Skip to content

Remove pad_max_tiles in CLIP inference #1853

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
lucylq merged 1 commit into from
Oct 16, 2024
Merged

Remove pad_max_tiles in CLIP inference #1853

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
9 changes: 0 additions & 9 deletions torchtune/models/clip/inference/_transform.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -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
Expand All @@ -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
Expand Down Expand Up @@ -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

Expand Down Expand Up @@ -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,
Expand Down Expand Up @@ -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
Expand Down Expand Up @@ -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]:
Expand Down