-
Notifications
You must be signed in to change notification settings - Fork 28.4k
/
Copy pathtest_processor_clip.py
177 lines (127 loc) · 6.89 KB
/
test_processor_clip.py
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
# 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 json
import os
import shutil
import tempfile
import unittest
import numpy as np
import pytest
from transformers import CLIPTokenizer
from transformers.file_utils import FEATURE_EXTRACTOR_NAME, is_vision_available
from transformers.models.clip.tokenization_clip import VOCAB_FILES_NAMES
from transformers.testing_utils import require_vision
if is_vision_available():
from PIL import Image
from transformers import CLIPFeatureExtractor, CLIPProcessor
@require_vision
class CLIPProcessorTest(unittest.TestCase):
def setUp(self):
self.tmpdirname = tempfile.mkdtemp()
# fmt: off
vocab = ["l", "o", "w", "e", "r", "s", "t", "i", "d", "n", "lo", "low</w>", "er</w>", "lowest</w>", "newer</w>", "wider", "<unk>", "<|endoftext|>"]
# fmt: on
vocab_tokens = dict(zip(vocab, range(len(vocab))))
merges = ["#version: 0.2", "l o", "lo w</w>", "e r</w>", ""]
self.special_tokens_map = {"unk_token": "<unk>"}
self.vocab_file = os.path.join(self.tmpdirname, VOCAB_FILES_NAMES["vocab_file"])
self.merges_file = os.path.join(self.tmpdirname, VOCAB_FILES_NAMES["merges_file"])
with open(self.vocab_file, "w", encoding="utf-8") as fp:
fp.write(json.dumps(vocab_tokens) + "\n")
with open(self.merges_file, "w", encoding="utf-8") as fp:
fp.write("\n".join(merges))
feature_extractor_map = {
"do_resize": True,
"size": 20,
"do_center_crop": True,
"crop_size": 18,
"do_normalize": True,
"image_mean": [0.48145466, 0.4578275, 0.40821073],
"image_std": [0.26862954, 0.26130258, 0.27577711],
}
self.feature_extractor_file = os.path.join(self.tmpdirname, FEATURE_EXTRACTOR_NAME)
with open(self.feature_extractor_file, "w", encoding="utf-8") as fp:
json.dump(feature_extractor_map, fp)
def get_tokenizer(self, **kwargs):
return CLIPTokenizer.from_pretrained(self.tmpdirname, **kwargs)
def get_feature_extractor(self, **kwargs):
return CLIPFeatureExtractor.from_pretrained(self.tmpdirname, **kwargs)
def tearDown(self):
shutil.rmtree(self.tmpdirname)
def prepare_image_inputs(self):
"""This function prepares a list of PIL images, or a list of numpy arrays if one specifies numpify=True,
or a list of PyTorch tensors if one specifies torchify=True.
"""
image_inputs = [np.random.randint(255, size=(3, 30, 400), dtype=np.uint8)]
image_inputs = [Image.fromarray(np.moveaxis(x, 0, -1)) for x in image_inputs]
return image_inputs
def test_save_load_pretrained_default(self):
tokenizer = self.get_tokenizer()
feature_extractor = self.get_feature_extractor()
processor = CLIPProcessor(tokenizer=tokenizer, feature_extractor=feature_extractor)
processor.save_pretrained(self.tmpdirname)
processor = CLIPProcessor.from_pretrained(self.tmpdirname)
self.assertEqual(processor.tokenizer.get_vocab(), tokenizer.get_vocab())
self.assertIsInstance(processor.tokenizer, CLIPTokenizer)
self.assertEqual(processor.feature_extractor.to_json_string(), feature_extractor.to_json_string())
self.assertIsInstance(processor.feature_extractor, CLIPFeatureExtractor)
def test_save_load_pretrained_additional_features(self):
processor = CLIPProcessor(tokenizer=self.get_tokenizer(), feature_extractor=self.get_feature_extractor())
processor.save_pretrained(self.tmpdirname)
tokenizer_add_kwargs = self.get_tokenizer(bos_token="(BOS)", eos_token="(EOS)")
feature_extractor_add_kwargs = self.get_feature_extractor(do_normalize=False, padding_value=1.0)
processor = CLIPProcessor.from_pretrained(
self.tmpdirname, bos_token="(BOS)", eos_token="(EOS)", do_normalize=False, padding_value=1.0
)
self.assertEqual(processor.tokenizer.get_vocab(), tokenizer_add_kwargs.get_vocab())
self.assertIsInstance(processor.tokenizer, CLIPTokenizer)
self.assertEqual(processor.feature_extractor.to_json_string(), feature_extractor_add_kwargs.to_json_string())
self.assertIsInstance(processor.feature_extractor, CLIPFeatureExtractor)
def test_feature_extractor(self):
feature_extractor = self.get_feature_extractor()
tokenizer = self.get_tokenizer()
processor = CLIPProcessor(tokenizer=tokenizer, feature_extractor=feature_extractor)
image_input = self.prepare_image_inputs()
input_feat_extract = feature_extractor(image_input, return_tensors="np")
input_processor = processor(images=image_input, return_tensors="np")
for key in input_feat_extract.keys():
self.assertAlmostEqual(input_feat_extract[key].sum(), input_processor[key].sum(), delta=1e-2)
def test_tokenizer(self):
feature_extractor = self.get_feature_extractor()
tokenizer = self.get_tokenizer()
processor = CLIPProcessor(tokenizer=tokenizer, feature_extractor=feature_extractor)
input_str = "lower newer"
encoded_processor = processor(text=input_str)
encoded_tok = tokenizer(input_str)
for key in encoded_tok.keys():
self.assertListEqual(encoded_tok[key], encoded_processor[key])
def test_processor(self):
feature_extractor = self.get_feature_extractor()
tokenizer = self.get_tokenizer()
processor = CLIPProcessor(tokenizer=tokenizer, feature_extractor=feature_extractor)
input_str = "lower newer"
image_input = self.prepare_image_inputs()
inputs = processor(text=input_str, images=image_input)
self.assertListEqual(list(inputs.keys()), ["input_ids", "attention_mask", "pixel_values"])
# test if it raises when no input is passed
with pytest.raises(ValueError):
processor()
def test_tokenizer_decode(self):
feature_extractor = self.get_feature_extractor()
tokenizer = self.get_tokenizer()
processor = CLIPProcessor(tokenizer=tokenizer, feature_extractor=feature_extractor)
predicted_ids = [[1, 4, 5, 8, 1, 0, 8], [3, 4, 3, 1, 1, 8, 9]]
decoded_processor = processor.batch_decode(predicted_ids)
decoded_tok = tokenizer.batch_decode(predicted_ids)
self.assertListEqual(decoded_tok, decoded_processor)