From be753a371d50d48229b4f451c46a419023e64cb3 Mon Sep 17 00:00:00 2001 From: Arthur Zucker Date: Tue, 27 Aug 2024 16:50:04 +0200 Subject: [PATCH 1/4] use a single for loop --- .../models/bit/image_processing_bit.py | 48 ++++++++++--------- .../chameleon/image_processing_chameleon.py | 47 +++++++++--------- .../image_processing_chinese_clip.py | 41 ++++++++-------- .../models/clip/image_processing_clip.py | 47 +++++++++--------- .../models/deit/image_processing_deit.py | 47 +++++++++--------- .../vit_hybrid/image_processing_vit_hybrid.py | 47 +++++++++--------- .../llava_next/image_processing_llava_next.py | 47 +++++++++--------- .../image_processing_llava_next_video.py | 47 +++++++++--------- .../image_processing_mobilenet_v1.py | 41 ++++++++-------- .../image_processing_mobilenet_v2.py | 47 +++++++++--------- 10 files changed, 245 insertions(+), 214 deletions(-) diff --git a/src/transformers/models/bit/image_processing_bit.py b/src/transformers/models/bit/image_processing_bit.py index a836d136ec96..45000f41829c 100644 --- a/src/transformers/models/bit/image_processing_bit.py +++ b/src/transformers/models/bit/image_processing_bit.py @@ -294,28 +294,32 @@ def preprocess( # We assume that all images have the same channel dimension format. input_data_format = infer_channel_dimension_format(images[0]) - if do_resize: - images = [ - self.resize(image=image, size=size, resample=resample, input_data_format=input_data_format) - for image in images - ] - - if do_center_crop: - images = [ - self.center_crop(image=image, size=crop_size, input_data_format=input_data_format) for image in images - ] - - if do_rescale: - images = [ - self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) - for image in images - ] - - if do_normalize: - images = [ - self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) - for image in images - ] + all_images = [] + for image in images: + if do_resize: + image = self.resize(image=image, size=size, resample=resample, input_data_format=input_data_format) + + if do_center_crop: + image = self.center_crop(image=image, size=crop_size, input_data_format=input_data_format) + + if do_rescale: + image = self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) + + if do_normalize: + image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) + + if do_resize: + image = self.resize(image=image, size=size, resample=resample, input_data_format=input_data_format) + + if do_center_crop: + image = self.center_crop(image=image, size=crop_size, input_data_format=input_data_format) + + if do_rescale: + image = self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) + + if do_normalize: + image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) + all_images.append(image) images = [ to_channel_dimension_format(image, data_format, input_channel_dim=input_data_format) for image in images diff --git a/src/transformers/models/chameleon/image_processing_chameleon.py b/src/transformers/models/chameleon/image_processing_chameleon.py index a23fdbed0288..78ebb53a5410 100644 --- a/src/transformers/models/chameleon/image_processing_chameleon.py +++ b/src/transformers/models/chameleon/image_processing_chameleon.py @@ -312,29 +312,32 @@ def preprocess( # We assume that all images have the same channel dimension format. input_data_format = infer_channel_dimension_format(images[0]) - if do_resize: - images = [ - self.resize(image=image, size=size, resample=resample, input_data_format=input_data_format) - for image in images - ] - - if do_center_crop: - images = [ - self.center_crop(image=image, size=crop_size, input_data_format=input_data_format) for image in images - ] - - if do_rescale: - images = [ - self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) - for image in images - ] - - if do_normalize: - images = [ - self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) - for image in images - ] + all_images = [] + for image in images: + if do_resize: + image = self.resize(image=image, size=size, resample=resample, input_data_format=input_data_format) + if do_center_crop: + image = self.center_crop(image=image, size=crop_size, input_data_format=input_data_format) + + if do_rescale: + image = self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) + + if do_normalize: + image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) + + if do_resize: + image = self.resize(image=image, size=size, resample=resample, input_data_format=input_data_format) + + if do_center_crop: + image = self.center_crop(image=image, size=crop_size, input_data_format=input_data_format) + + if do_rescale: + image = self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) + + if do_normalize: + image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) + all_images.append(image) images = [ to_channel_dimension_format(image, data_format, input_channel_dim=input_data_format) for image in images ] diff --git a/src/transformers/models/chinese_clip/image_processing_chinese_clip.py b/src/transformers/models/chinese_clip/image_processing_chinese_clip.py index b93bb81606a9..b8edd440a66e 100644 --- a/src/transformers/models/chinese_clip/image_processing_chinese_clip.py +++ b/src/transformers/models/chinese_clip/image_processing_chinese_clip.py @@ -280,29 +280,32 @@ def preprocess( # We assume that all images have the same channel dimension format. input_data_format = infer_channel_dimension_format(images[0]) - if do_resize: - images = [ - self.resize(image=image, size=size, resample=resample, input_data_format=input_data_format) - for image in images - ] + all_images = [] + for image in images: + if do_resize: + image = self.resize(image=image, size=size, resample=resample, input_data_format=input_data_format) - if do_center_crop: - images = [ - self.center_crop(image=image, size=crop_size, input_data_format=input_data_format) for image in images - ] + if