_ratio.change(
                lambda x: f"{x:.1%}",
                inputs=[downsample_ratio],
                outputs=[downsample_ratio_label],
            )

            num_samples_to_use.change(
                compute_ratio,
                inputs=[num_samples_to_use, num_samples],
                outputs=downsample_ratio,
            )
            num_samples.change(
                compute_ratio,
                inputs=[num_samples_to_use, num_samples],
                outputs=downsample_ratio,
            )

    # Download the dataset and show details
    def select_dataset(dataset_name):
        try:
            dataset = load_dataset(dataset_name, trust_remote_code=True)
            dataset_info = get_dataset_info(dataset)
        except ValueError:
            gr.Warning("Dataset collections are not supported. Please use a single dataset.")

        return (
            gr.update(value="Loaded", interactive=False, variant=DISABLED_BUTTON_VARIANT),
            gr.Accordion(open=True),
            dataset_name,
            dataset,
            *dataset_info
        )

    select_dataset_button.click(
        select_dataset,
        inputs=[dataset_name],
        outputs=[
            select_dataset_button,
            dataset_config,
            dataset_name_label,
            dataset,
            task_category,
            text_column,
            text_pair_column,
            label_column,
            num_samples,
        ],
        scroll_to_output=True,
    )

    ########## STEP 2 ##########

    gr.Markdown("## Step 2: Select a List of Language Models")
    with gr.Group():
        model_options = [
            (model_handle.split("/")[-1], model_handle)
            for model_handle in LANGUAGE_MODELS
        ]
        models = gr.CheckboxGroup(
            choices=model_options, label="Select Models", value=DEFAULT_MODELS
        )

    ########## STEP 3: Run Language Model Ranking ##########

    gr.Markdown("## Step 3: Rank LMs")

    with gr.Group():
        with gr.Accordion("Advanced settings", open=False):
            with gr.Row():
                estimator = gr.Dropdown(
                    choices=["hscore", "logme", "knn"],
                    label="Transferability metric",
                    value="hscore",
                )
                layer_pooling_options = ["lastlayer", "layermean", "bestlayer"]
                layer_pooling = gr.Dropdown(
                    choices=["lastlayer", "layermean", "bestlayer"],
                    label="Layer pooling",
                    value="layermean",
                )
        submit_button = gr.Button("Run Ranking", interactive=False, variant=DISABLED_BUTTON_VARIANT)

        # Make button active if the dataset is loaded
        dataset.change(
            check_dataset_is_loaded,
            inputs=[dataset, text_column, label_column, task_category],
            outputs=submit_button
        )

        label_column.change(
            check_dataset_is_loaded,
            inputs=[dataset, text_column, label_column, task_category],
            outputs=submit_button
        )

        text_column.change(
            check_dataset_is_loaded,
            inputs=[dataset, text_column, label_column, task_category],
            outputs=submit_button
        )

    def rank_models(
        dataset,
        downsample_ratio,
        selected_models,
        layer_pooling,
        estimator,
        text_column,
        text_pair_column,
        label_column,
        task_category,
        progress=gr.Progress(),
    ):

        if text_column == "-":
            raise gr.Error("Text column is not set.")

        if label_column == "-":
            raise gr.Error("Label column is not set.")

        if task_category == "-":
            raise gr.Error(
                "Task category is not set. The dataset must support classification or regression tasks."
            )

        if text_pair_column == "-":
            text_pair_column = None

        progress(0.0, "Starting")

        with EmbeddingProgressTracker(progress=progress, model_names=selected_models) as tracker:
            try:
                ranker = TransformerRanker(
                    dataset,
                    dataset_downsample=downsample_ratio,
                    text_column=text_column,
                    text_pair_column=text_pair_column,
                    label_column=label_column,
                    task_category=task_category,
                )

                results = ranker.run(
                    models=selected_models,
                    layer_aggregator=layer_pooling,
                    estimator=estimator,
                    batch_size=64,
                    tracker=tracker,
                )

                sorted_results = sorted(
                    results._results.items(), key=lambda item: item[1], reverse=True
                )
                return [
                    (i + 1, model, score) for i, (model, score) in enumerate(sorted_results)
                ]
            except Exception as e:
                gr.Error("The dataset is not supported.")

    gr.Markdown("## Results")
    ranking_results = gr.Dataframe(
        headers=["Rank", "Model", "Score"], datatype=["number", "str", "number"]
    )

    submit_button.click(
        rank_models,
        inputs=[
            dataset,
            downsample_ratio,
            models,
            layer_pooling,
            estimator,
            text_column,
            text_pair_column,
            label_column,
            task_category,
        ],
        outputs=ranking_results,
        scroll_to_output=True,
    )

    gr.Markdown(
        "*The results are ranked by their transferability score, with the most suitable model listed first. "
        "This ranking allows focusing on the higher-ranked models for further exploration and fine-tuning.*"
    )

    gr.Markdown(FOOTER)

if __name__ == "__main__":
    demo.queue(default_concurrency_limit=3)
    demo.launch(max_threads=6)