npm - @fugood/llama.node - Versions diffs - 0.3.0 → 0.3.2 - Mend

@fugood/llama.node 0.3.0 → 0.3.2

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.

Files changed (187) hide show

package/CMakeLists.txt +1 -10
package/bin/darwin/arm64/llama-node.node +0 -0
package/bin/darwin/x64/llama-node.node +0 -0
package/bin/linux/arm64/llama-node.node +0 -0
package/bin/linux/x64/llama-node.node +0 -0
package/bin/linux-vulkan/arm64/llama-node.node +0 -0
package/bin/linux-vulkan/x64/llama-node.node +0 -0
package/bin/win32/arm64/llama-node.node +0 -0
package/bin/win32/arm64/node.lib +0 -0
package/bin/win32/x64/llama-node.node +0 -0
package/bin/win32/x64/node.lib +0 -0
package/bin/win32-vulkan/arm64/llama-node.node +0 -0
package/bin/win32-vulkan/arm64/node.lib +0 -0
package/bin/win32-vulkan/x64/llama-node.node +0 -0
package/bin/win32-vulkan/x64/node.lib +0 -0
package/package.json +6 -4
package/src/LlamaCompletionWorker.cpp +6 -6
package/src/LlamaContext.cpp +7 -9
package/src/common.hpp +2 -1
package/src/llama.cpp/.github/workflows/build.yml +98 -24
package/src/llama.cpp/.github/workflows/close-issue.yml +5 -0
package/src/llama.cpp/.github/workflows/docker.yml +43 -34
package/src/llama.cpp/.github/workflows/nix-ci-aarch64.yml +7 -0
package/src/llama.cpp/.github/workflows/nix-ci.yml +7 -0
package/src/llama.cpp/.github/workflows/python-check-requirements.yml +2 -4
package/src/llama.cpp/.github/workflows/python-type-check.yml +3 -1
package/src/llama.cpp/.github/workflows/server.yml +7 -0
package/src/llama.cpp/CMakeLists.txt +20 -8
package/src/llama.cpp/common/CMakeLists.txt +12 -10
package/src/llama.cpp/common/arg.cpp +2006 -0
package/src/llama.cpp/common/arg.h +77 -0
package/src/llama.cpp/common/common.cpp +496 -1632
package/src/llama.cpp/common/common.h +161 -63
package/src/llama.cpp/common/console.cpp +3 -0
package/src/llama.cpp/common/log.cpp +401 -0
package/src/llama.cpp/common/log.h +66 -698
package/src/llama.cpp/common/ngram-cache.cpp +3 -0
package/src/llama.cpp/common/sampling.cpp +348 -350
package/src/llama.cpp/common/sampling.h +62 -139
package/src/llama.cpp/common/stb_image.h +5990 -6398
package/src/llama.cpp/common/train.cpp +2 -0
package/src/llama.cpp/docs/build.md +36 -1
package/src/llama.cpp/examples/CMakeLists.txt +0 -1
package/src/llama.cpp/examples/baby-llama/baby-llama.cpp +1 -2
package/src/llama.cpp/examples/batched/batched.cpp +39 -55
package/src/llama.cpp/examples/batched-bench/batched-bench.cpp +34 -44
package/src/llama.cpp/examples/convert-llama2c-to-ggml/convert-llama2c-to-ggml.cpp +55 -52
package/src/llama.cpp/examples/cvector-generator/cvector-generator.cpp +15 -15
package/src/llama.cpp/examples/cvector-generator/pca.hpp +3 -13
package/src/llama.cpp/examples/embedding/embedding.cpp +143 -87
