COMMITS
February 14, 2026
M
hexagon: further optimizations and refactoring for flash attention (#19583)
Max Krasnyansky committed
February 13, 2026
M
github : add missing backends to issue templates (#19603)
Mengsheng Wu committed
J
vulkan: restore -inf check in FA shaders (#19582)
Jeff Bolz committed
A
common : update download code (#19573)
Adrien Gallouët committed
X
model: support GLM MoE DSA arch (NOTE: indexer is not yet supported) (#19460)
Xuan-Son Nguyen committed
A
Fix wrong memcpy length for block_interleave == 4 (#19575)
Alberto Cabrera Pérez committed
Y
fix vulkan ggml_acc only works in 3d but not 4d (#19426)
ymcki committed
S
support --verbose-prompt (#19576)
Sigbjørn Skjæret committed
A
CUDA: loop over ne2*ne3 in case it overflows (#19538)
Aman Gupta committed
A
webui: UI and routing fixes (#19586)
Aleksander Grygier committed
O
CUDA: Do not mutate cgraph for fused ADDs (#19566)
Oliver Simons committed
P
docs : fix broken link and typo (#19560)
Pavan Shinde committed
Y
model : Kimi Linear fix conv state update (#19531)
ymcki committed
A
llama : remove deprecated codecvt (#19565)
Adrien Gallouët committed
A
vendor : update BoringSSL to 0.20260211.0 (#19562)
Adrien Gallouët committed
G
memory : fix kv cache size for hybrid models (#19559)
Georgi Gerganov committed
G
metal : improve concurrency (#19555)
Georgi Gerganov committed
G
metal : support GGML_OP_SET (#19548)
Georgi Gerganov committed
February 12, 2026
S
hexagon: fix typo in vtcm_needs_release (#19545)
Shupei Fan committed
L
opencl: add basic support for q4_1 (#19534)
lhez committed
G
args : add -kvu to llama-parallel (#19577)
Georgi Gerganov committed
A
webui: Add switcher to Chat Message UI to show raw LLM output (#19571)
Aleksander Grygier committed
A
vendor : update cpp-httplib (#19537)
Adrien Gallouët committed
C
llama : update outdated comment in llama.h (#19428)
Christian Schmitz committed
A
(webui) FEATURE: Enable adding or injecting System Message into chat (#19556)
Aleksander Grygier committed
D
scripts : add support for forks in pr2wt.sh (#19540)
Daniel Bevenius committed
A
(webui) REFACTOR: UI primitives and polish (#19551)
Aleksander Grygier committed
A
WebUI Architecture Cleanup (#19541)
Aleksander Grygier committed
G
metal : update sum_rows kernel to support float4 (#19524)
Georgi Gerganov committed
M
Add a workaround for compilation with ROCWMMA_FATTN and gfx9 (#19461)
Mario Limonciello committed