FM / transformers

/usr/local/lib/python3.6/site-packages/transformers/

/usr/local/lib/python3.6/site-packages/transformers

Name	Size	Mode	Actions
benchmark/	-	0755	rm
commands/	-	0755	rm
data/	-	0755	rm
models/	-	0755	rm
onnx/	-	0755	rm
pipelines/	-	0755	rm
sagemaker/	-	0755	rm
utils/	-	0755	rm
__pycache__/	-	0755	rm
activations.py	6612	0644	edit dl rm
activations_tf.py	4268	0644	edit dl rm
configuration_utils.py	46765	0644	edit dl rm
convert_graph_to_onnx.py	19575	0644	edit dl rm
convert_pytorch_checkpoint_to_tf2.py	16640	0644	edit dl rm
convert_slow_tokenizer.py	37626	0644	edit dl rm
convert_slow_tokenizers_checkpoints_to_fast.py	4954	0644	edit dl rm
convert_tf_hub_seq_to_seq_bert_to_pytorch.py	2899	0644	edit dl rm
debug_utils.py	12887	0644	edit dl rm
deepspeed.py	18883	0644	edit dl rm
dependency_versions_check.py	1767	0644	edit dl rm
dependency_versions_table.py	2390	0644	edit dl rm
dynamic_module_utils.py	19306	0644	edit dl rm
feature_extraction_sequence_utils.py	18074	0644	edit dl rm
feature_extraction_utils.py	26768	0644	edit dl rm
file_utils.py	3930	0644	edit dl rm
generation_beam_constraints.py	19078	0644	edit dl rm
generation_beam_search.py	40145	0644	edit dl rm
generation_flax_logits_process.py	10879	0644	edit dl rm
generation_flax_utils.py	38708	0644	edit dl rm
generation_logits_process.py	30336	0644	edit dl rm
generation_stopping_criteria.py	5509	0644	edit dl rm
generation_tf_logits_process.py	17503	0644	edit dl rm
generation_tf_utils.py	130671	0644	edit dl rm
generation_utils.py	177118	0644	edit dl rm
hf_argparser.py	11901	0644	edit dl rm
image_utils.py	12860	0644	edit dl rm
integrations.py	40203	0644	edit dl rm
keras_callbacks.py	18902	0644	edit dl rm
modelcard.py	35278	0644	edit dl rm
modeling_flax_outputs.py	35514	0644	edit dl rm
modeling_flax_pytorch_utils.py	12296	0644	edit dl rm
modeling_flax_utils.py	37661	0644	edit dl rm
modeling_outputs.py	60327	0644	edit dl rm
modeling_tf_outputs.py	46107	0644	edit dl rm
modeling_tf_pytorch_utils.py	19916	0644	edit dl rm
modeling_tf_utils.py	100832	0644	edit dl rm
modeling_utils.py	137800	0644	edit dl rm
optimization.py	27756	0644	edit dl rm
optimization_tf.py	15722	0644	edit dl rm
processing_utils.py	10559	0644	edit dl rm
py.typed	1	0644	edit dl rm
pytorch_utils.py	1649	0644	edit dl rm
testing_utils.py	48292	0644	edit dl rm
tf_utils.py	1568	0644	edit dl rm
tokenization_utils.py	39867	0644	edit dl rm
tokenization_utils_base.py	170159	0644	edit dl rm
tokenization_utils_fast.py	32909	0644	edit dl rm
trainer.py	143103	0644	edit dl rm
trainer_callback.py	23199	0644	edit dl rm
trainer_pt_utils.py	44764	0644	edit dl rm
trainer_seq2seq.py	10396	0644	edit dl rm
trainer_tf.py	34564	0644	edit dl rm
trainer_utils.py	18509	0644	edit dl rm
training_args.py	68042	0644	edit dl rm
training_args_seq2seq.py	2829	0644	edit dl rm
training_args_tf.py	14267	0644	edit dl rm
__init__.py	171952	0644	edit dl rm

Edit: /usr/local/lib/python3.6/site-packages/transformers/image_utils.py (12860B)

# coding=utf-8
# Copyright 2021 The HuggingFace Inc. team.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

import os
from typing import List, Union

import numpy as np
import PIL.Image
import PIL.ImageOps

import requests

from .utils import is_torch_available
from .utils.generic import _is_torch

IMAGENET_DEFAULT_MEAN = [0.485, 0.456, 0.406]
IMAGENET_DEFAULT_STD = [0.229, 0.224, 0.225]
IMAGENET_STANDARD_MEAN = [0.5, 0.5, 0.5]
IMAGENET_STANDARD_STD = [0.5, 0.5, 0.5]

ImageInput = Union[
    PIL.Image.Image, np.ndarray, "torch.Tensor", List[PIL.Image.Image], List[np.ndarray], List["torch.Tensor"]  # noqa
]

def is_torch_tensor(obj):
    return _is_torch(obj) if is_torch_available() else False

def load_image(image: Union[str, "PIL.Image.Image"]) -> "PIL.Image.Image":
    """
    Loads `image` to a PIL Image.

