COMMITS
September 30, 2026
P
ci : fix Fusion / metal by adding glm5-next to MTL.csv (#29712)
Pascal committed
R
musa : define __CUDA_ARCH__ for device passes (#29508)
R0CKSTAR committed
S
ci : add models backend check (#29651)
Sigbjørn Skjæret committed
C
SYCL: reduce tensor allreduce sync with pinned host buffers (#29604)
Captain-Tripps committed
T
add GLM-5.3-Flash (GLM5-Next) support (#27773)
timkhronos committed
A
vendor: update BoringSSL to 0.20260929.0 (#29669)
Alessandro de Oliveira Faria (A.K.A.CABELO) committed
A
ggml-zdnn: impl buffer reset, fix memory leaks (#29637)
Aaron Teo committed
September 29, 2026
C
Hexagon f16 activation ops (#29209)
cqderek committed
X
gguf : reject tensor size that wraps after padding (#26979)
Xiang Chen committed
A
ggml : check row bounds in get_rows_back (#29575)
Adrien Gallouët committed
B
model : support classifier_pooling for rerankers (#29627)
bosh committed
T
hexagon: optimize concat op (#29673)
Trivikram Reddy committed
P
CUDA: bitonic argsort handles rows wider than one block (#28957)
Pascal committed
T
ggml-cuda: HIP: optimize packed byte subtraction (`__vsubss4` -> `__vsub4`) (#29478)
thelittlefireman committed
A
ci: add zdnn backend build but not test (#29541)
Aaron Teo committed
R
vulkan: Tune GDN kernel, fix Intel performance (#29476)
Ruben Ortlam committed
M
vulkan : Load F32 A matrix 2 at a time when its 2-aligned (#29254)
Matt Corallo committed
A
vulkan: MOE aware mat_mul_id tile selection (#29182)
Ankit Khandelwal committed
X
T
vocab : keep </s> NORMAL in PLaMo-2 and PLaMo-3 (#29580)
Toki Nasin committed
A
ggml : accumulate f16 dot products in f32 on AVX512-FP16 (#29545)
Adrien Gallouët committed
G
ggml : require input tensors to be GGML_OP_NONE (#29647)
Georgi Gerganov committed
A
hexagon: add FP32 GELU_ERF and GEGLU_ERF support (#29631)
Aparna M P committed
E
common : stop accepting draft tokens at EOG (#29638)
Ethan Guo committed
E
server : remove the built-in UI's service worker when the UI is not served (#29565)
Emanuil Rusev committed
A
common : use fs::path for config dir (#29649)
Adrien Gallouët committed
G
tests : adjust server string regex to also match m2 utlra results (#29648)
Georgi Gerganov committed
A
common : add fs_write_atomic() (#29642)
Adrien Gallouët committed
G
ggml : collect all input tensors into graph_inputs (#29634)
Georgi Gerganov committed