File size: 12,209 Bytes
e0be88b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
# Copyright 2021 The HuggingFace Team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

import unittest

import datasets
from huggingface_hub import ImageClassificationOutputElement

from transformers import (
    MODEL_FOR_IMAGE_CLASSIFICATION_MAPPING,
    TF_MODEL_FOR_IMAGE_CLASSIFICATION_MAPPING,
    PreTrainedTokenizerBase,
    is_torch_available,
    is_vision_available,
)
from transformers.pipelines import ImageClassificationPipeline, pipeline
from transformers.testing_utils import (
    compare_pipeline_output_to_hub_spec,
    is_pipeline_test,
    nested_simplify,
    require_tf,
    require_torch,
    require_torch_or_tf,
    require_vision,
    slow,
)

from .test_pipelines_common import ANY


if is_torch_available():
    import torch

if is_vision_available():
    from PIL import Image
else:

    class Image:
        @staticmethod
        def open(*args, **kwargs):
            pass


@is_pipeline_test
@require_torch_or_tf
@require_vision
class ImageClassificationPipelineTests(unittest.TestCase):
    model_mapping = MODEL_FOR_IMAGE_CLASSIFICATION_MAPPING
    tf_model_mapping = TF_MODEL_FOR_IMAGE_CLASSIFICATION_MAPPING
    _dataset = None

    @classmethod
    def _load_dataset(cls):
        # Lazy loading of the dataset. Because it is a class method, it will only be loaded once per pytest process.
        if cls._dataset is None:
            # we use revision="refs/pr/1" until the PR is merged
            # https://hf.co/datasets/hf-internal-testing/fixtures_image_utils/discussions/1
            cls._dataset = datasets.load_dataset(
                "hf-internal-testing/fixtures_image_utils", split="test", revision="refs/pr/1"
            )

    def get_test_pipeline(
        self,
        model,
        tokenizer=None,
        image_processor=None,
        feature_extractor=None,
        processor=None,
        torch_dtype="float32",
    ):
        image_classifier = ImageClassificationPipeline(
            model=model,
            tokenizer=tokenizer,
            feature_extractor=feature_extractor,
            image_processor=image_processor,
            processor=processor,
            torch_dtype=torch_dtype,
            top_k=2,
        )
        examples = [
            Image.open("./tests/fixtures/tests_samples/COCO/000000039769.png"),
            "http://images.cocodataset.org/val2017/000000039769.jpg",
        ]
        return image_classifier, examples

    def run_pipeline_test(self, image_classifier, examples):
        self._load_dataset()
        outputs = image_classifier("./tests/fixtures/tests_samples/COCO/000000039769.png")

        self.assertEqual(
            outputs,
            [
                {"score": ANY(float), "label": ANY(str)},
                {"score": ANY(float), "label": ANY(str)},
            ],
        )

        # Accepts URL + PIL.Image + lists
        outputs = image_classifier(
            [
                Image.open("./tests/fixtures/tests_samples/COCO/000000039769.png"),
                "http://images.cocodataset.org/val2017/000000039769.jpg",
                # RGBA
                self._dataset[0]["image"],
                # LA
                self._dataset[1]["image"],
                # L
                self._dataset[2]["image"],
            ]
        )
        self.assertEqual(
            outputs,
            [
                [
                    {"score": ANY(float), "label": ANY(str)},
                    {"score": ANY(float), "label": ANY(str)},
                ],
                [
                    {"score": ANY(float), "label": ANY(str)},
                    {"score": ANY(float), "label": ANY(str)},
                ],
                [
                    {"score": ANY(float), "label": ANY(str)},
                    {"score": ANY(float), "label": ANY(str)},
                ],
                [
                    {"score": ANY(float), "label": ANY(str)},
                    {"score": ANY(float), "label": ANY(str)},
                ],
                [
                    {"score": ANY(float), "label": ANY(str)},
                    {"score": ANY(float), "label": ANY(str)},
                ],
            ],
        )

        for single_output in outputs:
            for output_element in single_output:
                compare_pipeline_output_to_hub_spec(output_element, ImageClassificationOutputElement)

    @require_torch
    def test_small_model_pt(self):
        small_model = "hf-internal-testing/tiny-random-vit"
        image_classifier = pipeline("image-classification", model=small_model)

        outputs = image_classifier("http://images.cocodataset.org/val2017/000000039769.jpg")
        self.assertEqual(
            nested_simplify(outputs, decimals=4),
            [{"label": "LABEL_1", "score": 0.574}, {"label": "LABEL_0", "score": 0.426}],
        )

        outputs = image_classifier(
            [
                "http://images.cocodataset.org/val2017/000000039769.jpg",
                "http://images.cocodataset.org/val2017/000000039769.jpg",
            ],
            top_k=2,
        )
        self.assertEqual(
            nested_simplify(outputs, decimals=4),
            [
                [{"label": "LABEL_1", "score": 0.574}, {"label": "LABEL_0", "score": 0.426}],
                [{"label": "LABEL_1", "score": 0.574}, {"label": "LABEL_0", "score": 0.426}],
            ],
        )

    @require_tf
    def test_small_model_tf(self):
        small_model = "hf-internal-testing/tiny-random-vit"
        image_classifier = pipeline("image-classification", model=small_model, framework="tf")

        outputs = image_classifier("http://images.cocodataset.org/val2017/000000039769.jpg")
        self.assertEqual(
            nested_simplify(outputs, decimals=4),
            [{"label": "LABEL_1", "score": 0.574}, {"label": "LABEL_0", "score": 0.426}],
        )

        outputs = image_classifier(
            [
                "http://images.cocodataset.org/val2017/000000039769.jpg",
                "http://images.cocodataset.org/val2017/000000039769.jpg",
            ],
            top_k=2,
        )
        self.assertEqual(
            nested_simplify(outputs, decimals=4),
            [
                [{"label": "LABEL_1", "score": 0.574}, {"label": "LABEL_0", "score": 0.426}],
                [{"label": "LABEL_1", "score": 0.574}, {"label": "LABEL_0", "score": 0.426}],
            ],
        )

    def test_custom_tokenizer(self):
        tokenizer = PreTrainedTokenizerBase()

        # Assert that the pipeline can be initialized with a feature extractor that is not in any mapping
        image_classifier = pipeline(
            "image-classification", model="hf-internal-testing/tiny-random-vit", tokenizer=tokenizer
        )

        self.assertIs(image_classifier.tokenizer, tokenizer)

    @require_torch
    def test_torch_float16_pipeline(self):
        image_classifier = pipeline(
            "image-classification", model="hf-internal-testing/tiny-random-vit", torch_dtype=torch.float16
        )
        outputs = image_classifier("http://images.cocodataset.org/val2017/000000039769.jpg")

        self.assertEqual(
            nested_simplify(outputs, decimals=3),
            [{"label": "LABEL_1", "score": 0.574}, {"label": "LABEL_0", "score": 0.426}],
        )

    @require_torch
    def test_torch_bfloat16_pipeline(self):
        image_classifier = pipeline(
            "image-classification", model="hf-internal-testing/tiny-random-vit", torch_dtype=torch.bfloat16
        )
        outputs = image_classifier("http://images.cocodataset.org/val2017/000000039769.jpg")

        self.assertEqual(
            nested_simplify(outputs, decimals=3),
            [{"label": "LABEL_1", "score": 0.574}, {"label": "LABEL_0", "score": 0.426}],
        )

    @slow
    @require_torch
    def test_perceiver(self):
        # Perceiver is not tested by `run_pipeline_test` properly.
        # That is because the type of feature_extractor and model preprocessor need to be kept
        # in sync, which is not the case in the current design
        image_classifier = pipeline("image-classification", model="deepmind/vision-perceiver-conv")
        outputs = image_classifier("http://images.cocodataset.org/val2017/000000039769.jpg")
        self.assertEqual(
            nested_simplify(outputs, decimals=4),
            [
                {"score": 0.4385, "label": "tabby, tabby cat"},
                {"score": 0.321, "label": "tiger cat"},
                {"score": 0.0502, "label": "Egyptian cat"},
                {"score": 0.0137, "label": "crib, cot"},
                {"score": 0.007, "label": "radiator"},
            ],
        )

        image_classifier = pipeline("image-classification", model="deepmind/vision-perceiver-fourier")
        outputs = image_classifier("http://images.cocodataset.org/val2017/000000039769.jpg")
        self.assertEqual(
            nested_simplify(outputs, decimals=4),
            [
                {"score": 0.5658, "label": "tabby, tabby cat"},
                {"score": 0.1309, "label": "tiger cat"},
                {"score": 0.0722, "label": "Egyptian cat"},
                {"score": 0.0707, "label": "remote control, remote"},
                {"score": 0.0082, "label": "computer keyboard, keypad"},
            ],
        )

        image_classifier = pipeline("image-classification", model="deepmind/vision-perceiver-learned")
        outputs = image_classifier("http://images.cocodataset.org/val2017/000000039769.jpg")
        self.assertEqual(
            nested_simplify(outputs, decimals=4),
            [
                {"score": 0.3022, "label": "tabby, tabby cat"},
                {"score": 0.2362, "label": "Egyptian cat"},
                {"score": 0.1856, "label": "tiger cat"},
                {"score": 0.0324, "label": "remote control, remote"},
                {"score": 0.0096, "label": "quilt, comforter, comfort, puff"},
            ],
        )

    @slow
    @require_torch
    def test_multilabel_classification(self):
        small_model = "hf-internal-testing/tiny-random-vit"

        # Sigmoid is applied for multi-label classification
        image_classifier = pipeline("image-classification", model=small_model)
        image_classifier.model.config.problem_type = "multi_label_classification"

        outputs = image_classifier("http://images.cocodataset.org/val2017/000000039769.jpg")
        self.assertEqual(
            nested_simplify(outputs, decimals=4),
            [{"label": "LABEL_1", "score": 0.5356}, {"label": "LABEL_0", "score": 0.4612}],
        )

        outputs = image_classifier(
            [
                "http://images.cocodataset.org/val2017/000000039769.jpg",
                "http://images.cocodataset.org/val2017/000000039769.jpg",
            ]
        )
        self.assertEqual(
            nested_simplify(outputs, decimals=4),
            [
                [{"label": "LABEL_1", "score": 0.5356}, {"label": "LABEL_0", "score": 0.4612}],
                [{"label": "LABEL_1", "score": 0.5356}, {"label": "LABEL_0", "score": 0.4612}],
            ],
        )

    @slow
    @require_torch
    def test_function_to_apply(self):
        small_model = "hf-internal-testing/tiny-random-vit"

        # Sigmoid is applied for multi-label classification
        image_classifier = pipeline("image-classification", model=small_model)

        outputs = image_classifier(
            "http://images.cocodataset.org/val2017/000000039769.jpg",
            function_to_apply="sigmoid",
        )
        self.assertEqual(
            nested_simplify(outputs, decimals=4),
            [{"label": "LABEL_1", "score": 0.5356}, {"label": "LABEL_0", "score": 0.4612}],
        )