Args:
        image (`str` or `PIL.Image.Image`):
            The image to convert to the PIL Image format.

Returns:
        `PIL.Image.Image`: A PIL Image.
    """
    if isinstance(image, str):
        if image.startswith("http://") or image.startswith("https://"):
            # We need to actually check for a real protocol, otherwise it's impossible to use a local file
            # like http_huggingface_co.png
            image = PIL.Image.open(requests.get(image, stream=True).raw)
        elif os.path.isfile(image):
            image = PIL.Image.open(image)
        else:
            raise ValueError(
                f"Incorrect path or url, URLs must start with `http://` or `https://`, and {image} is not a valid path"
            )
    elif isinstance(image, PIL.Image.Image):
        image = image
    else:
        raise ValueError(
            "Incorrect format used for image. Should be an url linking to an image, a local path, or a PIL image."
        )
    image = PIL.ImageOps.exif_transpose(image)
    image = image.convert("RGB")
    return image

# In the future we can add a TF implementation here when we have TF models.
class ImageFeatureExtractionMixin:
    """
    Mixin that contain utilities for preparing image features.
    """

def _ensure_format_supported(self, image):
        if not isinstance(image, (PIL.Image.Image, np.ndarray)) and not is_torch_tensor(image):
            raise ValueError(
                f"Got type {type(image)} which is not supported, only `PIL.Image.Image`, `np.array` and "
                "`torch.Tensor` are."
            )

def to_pil_image(self, image, rescale=None):
        """
        Converts `image` to a PIL Image. Optionally rescales it and puts the channel dimension back as the last axis if
        needed.

Args:
            image (`PIL.Image.Image` or `numpy.ndarray` or `torch.Tensor`):
                The image to convert to the PIL Image format.
            rescale (`bool`, *optional*):
                Whether or not to apply the scaling factor (to make pixel values integers between 0 and 255). Will
                default to `True` if the image type is a floating type, `False` otherwise.
        """
        self._ensure_format_supported(image)

if is_torch_tensor(image):
            image = image.numpy()

if isinstance(image, np.ndarray):
            if rescale is None:
                # rescale default to the array being of floating type.
                rescale = isinstance(image.flat[0], np.floating)
            # If the channel as been moved to first dim, we put it back at the end.
            if image.ndim == 3 and image.shape[0] in [1, 3]:
                image = image.transpose(1, 2, 0)
            if rescale:
                image = image * 255
            image = image.astype(np.uint8)
            return PIL.Image.fromarray(image)
        return image

def to_numpy_array(self, image, rescale=None, channel_first=True):
        """
        Converts `image` to a numpy array. Optionally rescales it and puts the channel dimension as the first
        dimension.

Args:
            image (`PIL.Image.Image` or `np.ndarray` or `torch.Tensor`):
                The image to convert to a NumPy array.
            rescale (`bool`, *optional*):
                Whether or not to apply the scaling factor (to make pixel values floats between 0. and 1.). Will
                default to `True` if the image is a PIL Image or an array/tensor of integers, `False` otherwise.
            channel_first (`bool`, *optional*, defaults to `True`):
                Whether or not to permute the dimensions of the image to put the channel dimension first.
        """
        self._ensure_format_supported(image)

if isinstance(image, PIL.Image.Image):
            image = np.array(image)

if is_torch_tensor(image):
            image = image.numpy()

if rescale is None:
            rescale = isinstance(image.flat[0], np.integer)

if rescale:
            image = image.astype(np.float32) / 255.0

if channel_first and image.ndim == 3:
            image = image.transpose(2, 0, 1)

return image

def normalize(self, image, mean, std):
        """
        Normalizes `image` with `mean` and `std`. Note that this will trigger a conversion of `image` to a NumPy array
        if it's a PIL Image.

Args:
            image (`PIL.Image.Image` or `np.ndarray` or `torch.Tensor`):
                The image to normalize.
            mean (`List[float]` or `np.ndarray` or `torch.Tensor`):
                The mean (per channel) to use for normalization.
            std (`List[float]` or `np.ndarray` or `torch.Tensor`):
                The standard deviation (per channel) to use for normalization.
        """
        self._ensure_format_supported(image)

if isinstance(image, PIL.Image.Image):
            image = self.to_numpy_array(image)

if isinstance(image, np.ndarray):
            if not isinstance(mean, np.ndarray):
                mean = np.array(mean).astype(image.dtype)
            if not isinstance(std, np.ndarray):
                std = np.array(std).astype(image.dtype)
        elif is_torch_tensor(image):
            import torch

if not isinstance(mean, torch.Tensor):
                mean = torch.tensor(mean)
            if not isinstance(std, torch.Tensor):
                std = torch.tensor(std)

if image.ndim == 3 and image.shape[0] in [1, 3]:
            return (image - mean[:, None, None]) / std[:, None, None]
        else:
            return (image - mean) / std

def resize(self, image, size, resample=PIL.Image.BILINEAR, default_to_square=True, max_size=None):
        """
        Resizes `image`. Note that this will trigger a conversion of `image` to a PIL Image.

