Skip to content
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

Epsilon change in normalise for stability #2421

Merged
merged 3 commits into from
Nov 5, 2024
Merged

Epsilon change in normalise for stability #2421

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
da2061d
epsilon change for stability
billera Apr 7, 2024
b600f7a
Change comment for eps
billera Apr 7, 2024
fbc1186
Merge branch 'master' into master
CarloLucibello Nov 5, 2024
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
11 changes: 6 additions & 5 deletions src/layers/stateless.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,10 +1,10 @@

"""
normalise(x; dims=ndims(x), eps=1e-5)
normalise(x; dims=ndims(x), eps=1f-5)

Normalise `x` to mean 0 and standard deviation 1 across the dimension(s) given by `dims`.
Per default, `dims` is the last dimension.
`eps` is a small term added to the denominator for numerical stability.
`eps` is a small term added to the variance for numerical stability.

# Examples
```jldoctest
Expand Down Expand Up @@ -34,10 +34,11 @@ julia> isapprox(std(y; dims=1, corrected=false), ones(1, 10), atol=1e-5)
true
```
"""
@inline function normalise(x::AbstractArray; dims=ndims(x), eps=ofeltype(x, 1e-5))
@inline function normalise(x::AbstractArray; dims=ndims(x), eps=1f-5)
Copy link
Member

Choose a reason for hiding this comment

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

Why does this now assume Float32? Elsewhere we try to allow for Float16 too.

μ = mean(x, dims=dims)
σ = std(x, dims=dims, mean=μ, corrected=false)
return @. (x - μ) / (σ + eps)
σ² = var(x, dims=dims, mean=μ, corrected=false)
ε = ofeltype(x, eps)
return @. (x - μ) / sqrt(σ² + ε^2)
end

"""
Expand Down
Loading