COMMITS
October 1, 2026
K
model : re-enable -sm tensor for qwen4exp (#28569)
Kevin Hopper committed
M
webgpu: fix SSM_SCAN binding aliasing (#29750)
Masashi Yoshimura committed
September 30, 2026
A
ggml-opencl : replace alloca() with std::vector (#29765)
Adrien Gallouët committed
R
cuda: guard the iq4_nl dequantize row kernel against short rows (#29683)
R0CKSTAR committed
E
Hexagon: optimize ALLREDUCE with support for safe scatter mode (#29757)
Ehsan Bateni committed
P
args: fix cli download mmproj arg (#28977)
pr3pony committed
H
llama : preserve original batch order for speculative decoding layer inputs (#29019)
Hrishith Thadicherla committed
S
test-llama-archs : toggle causal_attn to catch graph shape changes (#29724)
Sihan Yu committed
S
convert : fix LoRA conversion crash for Qwen3.5 V-head reorder (#28324)
Swigler committed
X
llama: properly handle KV on training (#28520)
Xuan-Son Nguyen committed
X
batch: migrate the rest of examples to llama_batch_ext (#29601)
Xuan-Son Nguyen committed
P
glm5-next: give dead indexer slots unique scatter rows (#29745)
Pascal committed
D
ggml/gguf : fix integer overflow (#29384)
Daniel Kuts committed
V
codeowners : remove former ZenDNN owner (#29747)
Vishal Singh committed
P
G
mimo : support dflash (convert + feature extraction) (#29650)
Georgi Gerganov committed
S
jinja : support coerced array attributes (#29574)
Sigbjørn Skjæret committed
A
ggml-et : remove useless alloca() (#29663)
Adrien Gallouët committed
P
A
llama: llama_prefetch_rows (#29599)
Aman Gupta committed
A
ggml : add BF16 unary, GLU, binary and scale ops (CPU, CUDA) (#29675)
Aman Gupta committed
P
cpu: accept BF16 in src1 of mul_mat (#28937)
Pascal committed
D
model-conversion : add --add-bos to run org model script (#29558)
Daniel Bevenius committed
A
ui : shared model display primitives (#29644)
Aleksander Grygier committed
A
ui : model download pipeline (#27959)
Aleksander Grygier committed
A
ui : model memory-fit estimation (#27957)
Aleksander Grygier committed
A
ui : Hugging Face Hub data layer (#27947)
Aleksander Grygier committed
A
ui : model id grammar for sidecars, quants and capability parsing (#27946)
Aleksander Grygier committed
A
ui : type-safe API types, fetch helpers and download-ready models store plumbing (#29582)
Aleksander Grygier committed
P