COMMITS
July 21, 2026
P
ggml: fork update
Pascal committed
July 20, 2026
P
vulkan: support misaligned get_rows operands
Pascal committed
P
P
ggml: fix conv_1d result layout for batched input
Pascal committed
P
graphs: express the SwiGLU MLPs through ggml_swiglu_split
Pascal committed
July 19, 2026
P
ggml: bump submodule for full CUDA GET_ROWS type coverage
Pascal committed
P
ggml: bump submodule for CUDA k-quant GET_ROWS support
Pascal committed
P
ggml: fork update
Pascal committed
July 17, 2026
P
loader: validate raw u32 counts before the (int) cast
Pascal committed
P
nit: realign endpoint tables on the widened voice routes
Pascal committed
P
clang-format
Pascal committed
P
July 9, 2026
K
server: compatible with OpenAI API, add model alias
karl committed
P
backend: dedup consecutive identical ggml log lines
Pascal committed
July 12, 2026
R
pipeline-codec: reset sched before encode alloc
RaduAdumitroaei committed
P
tts: reuse decode scratch buffers across streaming steps
Pascal committed
July 14, 2026
P
ggml: fork update
Pascal committed
July 8, 2026
P
ggml: fork update
Pascal committed
P
ggml: fork update
Pascal committed
July 6, 2026
P
codec: adaptive chunk width on the streaming decode
Pascal committed
P
talker: static decode graphs per attention window class
Pascal committed
P
predictor: static graphs replayed with a single id upload per step
Pascal committed
P
ggml: fork update
Pascal committed
July 5, 2026
P
server: sampling overrides on the speech endpoint
Pascal committed
P
codec: primed state snapshot LRU keyed by the reference content
Pascal committed
P
server: reject unknown voice names on base models
Pascal committed
P
qwen: lazy speaker encoder load in the voice ref extraction path
Pascal committed