Remove pad_max_tiles in CLIP inference (pytorch#1853)

torchtune/models/clip/inference/_transform.py

@@ -18,7 +18,6 @@
 from torchtune.modules.transforms.vision_utils.get_inscribed_size import (
  get_inscribed_size,
 )
-from torchtune.modules.transforms.vision_utils.pad_dim_to_size import pad_dim_to_size
 from torchtune.modules.transforms.vision_utils.tile_crop import tile_crop
 
 from torchvision.transforms.v2 import functional as F
@@ -35,15 +34,13 @@ def __init__(
  tile_size: int,
  max_num_tiles: int,
  antialias: bool,
- pad_max_tiles: bool = False,
  ):
  super().__init__()
  self.resample = resample
  self.image_mean = image_mean
  self.image_std = image_std
  self.tile_size = tile_size
  self.max_num_tiles = max_num_tiles
- self.pad_tile_size = max_num_tiles if pad_max_tiles else None
  self.antialias = antialias
  self.tile_crop = tile_crop
  self.pad = torch.nn.functional.pad
@@ -121,9 +118,6 @@ def forward(
  # Reshape.
  tiles = self.tile_crop(output, self.tile_size)
 
- if self.pad_tile_size:
- tiles = pad_dim_to_size(tiles, size=self.pad_tile_size, dim=0)
-
  # Calculate aspect ratio.
  aspect_ratio = canvas_size // self.tile_size
 
@@ -181,7 +175,6 @@ class CLIPImageTransform:
  If False, it will pick the resolution that minimizes downscaling, including no downscaling at all.
  In this case, the image will only be upscaled if it's size < tile_size. Default False.
  antialias (bool): Whether to apply antialiasing when resizing the image. Default True.
- pad_max_tiles (bool): If True, the image will be padded to have tiles == max_num_tiles.
  Examples:
  >>> image_transform = CLIPImageTransform(
  ... image_mean=None,
@@ -212,7 +205,6 @@ def __init__(
  resample: str = "bilinear",
  resize_to_max_canvas: bool = False,
  antialias: bool = True,
- pad_max_tiles: bool = False,
  ) -> None:
 
  # get_canvas_best_fit
@@ -258,7 +250,6 @@ def __init__(
  tile_size=self.tile_size,
  max_num_tiles=self.max_num_tiles,
  antialias=self.antialias,
- pad_max_tiles=pad_max_tiles,
  )
 
  def __call__(self, *, image: Image.Image, **kwargs) -> Mapping[str, Any]: