Merge branch 'layla-build' of https://github.com/l3utterfly/llama.cpp …

adjusted loading clip models to support file descriptor + offset

fix: use the correct mode when opening files with file descriptors

add support for ARM CPU variants on Android and Apple platforms

added ability to build different cpu variants for phones

Merge branch 'layla-build' of https://github.com/l3utterfly/llama.cpp …

Revert changes in ggml_backend_opencl_synchronize to prevent crashes …

Merge branch 'layla-build' of https://github.com/l3utterfly/llama.cpp …

Fix formatting and restore nrows assignment for ARM_FEATURE_MATMUL_IN…

Update optimization flags for ARM architecture in CMakeLists.txt

Uncomment ARM_FEATURE_MATMUL_INT8 assertions and related code in ggml…

Comment out nrows assignment for ARM_FEATURE_MATMUL_INT8 in CPU type …

merge changes

Refactor constructor and method parameters for consistency and readab…

Fix missing endif() for BLAS not found error message in CMakeLists.txt

Merge pull request #69 from l3utterfly/merge

Merge branch 'layla-build' into merge

opencl: add backend_synchronize (ggml-org#13939)

opencl: add backend_synchronize (ggml-org#13939)

- Fix include paths for nlohmann/json in minja headers

Merge branch 'ggml-hexagon' into layla-build

Merge pull request #68 from zhouwg/self-build

CANN: Support MOE Model MUL_MAT_ID (ggml-org#13042)

disable int8_matmul optimisations for non Q4_0 models

disable int8_matmul optimisations for non Q4_0 models

Enhance optimization flags for Hexagon backend to improve performance…

Enhance cross-compilation support for ARM by updating compiler comman…

Merge branch 'layla-build' of https://github.com/l3utterfly/llama.cpp …