Skip to content

GitLab

Unverified Commit a95fd354 authored by amyeroberts's avatar amyeroberts Committed by GitHub
Browse files

Vision processors - replace FE with IPs (#20590) 



* Replace FE references with IPs

* Update processor tests

* Update src/transformers/models/clip/processing_clip.py
Co-authored-by: default avatarSylvain Gugger <35901082+sgugger@users.noreply.github.com>

* Update src/transformers/models/clip/processing_clip.py
Co-authored-by: default avatarSylvain Gugger <35901082+sgugger@users.noreply.github.com>

* Update warning messages v4.27 -> v5

* Fixup

* Update Chinese CLIP processor

* Add feature_extractor property

* Add attributes

* Add tests
Co-authored-by: default avatarSylvain Gugger <35901082+sgugger@users.noreply.github.com>
parent 704027f0
Show whitespace changes
Inline Side-by-side
Showing with 65 additions and 65 deletions
tests/models/layoutlmv3/test_processor_layoutlmv3.py
View file @ a95fd354
......@@ -31,7 +31,7 @@ from transformers.utils import FEATURE_EXTRACTOR_NAME, cached_property, is_pytes
if is_pytesseract_available():
from PIL import Image
from transformers import LayoutLMv3FeatureExtractor, LayoutLMv3Processor
from transformers import LayoutLMv3ImageProcessor, LayoutLMv3Processor
@require_pytesseract
......@@ -76,7 +76,7 @@ class LayoutLMv3ProcessorTest(unittest.TestCase):
with open(self.merges_file, "w", encoding="utf-8") as fp:
fp.write("\n".join(merges))
feature_extractor_map = {
image_processor_map = {
"do_resize": True,
"size": 224,
"apply_ocr": True,
......@@ -84,7 +84,7 @@ class LayoutLMv3ProcessorTest(unittest.TestCase):
self.feature_extraction_file = os.path.join(self.tmpdirname, FEATURE_EXTRACTOR_NAME)
with open(self.feature_extraction_file, "w", encoding="utf-8") as fp:
fp.write(json.dumps(feature_extractor_map) + "\n")
fp.write(json.dumps(image_processor_map) + "\n")
def get_tokenizer(self, **kwargs) -> PreTrainedTokenizer:
return self.tokenizer_class.from_pretrained(self.tmpdirname, **kwargs)
......@@ -95,8 +95,8 @@ class LayoutLMv3ProcessorTest(unittest.TestCase):
def get_tokenizers(self, **kwargs) -> List[PreTrainedTokenizerBase]:
return [self.get_tokenizer(**kwargs), self.get_rust_tokenizer(**kwargs)]
def get_feature_extractor(self, **kwargs):
return LayoutLMv3FeatureExtractor.from_pretrained(self.tmpdirname, **kwargs)
def get_image_processor(self, **kwargs):
return LayoutLMv3ImageProcessor.from_pretrained(self.tmpdirname, **kwargs)
def tearDown(self):
shutil.rmtree(self.tmpdirname)
......@@ -113,10 +113,10 @@ class LayoutLMv3ProcessorTest(unittest.TestCase):
return image_inputs
def test_save_load_pretrained_default(self):
feature_extractor = self.get_feature_extractor()
image_processor = self.get_image_processor()
tokenizers = self.get_tokenizers()
for tokenizer in tokenizers:
processor = LayoutLMv3Processor(feature_extractor=feature_extractor, tokenizer=tokenizer)
processor = LayoutLMv3Processor(image_processor=image_processor, tokenizer=tokenizer)
processor.save_pretrained(self.tmpdirname)
processor = LayoutLMv3Processor.from_pretrained(self.tmpdirname)
......@@ -124,16 +124,16 @@ class LayoutLMv3ProcessorTest(unittest.TestCase):
self.assertEqual(processor.tokenizer.get_vocab(), tokenizer.get_vocab())
self.assertIsInstance(processor.tokenizer, (LayoutLMv3Tokenizer, LayoutLMv3TokenizerFast))
self.assertEqual(processor.feature_extractor.to_json_string(), feature_extractor.to_json_string())
self.assertIsInstance(processor.feature_extractor, LayoutLMv3FeatureExtractor)
self.assertEqual(processor.image_processor.to_json_string(), image_processor.to_json_string())
