Repository navigation

feat: Triton fused SoftSignGLU kernel for LYNXNet2 - #60

Closed

KakaruHayate wants to merge 23 commits into

mainfrom

feat/fused-softsign-glu

Closed

feat: Triton fused SoftSignGLU kernel for LYNXNet2#60
KakaruHayate wants to merge 23 commits into
mainfrom
feat/fused-softsign-glu

Commits on Jul 2, 2026
Commits on Jul 3, 2026
- fix: backward kernel dtype not hardcoded fp16
  KakaruHayate
  committed
  ade184c
Commits on Jul 7, 2026
- fix: optimize Triton fused SoftSignGLU kernel for training
  KakaruHayate
  committed
  cf25e36
- fix: warn when fused kernels are skipped due to unsupported glu_type
  KakaruHayate
  committed
  bce80cd
Commits on Jul 14, 2026
- review: fix hard blocks — TF32 removal, glu_type revert, robustness
  KakaruHayate
  committed
  b336d78
Commits on Jul 15, 2026
Commits on Jul 21, 2026
- fix: fail fast when Triton is unavailable
  KakaruHayate
  committed
  f8056c2
Commits on Aug 3, 2026
- fix: warm only patched variance backbones
  KakaruHayate
  committed
  c13588f
Commits on Aug 6, 2026
- fix: address CodeRabbit fused kernel review
  KakaruHayate
  committed
  222a72b
- fix: fail fast when fused kernels lack Triton
  KakaruHayate
  committed
  6133bb5
- refactor: finalize fused SoftSignGLU integration
  KakaruHayate
  committed
  8cd54bf
Commits on Aug 7, 2026
- fix: isolate warmup RNG, restore 1D input rank
  KakaruHayate
  committed
  6716397
Commits on Aug 13, 2026
- Harden LYNXNet2Block layout check in fusion wrapper
  KakaruHayate
  committed
  b30828d