SUMIN/HuggingFace_transformer
1
0
Fork 0
Code Issues Pull Requests Actions 7 Packages Projects Releases Wiki Activity

Remove repeated prepare_images in processor tests (#33163)

Browse Source 
* Remove repeated prepare_images

* Address comments - update docstring; explanatory comment
This commit is contained in:
amyeroberts
2024-09-09 13:20:27 +01:00
committed by GitHub
parent 0574fa668b
commit f745e7d3f9
21 changed files with 140 additions and 304 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
tests/models/align/test_processor_align.py
View File

@@ -18,7 +18,6 @@ import shutil
import tempfile
import unittest
import numpy as np
import pytest
from transformers import BertTokenizer, BertTokenizerFast
@@ -30,8 +29,6 @@ from ...test_processing_common import ProcessorTesterMixin
if is_vision_available():
from PIL import Image
from transformers import AlignProcessor, EfficientNetImageProcessor
@@ -86,15 +83,6 @@ class AlignProcessorTest(ProcessorTesterMixin, unittest.TestCase):
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_slow = self.get_tokenizer()
tokenizer_fast = self.get_rust_tokenizer()

20
tests/models/blip/test_processor_blip.py
View File

@@ -15,21 +15,22 @@ import shutil
import tempfile
import unittest
import numpy as np
import pytest
from transformers.testing_utils import require_vision
from transformers.utils import is_vision_available
from ...test_processing_common import ProcessorTesterMixin
if is_vision_available():
from PIL import Image
from transformers import AutoProcessor, BertTokenizer, BlipImageProcessor, BlipProcessor, PreTrainedTokenizerFast
@require_vision
class BlipProcessorTest(unittest.TestCase):
class BlipProcessorTest(ProcessorTesterMixin, unittest.TestCase):
processor_class = BlipProcessor
def setUp(self):
self.tmpdirname = tempfile.mkdtemp()
@@ -49,17 +50,6 @@ class BlipProcessorTest(unittest.TestCase):
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_additional_features(self):
processor = BlipProcessor(tokenizer=self.get_tokenizer(), image_processor=self.get_image_processor())
processor.save_pretrained(self.tmpdirname)

20
tests/models/blip_2/test_processor_blip_2.py
View File

@@ -15,21 +15,22 @@ import shutil
import tempfile
import unittest
import numpy as np
import pytest
from transformers.testing_utils import require_vision
from transformers.utils import is_vision_available
from ...test_processing_common import ProcessorTesterMixin
if is_vision_available():
from PIL import Image
from transformers import AutoProcessor, Blip2Processor, BlipImageProcessor, GPT2Tokenizer, PreTrainedTokenizerFast
@require_vision
class Blip2ProcessorTest(unittest.TestCase):
class Blip2ProcessorTest(ProcessorTesterMixin, unittest.TestCase):
processor_class = Blip2Processor
def setUp(self):
self.tmpdirname = tempfile.mkdtemp()
@@ -49,17 +50,6 @@ class Blip2ProcessorTest(unittest.TestCase):
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_additional_features(self):
processor = Blip2Processor(tokenizer=self.get_tokenizer(), image_processor=self.get_image_processor())
processor.save_pretrained(self.tmpdirname)