self.assertIsInstance(processor.image_processor, LayoutLMv3ImageProcessor)
def test_save_load_pretrained_additional_features(self):
processor = LayoutLMv3Processor(feature_extractor=self.get_feature_extractor(), tokenizer=self.get_tokenizer())
processor = LayoutLMv3Processor(image_processor=self.get_image_processor(), tokenizer=self.get_tokenizer())
processor.save_pretrained(self.tmpdirname)
# slow tokenizer
tokenizer_add_kwargs = self.get_tokenizer(bos_token="(BOS)", eos_token="(EOS)")
feature_extractor_add_kwargs = self.get_feature_extractor(do_resize=False, size=30)
image_processor_add_kwargs = self.get_image_processor(do_resize=False, size=30)
processor = LayoutLMv3Processor.from_pretrained(
self.tmpdirname, use_fast=False, bos_token="(BOS)", eos_token="(EOS)", do_resize=False, size=30
......@@ -142,12 +142,12 @@ class LayoutLMv3ProcessorTest(unittest.TestCase):
self.assertEqual(processor.tokenizer.get_vocab(), tokenizer_add_kwargs.get_vocab())
self.assertIsInstance(processor.tokenizer, LayoutLMv3Tokenizer)
self.assertEqual(processor.feature_extractor.to_json_string(), feature_extractor_add_kwargs.to_json_string())
self.assertIsInstance(processor.feature_extractor, LayoutLMv3FeatureExtractor)
self.assertEqual(processor.image_processor.to_json_string(), image_processor_add_kwargs.to_json_string())
self.assertIsInstance(processor.image_processor, LayoutLMv3ImageProcessor)
# fast tokenizer
tokenizer_add_kwargs = self.get_rust_tokenizer(bos_token="(BOS)", eos_token="(EOS)")
feature_extractor_add_kwargs = self.get_feature_extractor(do_resize=False, size=30)
image_processor_add_kwargs = self.get_image_processor(do_resize=False, size=30)
processor = LayoutLMv3Processor.from_pretrained(
self.tmpdirname, bos_token="(BOS)", eos_token="(EOS)", do_resize=False, size=30
......@@ -156,14 +156,14 @@ class LayoutLMv3ProcessorTest(unittest.TestCase):
self.assertEqual(processor.tokenizer.get_vocab(), tokenizer_add_kwargs.get_vocab())
self.assertIsInstance(processor.tokenizer, LayoutLMv3TokenizerFast)
self.assertEqual(processor.feature_extractor.to_json_string(), feature_extractor_add_kwargs.to_json_string())
self.assertIsInstance(processor.feature_extractor, LayoutLMv3FeatureExtractor)
self.assertEqual(processor.image_processor.to_json_string(), image_processor_add_kwargs.to_json_string())
self.assertIsInstance(processor.image_processor, LayoutLMv3ImageProcessor)
def test_model_input_names(self):
feature_extractor = self.get_feature_extractor()
image_processor = self.get_image_processor()
tokenizer = self.get_tokenizer()
processor = LayoutLMv3Processor(tokenizer=tokenizer, feature_extractor=feature_extractor)
processor = LayoutLMv3Processor(tokenizer=tokenizer, image_processor=image_processor)
input_str = "lower newer"
image_input = self.prepare_image_inputs()
......@@ -200,15 +200,15 @@ class LayoutLMv3ProcessorIntegrationTests(unittest.TestCase):
def test_processor_case_1(self):
# case 1: document image classification (training, inference) + token classification (inference), apply_ocr = True
feature_extractor = LayoutLMv3FeatureExtractor()
image_processor = LayoutLMv3ImageProcessor()
tokenizers = self.get_tokenizers
images = self.get_images
for tokenizer in tokenizers:
processor = LayoutLMv3Processor(feature_extractor=feature_extractor, tokenizer=tokenizer)
processor = LayoutLMv3Processor(image_processor=image_processor, tokenizer=tokenizer)
# not batched
input_feat_extract = feature_extractor(images[0], return_tensors="pt")
input_image_proc = image_processor(images[0], return_tensors="pt")
input_processor = processor(images[0], return_tensors="pt")
# verify keys
......@@ -218,7 +218,7 @@ class LayoutLMv3ProcessorIntegrationTests(unittest.TestCase):
