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

Add support for args to ProcessorMixin for backward compatibility (#33479)

Browse Source 
* add check and prepare args for BC to ProcessorMixin, improve ProcessorTesterMixin

* change size and crop_size in processor kwargs tests to do_rescale and rescale_factor

* remove unnecessary llava processor kwargs test overwrite

* nit

* change data_arg_name to input_name

* Remove unnecessary test override

* Remove unnecessary tests Paligemma

* Move test_prepare_and_validate_optional_call_args to TesterMixin, add docstring
This commit is contained in:
Yoni Gozlan
2024-09-20 11:40:59 -04:00
committed by GitHub
parent 31caf0b95f
commit c0c6815dc9
10 changed files with 173 additions and 812 deletions
Download Patch File Download Diff File Expand all files Collapse all files

115
tests/models/altclip/test_processor_altclip.py
View File

@@ -18,7 +18,7 @@ import tempfile
import unittest
from transformers import XLMRobertaTokenizer, XLMRobertaTokenizerFast
from transformers.testing_utils import require_torch, require_vision
from transformers.testing_utils import require_vision
from transformers.utils import is_vision_available
from ...test_processing_common import ProcessorTesterMixin
@@ -50,116 +50,3 @@ class AltClipProcessorTest(ProcessorTesterMixin, unittest.TestCase):
def get_image_processor(self, **kwargs):
return CLIPImageProcessor.from_pretrained(self.model_id, **kwargs)
@require_torch
@require_vision
def test_unstructured_kwargs_batched(self):
if "image_processor" not in self.processor_class.attributes:
self.skipTest(f"image_processor attribute not present in {self.processor_class}")
image_processor = self.get_component("image_processor")
tokenizer = self.get_component("tokenizer")
processor = self.processor_class(tokenizer=tokenizer, image_processor=image_processor)
self.skip_processor_without_typed_kwargs(processor)
input_str = ["lower newer", "upper older longer string"]
image_input = self.prepare_image_inputs() * 2
inputs = processor(
text=input_str,
images=image_input,
return_tensors="pt",
crop_size={"height": 214, "width": 214},
padding="longest",
max_length=76,
)
self.assertEqual(inputs["pixel_values"].shape[2], 214)
self.assertEqual(len(inputs["input_ids"][0]), 7)
def test_structured_kwargs_nested(self):
if "image_processor" not in self.processor_class.attributes:
self.skipTest(f"image_processor attribute not present in {self.processor_class}")
image_processor = self.get_component("image_processor")
tokenizer = self.get_component("tokenizer")
processor = self.processor_class(tokenizer=tokenizer, image_processor=image_processor)
self.skip_processor_without_typed_kwargs(processor)
input_str = "lower newer"
image_input = self.prepare_image_inputs()
# Define the kwargs for each modality
all_kwargs = {
"common_kwargs": {"return_tensors": "pt"},
"images_kwargs": {"crop_size": {"height": 214, "width": 214}},
"text_kwargs": {"padding": "max_length", "max_length": 76},
}
inputs = processor(text=input_str, images=image_input, **all_kwargs)
self.skip_processor_without_typed_kwargs(processor)
self.assertEqual(inputs["pixel_values"].shape[2], 214)
self.assertEqual(len(inputs["input_ids"][0]), 76)
def test_structured_kwargs_nested_from_dict(self):
if "image_processor" not in self.processor_class.attributes:
self.skipTest(f"image_processor attribute not present in {self.processor_class}")
image_processor = self.get_component("image_processor")
tokenizer = self.get_component("tokenizer")
processor = self.processor_class(tokenizer=tokenizer, image_processor=image_processor)
self.skip_processor_without_typed_kwargs(processor)
input_str = "lower newer"
image_input = self.prepare_image_inputs()
# Define the kwargs for each modality
all_kwargs = {
"common_kwargs": {"return_tensors": "pt"},
"images_kwargs": {"crop_size": {"height": 214, "width": 214}},
"text_kwargs": {"padding": "max_length", "max_length": 76},
}
inputs = processor(text=input_str, images=image_input, **all_kwargs)
self.assertEqual(inputs["pixel_values"].shape[2], 214)
self.assertEqual(len(inputs["input_ids"][0]), 76)
def test_unstructured_kwargs(self):
if "image_processor" not in self.processor_class.attributes:
self.skipTest(f"image_processor attribute not present in {self.processor_class}")
image_processor = self.get_component("image_processor")
tokenizer = self.get_component("tokenizer")
processor = self.processor_class(tokenizer=tokenizer, image_processor=image_processor)
self.skip_processor_without_typed_kwargs(processor)
input_str = "lower newer"
image_input = self.prepare_image_inputs()
inputs = processor(
text=input_str,
images=image_input,
return_tensors="pt",
crop_size={"height": 214, "width": 214},
padding="max_length",
max_length=76,
)
self.assertEqual(inputs["pixel_values"].shape[2], 214)
self.assertEqual(len(inputs["input_ids"][0]), 76)
def test_image_processor_defaults_preserved_by_image_kwargs(self):
if "image_processor" not in self.processor_class.attributes:
self.skipTest(f"image_processor attribute not present in {self.processor_class}")
image_processor = self.get_component("image_processor", crop_size=(234, 234))
tokenizer = self.get_component("tokenizer", max_length=117, padding="max_length")
processor = self.processor_class(tokenizer=tokenizer, image_processor=image_processor)
self.skip_processor_without_typed_kwargs(processor)
input_str = "lower newer"
image_input = self.prepare_image_inputs()
inputs = processor(text=input_str, images=image_input)
self.assertEqual(len(inputs["pixel_values"][0][0]), 234)

126
tests/models/chinese_clip/test_processor_chinese_clip.py
View File

@@ -206,129 +206,3 @@ class ChineseCLIPProcessorTest(ProcessorTesterMixin, unittest.TestCase):
inputs = processor(text=input_str, images=image_input)
self.assertListEqual(list(inputs.keys()), processor.model_input_names)
def test_unstructured_kwargs_batched(self):
if "image_processor" not in self.processor_class.attributes:
self.skipTest(f"image_processor attribute not present in {self.processor_class}")
image_processor = self.get_component("image_processor")
tokenizer = self.get_component("tokenizer")
processor = self.processor_class(tokenizer=tokenizer, image_processor=image_processor)
self.skip_processor_without_typed_kwargs(processor)
input_str = ["lower newer", "upper older longer string"]
image_input = self.prepare_image_inputs() * 2
inputs = processor(
text=input_str,
images=image_input,
return_tensors="pt",
crop_size={"height": 214, "width": 214},
padding="longest",
max_length=76,
)
self.assertEqual(inputs["pixel_values"].shape[2], 214)
self.assertEqual(len(inputs["input_ids"][0]), 6)
def test_structured_kwargs_nested(self):
if "image_processor" not in self.processor_class.attributes:
self.skipTest(f"image_processor attribute not present in {self.processor_class}")
image_processor = self.get_component("image_processor")
tokenizer = self.get_component("tokenizer")
processor = self.processor_class(tokenizer=tokenizer, image_processor=image_processor)
self.skip_processor_without_typed_kwargs(processor)
input_str = "lower newer"
image_input = self.prepare_image_inputs()
# Define the kwargs for each modality
all_kwargs = {
"common_kwargs": {"return_tensors": "pt"},
"images_kwargs": {"crop_size": {"height": 214, "width": 214}},
"text_kwargs": {"padding": "max_length", "max_length": 76},
}
inputs = processor(text=input_str, images=image_input, **all_kwargs)
self.skip_processor_without_typed_kwargs(processor)
self.assertEqual(inputs["pixel_values"].shape[2], 214)
self.assertEqual(len(inputs["input_ids"][0]), 76)
def test_structured_kwargs_nested_from_dict(self):
if "image_processor" not in self.processor_class.attributes:
self.skipTest(f"image_processor attribute not present in {self.processor_class}")
image_processor = self.get_component("image_processor")
tokenizer = self.get_component("tokenizer")
processor = self.processor_class(tokenizer=tokenizer, image_processor=image_processor)
self.skip_processor_without_typed_kwargs(processor)
input_str = "lower newer"
image_input = self.prepare_image_inputs()
# Define the kwargs for each modality
all_kwargs = {
"common_kwargs": {"return_tensors": "pt"},
"images_kwargs": {"crop_size": {"height": 214, "width": 214}},
"text_kwargs": {"padding": "max_length", "max_length": 76},
}
inputs = processor(text=input_str, images=image_input, **all_kwargs)
self.assertEqual(inputs["pixel_values"].shape[2], 214)
self.assertEqual(len(inputs["input_ids"][0]), 76)
def test_unstructured_kwargs(self):
if "image_processor" not in self.processor_class.attributes:
self.skipTest(f"image_processor attribute not present in {self.processor_class}")
image_processor = self.get_component("image_processor")
tokenizer = self.get_component("tokenizer")
processor = self.processor_class(tokenizer=tokenizer, image_processor=image_processor)
self.skip_processor_without_typed_kwargs(processor)
input_str = "lower newer"
image_input = self.prepare_image_inputs()
inputs = processor(
text=input_str,
images=image_input,
return_tensors="pt",
crop_size={"height": 214, "width": 214},
padding="max_length",
max_length=76,
)
self.assertEqual(inputs["pixel_values"].shape[2], 214)
self.assertEqual(len(inputs["input_ids"][0]), 76)
def test_image_processor_defaults_preserved_by_image_kwargs(self):
if "image_processor" not in self.processor_class.attributes:
self.skipTest(f"image_processor attribute not present in {self.processor_class}")
image_processor = self.get_component("image_processor", crop_size=(234, 234))
tokenizer = self.get_component("tokenizer", max_length=117, padding="max_length")
processor = self.processor_class(tokenizer=tokenizer, image_processor=image_processor)
self.skip_processor_without_typed_kwargs(processor)
input_str = "lower newer"
image_input = self.prepare_image_inputs()
inputs = processor(text=input_str, images=image_input)
self.assertEqual(len(inputs["pixel_values"][0][0]), 234)
def test_kwargs_overrides_default_image_processor_kwargs(self):
if "image_processor" not in self.processor_class.attributes:
self.skipTest(f"image_processor attribute not present in {self.processor_class}")
image_processor = self.get_component("image_processor", crop_size=(234, 234))
tokenizer = self.get_component("tokenizer", max_length=117, padding="max_length")
processor = self.processor_class(tokenizer=tokenizer, image_processor=image_processor)
self.skip_processor_without_typed_kwargs(processor)
input_str = "lower newer"
image_input = self.prepare_image_inputs()
inputs = processor(text=input_str, images=image_input, crop_size=[224, 224])
self.assertEqual(len(inputs["pixel_values"][0][0]), 224)

28
tests/models/llava/test_processor_llava.py
View File

@@ -17,7 +17,7 @@ import tempfile
import unittest
from transformers import AutoProcessor, AutoTokenizer, LlamaTokenizerFast, LlavaProcessor
from transformers.testing_utils import require_torch, require_vision
from transformers.testing_utils import require_vision
from transformers.utils import is_vision_available
from ...test_processing_common import ProcessorTesterMixin
@@ -93,29 +93,3 @@ class LlavaProcessorTest(ProcessorTesterMixin, unittest.TestCase):
formatted_prompt = processor.apply_chat_template(messages, add_generation_prompt=True)
self.assertEqual(expected_prompt, formatted_prompt)
@require_torch
@require_vision
def test_unstructured_kwargs_batched(self):
if "image_processor" not in self.processor_class.attributes:
self.skipTest(f"image_processor attribute not present in {self.processor_class}")
image_processor = self.get_component("image_processor")
tokenizer = self.get_component("tokenizer")
processor = self.processor_class(tokenizer=tokenizer, image_processor=image_processor)
self.skip_processor_without_typed_kwargs(processor)
input_str = ["lower newer", "upper older longer string"]
image_input = self.prepare_image_inputs() * 2
inputs = processor(
images=image_input,
text=input_str,
return_tensors="pt",
size={"height": 214, "width": 214},
padding="longest",
max_length=76,
)
self.assertEqual(inputs["pixel_values"].shape[2], 214)
self.assertEqual(len(inputs["input_ids"][0]), 5)

203
tests/models/llava_onevision/test_processing_llava_onevision.py
View File

@@ -16,7 +16,7 @@ import shutil
import tempfile
import unittest
from transformers.testing_utils import require_torch, require_vision
from transformers.testing_utils import require_vision
from transformers.utils import is_vision_available
from ...test_processing_common import ProcessorTesterMixin
@@ -100,204 +100,3 @@ class LlavaOnevisionProcessorTest(ProcessorTesterMixin, unittest.TestCase):
formatted_prompt = processor.apply_chat_template(messages, add_generation_prompt=True)
self.assertEqual(expected_prompt, formatted_prompt)
@require_torch
@require_vision
def test_image_processor_defaults_preserved_by_image_kwargs(self):
# Rewrite as llava-next image processor return pixel values with an added dimesion for image patches
if "image_processor" not in self.processor_class.attributes:
self.skipTest(f"image_processor attribute not present in {self.processor_class}")
image_processor = self.get_component("image_processor", size=(234, 234))
video_processor = self.get_component("video_processor", size=(234, 234))
tokenizer = self.get_component("tokenizer", max_length=117)
processor = self.processor_class(
tokenizer=tokenizer, image_processor=image_processor, video_processor=video_processor
)
self.skip_processor_without_typed_kwargs(processor)
input_str = "lower newer"
image_input = self.prepare_image_inputs()
inputs = processor(text=input_str, images=image_input)
# added dimension for image patches
self.assertEqual(len(inputs["pixel_values"][0][0][0]), 234)
@require_torch
@require_vision
def test_kwargs_overrides_default_image_processor_kwargs(self):
if "image_processor" not in self.processor_class.attributes:
self.skipTest(f"image_processor attribute not present in {self.processor_class}")
image_processor = self.get_component("image_processor", crop_size=(234, 234))
video_processor = self.get_component("video_processor", size=(234, 234))
tokenizer = self.get_component("tokenizer", max_length=117)
processor = self.processor_class(
tokenizer=tokenizer, image_processor=image_processor, video_processor=video_processor
)
self.skip_processor_without_typed_kwargs(processor)
input_str = "lower newer"
image_input = self.prepare_image_inputs()
inputs = processor(text=input_str, images=image_input, size=[224, 224])
# added dimension for image patches
self.assertEqual(len(inputs["pixel_values"][0][0][0]), 224)
@require_torch
@require_vision
def test_unstructured_kwargs(self):
image_processor = self.get_component("image_processor")
video_processor = self.get_component("video_processor")
tokenizer = self.get_component("tokenizer")
processor = self.processor_class(
tokenizer=tokenizer, image_processor=image_processor, video_processor=video_processor
)
self.skip_processor_without_typed_kwargs(processor)
input_str = "lower newer"
image_input = self.prepare_image_inputs()
inputs = processor(
text=input_str,
images=image_input,
return_tensors="pt",
size={"height": 214, "width": 214},
padding="max_length",
max_length=76,
)
# added dimension for image patches
self.assertEqual(inputs["pixel_values"].shape[3], 214)
self.assertEqual(len(inputs["input_ids"][0]), 76)
@require_torch
@require_vision
def test_unstructured_kwargs_batched(self):
image_processor = self.get_component("image_processor")
video_processor = self.get_component("video_processor")
tokenizer = self.get_component("tokenizer")
processor = self.processor_class(
tokenizer=tokenizer, image_processor=image_processor, video_processor=video_processor
)
self.skip_processor_without_typed_kwargs(processor)
input_str = ["lower newer", "upper older longer string"]
image_input = self.prepare_image_inputs() * 2
inputs = processor(
text=input_str,
images=image_input,
return_tensors="pt",
size={"height": 214, "width": 214},
padding="longest",
max_length=76,
)
self.assertEqual(inputs["pixel_values"].shape[3], 214)
self.assertEqual(len(inputs["input_ids"][0]), 4)
@require_torch
@require_vision
def test_structured_kwargs_nested(self):
image_processor = self.get_component("image_processor")
video_processor = self.get_component("video_processor")
tokenizer = self.get_component("tokenizer")
processor = self.processor_class(
tokenizer=tokenizer, image_processor=image_processor, video_processor=video_processor
)
self.skip_processor_without_typed_kwargs(processor)
input_str = "lower newer"
image_input = self.prepare_image_inputs()
# Define the kwargs for each modality
all_kwargs = {
"common_kwargs": {"return_tensors": "pt"},
"images_kwargs": {"size": {"height": 214, "width": 214}},
"text_kwargs": {"padding": "max_length", "max_length": 76},
}
inputs = processor(text=input_str, images=image_input, **all_kwargs)
self.skip_processor_without_typed_kwargs(processor)
self.assertEqual(inputs["pixel_values"].shape[3], 214)
self.assertEqual(len(inputs["input_ids"][0]), 76)
@require_torch
@require_vision
def test_structured_kwargs_nested_from_dict(self):
image_processor = self.get_component("image_processor")
video_processor = self.get_component("video_processor")
tokenizer = self.get_component("tokenizer")
processor = self.processor_class(
tokenizer=tokenizer, image_processor=image_processor, video_processor=video_processor
)
self.skip_processor_without_typed_kwargs(processor)
input_str = "lower newer"
image_input = self.prepare_image_inputs()
# Define the kwargs for each modality
all_kwargs = {
"common_kwargs": {"return_tensors": "pt"},
"images_kwargs": {"size": {"height": 214, "width": 214}},
"text_kwargs": {"padding": "max_length", "max_length": 76},
}
inputs = processor(text=input_str, images=image_input, **all_kwargs)
self.assertEqual(inputs["pixel_values"].shape[3], 214)
self.assertEqual(len(inputs["input_ids"][0]), 76)
@require_torch
@require_vision
def test_doubly_passed_kwargs(self):
image_processor = self.get_component("image_processor")
video_processor = self.get_component("video_processor")
tokenizer = self.get_component("tokenizer")
processor = self.processor_class(
tokenizer=tokenizer, image_processor=image_processor, video_processor=video_processor
)
self.skip_processor_without_typed_kwargs(processor)
input_str = ["lower newer"]
image_input = self.prepare_image_inputs()
with self.assertRaises(ValueError):
_ = processor(
text=input_str,
images=image_input,
images_kwargs={"size": {"height": 222, "width": 222}},
size={"height": 214, "width": 214},
)
@require_vision
@require_torch
def test_kwargs_overrides_default_tokenizer_kwargs(self):
image_processor = self.get_component("image_processor")
video_processor = self.get_component("video_processor")
tokenizer = self.get_component("tokenizer", max_length=117)
processor = self.processor_class(
tokenizer=tokenizer, image_processor=image_processor, video_processor=video_processor
)
self.skip_processor_without_typed_kwargs(processor)
input_str = "lower newer"
image_input = self.prepare_image_inputs()
inputs = processor(text=input_str, images=image_input, return_tensors="pt", max_length=112)
self.assertEqual(len(inputs["input_ids"][0]), 2)
@require_vision
@require_torch
def test_tokenizer_defaults_preserved_by_kwargs(self):
image_processor = self.get_component("image_processor")
video_processor = self.get_component("video_processor")
tokenizer = self.get_component("tokenizer", max_length=117)
processor = self.processor_class(
tokenizer=tokenizer, image_processor=image_processor, video_processor=video_processor
)
self.skip_processor_without_typed_kwargs(processor)
input_str = "lower newer"
image_input = self.prepare_image_inputs()
inputs = processor(text=input_str, images=image_input, return_tensors="pt")
self.assertEqual(len(inputs["input_ids"][0]), 2)

26
tests/models/paligemma/test_processor_paligemma.py
View File

@@ -61,29 +61,3 @@ class PaliGemmaProcessorTest(ProcessorTesterMixin, unittest.TestCase):
text=input_str, images=image_input, return_tensors="pt", max_length=112, padding="max_length"
)
self.assertEqual(len(inputs["input_ids"][0]), 112 + 14)
@require_torch
@require_vision
def test_unstructured_kwargs_batched(self):
if "image_processor" not in self.processor_class.attributes:
self.skipTest(f"image_processor attribute not present in {self.processor_class}")
image_processor = self.get_component("image_processor")
tokenizer = self.get_component("tokenizer")
processor = self.processor_class(tokenizer=tokenizer, image_processor=image_processor)
self.skip_processor_without_typed_kwargs(processor)
input_str = ["lower newer", "upper older longer string"]
image_input = self.prepare_image_inputs() * 2
inputs = processor(
text=input_str,
images=image_input,
return_tensors="pt",
size={"height": 214, "width": 214},
padding="longest",
max_length=76,
)
self.assertEqual(inputs["pixel_values"].shape[2], 214)
self.assertEqual(len(inputs["input_ids"][0]), 10)

137
tests/models/pixtral/test_processor_pixtral.py
View File

@@ -19,7 +19,6 @@ import requests
import torch
from transformers.testing_utils import (
require_torch,
require_vision,
)
from transformers.utils import is_vision_available
@@ -248,144 +247,28 @@ class PixtralProcessorTest(ProcessorTesterMixin, unittest.TestCase):
)
# fmt: on
# Override all tests requiring shape as returning tensor batches is not supported by PixtralProcessor
@require_torch
@require_vision
def test_image_processor_defaults_preserved_by_image_kwargs(self):
if "image_processor" not in self.processor_class.attributes:
self.skipTest(f"image_processor attribute not present in {self.processor_class}")
image_processor = self.get_component("image_processor", size={"height": 240, "width": 240})
tokenizer = self.get_component("tokenizer", max_length=117, padding="max_length")
processor = self.processor_class(tokenizer=tokenizer, image_processor=image_processor)
self.skip_processor_without_typed_kwargs(processor)
input_str = "lower newer"
image_input = self.prepare_image_inputs()
inputs = processor(text=input_str, images=image_input)
# Added dimension by pixtral image processor
self.assertEqual(len(inputs["pixel_values"][0][0][0][0]), 240)
@require_torch
@require_vision
def test_kwargs_overrides_default_image_processor_kwargs(self):
if "image_processor" not in self.processor_class.attributes:
self.skipTest(f"image_processor attribute not present in {self.processor_class}")
image_processor = self.get_component("image_processor", size={"height": 400, "width": 400})
tokenizer = self.get_component("tokenizer", max_length=117, padding="max_length")
processor = self.processor_class(tokenizer=tokenizer, image_processor=image_processor)
self.skip_processor_without_typed_kwargs(processor)
input_str = "lower newer"
image_input = self.prepare_image_inputs()
inputs = processor(text=input_str, images=image_input, size={"height": 240, "width": 240})
self.assertEqual(len(inputs["pixel_values"][0][0][0][0]), 240)
@require_torch
@require_vision
def test_structured_kwargs_nested(self):
if "image_processor" not in self.processor_class.attributes:
self.skipTest(f"image_processor attribute not present in {self.processor_class}")
image_processor = self.get_component("image_processor")
tokenizer = self.get_component("tokenizer")
processor = self.processor_class(tokenizer=tokenizer, image_processor=image_processor)
self.skip_processor_without_typed_kwargs(processor)