do_center_crop: + image = self.center_crop(image=image, size=crop_size, input_data_format=input_data_format) - if do_rescale: - images = [ - self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) - for image in images - ] + if do_rescale: + image = self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) - if do_normalize: - images = [ - self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) - for image in images - ] + if do_normalize: + image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) + if do_resize: + image = self.resize(image=image, size=size, resample=resample, input_data_format=input_data_format) + + if do_center_crop: + image = self.center_crop(image=image, size=crop_size, input_data_format=input_data_format) + + if do_rescale: + image = self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) + + if do_normalize: + image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) + all_images.append(image) images = [ to_channel_dimension_format(image, data_format, input_channel_dim=input_data_format) for image in images ] diff --git a/src/transformers/models/clip/image_processing_clip.py b/src/transformers/models/clip/image_processing_clip.py index bc545e08e20e..cff3875f43d6 100644 --- a/src/transformers/models/clip/image_processing_clip.py +++ b/src/transformers/models/clip/image_processing_clip.py @@ -319,29 +319,32 @@ def preprocess( # We assume that all images have the same channel dimension format. input_data_format = infer_channel_dimension_format(images[0]) - if do_resize: - images = [ - self.resize(image=image, size=size, resample=resample, input_data_format=input_data_format) - for image in images - ] - - if do_center_crop: - images = [ - self.center_crop(image=image, size=crop_size, input_data_format=input_data_format) for image in images - ] - - if do_rescale: - images = [ - self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) - for image in images - ] - - if do_normalize: - images = [ - self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) - for image in images - ] + all_images = [] + for image in images: + if do_resize: + image = self.resize(image=image, size=size, resample=resample, input_data_format=input_data_format) + if do_center_crop: + image = self.center_crop(image=image, size=crop_size, input_data_format=input_data_format) + + if do_rescale: + image = self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) + + if do_normalize: + image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) + + if do_resize: + image = self.resize(image=image, size=size, resample=resample, input_data_format=input_data_format) + + if do_center_crop: + image = self.center_crop(image=image, size=crop_size, input_data_format=input_data_format) + + if do_rescale: + image = self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) + + if do_normalize: + image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) + all_images.append(image) images = [ to_channel_dimension_format(image, data_format, input_channel_dim=input_data_format) for image in images ] diff --git a/src/transformers/models/deit/image_processing_deit.py b/src/transformers/models/deit/image_processing_deit.py index d5dfb211e03c..8f313a3e881f 100644 --- a/src/transformers/models/deit/image_processing_deit.py +++ b/src/transformers/models/deit/image_processing_deit.py @@ -270,29 +270,32 @@ def preprocess( # We assume that all images have the same channel dimension format. input_data_format = infer_channel_dimension_format(images[0]) - if do_resize: - images = [ - self.resize(image=image, size=size, resample=resample, input_data_format=input_data_format) - for image in images - ] - - if do_center_crop: - images = [ - self.center_crop(image=image, size=crop_size, input_data_format=input_data_format) for image in images - ] - - if do_rescale: - images = [ - self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) - for image in images - ] - - if do_normalize: - images = [ - self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) - for image in images - ] + all_images = [] + for image in images: + if do_resize: + image = self.resize(image=image, size=size, resample=resample, input_data_format=input_data_format) + if do_center_crop: + image = self.center_crop(image=image, size=crop_size, input_data_format=input_data_format) + + if do_rescale: + image = self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) + + if do_normalize: + image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) + + if do_resize: + image = self.resize(image=image, size=size, resample=resample, input_data_format=input_data_format) + + if do_center_crop: + image = self.center_crop(image=image, size=crop_size, input_data_format=input_data_format) + + if do_rescale: + image = self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) + + if do_normalize: + image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) + all_images.append(image) images = [ to_channel_dimension_format(image, data_format, input_channel_dim=input_data_format) for image in images ] diff --git a/src/transformers/models/deprecated/vit_hybrid/image_processing_vit_hybrid.py b/src/transformers/models/deprecated/vit_hybrid/image_processing_vit_hybrid.py index 89a8f9e676e8..40afc716d787 100644 --- a/src/transformers/models/deprecated/vit_hybrid/image_processing_vit_hybrid.py +++ b/src/transformers/models/deprecated/vit_hybrid/image_processing_vit_hybrid.py @@ -312,29 +312,32 @@ def preprocess( # We assume that all images have the same channel dimension format. input_data_format = infer_channel_dimension_format(images[0]) - if do_resize: - images = [ - self.resize(image=image, size=size, resample=resample, input_data_format=input_data_format) - for image in images - ] - - if do_center_crop: - images = [ - self.center_crop(image=image, size=crop_size, input_data_format=input_data_format) for