25
tests/models/chinese_clip/test_processor_chinese_clip.py
View File

@@ -18,7 +18,6 @@ import shutil
import tempfile
import unittest
import numpy as np
import pytest
from transformers import BertTokenizer, BertTokenizerFast
@@ -26,15 +25,17 @@ from transformers.models.bert.tokenization_bert import VOCAB_FILES_NAMES
from transformers.testing_utils import require_vision
from transformers.utils import FEATURE_EXTRACTOR_NAME, is_vision_available
from ...test_processing_common import ProcessorTesterMixin
if is_vision_available():
from PIL import Image
from transformers import ChineseCLIPImageProcessor, ChineseCLIPProcessor
@require_vision
class ChineseCLIPProcessorTest(unittest.TestCase):
class ChineseCLIPProcessorTest(ProcessorTesterMixin, unittest.TestCase):
processor_class = ChineseCLIPProcessor
def setUp(self):
self.tmpdirname = tempfile.mkdtemp()
@@ -76,6 +77,11 @@ class ChineseCLIPProcessorTest(unittest.TestCase):
with open(self.image_processor_file, "w", encoding="utf-8") as fp:
json.dump(image_processor_map, fp)
tokenizer = self.get_tokenizer()
image_processor = self.get_image_processor()
processor = ChineseCLIPProcessor(tokenizer=tokenizer, image_processor=image_processor)
processor.save_pretrained(self.tmpdirname)
def get_tokenizer(self, **kwargs):
return BertTokenizer.from_pretrained(self.tmpdirname, **kwargs)
@@ -88,17 +94,6 @@ class ChineseCLIPProcessorTest(unittest.TestCase):
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_slow = self.get_tokenizer()
tokenizer_fast = self.get_rust_tokenizer()

14
tests/models/clip/test_processor_clip.py
View File

@@ -18,7 +18,6 @@ import shutil
import tempfile
import unittest
import numpy as np
import pytest
from transformers import CLIPTokenizer, CLIPTokenizerFast
@@ -30,8 +29,6 @@ from ...test_processing_common import ProcessorTesterMixin
if is_vision_available():
from PIL import Image
from transformers import CLIPImageProcessor, CLIPProcessor
@@ -79,17 +76,6 @@ class CLIPProcessorTest(ProcessorTesterMixin, unittest.TestCase):
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_slow = self.get_tokenizer()
tokenizer_fast = self.get_rust_tokenizer()

19
tests/models/clipseg/test_processor_clipseg.py
View File

@@ -18,7 +18,6 @@ import shutil
import tempfile
import unittest
import numpy as np
import pytest
from transformers import CLIPTokenizer, CLIPTokenizerFast
@@ -26,15 +25,17 @@ from transformers.models.clip.tokenization_clip import VOCAB_FILES_NAMES
from transformers.testing_utils import require_vision
from transformers.utils import IMAGE_PROCESSOR_NAME, is_vision_available
from ...test_processing_common import ProcessorTesterMixin
if is_vision_available():
from PIL import Image
from transformers import CLIPSegProcessor, ViTImageProcessor
@require_vision
class CLIPSegProcessorTest(unittest.TestCase):
class CLIPSegProcessorTest(ProcessorTesterMixin, unittest.TestCase):
processor_class = CLIPSegProcessor
def setUp(self):
self.tmpdirname = tempfile.mkdtemp()
@@ -75,16 +76,6 @@ class CLIPSegProcessorTest(unittest.TestCase):
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_slow = self.get_tokenizer()
tokenizer_fast = self.get_rust_tokenizer()

20
tests/models/flava/test_processor_flava.py
View File

@@ -19,7 +19,6 @@ import shutil
import tempfile
import unittest
import numpy as np
import pytest
from transformers import BertTokenizer, BertTokenizerFast
@@ -27,10 +26,10 @@ from transformers.models.bert.tokenization_bert import VOCAB_FILES_NAMES
from transformers.testing_utils import require_vision
from transformers.utils import IMAGE_PROCESSOR_NAME, is_vision_available
from ...test_processing_common import ProcessorTesterMixin
if is_vision_available():
from PIL import Image
from transformers import FlavaImageProcessor, FlavaProcessor
from transformers.models.flava.image_processing_flava import (
FLAVA_CODEBOOK_MEAN,
@@ -41,7 +40,9 @@ if is_vision_available():
@require_vision
class FlavaProcessorTest(unittest.TestCase):
class FlavaProcessorTest(ProcessorTesterMixin, unittest.TestCase):
processor_class = FlavaProcessor
def setUp(self):
self.tmpdirname = tempfile.mkdtemp()
@@ -91,17 +92,6 @@ class FlavaProcessorTest(unittest.TestCase):
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_slow = self.get_tokenizer()
tokenizer_fast = self.get_rust_tokenizer()

