Skip to content

Update to new DimWit broadcast where #7

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
benikm91 merged 1 commit into from
Sep 18, 2026
Merged

Update to new DimWit broadcast where #7

Show file tree
Hide file tree
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
3 changes: 2 additions & 1 deletion core/src/main/scala/deepwit/attention/Attention.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,6 +1,7 @@
package deepwit.attention

import dimwit.*
import dimwit.Conversions.given
import deepwit.activation.softmax
import deepwit.base.LinearLayer
import dimwit.Label as Λ
Expand Down Expand Up @@ -38,7 +39,7 @@ abstract class Attention[Source: Λ, SourceEmbedding: Λ, Target: Λ, TargetEmbe
private final def calculateAttentionWeights(queries: Tensor2[Target, Query, V], keys: Tensor2[Source, Key, V]) =
val attentionScores = attentionScore(queries, keys)
val attentionMask = createAttentionMask(attentionScores.shape)
val maskedScores = where(attentionMask, attentionScores, Tensor.like(attentionScores).fill(Float.NegativeInfinity))
val maskedScores = where_!(attentionMask, attentionScores, Float.NegativeInfinity)
maskedScores.vapply(Axis[Source])(softmax)

/** Creates a boolean mask indicating which source positions each target position may attend to.
Expand Down
3 changes: 2 additions & 1 deletion core/src/main/scala/deepwit/regularization/Perturbation.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,6 +1,7 @@
package deepwit.regularization

import dimwit.*
import dimwit.Conversions.given
import dimwit.stats.Bernoulli
import dimwit.Label as Λ

Expand Down Expand Up @@ -57,4 +58,4 @@ object Perturbation:
val keepProbability = Tensor0(1f) - probability
val ones = Tensor(Shape1(extent)).fill(1f)
val kept = Bernoulli(Prob(ones.scale(keepProbability))).sample(key)
where(kept, ones / ones.scale(keepProbability), Tensor(Shape1(extent)).fill(0f))
where_!(kept, ones / ones.scale(keepProbability), 0f)
Loading