Feature/kimi linear support#17592

Closed

cacaview wants to merge 21 commits into

ggml-org:masterfrom

cacaview:feature/kimi-linear-support

Commits on Nov 28, 2025

Add Kimi Linear model conversion support
cacaview
committed
feat: complete Kimi-Linear inference implementation
cacaview
committed
kimi: fix MoE parameters and conversion script
cacaview
committed

Commits on Nov 29, 2025

feat(kimi): add KDA CUDA kernel and optimize recurrence implementation
cacaview
committed
Merge branch 'ggml-org:master' into feature/kimi-linear-support
cacaview
authored

Commits on Nov 30, 2025

Commits on Dec 8, 2025

Commits on Dec 9, 2025

feat(cuda): add MLA Flash Attention support for head_dim=192
cacaview
committed

Commits on Dec 13, 2025

fix: KDA numerical stability and shared expert input
cacaview
committed

Commits on Dec 31, 2025

Commits on Jan 1, 2026

vocab: fix tokenization for TikToken-based models (cherry-picked from pr-18381)
cacaview
committed