# verify image
self.assertAlmostEqual(
input_feat_extract["pixel_values"].sum(), input_processor["pixel_values"].sum(), delta=1e-2
input_image_proc["pixel_values"].sum(), input_processor["pixel_values"].sum(), delta=1e-2
)
# verify input_ids
......@@ -230,7 +230,7 @@ class LayoutLMv3ProcessorIntegrationTests(unittest.TestCase):
self.assertSequenceEqual(decoding, expected_decoding)
# batched
input_feat_extract = feature_extractor(images, return_tensors="pt")
input_image_proc = image_processor(images, return_tensors="pt")
input_processor = processor(images, padding=True, return_tensors="pt")
# verify keys
......@@ -240,7 +240,7 @@ class LayoutLMv3ProcessorIntegrationTests(unittest.TestCase):
# verify images
self.assertAlmostEqual(
input_feat_extract["pixel_values"].sum(), input_processor["pixel_values"].sum(), delta=1e-2
input_image_proc["pixel_values"].sum(), input_processor["pixel_values"].sum(), delta=1e-2
)
# verify input_ids
......@@ -255,12 +255,12 @@ class LayoutLMv3ProcessorIntegrationTests(unittest.TestCase):
def test_processor_case_2(self):
# case 2: document image classification (training, inference) + token classification (inference), apply_ocr=False
feature_extractor = LayoutLMv3FeatureExtractor(apply_ocr=False)
image_processor = LayoutLMv3ImageProcessor(apply_ocr=False)
tokenizers = self.get_tokenizers
images = self.get_images
for tokenizer in tokenizers:
processor = LayoutLMv3Processor(feature_extractor=feature_extractor, tokenizer=tokenizer)
processor = LayoutLMv3Processor(image_processor=image_processor, tokenizer=tokenizer)
# not batched
words = ["hello", "world"]
......@@ -309,12 +309,12 @@ class LayoutLMv3ProcessorIntegrationTests(unittest.TestCase):
def test_processor_case_3(self):
# case 3: token classification (training), apply_ocr=False
feature_extractor = LayoutLMv3FeatureExtractor(apply_ocr=False)
image_processor = LayoutLMv3ImageProcessor(apply_ocr=False)
tokenizers = self.get_tokenizers
images = self.get_images
for tokenizer in tokenizers:
processor = LayoutLMv3Processor(feature_extractor=feature_extractor, tokenizer=tokenizer)
processor = LayoutLMv3Processor(image_processor=image_processor, tokenizer=tokenizer)
# not batched
words = ["weirdly", "world"]
......@@ -374,12 +374,12 @@ class LayoutLMv3ProcessorIntegrationTests(unittest.TestCase):
def test_processor_case_4(self):
# case 4: visual question answering (inference), apply_ocr=True
feature_extractor = LayoutLMv3FeatureExtractor()
image_processor = LayoutLMv3ImageProcessor()
tokenizers = self.get_tokenizers
images = self.get_images
for tokenizer in tokenizers:
processor = LayoutLMv3Processor(feature_extractor=feature_extractor, tokenizer=tokenizer)
processor = LayoutLMv3Processor(image_processor=image_processor, tokenizer=tokenizer)
# not batched
question = "What's his name?"
......@@ -425,12 +425,12 @@ class LayoutLMv3ProcessorIntegrationTests(unittest.TestCase):
def test_processor_case_5(self):
# case 5: visual question answering (inference), apply_ocr=False
feature_extractor = LayoutLMv3FeatureExtractor(apply_ocr=False)
image_processor = LayoutLMv3ImageProcessor(apply_ocr=False)
tokenizers = self.get_tokenizers
images = self.get_images
for tokenizer in tokenizers:
processor = LayoutLMv3Processor(feature_extractor=feature_extractor, tokenizer=tokenizer)
processor = LayoutLMv3Processor(image_processor=image_processor, tokenizer=tokenizer)
# not batched
question = "What's his name?"
......
tests/models/owlvit/test_processor_owlvit.py
View file @ a95fd354
......@@ -24,13 +24,13 @@ import pytest
from transformers import CLIPTokenizer, CLIPTokenizerFast
