[processor] clean up mulitmodal tests (#37362)
* clkea up mulitmodal processor tests * fixup * fix tests * fix one last test * forgot
This commit is contained in:
Raushan Turganbay
committed by
GitHub
GitHub
parent
3c39c07939
commit
a563999a02
@@ -16,7 +16,6 @@ import shutil
|
||||
import tempfile
|
||||
import unittest
|
||||
from io import BytesIO
|
||||
from typing import Optional
|
||||
|
||||
import numpy as np
|
||||
import requests
|
||||
@@ -81,6 +80,10 @@ class Idefics3ProcessorTest(ProcessorTesterMixin, unittest.TestCase):
|
||||
def get_processor(self, **kwargs):
|
||||
return AutoProcessor.from_pretrained(self.tmpdirname, **kwargs)
|
||||
|
||||
@staticmethod
|
||||
def prepare_processor_dict():
|
||||
return {"image_seq_len": 2}
|
||||
|
||||
def get_split_image_expected_tokens(self, processor, image_rows, image_cols):
|
||||
text_split_images = []
|
||||
for n_h in range(image_rows):
|
||||
@@ -352,159 +355,6 @@ class Idefics3ProcessorTest(ProcessorTesterMixin, unittest.TestCase):
|
||||
)
|
||||
self.assertEqual(rendered, expected_rendered)
|
||||
|
||||
# Override as Idefics3Processor needs image tokens in prompts
|
||||
def prepare_text_inputs(self, batch_size: Optional[int] = None):
|
||||
if batch_size is None:
|
||||
return "lower newer <image>"
|
||||
|
||||
if batch_size < 1:
|
||||
raise ValueError("batch_size must be greater than 0")
|
||||
|
||||
if batch_size == 1:
|
||||
return ["lower newer <image>"]
|
||||
return ["lower newer <image>", "<image> upper older longer string"] + ["<image> lower newer"] * (
|
||||
batch_size - 2
|
||||
)
|
||||
|
||||
# Override tests as inputs_ids padded dimension is the second one but not the last one
|
||||
@require_vision
|
||||
@require_torch
|
||||
def test_kwargs_overrides_default_tokenizer_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", max_length=30)
|
||||
|
||||
processor = self.processor_class(tokenizer=tokenizer, image_processor=image_processor)
|
||||
self.skip_processor_without_typed_kwargs(processor)
|
||||
input_str = self.prepare_text_inputs()
|
||||
image_input = self.prepare_image_inputs()
|
||||
|
||||
inputs = processor(text=input_str, images=image_input, return_tensors="pt", max_length=30)
|
||||
self.assertEqual(len(inputs["input_ids"][0]), 30)
|
||||
|
||||
@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 = self.prepare_text_inputs()
|
||||
image_input = self.prepare_image_inputs()
|
||||
|
||||
# Define the kwargs for each modality
|
||||
inputs = processor(
|
||||
text=input_str,
|
||||
images=image_input,
|
||||
common_kwargs={"return_tensors": "pt"},
|
||||
images_kwargs={"max_image_size": {"longest_edge": 32}},
|
||||
text_kwargs={"padding": "max_length", "max_length": 120, "truncation": "longest_first"},
|
||||
)
|
||||
self.skip_processor_without_typed_kwargs(processor)
|
||||
|
||||
self.assertEqual(inputs["pixel_values"].shape[3], 32)
|
||||
|
||||
self.assertEqual(len(inputs["input_ids"][0]), 120)
|
||||
|
||||
@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 = self.prepare_text_inputs()
|
||||
image_input = self.prepare_image_inputs()
|
||||
|
||||
# Define the kwargs for each modality
|
||||
all_kwargs = {
|
||||
"common_kwargs": {"return_tensors": "pt"},
|
||||
"images_kwargs": {"max_image_size": {"longest_edge": 32}},
|
||||
"text_kwargs": {"padding": "max_length", "max_length": 120, "truncation": "longest_first"},
|
||||
}
|
||||
|
||||
inputs = processor(text=input_str, images=image_input, **all_kwargs)
|
||||
self.assertEqual(inputs["pixel_values"].shape[3], 32)
|
||||
self.assertEqual(len(inputs["input_ids"][0]), 120)
|
||||
|
||||
@require_vision
|
||||
@require_torch
|
||||
def test_tokenizer_defaults_preserved_by_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", max_length=30)
|
||||
|
||||
processor = self.processor_class(tokenizer=tokenizer, image_processor=image_processor)
|
||||
self.skip_processor_without_typed_kwargs(processor)
|
||||
input_str = self.prepare_text_inputs()
|
||||
image_input = self.prepare_image_inputs()
|
||||
|
||||
inputs = processor(text=input_str, images=image_input, return_tensors="pt")
|
||||
self.assertEqual(len(inputs["input_ids"][0]), 30)
|
||||
|
||||
@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 = self.prepare_text_inputs(batch_size=2)
|
||||
image_input = self.prepare_image_inputs(batch_size=2)
|
||||
inputs = processor(
|
||||
text=input_str,
|
||||
images=image_input,
|
||||
return_tensors="pt",
|
||||
padding="longest",
|
||||
max_length=76,
|
||||
truncation=True,
|
||||
max_image_size={"longest_edge": 30},
|
||||
)
|
||||
|
||||
self.assertEqual(inputs["pixel_values"].shape[2], 3)
|
||||
self.assertEqual(inputs["pixel_values"].shape[3], 30)
|
||||
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 = self.prepare_text_inputs()
|
||||
image_input = self.prepare_image_inputs()
|
||||
inputs = processor(
|
||||
text=input_str,
|
||||
images=image_input,
|
||||
return_tensors="pt",
|
||||
max_image_size={"longest_edge": 32},
|
||||
padding="max_length",
|
||||
max_length=120,
|
||||
truncation="longest_first",
|
||||
)
|
||||
|
||||
self.assertEqual(inputs["pixel_values"].shape[3], 32)
|
||||
self.assertEqual(len(inputs["input_ids"][0]), 120)
|
||||
|
||||
@require_torch
|
||||
@require_vision
|
||||
def test_text_only_inference(self):
|
||||
|
||||
Reference in New Issue
Block a user