merge from upstream#94

Merged

l3utterfly merged 82 commits intolayla-buildfrom

Apr 10, 2026

Commits on Apr 6, 2026

Commits on Apr 7, 2026

docs: fix typo in build.md (emdawbwebgpu -> emdawnwebgpu) (#21518)
CastelDazur
authored
[SYCL] Add Q8_0 reorder optimization (~3x tg speedup on Intel Arc) (#21527)
PMZFX
authored
Fix rtl text rendering (#21382)
Kabir08
authored
fix: Detect streaming state in reasoning content blocks (#21549)
allozaur
authored
ggml-cuda : fix CDNA2 compute capability constant for gfx90a (MI210) (#21519)
aviallon
authored
webui : store reasoning_content so it is sent back in subsequent requests (#21249)
aldehir
authored
vulkan: add FA dequant for q4_1, q5_0, q5_1, iq4_nl (#21029)
mkoker
authored
ggml: Vulkan build, Linux -- output error string for errno on fork failure (#20868) (#20904)
tomoverlund
authored
ggml : deprecate GGML_OP_ADD1 (#21363)
ggerganov
authored
server : fix restore for checkpoints with pos_min == 0 (#21510)
ggerganov
authored
llama: remove per-arch tensor name lists (#21531)
JohannesGaessler
authored
unicode : add custom Qwen2 regex handler to fix segfault on long input (#21257)

nhs000
and
aldehir
authored
llama-server: fix model params not propagated (#21509)
taronaeo
authored
CUDA: check for buffer overlap before fusing (#21566)
am17an
authored
ggml-webgpu: parameterize submission size and add iOS specific limits (#21533)
reeselevine
authored
kv-cache : support attention rotation for heterogeneous iSWA (#21513)
ggerganov
authored
gguf-py : fix missing comma after bad merge in tensor-mapping (#21558)
danbev
authored
ggml-cuda: ds_read_b128 for q4_0 and q4_1 mmq kernels (#21168)

authored

Commits on Apr 8, 2026

Commits on Apr 9, 2026

Commits on Apr 10, 2026