package/src/llama.cpp/examples/eval-callback/eval-callback.cpp +33 -33
package/src/llama.cpp/examples/export-lora/export-lora.cpp +36 -35
package/src/llama.cpp/examples/gbnf-validator/gbnf-validator.cpp +14 -39
package/src/llama.cpp/examples/gen-docs/CMakeLists.txt +5 -0
package/src/llama.cpp/examples/gen-docs/gen-docs.cpp +83 -0
package/src/llama.cpp/examples/gguf-split/gguf-split.cpp +58 -39
package/src/llama.cpp/examples/gritlm/gritlm.cpp +34 -27
package/src/llama.cpp/examples/imatrix/imatrix.cpp +59 -62
package/src/llama.cpp/examples/infill/infill.cpp +117 -132
package/src/llama.cpp/examples/llama-bench/llama-bench.cpp +265 -58
package/src/llama.cpp/examples/llama.android/llama/src/main/cpp/llama-android.cpp +29 -22
package/src/llama.cpp/examples/llava/CMakeLists.txt +7 -0
package/src/llama.cpp/examples/llava/clip.cpp +685 -150
package/src/llama.cpp/examples/llava/clip.h +11 -2
package/src/llama.cpp/examples/llava/llava-cli.cpp +47 -58
package/src/llama.cpp/examples/llava/llava.cpp +110 -24
package/src/llama.cpp/examples/llava/llava.h +2 -3
package/src/llama.cpp/examples/llava/minicpmv-cli.cpp +323 -0
package/src/llama.cpp/examples/llava/requirements.txt +1 -0
package/src/llama.cpp/examples/lookahead/lookahead.cpp +42 -43
package/src/llama.cpp/examples/lookup/lookup-create.cpp +10 -8
package/src/llama.cpp/examples/lookup/lookup-stats.cpp +23 -22
package/src/llama.cpp/examples/lookup/lookup.cpp +40 -43
package/src/llama.cpp/examples/main/main.cpp +210 -262
package/src/llama.cpp/examples/parallel/parallel.cpp +49 -49
package/src/llama.cpp/examples/passkey/passkey.cpp +42 -50
package/src/llama.cpp/examples/perplexity/perplexity.cpp +187 -200
package/src/llama.cpp/examples/quantize/CMakeLists.txt +1 -1
package/src/llama.cpp/examples/quantize/quantize.cpp +27 -9
package/src/llama.cpp/examples/quantize-stats/quantize-stats.cpp +2 -3
package/src/llama.cpp/examples/retrieval/retrieval.cpp +49 -44
package/src/llama.cpp/examples/rpc/rpc-server.cpp +24 -1
package/src/llama.cpp/examples/save-load-state/save-load-state.cpp +32 -35
package/src/llama.cpp/examples/server/CMakeLists.txt +3 -5
package/src/llama.cpp/examples/server/server.cpp +1027 -1073
package/src/llama.cpp/examples/server/tests/requirements.txt +2 -1
package/src/llama.cpp/examples/server/utils.hpp +107 -105
package/src/llama.cpp/examples/simple/simple.cpp +35 -41
package/src/llama.cpp/examples/speculative/speculative.cpp +129 -103
package/src/llama.cpp/examples/sycl/run-llama2.sh +10 -19
package/src/llama.cpp/examples/sycl/win-run-llama2.bat +1 -1
package/src/llama.cpp/examples/tokenize/tokenize.cpp +25 -27
package/src/llama.cpp/ggml/CMakeLists.txt +14 -3
package/src/llama.cpp/ggml/include/ggml-alloc.h +3 -3