image in images - ] - - if do_rescale: - images = [ - self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) - for image in images - ] - - if do_normalize: - images = [ - self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) - for image in images - ] + all_images = [] + for image in images: + if do_resize: + image = self.resize(image=image, size=size, resample=resample, input_data_format=input_data_format) + if do_center_crop: + image = self.center_crop(image=image, size=crop_size, input_data_format=input_data_format) + + if do_rescale: + image = self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) + + if do_normalize: + image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) + + if do_resize: + image = self.resize(image=image, size=size, resample=resample, input_data_format=input_data_format) + + if do_center_crop: + image = self.center_crop(image=image, size=crop_size, input_data_format=input_data_format) + + if do_rescale: + image = self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) + + if do_normalize: + image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) + all_images.append(image) images = [ to_channel_dimension_format(image, data_format, input_channel_dim=input_data_format) for image in images ] diff --git a/src/transformers/models/llava_next/image_processing_llava_next.py b/src/transformers/models/llava_next/image_processing_llava_next.py index f744b9fcf9c1..6a89a8edb3a9 100644 --- a/src/transformers/models/llava_next/image_processing_llava_next.py +++ b/src/transformers/models/llava_next/image_processing_llava_next.py @@ -409,29 +409,32 @@ def _preprocess( """ images = make_list_of_images(images) - if do_resize: - images = [ - self.resize(image=image, size=size, resample=resample, input_data_format=input_data_format) - for image in images - ] - - if do_center_crop: - images = [ - self.center_crop(image=image, size=crop_size, input_data_format=input_data_format) for image in images - ] - - if do_rescale: - images = [ - self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) - for image in images - ] - - if do_normalize: - images = [ - self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) - for image in images - ] + all_images = [] + for image in images: + if do_resize: + image = self.resize(image=image, size=size, resample=resample, input_data_format=input_data_format) + + if do_center_crop: + image = self.center_crop(image=image, size=crop_size, input_data_format=input_data_format) + + if do_rescale: + image = self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) + + if do_normalize: + image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) + + if do_resize: + image = self.resize(image=image, size=size, resample=resample, input_data_format=input_data_format) + + if do_center_crop: + image = self.center_crop(image=image, size=crop_size, input_data_format=input_data_format) + + if do_rescale: + image = self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) + if do_normalize: + image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) + all_images.append(image) images = [ to_channel_dimension_format(image, data_format, input_channel_dim=input_data_format) for image in images ] diff --git a/src/transformers/models/llava_next_video/image_processing_llava_next_video.py b/src/transformers/models/llava_next_video/image_processing_llava_next_video.py index e16e71875bb2..0150cb72d22c 100644 --- a/src/transformers/models/llava_next_video/image_processing_llava_next_video.py +++ b/src/transformers/models/llava_next_video/image_processing_llava_next_video.py @@ -272,29 +272,32 @@ def _preprocess( # We assume that all images have the same channel dimension format. input_data_format = infer_channel_dimension_format(images[0]) - if do_resize: - images = [ - self.resize(image=image, size=size, resample=resample, input_data_format=input_data_format) - for image in images - ] - - if do_center_crop: - images = [ - self.center_crop(image=image, size=crop_size, input_data_format=input_data_format) for image in images - ] - - if do_rescale: - images = [ - self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) - for image in images - ] - - if do_normalize: - images = [ - self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) - for image in images - ] + all_images = [] + for image in images: + if do_resize: + image = self.resize(image=image, size=size, resample=resample, input_data_format=input_data_format) + if do_center_crop: + image = self.center_crop(image=image, size=crop_size, input_data_format=input_data_format) + + if do_rescale: + image = self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) + + if do_normalize: + image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) + + if do_resize: + image = self.resize(image=image, size=size, resample=resample, input_data_format=input_data_format) + + if do_center_crop: + image = self.center_crop(image=image, size=crop_size, input_data_format=input_data_format) + + if do_rescale: + image = self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) + + if do_normalize: + image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) + all_images.append(image) images = [ to_channel_dimension_format(image, data_format, input_channel_dim=input_data_format) for image in images ] diff --git a/src/transformers/models/mobilenet_v1/image_processing_mobilenet_v1.py b/src/transformers/models/mobilenet_v1/image_processing_mobilenet_v1.py index 967d17929f82..54a60fc0c54a 100644 --- a/src/transformers/models/mobilenet_v1/image_processing_mobilenet_v1.py +++ b/src/transformers/models/mobilenet_v1/image_processing_mobilenet_v1.py @@ -276,29 +276,32 @@ def preprocess( # We assume that all images have the same channel dimension format. input_data_format = infer_channel_dimension_format(images[0]) - if do_resize: - images = [ - self.resize(image=image, size=size, resample=resample, input_data_format=input_data_format) - for image in images - ] + all_images = [] + for image in images: + if do_resize: + image = self.resize(image=image, size=size, resample=resample, input_data_format=input_data_format) - if do_center_crop: - images = [ - self.center_crop(image=image, size=crop_size, input_data_format=input_data_format) for image in images - ] + if do_center_crop: + image = self.center_crop(image=image, size=crop_size, input_data_format=input_data_format) - if do_rescale: - images = [ - self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) - for image in images - ] + if do_rescale: + image = self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) - if do_normalize: - images = [ - self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) - for image in images - ] + if do_normalize: + image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) + if do_resize: + image = self.resize(image=image, size=size, resample=resample, input_data_format=input_data_format) + + if do_center_crop: + image = self.center_crop(image=image, size=crop_size, input_data_format=input_data_format) + + if do_rescale: + image = self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) + + if do_normalize: + image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) + all_images.append(image) images = [ to_channel_dimension_format(image, data_format, input_channel_dim=input_data_format) for image in images ] diff --git a/src/transformers/models/mobilenet_v2/image_processing_mobilenet_v2.py b/src/transformers/models/mobilenet_v2/image_processing_mobilenet_v2.py index 072295a4ff77..cbb79364f4b4 100644 --- a/src/transformers/models/mobilenet_v2/image_processing_mobilenet_v2.py +++ b/src/transformers/models/mobilenet_v2/image_processing_mobilenet_v2.py @@ -279,29 +279,32 @@ def preprocess( # We assume that all images have the same channel dimension format. input_data_format = infer_channel_dimension_format(images[0]) - if do_resize: - images = [ - self.resize(image=image, size=size, resample=resample, input_data_format=input_data_format) - for image in images - ] - - if do_center_crop: - images = [ - self.center_crop(image=image, size=crop_size, input_data_format=input_data_format) for image in images - ] - - if do_rescale: - images = [ - self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) - for image in images - ] - - if do_normalize: - images = [ - self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) - for image in images - ] + all_images = [] + for image in images: + if do_resize: + image = self.resize(image=image, size=size, resample=resample, input_data_format=input_data_format) + if do_center_crop: + image = self.center_crop(image=image, size=crop_size, input_data_format=input_data_format) + + if do_rescale: + image = self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) + + if do_normalize: + image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) + + if do_resize: + image = self.resize(image=image, size=size, resample=resample, input_data_format=input_data_format) + + if do_center_crop: + image = self.center_crop(image=image, size=crop_size, input_data_format=input_data_format) + + if do_rescale: + image = self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) + + if do_normalize: + image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) + all_images.append(image) images = [ to_channel_dimension_format(image, data_format, input_channel_dim=input_data_format) for image in images ] From aa698e7349e639aff3f1b2aedb765c96cdbccb93 Mon Sep 17 00:00:00 2001 From: Arthur Zucker Date: Tue, 27 Aug 2024 16:56:08 +0200 Subject: [PATCH 2/4] oups --- src/transformers/models/bit/image_processing_bit.py | 11 ----------- .../models/chameleon/image_processing_chameleon.py | 12 ------------ .../chinese_clip/image_processing_chinese_clip.py | 11 ----------- .../models/clip/image_processing_clip.py | 11 ----------- .../models/deit/image_processing_deit.py | 11 ----------- .../vit_hybrid/image_processing_vit_hybrid.py | 11 ----------- .../models/llava_next/image_processing_llava_next.py | 11 ----------- .../image_processing_llava_next_video.py | 11 ----------- .../mobilenet_v1/image_processing_mobilenet_v1.py | 11 ----------- .../mobilenet_v2/image_processing_mobilenet_v2.py | 11 ----------- 10 files changed, 111 deletions(-) diff --git a/src/transformers/models/bit/image_processing_bit.py b/src/transformers/models/bit/image_processing_bit.py index 45000f41829c..84302e695c21 100644 --- a/src/transformers/models/bit/image_processing_bit.py +++ b/src/transformers/models/bit/image_processing_bit.py @@ -308,17 +308,6 @@ def preprocess( if do_normalize: image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) - if do_resize: - image = self.resize(image=image, size=size, resample=resample, input_data_format=input_data_format) - - if do_center_crop: - image = self.center_crop(image=image, size=crop_size, input_data_format=input_data_format) - - if do_rescale: - image = self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) - - if do_normalize: - image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) all_images.append(image) images = [ diff --git a/src/transformers/models/chameleon/image_processing_chameleon.py b/src/transformers/models/chameleon/image_processing_chameleon.py index 78ebb53a5410..ca6b72410194 100644 --- a/src/transformers/models/chameleon/image_processing_chameleon.py +++ b/src/transformers/models/chameleon/image_processing_chameleon.py @@ -311,7 +311,6 @@ def preprocess( if input_data_format is None: # We assume that all images have the same channel dimension format. input_data_format = infer_channel_dimension_format(images[0]) - all_images = [] for image in images: if do_resize: @@ -326,17 +325,6 @@ def preprocess( if do_normalize: image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) - if