20
tests/models/git/test_processor_git.py
View File

@@ -15,21 +15,22 @@ import shutil
import tempfile
import unittest
import numpy as np
import pytest
from transformers.testing_utils import require_vision
from transformers.utils import is_vision_available
from ...test_processing_common import ProcessorTesterMixin
if is_vision_available():
from PIL import Image
from transformers import AutoProcessor, BertTokenizer, CLIPImageProcessor, GitProcessor, PreTrainedTokenizerFast
@require_vision
class GitProcessorTest(unittest.TestCase):
class GitProcessorTest(ProcessorTesterMixin, unittest.TestCase):
processor_class = GitProcessor
def setUp(self):
self.tmpdirname = tempfile.mkdtemp()
@@ -51,17 +52,6 @@ class GitProcessorTest(unittest.TestCase):
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_additional_features(self):
processor = GitProcessor(tokenizer=self.get_tokenizer(), image_processor=self.get_image_processor())
processor.save_pretrained(self.tmpdirname)

15
tests/models/grounding_dino/test_processor_grounding_dino.py
View File

@@ -18,7 +18,6 @@ import shutil
import tempfile
import unittest
import numpy as np
import pytest
from transformers import BertTokenizer, BertTokenizerFast, GroundingDinoProcessor
@@ -35,8 +34,6 @@ if is_torch_available():
from transformers.models.grounding_dino.modeling_grounding_dino import GroundingDinoObjectDetectionOutput
if is_vision_available():
from PIL import Image
from transformers import GroundingDinoImageProcessor
@@ -96,18 +93,6 @@ class GroundingDinoProcessorTest(ProcessorTesterMixin, unittest.TestCase):
def tearDown(self):
shutil.rmtree(self.tmpdirname)
# Copied from tests.models.clip.test_processor_clip.CLIPProcessorTest.prepare_image_inputs
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 get_fake_grounding_dino_output(self):
torch.manual_seed(42)
return GroundingDinoObjectDetectionOutput(

14
tests/models/instructblip/test_processor_instructblip.py
View File

@@ -15,7 +15,6 @@ import shutil
import tempfile
import unittest
import numpy as np
import pytest
from transformers.testing_utils import require_torch, require_vision
@@ -25,8 +24,6 @@ from ...test_processing_common import ProcessorTesterMixin
if is_vision_available():
from PIL import Image
from transformers import (
AutoProcessor,
BertTokenizerFast,
@@ -64,17 +61,6 @@ class InstructBlipProcessorTest(ProcessorTesterMixin, unittest.TestCase):
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_additional_features(self):
processor = InstructBlipProcessor(
tokenizer=self.get_tokenizer(),

32
tests/models/kosmos2/test_processor_kosmos2.py
View File

@@ -23,6 +23,7 @@ import numpy as np
import pytest
import requests
from transformers.models.auto.processing_auto import processor_class_from_name
from transformers.testing_utils import (
get_tests_dir,
require_sentencepiece,
@@ -32,6 +33,8 @@ from transformers.testing_utils import (
)
from transformers.utils import is_vision_available
from ...test_processing_common import ProcessorTesterMixin
if is_vision_available():
from PIL import Image
@@ -52,7 +55,9 @@ SAMPLE_VOCAB = get_tests_dir("fixtures/test_sentencepiece.model")
@require_sentencepiece
@require_tokenizers
@require_vision
class Kosmos2ProcessorTest(unittest.TestCase):
class Kosmos2ProcessorTest(ProcessorTesterMixin, unittest.TestCase):
processor_class = Kosmos2Processor
def setUp(self):
self.tmpdirname = tempfile.mkdtemp()
@@ -65,6 +70,20 @@ class Kosmos2ProcessorTest(unittest.TestCase):
processor = Kosmos2Processor(image_processor, fast_tokenizer)
processor.save_pretrained(self.tmpdirname)
# We override this method to take the fast tokenizer or image processor by default
def get_component(self, attribute, **kwargs):
assert attribute in self.processor_class.attributes
component_class_name = getattr(self.processor_class, f"{attribute}_class")
if isinstance(component_class_name, tuple):
component_class_name = component_class_name[-1]
component_class = processor_class_from_name(component_class_name)
component = component_class.from_pretrained(self.tmpdirname, **kwargs) # noqa
if attribute == "tokenizer" and not component.pad_token:
component.pad_token = "[TEST_PAD]"
return component
def get_tokenizer(self, **kwargs):
return AutoProcessor.from_pretrained(self.tmpdirname, **kwargs).tokenizer
@@ -74,17 +93,6 @@ class Kosmos2ProcessorTest(unittest.TestCase):
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_image_procesor_load_save_reload(self):
# make sure load from Hub repo. -> save -> reload locally work
image_processor = CLIPImageProcessor.from_pretrained("microsoft/kosmos-2-patch14-224")

