vulkan: optimize flash attention split_k_reduce (llama/14554)

vulkan : fix rope with partial rotation and non-cont src (llama/14582)

cuda : fix rope with partial rotation and non-cont src (llama/14580)

CUDA: add bilinear interpolation for upscale (llama/14563)

musa: fix build warnings (unused variable) (llama/14561)

CUDA: add bf16 and i32 to getrows (llama/14529)

vulkan: increase LOAD_VEC_A to 8 (IQ1/IQ2) or 4 (IQ3) (llama/14485)

vulkan: fix rms_norm+mul fusion (llama/14545)

vulkan: Handle updated FA dim2/3 definition (llama/14518)

opencl: add GELU_ERF (llama/14476)

metal : disable fast math in all quantize kernels (llama/14528)

CANN: Replace aclrtMemsetSync with aclnnInplaceZero operator (llama/14002)

ggml : implement GEGLU_ERF and GEGLU_QUICK ops (llama/14445)

opencl : broadcast for soft_max (llama/14510)

vulkan: support mixed/deepseekR1 FA head sizes (llama/14509)

ggml: backward pass for split swiglu (llama/14483)

Fix conditional enabling following arch checks for ggml-sycl (llama/14504)

kv-cache : use ggml_set_rows (llama/14285)

ggml : fix FA mask dim 2 and 3 (llama/14505)

CUDA: add dynamic shared mem to softmax, refactor general usage (llama/14497)

llama : initial Mamba-2 support (llama/9126)

CUDA: add softmax broadcast (llama/14475)

CUDA: broadcasting for FlashAttention mask (llama/14500)

vulkan: support softmax/FA batch and broadcast (llama/14449)

ggml : support bcast ggml_soft_max_ext, ggml_flash_attn_ext (llama/14435)

opencl : fix possible buffer overflow in dump_tensor (llama/14490)

opencl : skip empty nodes on cgraph compute (llama/14491)

opencl : update upscale to support align corners (llama/14488)

ggml : Callback before abort (llama/14481)

ci : disable fast-math for Metal GHA CI (llama/14478)

CANN: update aclnnGroupedMatmulV2 to aclnnGroupedMatmulV3 (llama/14411)

vulkan: Split large mul_mat_id to fit in shared memory (llama/14451)

add GELU_ERF (llama/14455)

vulkan : implement bilinear interpolation for ggml_upscale/ggml_interpolate (ggml/1291)

vulkan : implement ggml_roll (ggml/1290)

ggml : add version function to get lib version (ggml/1286)

server : add dtw.params for v3-large-turbo (#3307)

feat: support vad for addon.node (#3301)

sync : ggml

talk-llama : sync llama.cpp

ggml : remove trailing whitespace (llama/0)

opencl : add GEGLU, REGLU, SWIGLU (llama/14456)

Add Conv2d for CPU (llama/14388)

metal : disable fast-math for some cpy kernels (llama/14460)

ggml-cpu: sycl: Re-enable exp f16 (llama/14462)

cmake : Remove redundant include path in CMakeLists.txt (llama/14452)

scripts : make the shell scripts cross-platform (llama/14341)

SYCL: disable faulty fp16 exp kernel (llama/14395)

ggml : fix unmerged GGML_FPxx_TO_FPxx refactoring (llama/14443)

ggml : implement REGLU/GEGLU/SWIGLU ops (llama/14158)