from transformers.models.clip.tokenization_clip import VOCAB_FILES_NAMES
from transformers.testing_utils import require_vision
from transformers.utils import FEATURE_EXTRACTOR_NAME, is_vision_available
from transformers.utils import IMAGE_PROCESSOR_NAME, is_vision_available
if is_vision_available():
from PIL import Image
from transformers import OwlViTFeatureExtractor, OwlViTProcessor
from transformers import OwlViTImageProcessor, OwlViTProcessor
@require_vision
......@@ -52,7 +52,7 @@ class OwlViTProcessorTest(unittest.TestCase):
with open(self.merges_file, "w", encoding="utf-8") as fp:
fp.write("\n".join(merges))
feature_extractor_map = {
image_processor_map = {
"do_resize": True,
"size": 20,
"do_center_crop": True,
......@@ -61,9 +61,9 @@ class OwlViTProcessorTest(unittest.TestCase):
"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)
self.image_processor_file = os.path.join(self.tmpdirname, IMAGE_PROCESSOR_NAME)
with open(self.image_processor_file, "w", encoding="utf-8") as fp:
json.dump(image_processor_map, fp)
def get_tokenizer(self, **kwargs):
return CLIPTokenizer.from_pretrained(self.tmpdirname, pad_token="!", **kwargs)
......@@ -71,8 +71,8 @@ class OwlViTProcessorTest(unittest.TestCase):
def get_rust_tokenizer(self, **kwargs):
return CLIPTokenizerFast.from_pretrained(self.tmpdirname, pad_token="!", **kwargs)
def get_feature_extractor(self, **kwargs):
return OwlViTFeatureExtractor.from_pretrained(self.tmpdirname, **kwargs)
def get_image_processor(self, **kwargs):
return OwlViTImageProcessor.from_pretrained(self.tmpdirname, **kwargs)
def tearDown(self):
shutil.rmtree(self.tmpdirname)
......@@ -91,13 +91,13 @@ class OwlViTProcessorTest(unittest.TestCase):
def test_save_load_pretrained_default(self):
tokenizer_slow = self.get_tokenizer()
tokenizer_fast = self.get_rust_tokenizer()
feature_extractor = self.get_feature_extractor()
image_processor = self.get_image_processor()
processor_slow = OwlViTProcessor(tokenizer=tokenizer_slow, feature_extractor=feature_extractor)
processor_slow = OwlViTProcessor(tokenizer=tokenizer_slow, image_processor=image_processor)
processor_slow.save_pretrained(self.tmpdirname)
processor_slow = OwlViTProcessor.from_pretrained(self.tmpdirname, use_fast=False)
processor_fast = OwlViTProcessor(tokenizer=tokenizer_fast, feature_extractor=feature_extractor)
processor_fast = OwlViTProcessor(tokenizer=tokenizer_fast, image_processor=image_processor)
processor_fast.save_pretrained(self.tmpdirname)
processor_fast = OwlViTProcessor.from_pretrained(self.tmpdirname)
......@@ -107,17 +107,17 @@ class OwlViTProcessorTest(unittest.TestCase):
self.assertIsInstance(processor_slow.tokenizer, CLIPTokenizer)
self.assertIsInstance(processor_fast.tokenizer, CLIPTokenizerFast)
self.assertEqual(processor_slow.feature_extractor.to_json_string(), feature_extractor.to_json_string())
self.assertEqual(processor_fast.feature_extractor.to_json_string(), feature_extractor.to_json_string())
self.assertIsInstance(processor_slow.feature_extractor, OwlViTFeatureExtractor)
self.assertIsInstance(processor_fast.feature_extractor, OwlViTFeatureExtractor)
self.assertEqual(processor_slow.image_processor.to_json_string(), image_processor.to_json_string())
self.assertEqual(processor_fast.image_processor.to_json_string(), image_processor.to_json_string())
self.assertIsInstance(processor_slow.image_processor, OwlViTImageProcessor)
self.assertIsInstance(processor_fast.image_processor, OwlViTImageProcessor)
def test_save_load_pretrained_additional_features(self):
processor = OwlViTProcessor(tokenizer=self.get_tokenizer(), feature_extractor=self.get_feature_extractor())