22
tests/models/layoutlmv2/test_processor_layoutlmv2.py
View File

@@ -19,26 +19,27 @@ import tempfile
import unittest
from typing import List
import numpy as np
from transformers import PreTrainedTokenizer, PreTrainedTokenizerBase, PreTrainedTokenizerFast
from transformers.models.layoutlmv2 import LayoutLMv2Tokenizer, LayoutLMv2TokenizerFast
from transformers.models.layoutlmv2 import LayoutLMv2Processor, LayoutLMv2Tokenizer, LayoutLMv2TokenizerFast
from transformers.models.layoutlmv2.tokenization_layoutlmv2 import VOCAB_FILES_NAMES
from transformers.testing_utils import require_pytesseract, require_tokenizers, require_torch, slow
from transformers.utils import FEATURE_EXTRACTOR_NAME, cached_property, is_pytesseract_available
from ...test_processing_common import ProcessorTesterMixin
if is_pytesseract_available():
from PIL import Image
from transformers import LayoutLMv2ImageProcessor, LayoutLMv2Processor
from transformers import LayoutLMv2ImageProcessor
@require_pytesseract
@require_tokenizers
class LayoutLMv2ProcessorTest(unittest.TestCase):
class LayoutLMv2ProcessorTest(ProcessorTesterMixin, unittest.TestCase):
tokenizer_class = LayoutLMv2Tokenizer
rust_tokenizer_class = LayoutLMv2TokenizerFast
processor_class = LayoutLMv2Processor
def setUp(self):
vocab_tokens = [
@@ -88,17 +89,6 @@ class LayoutLMv2ProcessorTest(unittest.TestCase):
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):
image_processor = self.get_image_processor()
tokenizers = self.get_tokenizers()

22
tests/models/layoutlmv3/test_processor_layoutlmv3.py
View File

@@ -19,26 +19,27 @@ import tempfile
import unittest
from typing import List
import numpy as np
from transformers import PreTrainedTokenizer, PreTrainedTokenizerBase, PreTrainedTokenizerFast
from transformers.models.layoutlmv3 import LayoutLMv3Tokenizer, LayoutLMv3TokenizerFast
from transformers.models.layoutlmv3 import LayoutLMv3Processor, LayoutLMv3Tokenizer, LayoutLMv3TokenizerFast
from transformers.models.layoutlmv3.tokenization_layoutlmv3 import VOCAB_FILES_NAMES
from transformers.testing_utils import require_pytesseract, require_tokenizers, require_torch, slow
from transformers.utils import FEATURE_EXTRACTOR_NAME, cached_property, is_pytesseract_available
from ...test_processing_common import ProcessorTesterMixin
if is_pytesseract_available():
from PIL import Image
from transformers import LayoutLMv3ImageProcessor, LayoutLMv3Processor
from transformers import LayoutLMv3ImageProcessor
@require_pytesseract
@require_tokenizers
class LayoutLMv3ProcessorTest(unittest.TestCase):
class LayoutLMv3ProcessorTest(ProcessorTesterMixin, unittest.TestCase):
tokenizer_class = LayoutLMv3Tokenizer
rust_tokenizer_class = LayoutLMv3TokenizerFast
processor_class = LayoutLMv3Processor
def setUp(self):
# Adapted from Sennrich et al. 2015 and https://github.com/rsennrich/subword-nmt
@@ -101,17 +102,6 @@ class LayoutLMv3ProcessorTest(unittest.TestCase):
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):
image_processor = self.get_image_processor()
tokenizers = self.get_tokenizers()

27
tests/models/layoutxlm/test_processor_layoutxlm.py
View File

@@ -19,10 +19,8 @@ import tempfile
import unittest
from typing import List
import numpy as np
from transformers import PreTrainedTokenizer, PreTrainedTokenizerBase, PreTrainedTokenizerFast
from transformers.models.layoutxlm import LayoutXLMTokenizer, LayoutXLMTokenizerFast
from transformers.models.layoutxlm import LayoutXLMProcessor, LayoutXLMTokenizer, LayoutXLMTokenizerFast
from transformers.testing_utils import (
require_pytesseract,
require_sentencepiece,
@@ -32,19 +30,22 @@ from transformers.testing_utils import (
)
from transformers.utils import FEATURE_EXTRACTOR_NAME, cached_property, is_pytesseract_available
from ...test_processing_common import ProcessorTesterMixin
if is_pytesseract_available():
from PIL import Image
from transformers import LayoutLMv2ImageProcessor, LayoutXLMProcessor
from transformers import LayoutLMv2ImageProcessor
@require_pytesseract
@require_sentencepiece
@require_tokenizers
class LayoutXLMProcessorTest(unittest.TestCase):
class LayoutXLMProcessorTest(ProcessorTesterMixin, unittest.TestCase):
tokenizer_class = LayoutXLMTokenizer
rust_tokenizer_class = LayoutXLMTokenizerFast
processor_class = LayoutXLMProcessor
def setUp(self):
image_processor_map = {
@@ -61,6 +62,11 @@ class LayoutXLMProcessorTest(unittest.TestCase):
# taken from `test_tokenization_layoutxlm.LayoutXLMTokenizationTest.test_save_pretrained`
self.tokenizer_pretrained_name = "hf-internal-testing/tiny-random-layoutxlm"
tokenizer = self.get_tokenizer()
image_processor = self.get_image_processor()
processor = LayoutXLMProcessor(tokenizer=tokenizer, image_processor=image_processor)
processor.save_pretrained(self.tmpdirname)
def get_tokenizer(self, **kwargs) -> PreTrainedTokenizer:
return self.tokenizer_class.from_pretrained(self.tokenizer_pretrained_name, **kwargs)
@@ -76,17 +82,6 @@ class LayoutXLMProcessorTest(unittest.TestCase):
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):
image_processor = self.get_image_processor()
tokenizers = self.get_tokenizers()

20
tests/models/mgp_str/test_processor_mgp_str.py
View File

@@ -70,6 +70,17 @@ class MgpstrProcessorTest(unittest.TestCase):
with open(self.image_processor_file, "w", encoding="utf-8") as fp:
json.dump(image_processor_map, fp)
# We copy here rather than use the ProcessorTesterMixin as this processor has a `char_tokenizer` instad of a
# tokenizer attribute, which means all the tests would need to be overridden.
@require_vision
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 get_tokenizer(self, **kwargs):
return MgpstrTokenizer.from_pretrained(self.tmpdirname, **kwargs)
@@ -79,15 +90,6 @@ class MgpstrProcessorTest(unittest.TestCase):
def tearDown(self):
shutil.rmtree(self.tmpdirname)
def prepare_image_inputs(self):
"""This function prepares a list of PIL images."""
image_input = np.random.randint(255, size=(3, 30, 400), dtype=np.uint8)
image_input = Image.fromarray(np.moveaxis(image_input, 0, -1))
return image_input
def test_save_load_pretrained_default(self):
tokenizer = self.get_tokenizer()
image_processor = self.get_image_processor()

