Repository navigation

perf(gated_delta): chunk-parallel differentiable training path for Ga… - #2021

Draft

kru2710shna wants to merge 5 commits into

ml-explore:mainfrom

kru2710shna:fix/gated-delta-chunked-training

Draft

perf(gated_delta): chunk-parallel differentiable training path for Ga…#2021
kru2710shna wants to merge 5 commits into
ml-explore:mainfrom
kru2710shna:fix/gated-delta-chunked-training

Commits on Oct 29, 2026
- perf(gated_delta): chunk-parallel differentiable training path for GatedDeltaNet
  kru2710shna
  committed
  0d5d488
Commits on Oct 30, 2026
- Merge branch 'main' into fix/gated-delta-chunked-training
  kru2710shna
  authored
  060a0f9
Commits on Nov 2, 2026
- Merge branch 'main' into fix/gated-delta-chunked-training
  kru2710shna
  authored
  a06ca01
Commits on Nov 7, 2026
- fix(gated_delta): clamp decay_ratio exponent to avoid NaN gradients
  kru2710shna
  committed
  f824cc1
Commits on Nov 8, 2026
- fix(gated_delta): block-Schur inverse for near-singular chunk matrices
  kru2710shna
  committed
  5f40854