package/src/llama.cpp/ggml/include/ggml-backend.h +145 -60
package/src/llama.cpp/ggml/include/ggml-blas.h +3 -3
package/src/llama.cpp/ggml/include/ggml-cann.h +15 -19
package/src/llama.cpp/ggml/include/ggml-cuda.h +16 -16
package/src/llama.cpp/ggml/include/ggml-metal.h +5 -8
package/src/llama.cpp/ggml/include/ggml-rpc.h +5 -5
package/src/llama.cpp/ggml/include/ggml-sycl.h +8 -8
package/src/llama.cpp/ggml/include/ggml-vulkan.h +7 -7
package/src/llama.cpp/ggml/include/ggml.h +293 -186
package/src/llama.cpp/ggml/src/CMakeLists.txt +86 -44
package/src/llama.cpp/ggml/src/ggml-aarch64.c +2135 -1119
package/src/llama.cpp/ggml/src/ggml-alloc.c +6 -0
package/src/llama.cpp/ggml/src/ggml-backend-impl.h +152 -70
package/src/llama.cpp/ggml/src/{ggml-backend.c → ggml-backend.cpp} +606 -286
package/src/llama.cpp/ggml/src/ggml-blas.cpp +9 -10
package/src/llama.cpp/ggml/src/ggml-cann/acl_tensor.cpp +4 -27
package/src/llama.cpp/ggml/src/ggml-cann/acl_tensor.h +32 -4
package/src/llama.cpp/ggml/src/ggml-cann/aclnn_ops.cpp +179 -41
package/src/llama.cpp/ggml/src/ggml-cann/common.h +1 -0
package/src/llama.cpp/ggml/src/ggml-cann/kernels/CMakeLists.txt +2 -1
package/src/llama.cpp/ggml/src/ggml-cann/kernels/ascendc_kernels.h +2 -0
package/src/llama.cpp/ggml/src/ggml-cann/kernels/quantize_float_to_q4_0.cpp +278 -0
package/src/llama.cpp/ggml/src/ggml-cann.cpp +215 -216
package/src/llama.cpp/ggml/src/ggml-common.h +20 -0
package/src/llama.cpp/ggml/src/ggml-cpu-impl.h +614 -0
package/src/llama.cpp/ggml/src/ggml-cuda/vendors/cuda.h +14 -0
package/src/llama.cpp/ggml/src/ggml-cuda/vendors/hip.h +178 -0
package/src/llama.cpp/ggml/src/ggml-cuda/vendors/musa.h +134 -0
package/src/llama.cpp/ggml/src/ggml-impl.h +49 -603
package/src/llama.cpp/ggml/src/ggml-kompute.cpp +4 -24
package/src/llama.cpp/ggml/src/ggml-quants.c +972 -92
package/src/llama.cpp/ggml/src/ggml-quants.h +15 -0
package/src/llama.cpp/ggml/src/ggml-rpc.cpp +116 -66
package/src/llama.cpp/ggml/src/ggml-sycl/backend.hpp +3 -0
package/src/llama.cpp/ggml/src/ggml-sycl/common.cpp +11 -0
package/src/llama.cpp/ggml/src/ggml-sycl/common.hpp +52 -0
package/src/llama.cpp/ggml/src/ggml-sycl/conv.cpp +99 -0
package/src/llama.cpp/ggml/src/ggml-sycl/conv.hpp +21 -0
package/src/llama.cpp/ggml/src/ggml-sycl/convert.cpp +57 -57
package/src/llama.cpp/ggml/src/ggml-sycl/convert.hpp +1 -1
package/src/llama.cpp/ggml/src/ggml-sycl/dequantize.hpp +106 -106
package/src/llama.cpp/ggml/src/ggml-sycl/dmmv.cpp +4 -4
package/src/llama.cpp/ggml/src/ggml-sycl/dpct/helper.hpp +16 -3
package/src/llama.cpp/ggml/src/ggml-sycl/gemm.hpp +101 -0
package/src/llama.cpp/ggml/src/ggml-sycl/im2col.cpp +125 -0
package/src/llama.cpp/ggml/src/ggml-sycl/im2col.hpp +23 -0
package/src/llama.cpp/ggml/src/ggml-sycl/mmvq.cpp +1 -1
package/src/llama.cpp/ggml/src/ggml-sycl/norm.cpp +6 -3
package/src/llama.cpp/ggml/src/ggml-sycl/presets.hpp +2 -0
package/src/llama.cpp/ggml/src/ggml-sycl/rope.cpp +1 -1
package/src/llama.cpp/ggml/src/ggml-sycl/tsembd.cpp +71 -0
package/src/llama.cpp/ggml/src/ggml-sycl/tsembd.hpp +21 -0
package/src/llama.cpp/ggml/src/ggml-sycl.cpp +97 -169
package/src/llama.cpp/ggml/src/ggml-vulkan.cpp +1508 -1124
package/src/llama.cpp/ggml/src/ggml.c +3001 -1647
package/src/llama.cpp/ggml/src/llamafile/sgemm.cpp +192 -0
package/src/llama.cpp/ggml/src/vulkan-shaders/CMakeLists.txt +2 -0
package/src/llama.cpp/ggml/src/vulkan-shaders/vulkan-shaders-gen.cpp +88 -40
package/src/llama.cpp/include/llama.h +241 -264
package/src/llama.cpp/models/ggml-vocab-chameleon.gguf.inp +112 -0
package/src/llama.cpp/models/ggml-vocab-chameleon.gguf.out +46 -0
package/src/llama.cpp/requirements/requirements-convert_legacy_llama.txt +1 -1
package/src/llama.cpp/src/llama-grammar.cpp +721 -122
package/src/llama.cpp/src/llama-grammar.h +120 -15
package/src/llama.cpp/src/llama-impl.h +156 -1
package/src/llama.cpp/src/llama-sampling.cpp +1375 -303
package/src/llama.cpp/src/llama-sampling.h +20 -47
package/src/llama.cpp/src/llama-vocab.cpp +343 -120
package/src/llama.cpp/src/llama-vocab.h +33 -17
package/src/llama.cpp/src/llama.cpp +4247 -1525
package/src/llama.cpp/src/unicode-data.cpp +6 -4
package/src/llama.cpp/src/unicode-data.h +4 -4
package/src/llama.cpp/src/unicode.cpp +15 -7
package/src/llama.cpp/tests/CMakeLists.txt +3 -0
package/src/llama.cpp/tests/test-arg-parser.cpp +131 -0
package/src/llama.cpp/tests/test-backend-ops.cpp +1592 -289
package/src/llama.cpp/tests/test-barrier.cpp +93 -0
package/src/llama.cpp/tests/test-grad0.cpp +187 -70
package/src/llama.cpp/tests/test-grammar-integration.cpp +23 -38
package/src/llama.cpp/tests/test-grammar-parser.cpp +6 -4
package/src/llama.cpp/tests/test-json-schema-to-grammar.cpp +6 -4
package/src/llama.cpp/tests/test-llama-grammar.cpp +9 -8
package/src/llama.cpp/tests/test-log.cpp +39 -0
package/src/llama.cpp/tests/test-quantize-fns.cpp +6 -0
package/src/llama.cpp/tests/test-rope.cpp +1 -1
package/src/llama.cpp/tests/test-sampling.cpp +157 -98
package/src/llama.cpp/tests/test-tokenizer-0.cpp +55 -35
package/patches/llama.patch +0 -22
package/src/llama.cpp/.github/workflows/bench.yml +0 -310
package/src/llama.cpp/common/grammar-parser.cpp +0 -536
package/src/llama.cpp/common/grammar-parser.h +0 -29
package/src/llama.cpp/examples/benchmark/CMakeLists.txt +0 -6
package/src/llama.cpp/examples/benchmark/benchmark-matmult.cpp +0 -275

package/src/llama.cpp/ggml/src/CMakeLists.txt CHANGED Viewed

@@ -26,6 +26,9 @@ if (NOT MSVC)
     endif()
 endif()
+unset(GGML_EXTRA_LIBS_PRIVATE)
