File size: 2,276 Bytes
c69c4af |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 |
import torchvision.transforms as T
def get_image_transform(
image_size: int,
center_crop: bool = False,
interpolation: T.InterpolationMode = T.InterpolationMode.BILINEAR # We used bilinear during training
):
if center_crop:
crop = [
T.Resize(image_size, interpolation=interpolation),
T.CenterCrop(image_size)
]
else:
# "Squash": most versatile
crop = [
T.Resize((image_size, image_size), interpolation=interpolation)
]
return T.Compose(crop + [
T.Lambda(lambda x: x.convert("RGB")),
T.ToTensor(),
T.Normalize([0.5, 0.5, 0.5], [0.5, 0.5, 0.5], inplace=True),
])
import torchvision.transforms as T
from PIL import Image
def get_image_transform(
image_size: int,
center_crop: bool = False,
interpolation: T.InterpolationMode = T.InterpolationMode.BILINEAR # We used bilinear during training
):
if center_crop:
crop = [
T.Resize(image_size, interpolation=interpolation),
T.CenterCrop(image_size)
]
else:
# "Squash": most versatile
crop = [
T.Resize((image_size, image_size), interpolation=interpolation)
]
return T.Compose(crop + [
T.Lambda(lambda x: x.convert("RGB")),
T.ToTensor(),
T.Normalize([0.5, 0.5, 0.5], [0.5, 0.5, 0.5], inplace=True),
])
def _convert_to_rgb(image: Image.Image) -> Image.Image:
"""Converts a PIL Image to RGB format."""
return image.convert("RGB")
def get_image_transform_fix(
image_size: int,
center_crop: bool = False,
interpolation: T.InterpolationMode = T.InterpolationMode.BILINEAR
):
if center_crop:
crop = [
T.Resize(image_size, interpolation=interpolation),
T.CenterCrop(image_size)
]
else:
# "Squash": most versatile
crop = [
T.Resize((image_size, image_size), interpolation=interpolation)
]
return T.Compose(crop + [
T.Lambda(_convert_to_rgb),
T.ToTensor(),
T.Normalize([0.5, 0.5, 0.5], [0.5, 0.5, 0.5], inplace=True),
])
def get_text_tokenizer(context_length: int):
return SimpleTokenizer(context_length=context_length) |