input_str = "lower newer"
image_input = self.prepare_image_inputs()
# Define the kwargs for each modality
all_kwargs = {
"common_kwargs": {"return_tensors": "pt"},
"images_kwargs": {"size": {"height": 240, "width": 240}},
"text_kwargs": {"padding": "max_length", "max_length": 76},
}
inputs = processor(text=input_str, images=image_input, **all_kwargs)
self.skip_processor_without_typed_kwargs(processor)
self.assertEqual(inputs["pixel_values"][0][0].shape[-1], 240)
self.assertEqual(len(inputs["input_ids"][0]), 76)
@require_torch
@require_vision
def test_structured_kwargs_nested_from_dict(self):
if "image_processor" not in self.processor_class.attributes:
self.skipTest(f"image_processor attribute not present in {self.processor_class}")
image_processor = self.get_component("image_processor")
tokenizer = self.get_component("tokenizer")
processor = self.processor_class(tokenizer=tokenizer, image_processor=image_processor)
self.skip_processor_without_typed_kwargs(processor)
input_str = "lower newer"
image_input = self.prepare_image_inputs()
# Define the kwargs for each modality
all_kwargs = {
"common_kwargs": {"return_tensors": "pt"},
"images_kwargs": {"size": {"height": 240, "width": 240}},
"text_kwargs": {"padding": "max_length", "max_length": 76},
}
inputs = processor(text=input_str, images=image_input, **all_kwargs)
self.assertEqual(inputs["pixel_values"][0][0].shape[-1], 240)
self.assertEqual(len(inputs["input_ids"][0]), 76)
@require_torch
@require_vision
def test_unstructured_kwargs(self):
if "image_processor" not in self.processor_class.attributes:
self.skipTest(f"image_processor attribute not present in {self.processor_class}")
image_processor = self.get_component("image_processor")
tokenizer = self.get_component("tokenizer")
processor = self.processor_class(tokenizer=tokenizer, image_processor=image_processor)
self.skip_processor_without_typed_kwargs(processor)
input_str = "lower newer"
image_input = self.prepare_image_inputs()
inputs = processor(
text=input_str,
images=image_input,
return_tensors="pt",
size={"height": 240, "width": 240},
padding="max_length",
max_length=76,
)
self.assertEqual(inputs["pixel_values"][0][0].shape[-1], 240)
self.assertEqual(len(inputs["input_ids"][0]), 76)
@require_torch
@require_vision
# Override as PixtralProcessor needs nested images to work properly with batched inputs
def test_unstructured_kwargs_batched(self):
if "image_processor" not in self.processor_class.attributes:
self.skipTest(f"image_processor attribute not present in {self.processor_class}")
image_processor = self.get_component("image_processor")
tokenizer = self.get_component("tokenizer")
processor = self.processor_class(tokenizer=tokenizer, image_processor=image_processor)
processor_components = self.prepare_components()
processor = self.processor_class(**processor_components)
self.skip_processor_without_typed_kwargs(processor)
input_str = ["lower newer", "upper older longer string"]
# images needs to be nested to detect multiple prompts
image_input = [self.prepare_image_inputs()] * 2
inputs = processor(
text=input_str,
images=image_input,
return_tensors="pt",
size={"height": 240, "width": 240},
do_rescale=True,
rescale_factor=-1,
padding="longest",
max_length=76,
)
self.assertEqual(inputs["pixel_values"][0][0].shape[-1], 240)
self.assertEqual(len(inputs["input_ids"][0]), 4)
self.assertLessEqual(inputs[self.images_input_name][0][0].mean(), 0)
self.assertTrue(
len(inputs[self.text_input_name][0]) == len(inputs[self.text_input_name][1])
and len(inputs[self.text_input_name][1]) < 76
)

127
tests/models/qwen2_vl/test_processing_qwen2_vl.py
View File

@@ -108,130 +108,3 @@ class Qwen2VLProcessorTest(ProcessorTesterMixin, unittest.TestCase):
inputs = processor(text=input_str, images=image_input, videos=video_inputs)
self.assertListEqual(list(inputs.keys()), processor.model_input_names)
# Qwen2-VL doesn't accept `size` and resized to an optimal size using image_processor attrbutes
# defined at `init`. Therefore, all tests are overwritten and don't actually test if kwargs are passed
# to image processors
def test_image_processor_defaults_preserved_by_image_kwargs(self):
image_processor = self.get_component("image_processor")
tokenizer = self.get_component("tokenizer", max_length=117, padding="max_length")
processor = self.processor_class(tokenizer=tokenizer, image_processor=image_processor)
self.skip_processor_without_typed_kwargs(processor)
input_str = "lower newer"
image_input = self.prepare_image_inputs()
inputs = processor(text=input_str, images=image_input)
self.assertEqual(inputs["pixel_values"].shape[0], 800)
def test_kwargs_overrides_default_image_processor_kwargs(self):
image_processor = self.get_component(
"image_processor",
)
tokenizer = self.get_component("tokenizer", max_length=117, padding="max_length")
processor = self.processor_class(tokenizer=tokenizer, image_processor=image_processor)
self.skip_processor_without_typed_kwargs(processor)
input_str = "lower newer"
image_input = self.prepare_image_inputs()
inputs = processor(text=input_str, images=image_input)
self.assertEqual(inputs["pixel_values"].shape[0], 800)
def test_unstructured_kwargs(self):
image_processor = self.get_component("image_processor")
tokenizer = self.get_component("tokenizer")
processor = self.processor_class(tokenizer=tokenizer, image_processor=image_processor)
self.skip_processor_without_typed_kwargs(processor)
input_str = "lower newer"
image_input = self.prepare_image_inputs()
inputs = processor(
text=input_str,
images=image_input,
return_tensors="pt",
padding="max_length",
max_length=76,
)
self.assertEqual(inputs["pixel_values"].shape[0], 800)
self.assertEqual(len(inputs["input_ids"][0]), 76)
def test_unstructured_kwargs_batched(self):
if "image_processor" not in self.processor_class.attributes:
self.skipTest(f"image_processor attribute not present in {self.processor_class}")
image_processor = self.get_component("image_processor")
tokenizer = self.get_component("tokenizer")
processor = self.processor_class(tokenizer=tokenizer, image_processor=image_processor)
self.skip_processor_without_typed_kwargs(processor)
input_str = ["lower newer", "upper older longer string"]
image_input = self.prepare_image_inputs() * 2
inputs = processor(
text=input_str,
images=image_input,
return_tensors="pt",
padding="longest",
max_length=76,
)
self.assertEqual(inputs["pixel_values"].shape[0], 1600)
self.assertEqual(len(inputs["input_ids"][0]), 4)
def test_structured_kwargs_nested(self):
image_processor = self.get_component("image_processor")
tokenizer = self.get_component("tokenizer")
processor = self.processor_class(tokenizer=tokenizer, image_processor=image_processor)
self.skip_processor_without_typed_kwargs(processor)
input_str = "lower newer"
image_input = self.prepare_image_inputs()
# Define the kwargs for each modality
all_kwargs = {
"common_kwargs": {"return_tensors": "pt"},
"text_kwargs": {"padding": "max_length", "max_length": 76},
}
inputs = processor(text=input_str, images=image_input, **all_kwargs)
self.skip_processor_without_typed_kwargs(processor)
self.assertEqual(inputs["pixel_values"].shape[0], 800)
self.assertEqual(len(inputs["input_ids"][0]), 76)
def test_structured_kwargs_nested_from_dict(self):
image_processor = self.get_component("image_processor")
tokenizer = self.get_component("tokenizer")
processor = self.processor_class(tokenizer=tokenizer, image_processor=image_processor)
self.skip_processor_without_typed_kwargs(processor)
input_str = "lower newer"
image_input = self.prepare_image_inputs()
# Define the kwargs for each modality
all_kwargs = {
"common_kwargs": {"return_tensors": "pt"},
"text_kwargs": {"padding": "max_length", "max_length": 76},
}
inputs = processor(text=input_str, images=image_input, **all_kwargs)
self.assertEqual(inputs["pixel_values"].shape[0], 800)
self.assertEqual(len(inputs["input_ids"][0]), 76)
def test_image_processor_defaults_preserved_by_video_kwargs(self):
image_processor = self.get_component("image_processor")
tokenizer = self.get_component("tokenizer", max_length=117, padding="max_length")
processor = self.processor_class(tokenizer=tokenizer, image_processor=image_processor)
self.skip_processor_without_typed_kwargs(processor)
input_str = "lower newer"
video_input = self.prepare_video_inputs()
inputs = processor(text=input_str, videos=video_input)
self.assertEqual(inputs["pixel_values_videos"].shape[0], 9600)