Fix attention calculation on mps for torch 2.8.0#1068

Merged

jlarson4 merged 9 commits intoTransformerLensOrg:devfrom

BrownianNotion:fix-attn-mps

Mar 22, 2026

Commits on Mar 16, 2026

Commits on Mar 20, 2026

Merge remote-tracking branch 'origin/dev' into fix-attn-mps
jlarson4
committed

Commits on Mar 21, 2026

Adjusted to reduce drift on non-mps computations
jlarson4
committed

Commits on Mar 22, 2026

Very slightly reducing tolerance to account for accumulated floating point difference across random seeds.
jlarson4
committed