+unset(GGML_EXTRA_LIBS_PUBLIC)
 if (APPLE AND GGML_ACCELERATE)
     find_library(ACCELERATE_FRAMEWORK Accelerate)
     if (ACCELERATE_FRAMEWORK)
@@ -35,7 +38,7 @@ if (APPLE AND GGML_ACCELERATE)
         add_compile_definitions(ACCELERATE_NEW_LAPACK)
         add_compile_definitions(ACCELERATE_LAPACK_ILP64)
-        set(GGML_EXTRA_LIBS ${GGML_EXTRA_LIBS} ${ACCELERATE_FRAMEWORK})
+        list(APPEND GGML_EXTRA_LIBS_PRIVATE ${ACCELERATE_FRAMEWORK})
     else()
         message(WARNING "Accelerate framework not found")
     endif()
@@ -87,7 +90,7 @@ if (GGML_METAL)
             COMMENT "Generate assembly for embedded Metal library"
         )
-        set(GGML_SOURCES_METAL ${GGML_SOURCES_METAL} ${METALLIB_EMBED_ASM})
+        list(APPEND GGML_SOURCES_METAL ${METALLIB_EMBED_ASM})
     else()
         if (GGML_METAL_SHADER_DEBUG)
             # custom command to do the following:
@@ -132,7 +135,7 @@ if (GGML_METAL)
             )
     endif() # GGML_METAL_EMBED_LIBRARY
-    set(GGML_EXTRA_LIBS ${GGML_EXTRA_LIBS}
+    list(APPEND GGML_EXTRA_LIBS_PRIVATE
         ${FOUNDATION_LIBRARY}
         ${METAL_FRAMEWORK}
         ${METALKIT_FRAMEWORK}
@@ -157,11 +160,11 @@ if (GGML_OPENMP)
         add_compile_definitions(GGML_USE_OPENMP)
-        set(GGML_EXTRA_LIBS ${GGML_EXTRA_LIBS} OpenMP::OpenMP_C OpenMP::OpenMP_CXX)
+        list(APPEND GGML_EXTRA_LIBS_PRIVATE OpenMP::OpenMP_C OpenMP::OpenMP_CXX)
         if (GGML_MUSA)
-            set(GGML_EXTRA_INCLUDES ${GGML_EXTRA_INCLUDES} "/usr/lib/llvm-10/include/openmp")
-            set(GGML_EXTRA_LIBS ${GGML_EXTRA_LIBS} "/usr/lib/llvm-10/lib/libomp.so")
+            list(APPEND GGML_EXTRA_INCLUDES     "/usr/lib/llvm-10/include/openmp")
+            list(APPEND GGML_EXTRA_LIBS_PRIVATE "/usr/lib/llvm-10/lib/libomp.so")
         endif()
     else()
         message(WARNING "OpenMP not found")
@@ -244,8 +247,8 @@ if (GGML_BLAS)
         set(GGML_HEADERS_BLAS ../include/ggml-blas.h)
         set(GGML_SOURCES_BLAS ggml-blas.cpp)
-        set(GGML_EXTRA_LIBS     ${GGML_EXTRA_LIBS}     ${BLAS_LIBRARIES})
-        set(GGML_EXTRA_INCLUDES ${GGML_EXTRA_INCLUDES} ${BLAS_INCLUDE_DIRS})
+        list(APPEND GGML_EXTRA_LIBS_PRIVATE ${BLAS_LIBRARIES})
+        list(APPEND GGML_EXTRA_INCLUDES     ${BLAS_INCLUDE_DIRS})
     else()
         message(WARNING "BLAS not found, please refer to "
         "https://cmake.org/cmake/help/latest/module/FindBLAS.html#blas-lapack-vendors"
@@ -326,7 +329,7 @@ if (GGML_CUDA)
         add_compile_definitions(K_QUANTS_PER_ITERATION=${GGML_CUDA_KQUANTS_ITER})
         add_compile_definitions(GGML_CUDA_PEER_MAX_BATCH_SIZE=${GGML_CUDA_PEER_MAX_BATCH_SIZE})
-        if (GGML_CUDA_USE_GRAPHS)
+        if (GGML_CUDA_GRAPHS)
