vulkan: implement initial support for IQ2 and IQ3 quantizations (llama/11360)

vulkan: Catch pipeline creation failure and print an error message (llama/11436)

HIP: Supress transformation warning in softmax.cu

HIP: Only call rocblas_initialize on rocblas versions with the multiple instantation bug (llama/11080)

cmake : don't fail on `GGML_CPU=OFF` (llama/11457)

SYCL : SOFTMAX F16 mask support and other fixes (llama/11261)

AMD: parse the architecture as supplied by gcnArchName (llama/11244)

metal: Handle null returned from MTLCreateSystemDefaultDevice() (llama/11441)

metal : use residency sets (llama/11427)

cmake: add ggml find package (llama/11369)

vulkan: compile shaders on-demand (llama/11406)

Hip: disable VMM on hip as it seams that it dosent work in some configurations (llama/11420)

hip : Add hipGraph and VMM support to ROCM (llama/11362)

CUDA: fix FP16 cuBLAS GEMM (llama/11396)

rocBLAS: Avoid fp32->fp16->fp32 conversion on cdna (llama/11356)

CPU/CUDA: fix (GQA) mul mat back, add CUDA support (llama/11380)

Vulkan-run-test: fix mmq_wg_denoms (llama/11343)

vulkan: sort shaders for more deterministic binary (llama/11315)

vulkan: fix diag_mask_inf (llama/11323)

rpc : better caching of the base buffer pointer (llama/11331)

metal : fix out-of-bounds write (llama/11314)

vulkan: fix coopmat2 validation failures (llama/11284)

SYCL: Introducing memory host pool (llama/11251)

cmake : add sanitizer flags for llama.cpp (llama/11279)

vulkan: fix coopmat2 flash attention for non-contiguous inputs (llama/11281)

rpc : early register backend devices (llama/11262)

vulkan: support copy from f32 to q4_0/q4_1/q5_0/q5_1/q8_0/iq4_nl (llama/11166)

vulkan: optimize coopmat2 q4_k/q5_k dequant functions. (llama/11206)

vulkan: optimize coopmat2 q2_k dequant function (llama/11130)

CUDA: backwards pass for misc. ops, add tests (llama/11257)

ggml: aarch64: implement SVE kernels for q4_K_q8_K vector dot (llama/11227)

vulkan: scale caching for k quants + misc fixes (llama/11081)

fix: ggml: fix vulkan-shaders-gen build (llama/10448)

RoPE: fix back, CUDA support for back + noncont. (llama/11240)

SYCL: Add gated linear attention kernel (llama/11175)

ggml : add option to not print stack on abort (ggml/1081)

ggml-cpu : fix ggml_graph_compute_thread did not terminate on abort. (ggml/1065)

GGUF: C++ refactor, backend support, misc fixes (skip) (llama/11030)

ggml : add opencl backend (skip) (llama/10693)

cuda : CUDA Graph Compute Function Refactor (precursor for performance improvements) (llama/11042)

ggml : do not define GGML_USE_CUDA when building with GGML_BACKEND_DL (llama/11211)

Vulkan: Fix float16 use on devices without float16 support + fix subgroup_size_control validation error (llama/11161)

llama: add support for QRWKV6 model architecture (llama/11001)

SYCL: Refactor ggml_sycl_compute_forward (llama/11121)

fix: add missing msg in static_assert (llama/11143)

llamafile : ppc64le MMA INT8 implementation (llama/10912)

Disable GL_KHR_cooperative_matrix Vulkan extension if not available. (llama/11117)

fix: Vulkan shader gen binary path when Cross-compiling (llama/11096)

GGUF: C++ refactor, backend support, misc fixes (llama/11030)

ggml-backend : only offload from host buffers (fix) (llama/11124)