fix scratch size of softmax (llama/8642)

ggml: fix compile error for RISC-V (llama/8623)

CUDA: MMQ code deduplication + iquant support (llama/8495)

gguf : handle null name during init (llama/8587)

ggml : fix quant dot product with odd number of blocks (llama/8549)

ggml : add friendlier error message to fopen errors (llama/8575)

CUDA: fix partial offloading for ne0 % 256 != 0 (llama/8572)

Add Ascend NPU backend (llama/6035)

make/cmake: add missing force MMQ/cuBLAS for HIP (llama/8515)

Refactor lora adapter support (llama/8332)

add concat through dim 1/2 (llama/8483)

Vulkan MMQ Fix (llama/8479)

vulkan : cmake integration (llama/8119)

metal : template-ify some of the kernels (llama/8447)

ggml : minor naming changes (llama/8433)

fix the mul_mat_id ut issues (llama/8427)

ggml : add NVPL BLAS support (ggml/8329) (llama/8425)

cuda : suppress 'noreturn' warn in no_device_code (llama/8414)

CUDA: optimize and refactor MMQ (llama/8416)

Use multi_ptr to clean up deprecated warnings (llama/8256)

ggml : move sgemm sources to llamafile subfolder (llama/8394)

ggml : add AArch64 optimized GEMV and GEMM Q4 kernels (llama/5780)

sycl : Reenabled mmvq path for the SYCL Nvidia Backend (llama/8372)

sycl : fix powf call in device code (llama/8368)

ggml : loop tiling optimizations for scalar path (ggml/898)

ggml: add support for float16 input tensors in pooling operations (ggml/895)

vulkan : initialize vk_buffer_struct members to VK_NULL_HANDLE (ggml/893)

whisper : use vulkan as gpu backend when available (#2302)

ggml : sync sycl (skip) (#0)

ggml : remove unnecessary UNUSED macro call (ggml/880)

cmake : add GGML_BUILD and GGML_SHARED macro definitions (llama/8281)

Enabled more data types for oneMKL gemm_batch (llama/8236)

CUDA: MMQ support for iq4_nl, iq4_xs (llama/8278)

CUDA: revert part of the RDNA1 optimizations (llama/8309)

CUDA: fix MMQ stream-k rounding if ne00 % 128 != 0 (llama/8311)

Fix WARP_SIZE=16 bug of Intel GPU (llama/8266)

rm get_work_group_size() by local cache for performance (llama/8286)

Define and optimize RDNA1 (llama/8085)

fix typo (llama/8267)

Removes multiple newlines at the end of files that is breaking the editorconfig step of CI. (llama/8258)

cuda : update supports_op for matrix multiplication (llama/8245)

Fix win build conflict of math library (llama/8230)

Fix the sub group size of Intel (llama/8106)

CUDA: refactor and optimize IQ MMVQ (llama/8215)

Update SYCL-Rope op and Refactor (llama/8157)

CUDA: fix MMQ stream-k for --split-mode row (llama/8167)

feat: cuda implementation for `ggml_conv_transpose_1d` (ggml/854)

ggml : add GGML_CUDA_USE_GRAPHS option, restore GGML_CUDA_FORCE_CUBLAS (cmake) (llama/8140)

whisper : reorganize source code + improve CMake (#2256)