20
tests/models/owlvit/test_processor_owlvit.py
View File

@@ -18,7 +18,6 @@ import shutil
import tempfile
import unittest
import numpy as np
import pytest
from transformers import CLIPTokenizer, CLIPTokenizerFast
@@ -26,15 +25,17 @@ from transformers.models.clip.tokenization_clip import VOCAB_FILES_NAMES
from transformers.testing_utils import require_vision
from transformers.utils import IMAGE_PROCESSOR_NAME, is_vision_available
from ...test_processing_common import ProcessorTesterMixin
if is_vision_available():
from PIL import Image
from transformers import OwlViTImageProcessor, OwlViTProcessor
@require_vision
class OwlViTProcessorTest(unittest.TestCase):
class OwlViTProcessorTest(ProcessorTesterMixin, unittest.TestCase):
processor_class = OwlViTProcessor
def setUp(self):
self.tmpdirname = tempfile.mkdtemp()
@@ -75,17 +76,6 @@ class OwlViTProcessorTest(unittest.TestCase):
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_slow = self.get_tokenizer()
tokenizer_fast = self.get_rust_tokenizer()

20
tests/models/pix2struct/test_processor_pix2struct.py
View File

@@ -15,16 +15,15 @@ import shutil
import tempfile
import unittest
import numpy as np
import pytest
from transformers.testing_utils import require_torch, require_vision
from transformers.utils import is_vision_available
from ...test_processing_common import ProcessorTesterMixin
if is_vision_available():
from PIL import Image
from transformers import (
AutoProcessor,
Pix2StructImageProcessor,
@@ -36,7 +35,9 @@ if is_vision_available():
@require_vision
@require_torch
class Pix2StructProcessorTest(unittest.TestCase):
class Pix2StructProcessorTest(ProcessorTesterMixin, unittest.TestCase):
processor_class = Pix2StructProcessor
def setUp(self):
self.tmpdirname = tempfile.mkdtemp()
@@ -56,17 +57,6 @@ class Pix2StructProcessorTest(unittest.TestCase):
def tearDown(self):
shutil.rmtree(self.tmpdirname)
def prepare_image_inputs(self):
"""
This function prepares a list of random PIL images of the same fixed size.
"""
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_additional_features(self):
processor = Pix2StructProcessor(tokenizer=self.get_tokenizer(), image_processor=self.get_image_processor())
processor.save_pretrained(self.tmpdirname)

35
tests/models/sam/test_processor_sam.py
View File

@@ -26,6 +26,8 @@ from transformers.testing_utils import (
)
from transformers.utils import is_tf_available, is_torch_available, is_vision_available
from ...test_processing_common import prepare_image_inputs
if is_vision_available():
from PIL import Image
@@ -54,13 +56,10 @@ class SamProcessorTest(unittest.TestCase):
def tearDown(self):
shutil.rmtree(self.tmpdirname)
# Processor tester class can't use ProcessorTesterMixin atm because the processor is atypical e.g. only contains an image processor
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
"""This function prepares a list of PIL images."""
return prepare_image_inputs()
def prepare_mask_inputs(self):
"""This function prepares a list of PIL images, or a list of numpy arrays if one specifies numpify=True,
@@ -166,16 +165,10 @@ class TFSamProcessorTest(unittest.TestCase):
def tearDown(self):
shutil.rmtree(self.tmpdirname)
# Processor tester class can't use ProcessorTesterMixin as processor is atypical e.g. only contains an image processor and it assumes torch
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
"""This function prepares a list of PIL images."""
return prepare_image_inputs()
def test_save_load_pretrained_additional_features(self):
processor = SamProcessor(image_processor=self.get_image_processor())
@@ -255,16 +248,10 @@ class SamProcessorEquivalenceTest(unittest.TestCase):
def tearDown(self):
shutil.rmtree(self.tmpdirname)
# Processor tester class can't use ProcessorTesterMixin atm because the processor is atypical e.g. only contains an image processor
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
"""This function prepares a list of PIL images."""
return prepare_image_inputs()
@is_pt_tf_cross_test
def test_post_process_masks_equivalence(self):

