Sync master with upstream release b6848#305

Merged

jan-service-account merged 10 commits intodevfrom

update-dev-from-master-2025-10-27-00-38

Oct 27, 2025

Commits on Oct 25, 2025

convert : handle mmproj filename/path properly (#16760)

Galunid
and
CISC
authored

Commits on Oct 26, 2025

vulkan: deduplicate Microsoft Direct3D12 devices (#16689)
giladgd
authored
CUDA: General GEMV fusion (#16715)
am17an
authored
docs : add Jamba to Text-only models list (#16778)
amirai21
authored
model : set res->t_embd in SmallThinker models (#16782)
CISC
authored
graph : add clamping to ffn_moe_weights_sum to avoid div-by-zero (#16655)
CISC
authored
convert : enable expert group selection for all models with it (#16691)
CISC
authored
ggml: fix cuda kernel launch configuration for k_compute_batched_ptrs to support large batch (#16744)

leejet
and
JohannesGaessler
authored
cuda : use fast copy when src and dst are of different type and contiguous (#16789)
CISC
authored
ggml-alloc : make gallocr prefer chunks that allow memory reuse (#16788)
Acly
authored