do_resize: - image = self.resize(image=image, size=size, resample=resample, input_data_format=input_data_format) - - if do_center_crop: - image = self.center_crop(image=image, size=crop_size, input_data_format=input_data_format) - - if do_rescale: - image = self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) - - if do_normalize: - image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) all_images.append(image) images = [ to_channel_dimension_format(image, data_format, input_channel_dim=input_data_format) for image in images diff --git a/src/transformers/models/chinese_clip/image_processing_chinese_clip.py b/src/transformers/models/chinese_clip/image_processing_chinese_clip.py index b8edd440a66e..bb5dac68ca00 100644 --- a/src/transformers/models/chinese_clip/image_processing_chinese_clip.py +++ b/src/transformers/models/chinese_clip/image_processing_chinese_clip.py @@ -294,17 +294,6 @@ def preprocess( if do_normalize: image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) - if do_resize: - image = self.resize(image=image, size=size, resample=resample, input_data_format=input_data_format) - - if do_center_crop: - image = self.center_crop(image=image, size=crop_size, input_data_format=input_data_format) - - if do_rescale: - image = self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) - - if do_normalize: - image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) all_images.append(image) images = [ to_channel_dimension_format(image, data_format, input_channel_dim=input_data_format) for image in images diff --git a/src/transformers/models/clip/image_processing_clip.py b/src/transformers/models/clip/image_processing_clip.py index cff3875f43d6..46092b359492 100644 --- a/src/transformers/models/clip/image_processing_clip.py +++ b/src/transformers/models/clip/image_processing_clip.py @@ -333,17 +333,6 @@ def preprocess( if do_normalize: image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) - if do_resize: - image = self.resize(image=image, size=size, resample=resample, input_data_format=input_data_format) - - if do_center_crop: - image = self.center_crop(image=image, size=crop_size, input_data_format=input_data_format) - - if do_rescale: - image = self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) - - if do_normalize: - image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) all_images.append(image) images = [ to_channel_dimension_format(image, data_format, input_channel_dim=input_data_format) for image in images diff --git a/src/transformers/models/deit/image_processing_deit.py b/src/transformers/models/deit/image_processing_deit.py index 8f313a3e881f..70a975573d37 100644 --- a/src/transformers/models/deit/image_processing_deit.py +++ b/src/transformers/models/deit/image_processing_deit.py @@ -284,17 +284,6 @@ def preprocess( if do_normalize: image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) - if do_resize: - image = self.resize(image=image, size=size, resample=resample, input_data_format=input_data_format) - - if do_center_crop: - image = self.center_crop(image=image, size=crop_size, input_data_format=input_data_format) - - if do_rescale: - image = self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) - - if do_normalize: - image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) all_images.append(image) images = [ to_channel_dimension_format(image, data_format, input_channel_dim=input_data_format) for image in images diff --git a/src/transformers/models/deprecated/vit_hybrid/image_processing_vit_hybrid.py b/src/transformers/models/deprecated/vit_hybrid/image_processing_vit_hybrid.py index 40afc716d787..f929c766943d 100644 --- a/src/transformers/models/deprecated/vit_hybrid/image_processing_vit_hybrid.py +++ b/src/transformers/models/deprecated/vit_hybrid/image_processing_vit_hybrid.py @@ -326,17 +326,6 @@ def preprocess( if do_normalize: image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) - if do_resize: - image = self.resize(image=image, size=size, resample=resample, input_data_format=input_data_format) - - if do_center_crop: - image = self.center_crop(image=image, size=crop_size, input_data_format=input_data_format) - - if do_rescale: - image = self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) - - if do_normalize: - image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) all_images.append(image) images = [ to_channel_dimension_format(image, data_format, input_channel_dim=input_data_format) for image in images diff --git a/src/transformers/models/llava_next/image_processing_llava_next.py b/src/transformers/models/llava_next/image_processing_llava_next.py index 6a89a8edb3a9..4f4a1a1125ae 100644 --- a/src/transformers/models/llava_next/image_processing_llava_next.py +++ b/src/transformers/models/llava_next/image_processing_llava_next.py @@ -423,17 +423,6 @@ def _preprocess( if do_normalize: image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) - if do_resize: - image = self.resize(image=image, size=size, resample=resample, input_data_format=input_data_format) - - if do_center_crop: - image = self.center_crop(image=image, size=crop_size, input_data_format=input_data_format) - - if do_rescale: - image = self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) - - if do_normalize: - image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) all_images.append(image) images = [ to_channel_dimension_format(image, data_format, input_channel_dim=input_data_format) for image in images diff --git a/src/transformers/models/llava_next_video/image_processing_llava_next_video.py b/src/transformers/models/llava_next_video/image_processing_llava_next_video.py index 0150cb72d22c..5ae61048a2d1 100644 --- a/src/transformers/models/llava_next_video/image_processing_llava_next_video.py +++ b/src/transformers/models/llava_next_video/image_processing_llava_next_video.py @@ -286,17 +286,6 @@ def _preprocess( if do_normalize: image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) - if do_resize: - image = self.resize(image=image, size=size, resample=resample, input_data_format=input_data_format) - - if do_center_crop: - image = self.center_crop(image=image, size=crop_size, input_data_format=input_data_format) - - if do_rescale: - image = self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) - - if do_normalize: - image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) all_images.append(image) images = [ to_channel_dimension_format(image, data_format, input_channel_dim=input_data_format) for image in images diff --git a/src/transformers/models/mobilenet_v1/image_processing_mobilenet_v1.py b/src/transformers/models/mobilenet_v1/image_processing_mobilenet_v1.py index 54a60fc0c54a..87471f07e10a 100644 --- a/src/transformers/models/mobilenet_v1/image_processing_mobilenet_v1.py +++ b/src/transformers/models/mobilenet_v1/image_processing_mobilenet_v1.py @@ -290,17 +290,6 @@ def preprocess( if do_normalize: image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) - if do_resize: - image = self.resize(image=image, size=size, resample=resample, input_data_format=input_data_format) - - if do_center_crop: - image = self.center_crop(image=image, size=crop_size, input_data_format=input_data_format) - - if do_rescale: - image = self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) - - if do_normalize: - image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) all_images.append(image) images = [ to_channel_dimension_format(image, data_format, input_channel_dim=input_data_format) for image in images diff --git a/src/transformers/models/mobilenet_v2/image_processing_mobilenet_v2.py b/src/transformers/models/mobilenet_v2/image_processing_mobilenet_v2.py index cbb79364f4b4..1859436d7004 100644 --- a/src/transformers/models/mobilenet_v2/image_processing_mobilenet_v2.py +++ b/src/transformers/models/mobilenet_v2/image_processing_mobilenet_v2.py @@ -293,17 +293,6 @@ def preprocess( if do_normalize: image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) - if do_resize: - image = self.resize(image=image, size=size, resample=resample, input_data_format=input_data_format) - - if do_center_crop: - image = self.center_crop(image=image, size=crop_size, input_data_format=input_data_format) - - if do_rescale: - image = self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) - - if do_normalize: - image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) all_images.append(image) images = [ to_channel_dimension_format(image, data_format, input_channel_dim=input_data_format) for image in images From c18c4ea6214d4c97e27f497d951b62331276b1a5 Mon Sep 17 00:00:00 2001 From: Arthur Zucker Date: Thu, 29 Aug 2024 15:44:39 +0200 Subject: [PATCH 3/4] fixup --- src/transformers/models/bit/image_processing_bit.py | 4 +++- .../models/chameleon/image_processing_chameleon.py | 4 +++- .../models/chinese_clip/image_processing_chinese_clip.py | 4 +++- src/transformers/models/clip/image_processing_clip.py | 4 +++- src/transformers/models/deit/image_processing_deit.py | 4 +++- .../deprecated/vit_hybrid/image_processing_vit_hybrid.py | 4 +++- .../models/llava_next/image_processing_llava_next.py | 4 +++- .../llava_next_video/image_processing_llava_next_video.py | 4 +++- .../models/mobilenet_v1/image_processing_mobilenet_v1.py | 4 +++- .../models/mobilenet_v2/image_processing_mobilenet_v2.py | 4 +++- 10 files changed, 30 insertions(+), 10 deletions(-) diff --git a/src/transformers/models/bit/image_processing_bit.py b/src/transformers/models/bit/image_processing_bit.py index 84302e695c21..f18ad0c7f3a4 100644 --- a/src/transformers/models/bit/image_processing_bit.py +++ b/src/transformers/models/bit/image_processing_bit.py @@ -306,7 +306,9 @@ def preprocess( image = self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) if do_normalize: - image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) + image = self.normalize( + image=image, mean=image_mean, std=image_std, input_data_format=input_data_format + ) all_images.append(image) diff --git a/src/transformers/models/chameleon/image_processing_chameleon.py b/src/transformers/models/chameleon/image_processing_chameleon.py index ca6b72410194..aed515eeb89d 100644 --- a/src/transformers/models/chameleon/image_processing_chameleon.py +++ b/src/transformers/models/chameleon/image_processing_chameleon.py @@ -323,7 +323,9 @@ def preprocess( image = self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) if do_normalize: - image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) + image = self.normalize( + image=image, mean=image_mean, std=image_std, input_data_format=input_data_format + ) all_images.append(image) images = [ diff --git a/src/transformers/models/chinese_clip/image_processing_chinese_clip.py b/src/transformers/models/chinese_clip/image_processing_chinese_clip.py index bb5dac68ca00..e0fae0bc1e6a 100644 --- a/src/transformers/models/chinese_clip/image_processing_chinese_clip.py +++ b/src/transformers/models/chinese_clip/image_processing_chinese_clip.py @@ -292,7 +292,9 @@ def preprocess( image = self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) if do_normalize: - image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) + image = self.normalize( + image=image, mean=image_mean, std=image_std, input_data_format=input_data_format + ) all_images.append(image) images = [ diff --git a/src/transformers/models/clip/image_processing_clip.py b/src/transformers/models/clip/image_processing_clip.py index 46092b359492..8758b1110667 100644 --- a/src/transformers/models/clip/image_processing_clip.py +++ b/src/transformers/models/clip/image_processing_clip.py @@ -331,7 +331,9 @@ def preprocess( image = self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) if