Skip to content

Fixed the lpnormpool implementation #639

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
wants to merge 2 commits into
base: master
Choose a base branch
from
Open

Fixed the lpnormpool implementation #639

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
df035ec
Fix lpnormpool
IBArbitrary Jun 6, 2025
fa86b16
Added comments to lpnormpool
IBArbitrary Jun 6, 2025
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
16 changes: 9 additions & 7 deletions src/impl/pooling_direct.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -95,14 +95,14 @@ for name in (:max, :mean, :lpnorm)
elseif $(name == :mean)
m += x[input_kw, input_kh, input_kd, c, batch_idx]
elseif $(name == :lpnorm)
# y = (∑ᵢ xᵢ^p)^(1 / p), here to calculate ∑ᵢ xᵢ^p
m += x[input_kw, input_kh, input_kd, c, batch_idx]^p
# y = (∑ᵢ |xᵢ|^p)^(1 / p), here to calculate ∑ᵢ |xᵢ|^p
m += abs(x[input_kw, input_kh, input_kd, c, batch_idx])^p
else
error("Unimplemented codegen path")
end
end

# for lpnormpool, y = (∑ᵢ xᵢ^p)^(1 / p)
# for lpnormpool, y = (∑ᵢ |xᵢ|^p)^(1 / p)
m = $(name == :lpnorm) ? m^(T(1) / p) : m

y[w, h, d, c, batch_idx] = _alpha * m # + _beta * y[w, h, d, c, batch_idx]
Expand Down Expand Up @@ -151,7 +151,7 @@ for name in (:max, :mean, :lpnorm)
elseif $(name == :mean)
m += x[input_kw, input_kh, input_kd, c, batch_idx]
elseif $(name == :lpnorm)
m += x[input_kw, input_kh, input_kd, c, batch_idx]^p
m += abs(x[input_kw, input_kh, input_kd, c, batch_idx])^p
else
error("Unimplemented codegen path")
end
Expand Down Expand Up @@ -263,8 +263,9 @@ for name in (:max, :mean, :lpnorm)
# Either does meanpool :(
dx[input_kw, input_kh, input_kd, c, batch_idx] += dy_idx * _alpha
elseif $(name == :lpnorm)
# y = (∑ᵢ xᵢ^p)^(1 / p), ∂y/∂xᵢ = xᵢ^(p-1) × y^(1-p)
grad = x[input_kw, input_kh, input_kd, c, batch_idx]^(p-1) * y_idx^(1-p)
# y = (∑ᵢ |xᵢ|^p)^(1 / p), ∂y/∂xᵢ = |xᵢ|^(p-1) × y^(1-p) × sign(xᵢ)
xv = x[input_kw, input_kh, input_kd, c, batch_idx]
grad = abs(xv)^(p-1) * y_idx^(1-p) * sign(xv)
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

why the sign here?

Copy link
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

That comes from the derivative of abs

dx[input_kw, input_kh, input_kd, c, batch_idx] += dy_idx * grad
else
error("Unimplemented codegen path")
Expand Down Expand Up @@ -327,7 +328,8 @@ for name in (:max, :mean, :lpnorm)
elseif $(name == :mean)
dx[input_kw, input_kh, input_kd, c, batch_idx] += dy_idx * _alpha #+ _beta * dx[x_idxs...]
elseif $(name == :lpnorm)
grad = x[input_kw, input_kh, input_kd, c, batch_idx]^(p-1) * y_idx^(1-p)
xv = x[input_kw, input_kh, input_kd, c, batch_idx]
grad = abs(xv)^(p-1) * y_idx^(1-p) * sign(xv)
dx[input_kw, input_kh, input_kd, c, batch_idx] += dy_idx * grad
else
error("Unimplemented codegen path")
Expand Down
Loading