Args:
            image (`PIL.Image.Image` or `np.ndarray` or `torch.Tensor`):
                The image to resize.
            size (`int` or `Tuple[int, int]`):
                The size to use for resizing the image. If `size` is a sequence like (h, w), output size will be
                matched to this.

If `size` is an int and `default_to_square` is `True`, then image will be resized to (size, size). If
                `size` is an int and `default_to_square` is `False`, then smaller edge of the image will be matched to
                this number. i.e, if height > width, then image will be rescaled to (size * height / width, size).
            resample (`int`, *optional*, defaults to `PIL.Image.BILINEAR`):
                The filter to user for resampling.
            default_to_square (`bool`, *optional*, defaults to `True`):
                How to convert `size` when it is a single int. If set to `True`, the `size` will be converted to a
                square (`size`,`size`). If set to `False`, will replicate
                [`torchvision.transforms.Resize`](https://pytorch.org/vision/stable/transforms.html#torchvision.transforms.Resize)
                with support for resizing only the smallest edge and providing an optional `max_size`.
            max_size (`int`, *optional*, defaults to `None`):
                The maximum allowed for the longer edge of the resized image: if the longer edge of the image is
                greater than `max_size` after being resized according to `size`, then the image is resized again so
                that the longer edge is equal to `max_size`. As a result, `size` might be overruled, i.e the smaller
                edge may be shorter than `size`. Only used if `default_to_square` is `False`.
        """
        self._ensure_format_supported(image)

if not isinstance(image, PIL.Image.Image):
            image = self.to_pil_image(image)

if isinstance(size, list):
            size = tuple(size)

if isinstance(size, int) or len(size) == 1:
            if default_to_square:
                size = (size, size) if isinstance(size, int) else (size[0], size[0])
            else:
                width, height = image.size
                # specified size only for the smallest edge
                short, long = (width, height) if width <= height else (height, width)
                requested_new_short = size if isinstance(size, int) else size[0]

if short == requested_new_short:
                    return image

new_short, new_long = requested_new_short, int(requested_new_short * long / short)

if max_size is not None:
                    if max_size <= requested_new_short:
                        raise ValueError(
                            f"max_size = {max_size} must be strictly greater than the requested "
                            f"size for the smaller edge size = {size}"
                        )
                    if new_long > max_size:
                        new_short, new_long = int(max_size * new_short / new_long), max_size

size = (new_short, new_long) if width <= height else (new_long, new_short)

return image.resize(size, resample=resample)

def center_crop(self, image, size):
        """
        Crops `image` to the given size using a center crop. Note that if the image is too small to be cropped to the
        size given, it will be padded (so the returned result has the size asked).

Args:
            image (`PIL.Image.Image` or `np.ndarray` or `torch.Tensor`):
                The image to resize.
            size (`int` or `Tuple[int, int]`):
                The size to which crop the image.
        """
        self._ensure_format_supported(image)
        if not isinstance(size, tuple):
            size = (size, size)

# PIL Image.size is (width, height) but NumPy array and torch Tensors have (height, width)
        image_shape = (image.size[1], image.size[0]) if isinstance(image, PIL.Image.Image) else image.shape[-2:]
        top = (image_shape[0] - size[0]) // 2
        bottom = top + size[0]  # In case size is odd, (image_shape[0] + size[0]) // 2 won't give the proper result.
        left = (image_shape[1] - size[1]) // 2
        right = left + size[1]  # In case size is odd, (image_shape[1] + size[1]) // 2 won't give the proper result.

# For PIL Images we have a method to crop directly.
        if isinstance(image, PIL.Image.Image):
            return image.crop((left, top, right, bottom))

# Check if all the dimensions are inside the image.
        if top >= 0 and bottom <= image_shape[0] and left >= 0 and right <= image_shape[1]:
            return image[..., top:bottom, left:right]

# Otherwise, we may need to pad if the image is too small. Oh joy...
        new_shape = image.shape[:-2] + (max(size[0], image_shape[0]), max(size[1], image_shape[1]))
        if isinstance(image, np.ndarray):
            new_image = np.zeros_like(image, shape=new_shape)
        elif is_torch_tensor(image):
            new_image = image.new_zeros(new_shape)

top_pad = (new_shape[-2] - image_shape[0]) // 2
        bottom_pad = top_pad + image_shape[0]
        left_pad = (new_shape[-1] - image_shape[1]) // 2
        right_pad = left_pad + image_shape[1]
        new_image[..., top_pad:bottom_pad, left_pad:right_pad] = image

top += top_pad
        bottom += top_pad
        left += left_pad
        right += left_pad

return new_image[
            ..., max(0, top) : min(new_image.shape[-2], bottom), max(0, left) : min(new_image.shape[-1], right)
        ]

cmd: