HIP/CUDA: set the paramerter value in maintain_cuda_graph instead of replaceing it. (llama/12209)

ggml : upgrade init_tensor API to return a ggml_status (llama/11854)

cuda/vulkan: specify fp32-only support for some operations in supports_op (ggml/1129)

cuda/cpu: Increase support for fp16 unary operations (ggml/1125)

CUDA: app option to compile without FlashAttention (llama/12025)

cuda: Add Q5_1, Q5_0, Q4_1 and Q4_0 to F32 conversion support. (llama/12000)

MUSA: support ARM64 and enable dp4a .etc (llama/11843)

musa: bump MUSA SDK version to rc3.1.1 (llama/11822)

HIP: Switch to std::vector in rocblas version check (llama/11820)

CUDA: fix CUDART_VERSION checks (llama/11821)

CUDA: use arch list for compatibility check (llama/11775)

CUDA: support for mat. mul. with ne03 != ne13 (llama/11656)

CUDA: non-contiguous (RMS) norm support (llama/11659)

HIP: add GGML_CUDA_CC_IS_* for amd familys as increasing cc archtectures for amd gpus are not supersets of eatch other (llama/11601)

HIP: Prepare reduction operators for wave 64

CUDA/HIP: add warp_size to cuda_device_info

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

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

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)

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

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

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

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

CUDA: add BF16 support (llama/11093)

CUDA: rename macros to avoid conflicts with WinAPI (llama/10736)

ggml : refactor online repacking (llama/10446)

Add some minimal optimizations for CDNA (llama/10498)

ggml : add support for dynamic loading of backends (llama/10469)

CUDA: fix MMV kernel being used for FP16 src1 (llama/10357)

CUDA: remove DMMV, consolidate F16 mult mat vec (llama/10318)

ggml : build backends as libraries (llama/10256)