processor = OwlViTProcessor(tokenizer=self.get_tokenizer(), image_processor=self.get_image_processor())
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)
image_processor_add_kwargs = self.get_image_processor(do_normalize=False)
processor = OwlViTProcessor.from_pretrained(
self.tmpdirname, bos_token="(BOS)", eos_token="(EOS)", do_normalize=False
......@@ -126,28 +126,28 @@ class OwlViTProcessorTest(unittest.TestCase):
self.assertEqual(processor.tokenizer.get_vocab(), tokenizer_add_kwargs.get_vocab())
self.assertIsInstance(processor.tokenizer, CLIPTokenizerFast)
self.assertEqual(processor.feature_extractor.to_json_string(), feature_extractor_add_kwargs.to_json_string())
self.assertIsInstance(processor.feature_extractor, OwlViTFeatureExtractor)
self.assertEqual(processor.image_processor.to_json_string(), image_processor_add_kwargs.to_json_string())
self.assertIsInstance(processor.image_processor, OwlViTImageProcessor)
def test_feature_extractor(self):
feature_extractor = self.get_feature_extractor()
def test_image_processor(self):
image_processor = self.get_image_processor()
tokenizer = self.get_tokenizer()
processor = OwlViTProcessor(tokenizer=tokenizer, feature_extractor=feature_extractor)
processor = OwlViTProcessor(tokenizer=tokenizer, image_processor=image_processor)
image_input = self.prepare_image_inputs()
input_feat_extract = feature_extractor(image_input, return_tensors="np")
input_image_proc = image_processor(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)
for key in input_image_proc.keys():
self.assertAlmostEqual(input_image_proc[key].sum(), input_processor[key].sum(), delta=1e-2)
def test_tokenizer(self):
feature_extractor = self.get_feature_extractor()
image_processor = self.get_image_processor()
tokenizer = self.get_tokenizer()
processor = OwlViTProcessor(tokenizer=tokenizer, feature_extractor=feature_extractor)
processor = OwlViTProcessor(tokenizer=tokenizer, image_processor=image_processor)
input_str = "lower newer"
......@@ -159,10 +159,10 @@ class OwlViTProcessorTest(unittest.TestCase):
self.assertListEqual(encoded_tok[key][0].tolist(), encoded_processor[key][0].tolist())
def test_processor(self):
feature_extractor = self.get_feature_extractor()
image_processor = self.get_image_processor()
tokenizer = self.get_tokenizer()
processor = OwlViTProcessor(tokenizer=tokenizer, feature_extractor=feature_extractor)
processor = OwlViTProcessor(tokenizer=tokenizer, image_processor=image_processor)
input_str = "lower newer"
image_input = self.prepare_image_inputs()
......@@ -228,10 +228,10 @@ class OwlViTProcessorTest(unittest.TestCase):
self.assertListEqual(list(input_ids[1]), predicted_ids[1])
def test_processor_case2(self):
feature_extractor = self.get_feature_extractor()
image_processor = self.get_image_processor()
tokenizer = self.get_tokenizer()
processor = OwlViTProcessor(tokenizer=tokenizer, feature_extractor=feature_extractor)
processor = OwlViTProcessor(tokenizer=tokenizer, image_processor=image_processor)
image_input = self.prepare_image_inputs()
query_input = self.prepare_image_inputs()
......@@ -245,10 +245,10 @@ class OwlViTProcessorTest(unittest.TestCase):
processor()
def test_tokenizer_decode(self):
feature_extractor = self.get_feature_extractor()
image_processor = self.get_image_processor()
tokenizer = self.get_tokenizer()
processor = OwlViTProcessor(tokenizer=tokenizer, feature_extractor=feature_extractor)
processor = OwlViTProcessor(tokenizer=tokenizer, image_processor=image_processor)
predicted_ids = [[1, 4, 5, 8, 1, 0, 8], [3, 4, 3, 1, 1, 8, 9]]
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment