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

Add Idefics2/3 and SmolVLM Fast image processors + improvements for fast image processors (#38157)

Browse Source 
* add working idefics2 fast and improvements for fast nested images processing

* add fast image processors idefics 3 and smolvlm

* cleanup tests

* fic doc idefics2

* PR review and fix issues after merge

* Force providing disable_grouping to group_images_by_shape

* simplify group_images_by_shape

* fix modular

* Fix nits after review
This commit is contained in:
Yoni Gozlan
2025-06-23 10:17:25 -04:00
committed by GitHub
parent 1a96127e46
commit d29482cc91
61 changed files with 2023 additions and 425 deletions
Download Patch File Download Diff File Expand all files Collapse all files

16
tests/models/siglip2/test_image_processing_siglip2.py
View File

@@ -18,14 +18,11 @@ import unittest
import requests
from transformers.testing_utils import require_torch, require_vision
from transformers.utils import is_torch_available, is_torchvision_available, is_vision_available
from transformers.utils import is_torchvision_available, is_vision_available
from ...test_image_processing_common import ImageProcessingTestMixin, prepare_image_inputs
if is_torch_available():
import torch
if is_vision_available():
from PIL import Image
@@ -150,7 +147,6 @@ class Siglip2ImageProcessingTest(ImageProcessingTestMixin, unittest.TestCase):
def test_call_numpy_4_channels(self):
pass
# increase mean tolerance to 1e-3 -> 2e-3
# Ignore copy
def test_slow_fast_equivalence(self):
if not self.test_slow_image_processor or not self.test_fast_image_processor:
@@ -167,10 +163,7 @@ class Siglip2ImageProcessingTest(ImageProcessingTestMixin, unittest.TestCase):
encoding_slow = image_processor_slow(dummy_image, return_tensors="pt")
encoding_fast = image_processor_fast(dummy_image, return_tensors="pt")
torch.testing.assert_close(encoding_slow.pixel_values, encoding_fast.pixel_values, atol=1e-1, rtol=1e-1)
self.assertLessEqual(
torch.mean(torch.abs(encoding_slow.pixel_values - encoding_fast.pixel_values)).item(), 2e-3
)
self._assert_slow_fast_tensors_equivalence(encoding_slow.pixel_values, encoding_fast.pixel_values)
# increase mean tolerance to 1e-3 -> 2e-3
# Ignore copy
@@ -193,7 +186,4 @@ class Siglip2ImageProcessingTest(ImageProcessingTestMixin, unittest.TestCase):
encoding_slow = image_processor_slow(dummy_images, return_tensors="pt")
encoding_fast = image_processor_fast(dummy_images, return_tensors="pt")
torch.testing.assert_close(encoding_slow.pixel_values, encoding_fast.pixel_values, atol=1e-1, rtol=1e-1)
self.assertLessEqual(
torch.mean(torch.abs(encoding_slow.pixel_values - encoding_fast.pixel_values)).item(), 2e-3
)
self._assert_slow_fast_tensors_equivalence(encoding_slow.pixel_values, encoding_fast.pixel_values)