             add_compile_definitions(GGML_CUDA_USE_GRAPHS)
         endif()
@@ -361,26 +364,26 @@ if (GGML_CUDA)
         if (GGML_MUSA)
             set_source_files_properties(${GGML_SOURCES_CUDA} PROPERTIES LANGUAGE CXX)
             foreach(SOURCE ${GGML_SOURCES_CUDA})
-                set_property(SOURCE ${SOURCE} PROPERTY COMPILE_FLAGS "-x musa -mtgpu --cuda-gpu-arch=mp_22")
+                set_property(SOURCE ${SOURCE} PROPERTY COMPILE_FLAGS "-x musa -mtgpu --cuda-gpu-arch=mp_21 --cuda-gpu-arch=mp_22")
             endforeach()
         endif()
         if (GGML_STATIC)
             if (WIN32)
                 # As of 12.3.1 CUDA Toolkit for Windows does not offer a static cublas library
-                set(GGML_EXTRA_LIBS ${GGML_EXTRA_LIBS} CUDA::cudart_static CUDA::cublas CUDA::cublasLt)
+                list(APPEND GGML_EXTRA_LIBS_PRIVATE CUDA::cudart_static CUDA::cublas CUDA::cublasLt)
             else ()
                 if (GGML_MUSA)
-                    set(GGML_EXTRA_LIBS ${GGML_EXTRA_LIBS} MUSA::musart_static MUSA::mublas_static)
+                    list(APPEND GGML_EXTRA_LIBS_PRIVATE MUSA::musart_static MUSA::mublas_static)
                 else()
-                    set(GGML_EXTRA_LIBS ${GGML_EXTRA_LIBS} CUDA::cudart_static CUDA::cublas_static CUDA::cublasLt_static)
+                    list(APPEND GGML_EXTRA_LIBS_PRIVATE CUDA::cudart_static CUDA::cublas_static CUDA::cublasLt_static)
                 endif()
             endif()
         else()
             if (GGML_MUSA)
-                set(GGML_EXTRA_LIBS ${GGML_EXTRA_LIBS} MUSA::musart MUSA::mublas)
+                list(APPEND GGML_EXTRA_LIBS_PRIVATE MUSA::musart MUSA::mublas)
             else()
-                set(GGML_EXTRA_LIBS ${GGML_EXTRA_LIBS} CUDA::cudart CUDA::cublas CUDA::cublasLt)
+                list(APPEND GGML_EXTRA_LIBS_PRIVATE CUDA::cudart CUDA::cublas CUDA::cublasLt)
             endif()
         endif()
@@ -388,9 +391,9 @@ if (GGML_CUDA)
             # No VMM requested, no need to link directly with the cuda driver lib (libcuda.so)
         else()
             if (GGML_MUSA)
-                set(GGML_EXTRA_LIBS ${GGML_EXTRA_LIBS} MUSA::musa_driver) # required by muDeviceGetAttribute(), muMemGetAllocationGranularity(...), ...
+                list(APPEND GGML_EXTRA_LIBS_PRIVATE MUSA::musa_driver) # required by muDeviceGetAttribute(), muMemGetAllocationGranularity(...), ...
             else()
-                set(GGML_EXTRA_LIBS ${GGML_EXTRA_LIBS} CUDA::cuda_driver) # required by cuDeviceGetAttribute(), cuMemGetAllocationGranularity(...), ...
+                list(APPEND GGML_EXTRA_LIBS_PRIVATE CUDA::cuda_driver) # required by cuDeviceGetAttribute(), cuMemGetAllocationGranularity(...), ...
