Skip to content

Improve readability of PixelShuffle operation in Llama4 #304

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

Open
Yuki-Imajuku wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Improve readability of PixelShuffle operation in Llama4 #304

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
12 changes: 6 additions & 6 deletions models/llama4/vision/embedding.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,11 +18,11 @@


class PixelShuffle(nn.Module):
def __init__(self, ps_ratio):
def __init__(self, ps_ratio: float):
super().__init__()
self.ps_ratio = ps_ratio

def forward(self, x):
def forward(self, x: torch.Tensor) -> torch.Tensor:
# x: [B, N, C], N = number of patches
assert self.ps_ratio is not None, "ps_ratio is required for pixel shuffle"
assert x.dim() == 3, "pixel shuffle requires encoded patches [B, N, C]"
Expand All @@ -33,14 +33,14 @@ def forward(self, x):
return pixel_shuffle_patches


def pixel_shuffle_op(input_x, ps_ratio):
n, w, h, c = input_x.size()
input_x = input_x.view(n, w, int(h * ps_ratio), int(c / ps_ratio))
def pixel_shuffle_op(input_x: torch.Tensor, ps_ratio: float) -> torch.Tensor:
n, h, w, c = input_x.size()
input_x = input_x.view(n, h, int(w * ps_ratio), int(c / ps_ratio))
input_x = input_x.permute(0, 2, 1, 3).contiguous()
input_x = input_x.view(
n,
int(h * ps_ratio),
int(w * ps_ratio),
int(h * ps_ratio),
int(c / (ps_ratio * ps_ratio)),
)
input_x = input_x.permute(0, 2, 1, 3).contiguous()
Expand Down