test_image_processing_imagegpt.py 6.57 KB
Newer Older
NielsRogge's avatar
NielsRogge committed
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
# coding=utf-8
# Copyright 2021 HuggingFace Inc.
#
# 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 json
import os
import tempfile
import unittest

import numpy as np
from datasets import load_dataset

from transformers.testing_utils import require_torch, require_vision, slow
26
from transformers.utils import is_torch_available, is_vision_available
NielsRogge's avatar
NielsRogge committed
27

28
from ...test_image_processing_common import ImageProcessingSavingTestMixin
NielsRogge's avatar
NielsRogge committed
29
30
31
32
33
34
35
36


if is_torch_available():
    import torch

if is_vision_available():
    from PIL import Image

37
    from transformers import ImageGPTImageProcessor
NielsRogge's avatar
NielsRogge committed
38
39


40
class ImageGPTImageProcessingTester(unittest.TestCase):
NielsRogge's avatar
NielsRogge committed
41
42
43
44
45
46
47
48
49
    def __init__(
        self,
        parent,
        batch_size=7,
        num_channels=3,
        image_size=18,
        min_resolution=30,
        max_resolution=400,
        do_resize=True,
amyeroberts's avatar
amyeroberts committed
50
        size=None,
NielsRogge's avatar
NielsRogge committed
51
52
        do_normalize=True,
    ):
amyeroberts's avatar
amyeroberts committed
53
        size = size if size is not None else {"height": 18, "width": 18}
NielsRogge's avatar
NielsRogge committed
54
55
56
57
58
59
60
61
62
63
        self.parent = parent
        self.batch_size = batch_size
        self.num_channels = num_channels
        self.image_size = image_size
        self.min_resolution = min_resolution
        self.max_resolution = max_resolution
        self.do_resize = do_resize
        self.size = size
        self.do_normalize = do_normalize

64
    def prepare_image_processor_dict(self):
NielsRogge's avatar
NielsRogge committed
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
        return {
            # here we create 2 clusters for the sake of simplicity
            "clusters": np.asarray(
                [
                    [0.8866443634033203, 0.6618829369544983, 0.3891746401786804],
                    [-0.6042559146881104, -0.02295008860528469, 0.5423797369003296],
                ]
            ),
            "do_resize": self.do_resize,
            "size": self.size,
            "do_normalize": self.do_normalize,
        }


@require_torch
@require_vision
81
class ImageGPTImageProcessingTest(ImageProcessingSavingTestMixin, unittest.TestCase):
NielsRogge's avatar
NielsRogge committed
82

83
    image_processing_class = ImageGPTImageProcessor if is_vision_available() else None
NielsRogge's avatar
NielsRogge committed
84
85

    def setUp(self):
86
        self.image_processor_tester = ImageGPTImageProcessingTester(self)
NielsRogge's avatar
NielsRogge committed
87
88

    @property
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
    def image_processor_dict(self):
        return self.image_processor_tester.prepare_image_processor_dict()

    def test_image_processor_properties(self):
        image_processing = self.image_processing_class(**self.image_processor_dict)
        self.assertTrue(hasattr(image_processing, "clusters"))
        self.assertTrue(hasattr(image_processing, "do_resize"))
        self.assertTrue(hasattr(image_processing, "size"))
        self.assertTrue(hasattr(image_processing, "do_normalize"))

    def test_image_processor_from_dict_with_kwargs(self):
        image_processor = self.image_processing_class.from_dict(self.image_processor_dict)
        self.assertEqual(image_processor.size, {"height": 18, "width": 18})

        image_processor = self.image_processing_class.from_dict(self.image_processor_dict, size=42)
        self.assertEqual(image_processor.size, {"height": 42, "width": 42})

    def test_image_processor_to_json_string(self):
        image_processor = self.image_processing_class(**self.image_processor_dict)
        obj = json.loads(image_processor.to_json_string())
        for key, value in self.image_processor_dict.items():
NielsRogge's avatar
NielsRogge committed
110
111
112
113
114
            if key == "clusters":
                self.assertTrue(np.array_equal(value, obj[key]))
            else:
                self.assertEqual(obj[key], value)

115
116
    def test_image_processor_to_json_file(self):
        image_processor_first = self.image_processing_class(**self.image_processor_dict)
NielsRogge's avatar
NielsRogge committed
117
118

        with tempfile.TemporaryDirectory() as tmpdirname:
119
120
121
            json_file_path = os.path.join(tmpdirname, "image_processor.json")
            image_processor_first.to_json_file(json_file_path)
            image_processor_second = self.image_processing_class.from_json_file(json_file_path).to_dict()
NielsRogge's avatar
NielsRogge committed
122

123
124
        image_processor_first = image_processor_first.to_dict()
        for key, value in image_processor_first.items():
NielsRogge's avatar
NielsRogge committed
125
            if key == "clusters":
126
                self.assertTrue(np.array_equal(value, image_processor_second[key]))
NielsRogge's avatar
NielsRogge committed
127
            else:
128
                self.assertEqual(image_processor_first[key], value)
NielsRogge's avatar
NielsRogge committed
129

130
131
    def test_image_processor_from_and_save_pretrained(self):
        image_processor_first = self.image_processing_class(**self.image_processor_dict)
NielsRogge's avatar
NielsRogge committed
132
133

        with tempfile.TemporaryDirectory() as tmpdirname:
134
135
            image_processor_first.save_pretrained(tmpdirname)
            image_processor_second = self.image_processing_class.from_pretrained(tmpdirname).to_dict()
NielsRogge's avatar
NielsRogge committed
136

137
138
        image_processor_first = image_processor_first.to_dict()
        for key, value in image_processor_first.items():
NielsRogge's avatar
NielsRogge committed
139
            if key == "clusters":
140
                self.assertTrue(np.array_equal(value, image_processor_second[key]))
NielsRogge's avatar
NielsRogge committed
141
            else:
142
                self.assertEqual(image_processor_first[key], value)
NielsRogge's avatar
NielsRogge committed
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161

    @unittest.skip("ImageGPT requires clusters at initialization")
    def test_init_without_params(self):
        pass


def prepare_images():
    dataset = load_dataset("hf-internal-testing/fixtures_image_utils", split="test")

    image1 = Image.open(dataset[4]["file"])
    image2 = Image.open(dataset[5]["file"])

    images = [image1, image2]

    return images


@require_vision
@require_torch
162
class ImageGPTImageProcessorIntegrationTest(unittest.TestCase):
NielsRogge's avatar
NielsRogge committed
163
164
    @slow
    def test_image(self):
165
        image_processing = ImageGPTImageProcessor.from_pretrained("openai/imagegpt-small")
NielsRogge's avatar
NielsRogge committed
166
167
168
169

        images = prepare_images()

        # test non-batched
170
        encoding = image_processing(images[0], return_tensors="pt")
NielsRogge's avatar
NielsRogge committed
171

172
173
        self.assertIsInstance(encoding.input_ids, torch.LongTensor)
        self.assertEqual(encoding.input_ids.shape, (1, 1024))
NielsRogge's avatar
NielsRogge committed
174
175

        expected_slice = [306, 191, 191]
176
        self.assertEqual(encoding.input_ids[0, :3].tolist(), expected_slice)
NielsRogge's avatar
NielsRogge committed
177
178

        # test batched
179
        encoding = image_processing(images, return_tensors="pt")
NielsRogge's avatar
NielsRogge committed
180

181
182
        self.assertIsInstance(encoding.input_ids, torch.LongTensor)
        self.assertEqual(encoding.input_ids.shape, (2, 1024))
NielsRogge's avatar
NielsRogge committed
183
184

        expected_slice = [303, 13, 13]
185
        self.assertEqual(encoding.input_ids[1, -3:].tolist(), expected_slice)