26
tests/models/udop/test_processor_udop.py
View File

@@ -19,12 +19,11 @@ import tempfile
import unittest
from typing import List
import numpy as np
from transformers import (
PreTrainedTokenizer,
PreTrainedTokenizerBase,
PreTrainedTokenizerFast,
UdopProcessor,
UdopTokenizer,
UdopTokenizerFast,
)
@@ -37,6 +36,8 @@ from transformers.testing_utils import (
)
from transformers.utils import FEATURE_EXTRACTOR_NAME, cached_property, is_pytesseract_available, is_torch_available
from ...test_processing_common import ProcessorTesterMixin
if is_torch_available():
import torch
@@ -45,16 +46,17 @@ if is_torch_available():
if is_pytesseract_available():
from PIL import Image
from transformers import LayoutLMv3ImageProcessor, UdopProcessor
from transformers import LayoutLMv3ImageProcessor
@require_pytesseract
@require_sentencepiece
@require_tokenizers
class UdopProcessorTest(unittest.TestCase):
class UdopProcessorTest(ProcessorTesterMixin, unittest.TestCase):
tokenizer_class = UdopTokenizer
rust_tokenizer_class = UdopTokenizerFast
maxDiff = None
processor_class = UdopProcessor
def setUp(self):
image_processor_map = {
@@ -70,6 +72,11 @@ class UdopProcessorTest(unittest.TestCase):
self.tokenizer_pretrained_name = "microsoft/udop-large"
image_processor = self.get_image_processor()
tokenizer = self.get_tokenizers()[0]
processor = UdopProcessor(image_processor=image_processor, tokenizer=tokenizer)
processor.save_pretrained(self.tmpdirname)
def get_tokenizer(self, **kwargs) -> PreTrainedTokenizer:
return self.tokenizer_class.from_pretrained(self.tokenizer_pretrained_name, **kwargs)
@@ -85,17 +92,6 @@ class UdopProcessorTest(unittest.TestCase):
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):
image_processor = self.get_image_processor()
tokenizers = self.get_tokenizers()

26
tests/models/vision_text_dual_encoder/test_processor_vision_text_dual_encoder.py
View File

@@ -18,23 +18,23 @@ import shutil
import tempfile
import unittest
import numpy as np
from transformers import BertTokenizerFast
from transformers.models.bert.tokenization_bert import VOCAB_FILES_NAMES, BertTokenizer
from transformers.testing_utils import require_tokenizers, require_vision
from transformers.utils import IMAGE_PROCESSOR_NAME, is_vision_available
from ...test_processing_common import ProcessorTesterMixin
if is_vision_available():
from PIL import Image
from transformers import VisionTextDualEncoderProcessor, ViTImageProcessor
@require_tokenizers
@require_vision
class VisionTextDualEncoderProcessorTest(unittest.TestCase):
class VisionTextDualEncoderProcessorTest(ProcessorTesterMixin, unittest.TestCase):
processor_class = VisionTextDualEncoderProcessor
def setUp(self):
self.tmpdirname = tempfile.mkdtemp()
@@ -54,6 +54,11 @@ class VisionTextDualEncoderProcessorTest(unittest.TestCase):
with open(self.image_processor_file, "w", encoding="utf-8") as fp:
json.dump(image_processor_map, fp)
tokenizer = self.get_tokenizer()
image_processor = self.get_image_processor()
processor = VisionTextDualEncoderProcessor(tokenizer=tokenizer, image_processor=image_processor)
processor.save_pretrained(self.tmpdirname)
def get_tokenizer(self, **kwargs):
return BertTokenizer.from_pretrained(self.tmpdirname, **kwargs)
@@ -63,17 +68,6 @@ class VisionTextDualEncoderProcessorTest(unittest.TestCase):
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()
image_processor = self.get_image_processor()

15
tests/test_processing_common.py
View File

@@ -44,6 +44,13 @@ if is_vision_available():
from transformers import CLIPImageProcessor
def prepare_image_inputs():
"""This function prepares a list of PIL images"""
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
@require_torch
@require_vision
@require_torch
@@ -81,12 +88,8 @@ class ProcessorTesterMixin:
@require_vision
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
"""This function prepares a list of PIL images for testing"""
return prepare_image_inputs()
def test_processor_to_json_string(self):
processor = self.get_processor()