do_normalize: - image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) + image = self.normalize( + image=image, mean=image_mean, std=image_std, input_data_format=input_data_format + ) all_images.append(image) images = [ diff --git a/src/transformers/models/deit/image_processing_deit.py b/src/transformers/models/deit/image_processing_deit.py index 70a975573d37..1a880b883eeb 100644 --- a/src/transformers/models/deit/image_processing_deit.py +++ b/src/transformers/models/deit/image_processing_deit.py @@ -282,7 +282,9 @@ def preprocess( image = self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) if do_normalize: - image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) + image = self.normalize( + image=image, mean=image_mean, std=image_std, input_data_format=input_data_format + ) all_images.append(image) images = [ diff --git a/src/transformers/models/deprecated/vit_hybrid/image_processing_vit_hybrid.py b/src/transformers/models/deprecated/vit_hybrid/image_processing_vit_hybrid.py index f929c766943d..ebf358af8a94 100644 --- a/src/transformers/models/deprecated/vit_hybrid/image_processing_vit_hybrid.py +++ b/src/transformers/models/deprecated/vit_hybrid/image_processing_vit_hybrid.py @@ -324,7 +324,9 @@ def preprocess( image = self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) if do_normalize: - image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) + image = self.normalize( + image=image, mean=image_mean, std=image_std, input_data_format=input_data_format + ) all_images.append(image) images = [ diff --git a/src/transformers/models/llava_next/image_processing_llava_next.py b/src/transformers/models/llava_next/image_processing_llava_next.py index 4f4a1a1125ae..9b4d826664ac 100644 --- a/src/transformers/models/llava_next/image_processing_llava_next.py +++ b/src/transformers/models/llava_next/image_processing_llava_next.py @@ -421,7 +421,9 @@ def _preprocess( image = self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) if do_normalize: - image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) + image = self.normalize( + image=image, mean=image_mean, std=image_std, input_data_format=input_data_format + ) all_images.append(image) images = [ diff --git a/src/transformers/models/llava_next_video/image_processing_llava_next_video.py b/src/transformers/models/llava_next_video/image_processing_llava_next_video.py index 5ae61048a2d1..274cfa1cec89 100644 --- a/src/transformers/models/llava_next_video/image_processing_llava_next_video.py +++ b/src/transformers/models/llava_next_video/image_processing_llava_next_video.py @@ -284,7 +284,9 @@ def _preprocess( image = self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) if do_normalize: - image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) + image = self.normalize( + image=image, mean=image_mean, std=image_std, input_data_format=input_data_format + ) all_images.append(image) images = [ diff --git a/src/transformers/models/mobilenet_v1/image_processing_mobilenet_v1.py b/src/transformers/models/mobilenet_v1/image_processing_mobilenet_v1.py index 87471f07e10a..cd40371e088d 100644 --- a/src/transformers/models/mobilenet_v1/image_processing_mobilenet_v1.py +++ b/src/transformers/models/mobilenet_v1/image_processing_mobilenet_v1.py @@ -288,7 +288,9 @@ def preprocess( image = self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) if do_normalize: - image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) + image = self.normalize( + image=image, mean=image_mean, std=image_std, input_data_format=input_data_format + ) all_images.append(image) images = [ diff --git a/src/transformers/models/mobilenet_v2/image_processing_mobilenet_v2.py b/src/transformers/models/mobilenet_v2/image_processing_mobilenet_v2.py index 1859436d7004..c3745b4581e7 100644 --- a/src/transformers/models/mobilenet_v2/image_processing_mobilenet_v2.py +++ b/src/transformers/models/mobilenet_v2/image_processing_mobilenet_v2.py @@ -291,7 +291,9 @@ def preprocess( image = self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format) if do_normalize: - image = self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format) + image = self.normalize( + image=image, mean=image_mean, std=image_std, input_data_format=input_data_format + ) all_images.append(image) images = [ From 4bc943a722a39c75b0f4c4dabdbe98ef03d8299b Mon Sep 17 00:00:00 2001 From: Arthur Zucker Date: Thu, 29 Aug 2024 15:46:04 +0200 Subject: [PATCH 4/4] fix typo --- src/transformers/models/bit/image_processing_bit.py | 3 ++- .../models/chameleon/image_processing_chameleon.py | 3 ++- .../models/chinese_clip/image_processing_chinese_clip.py | 3 ++- src/transformers/models/clip/image_processing_clip.py | 3 ++- src/transformers/models/deit/image_processing_deit.py | 3 ++- .../deprecated/vit_hybrid/image_processing_vit_hybrid.py | 3 ++- .../models/llava_next/image_processing_llava_next.py | 3 ++- .../llava_next_video/image_processing_llava_next_video.py | 3 ++- .../models/mobilenet_v1/image_processing_mobilenet_v1.py | 3 ++- .../models/mobilenet_v2/image_processing_mobilenet_v2.py | 3 ++- 10 files changed, 20 insertions(+), 10 deletions(-) diff --git a/src/transformers/models/bit/image_processing_bit.py b/src/transformers/models/bit/image_processing_bit.py index f18ad0c7f3a4..ba2340789970 100644 --- a/src/transformers/models/bit/image_processing_bit.py +++ b/src/transformers/models/bit/image_processing_bit.py @@ -313,7 +313,8 @@ def preprocess( all_images.append(image) images = [ - to_channel_dimension_format(image, data_format, input_channel_dim=input_data_format) for image in images + to_channel_dimension_format(image, data_format, input_channel_dim=input_data_format) + for image in all_images ] data = {"pixel_values": images} diff --git a/src/transformers/models/chameleon/image_processing_chameleon.py