CPU/CUDA: Gemma 2 FlashAttention support (llama/8542)

llama : simplify Mamba with advanced batch splits (llama/8526)

ggml : move rope type enum to ggml.h (llama/8949)

feat: ref. cross entropy, add CUDA, fix grad test (ggml/929)

examples: add MNIST training + missing ops

feat: add new `sin` and `cos` operators (ggml/919)

ggml : add CANN backend (llama/0)

ggml : add epsilon as a parameter for group_norm (llama/8818)

Fix conversion of unnormalized BF16->BF16 weights (llama/7843)

feat: Support Moore Threads GPU (llama/8383)

metal : add abort callback (ggml/905)

ggml : move c parameter comment to ggml_rope_ext (ggml/901)

ggml : reduce hash table reset cost (llama/8698)

ggml : add and use ggml_cpu_has_llamafile() (llama/8664)

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

Add Ascend NPU backend (llama/6035)

ggml : minor naming changes (llama/8433)

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

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

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