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

Vision processors - replace FE with IPs (#20590)

Browse Source 
* Replace FE references with IPs

* Update processor tests

* Update src/transformers/models/clip/processing_clip.py

Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>

* Update src/transformers/models/clip/processing_clip.py

Co-authored-by: Sylvain 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: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>
This commit is contained in:
amyeroberts
2022-12-09 10:48:34 +00:00
committed by GitHub
parent 704027f0ef
commit a95fd35426
22 changed files with 681 additions and 375 deletions
Download Patch File Download Diff File Expand all files Collapse all files

58
tests/models/clipseg/test_processor_clipseg.py
View File

@@ -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 CLIPSegProcessor, ViTFeatureExtractor
from transformers import CLIPSegProcessor, ViTImageProcessor
@require_vision
@@ -52,7 +52,7 @@ class CLIPSegProcessorTest(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 CLIPSegProcessorTest(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, **kwargs)
@@ -71,8 +71,8 @@ class CLIPSegProcessorTest(unittest.TestCase):
def get_rust_tokenizer(self, **kwargs):
return CLIPTokenizerFast.from_pretrained(self.tmpdirname, **kwargs)
def get_feature_extractor(self, **kwargs):
return ViTFeatureExtractor.from_pretrained(self.tmpdirname, **kwargs)
def get_image_processor(self, **kwargs):
return ViTImageProcessor.from_pretrained(self.tmpdirname, **kwargs)
def tearDown(self):
shutil.rmtree(self.tmpdirname)
@@ -90,13 +90,13 @@ class CLIPSegProcessorTest(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 = CLIPSegProcessor(tokenizer=tokenizer_slow, feature_extractor=feature_extractor)
processor_slow = CLIPSegProcessor(tokenizer=tokenizer_slow, image_processor=image_processor)
processor_slow.save_pretrained(self.tmpdirname)
processor_slow = CLIPSegProcessor.from_pretrained(self.tmpdirname, use_fast=False)
processor_fast = CLIPSegProcessor(tokenizer=tokenizer_fast, feature_extractor=feature_extractor)
processor_fast = CLIPSegProcessor(tokenizer=tokenizer_fast, image_processor=image_processor)
processor_fast.save_pretrained(self.tmpdirname)
processor_fast = CLIPSegProcessor.from_pretrained(self.tmpdirname)
@@ -106,17 +106,17 @@ class CLIPSegProcessorTest(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, ViTFeatureExtractor)
self.assertIsInstance(processor_fast.feature_extractor, ViTFeatureExtractor)
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, ViTImageProcessor)
self.assertIsInstance(processor_fast.image_processor, ViTImageProcessor)
def test_save_load_pretrained_additional_features(self):
processor = CLIPSegProcessor(tokenizer=self.get_tokenizer(), feature_extractor=self.get_feature_extractor())
processor = CLIPSegProcessor(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, padding_value=1.0)
image_processor_add_kwargs = self.get_image_processor(do_normalize=False, padding_value=1.0)
processor = CLIPSegProcessor.from_pretrained(
self.tmpdirname, bos_token="(BOS)", eos_token="(EOS)", do_normalize=False, padding_value=1.0
@@ -125,28 +125,28 @@ class CLIPSegProcessorTest(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, ViTFeatureExtractor)
self.assertEqual(processor.image_processor.to_json_string(), image_processor_add_kwargs.to_json_string())
self.assertIsInstance(processor.image_processor, ViTImageProcessor)
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 = CLIPSegProcessor(tokenizer=tokenizer, feature_extractor=feature_extractor)
processor = CLIPSegProcessor(tokenizer=tokenizer, image_processor=image_processor)
image_input = self.prepare_image_inputs()
input_feat_extract = feature_extractor(image_input, return_tensors="np")
input_feat_extract = 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)
def test_tokenizer(self):
feature_extractor = self.get_feature_extractor()
image_processor = self.get_image_processor()
tokenizer = self.get_tokenizer()
processor = CLIPSegProcessor(tokenizer=tokenizer, feature_extractor=feature_extractor)
processor = CLIPSegProcessor(tokenizer=tokenizer, image_processor=image_processor)
input_str = "lower newer"
@@ -158,10 +158,10 @@ class CLIPSegProcessorTest(unittest.TestCase):
self.assertListEqual(encoded_tok[key], encoded_processor[key])
def test_processor(self):
feature_extractor = self.get_feature_extractor()
image_processor = self.get_image_processor()
tokenizer = self.get_tokenizer()
processor = CLIPSegProcessor(tokenizer=tokenizer, feature_extractor=feature_extractor)
processor = CLIPSegProcessor(tokenizer=tokenizer, image_processor=image_processor)
input_str = "lower newer"
image_input = self.prepare_image_inputs()
@@ -175,10 +175,10 @@ class CLIPSegProcessorTest(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 = CLIPSegProcessor(tokenizer=tokenizer, feature_extractor=feature_extractor)
processor = CLIPSegProcessor(tokenizer=tokenizer, image_processor=image_processor)
predicted_ids = [[1, 4, 5, 8, 1, 0, 8], [3, 4, 3, 1, 1, 8, 9]]