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

vulkan: use smaller combined allocations to avoid fragmentation (llama/11551)

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

ggml : remove return from ggml_gallocr_allocate_node (ggml/1048)

ggml: new optimization interface (ggml/988)

ggml : move more prints to the ggml log system (llama/9839)

ggml-alloc : remove buffer_id from leaf_alloc (ggml/987)

ggml-alloc : fix list of allocated tensors with GGML_ALLOCATOR_DEBUG (llama/9573)

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

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

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