b/src/transformers/models/chameleon/image_processing_chameleon.py index aed515eeb89d..46d081973bb4 100644 --- a/src/transformers/models/chameleon/image_processing_chameleon.py +++ b/src/transformers/models/chameleon/image_processing_chameleon.py @@ -329,7 +329,8 @@ def preprocess( all_images.append(image) images = [ - to_channel_dimension_format(image, data_format, input_channel_dim=input_data_format) for image in images + to_channel_dimension_format(image, data_format, input_channel_dim=input_data_format) + for image in all_images ] data = {"pixel_values": images} diff --git a/src/transformers/models/chinese_clip/image_processing_chinese_clip.py b/src/transformers/models/chinese_clip/image_processing_chinese_clip.py index e0fae0bc1e6a..515c2de0cfc3 100644 --- a/src/transformers/models/chinese_clip/image_processing_chinese_clip.py +++ b/src/transformers/models/chinese_clip/image_processing_chinese_clip.py @@ -298,7 +298,8 @@ def preprocess( all_images.append(image) images = [ - to_channel_dimension_format(image, data_format, input_channel_dim=input_data_format) for image in images + to_channel_dimension_format(image, data_format, input_channel_dim=input_data_format) + for image in all_images ] data = {"pixel_values": images} diff --git a/src/transformers/models/clip/image_processing_clip.py b/src/transformers/models/clip/image_processing_clip.py index 8758b1110667..fa398821ca61 100644 --- a/src/transformers/models/clip/image_processing_clip.py +++ b/src/transformers/models/clip/image_processing_clip.py @@ -337,7 +337,8 @@ def preprocess( all_images.append(image) images = [ - to_channel_dimension_format(image, data_format, input_channel_dim=input_data_format) for image in images + to_channel_dimension_format(image, data_format, input_channel_dim=input_data_format) + for image in all_images ] data = {"pixel_values": images} diff --git a/src/transformers/models/deit/image_processing_deit.py b/src/transformers/models/deit/image_processing_deit.py index 1a880b883eeb..bafb5f6e71ad 100644 --- a/src/transformers/models/deit/image_processing_deit.py +++ b/src/transformers/models/deit/image_processing_deit.py @@ -288,7 +288,8 @@ def preprocess( all_images.append(image) images = [ - to_channel_dimension_format(image, data_format, input_channel_dim=input_data_format) for image in images + to_channel_dimension_format(image, data_format, input_channel_dim=input_data_format) + for image in all_images ] data = {"pixel_values": images} diff --git a/src/transformers/models/deprecated/vit_hybrid/image_processing_vit_hybrid.py b/src/transformers/models/deprecated/vit_hybrid/image_processing_vit_hybrid.py index ebf358af8a94..e7c3193ceab4 100644 --- a/src/transformers/models/deprecated/vit_hybrid/image_processing_vit_hybrid.py +++ b/src/transformers/models/deprecated/vit_hybrid/image_processing_vit_hybrid.py @@ -330,7 +330,8 @@ def preprocess( all_images.append(image) images = [ - to_channel_dimension_format(image, data_format, input_channel_dim=input_data_format) for image in images + to_channel_dimension_format(image, data_format, input_channel_dim=input_data_format) + for image in all_images ] data = {"pixel_values": images} diff --git a/src/transformers/models/llava_next/image_processing_llava_next.py b/src/transformers/models/llava_next/image_processing_llava_next.py index 9b4d826664ac..579e6d44c143 100644 --- a/src/transformers/models/llava_next/image_processing_llava_next.py +++ b/src/transformers/models/llava_next/image_processing_llava_next.py @@ -427,7 +427,8 @@ def _preprocess( all_images.append(image) images = [ - to_channel_dimension_format(image, data_format, input_channel_dim=input_data_format) for image in images + to_channel_dimension_format(image, data_format, input_channel_dim=input_data_format) + for image in all_images ] return images diff --git a/src/transformers/models/llava_next_video/image_processing_llava_next_video.py b/src/transformers/models/llava_next_video/image_processing_llava_next_video.py index 274cfa1cec89..59d0d9d94472 100644 --- a/src/transformers/models/llava_next_video/image_processing_llava_next_video.py +++ b/src/transformers/models/llava_next_video/image_processing_llava_next_video.py @@ -290,7 +290,8 @@ def _preprocess( all_images.append(image) images = [ - to_channel_dimension_format(image, data_format, input_channel_dim=input_data_format) for image in images + to_channel_dimension_format(image, data_format, input_channel_dim=input_data_format) + for image in all_images ] return images diff --git a/src/transformers/models/mobilenet_v1/image_processing_mobilenet_v1.py b/src/transformers/models/mobilenet_v1/image_processing_mobilenet_v1.py index cd40371e088d..7981947307fd 100644 --- a/src/transformers/models/mobilenet_v1/image_processing_mobilenet_v1.py +++ b/src/transformers/models/mobilenet_v1/image_processing_mobilenet_v1.py @@ -294,7 +294,8 @@ def preprocess( all_images.append(image) images = [ - to_channel_dimension_format(image, data_format, input_channel_dim=input_data_format) for image in images + to_channel_dimension_format(image, data_format, input_channel_dim=input_data_format) + for image in all_images ] data = {"pixel_values": images} diff --git a/src/transformers/models/mobilenet_v2/image_processing_mobilenet_v2.py b/src/transformers/models/mobilenet_v2/image_processing_mobilenet_v2.py index c3745b4581e7..25d227bd582f 100644 --- a/src/transformers/models/mobilenet_v2/image_processing_mobilenet_v2.py +++ b/src/transformers/models/mobilenet_v2/image_processing_mobilenet_v2.py @@ -297,7 +297,8 @@ def preprocess( all_images.append(image) images = [ - to_channel_dimension_format(image, data_format, input_channel_dim=input_data_format) for image in images + to_channel_dimension_format(image, data_format, input_channel_dim=input_data_format) + for image in all_images ] data = {"pixel_values": images}