             endif()
         endif()
     else()
@@ -495,7 +498,7 @@ if (GGML_HIPBLAS)
     if (CXX_IS_HIPCC)
         set_source_files_properties(${GGML_SOURCES_ROCM} PROPERTIES LANGUAGE CXX)
-        set(GGML_EXTRA_LIBS ${GGML_EXTRA_LIBS} hip::device)
+        list(APPEND GGML_EXTRA_LIBS_PRIVATE hip::device)
     else()
         set_source_files_properties(${GGML_SOURCES_ROCM} PROPERTIES LANGUAGE HIP)
     endif()
@@ -504,16 +507,17 @@ if (GGML_HIPBLAS)
         message(FATAL_ERROR "Static linking not supported for HIP/ROCm")
     endif()
-    set(GGML_EXTRA_LIBS ${GGML_EXTRA_LIBS} PUBLIC hip::host roc::rocblas roc::hipblas)
+    list(APPEND GGML_EXTRA_LIBS_PUBLIC hip::host roc::rocblas roc::hipblas)
 endif()
 if (GGML_SYCL)
-    if (NOT GGML_SYCL_TARGET MATCHES "^(INTEL|NVIDIA)$")
-        message(FATAL_ERROR "Invalid backend chosen, supported options are INTEL or NVIDIA")
+    if (NOT GGML_SYCL_TARGET MATCHES "^(INTEL|NVIDIA|AMD)$")
+        message(FATAL_ERROR "Invalid backend chosen, supported options are INTEL, NVIDIA, or AMD")
     endif()
     check_cxx_compiler_flag("-fsycl" SUPPORTS_SYCL)
-    if ( DEFINED ENV{ONEAPI_ROOT})
+    if (DEFINED ENV{ONEAPI_ROOT})
         message(STATUS "Using oneAPI Release SYCL compiler (icpx).")
     elseif(SUPPORTS_SYCL)
         message(WARNING "Using open-source SYCL compiler (clang++). Didn't detect ENV {ONEAPI_ROOT}.
@@ -528,6 +532,9 @@ if (GGML_SYCL)
     list(APPEND GGML_CDEF_PUBLIC GGML_USE_SYCL)
     if (GGML_SYCL_F16)
+        if (GGML_SYCL_TARGET STREQUAL "AMD")
+            message(WARNING "AMD target does not entirely support FP16 in the SYCL backend.")
+        endif()
         add_compile_definitions(GGML_SYCL_F16)
     endif()
@@ -539,6 +546,12 @@ if (GGML_SYCL)
     if (GGML_SYCL_TARGET STREQUAL "NVIDIA")
         add_compile_definitions(GGML_SYCL_WARP_SIZE=32)
+    elseif (GGML_SYCL_TARGET STREQUAL "AMD")
+        # INFO: Allowed Sub_group_sizes are not consistent through all
+        # hip targets. For example, 64 is used for certain models, but the backend
+        # does not support it.
+        # Target archs tested working: gfx1030, gfx1031, (Only tested sub_group_size = 32)
+        add_compile_definitions(GGML_SYCL_WARP_SIZE=32)
     else()
         add_compile_definitions(GGML_SYCL_WARP_SIZE=16)
     endif()
@@ -549,16 +562,35 @@ if (GGML_SYCL)
     file(GLOB   GGML_SOURCES_SYCL "ggml-sycl/*.cpp")
     list(APPEND GGML_SOURCES_SYCL "ggml-sycl.cpp")
+    find_package(DNNL)
+    message("-- DNNL found:" ${DNNL_FOUND})
+    if (GGML_SYCL_TARGET STREQUAL "INTEL")
+        add_compile_definitions(GGML_SYCL_DNNL=${DNNL_FOUND})
+    else()
+        add_compile_definitions(GGML_SYCL_DNNL=0)
+    endif()
+    if (${DNNL_FOUND} AND GGML_SYCL_TARGET STREQUAL "INTEL")
+        list(APPEND GGML_EXTRA_LIBS_PRIVATE DNNL::dnnl)
+    endif()
     if (WIN32)
         find_package(IntelSYCL REQUIRED)
         find_package(MKL REQUIRED)
-        set(GGML_EXTRA_LIBS ${GGML_EXTRA_LIBS} IntelSYCL::SYCL_CXX MKL::MKL MKL::MKL_SYCL)
+        list(APPEND GGML_EXTRA_LIBS_PRIVATE IntelSYCL::SYCL_CXX MKL::MKL MKL::MKL_SYCL)
     else()
         if (GGML_SYCL_TARGET STREQUAL "INTEL")
-            set(GGML_EXTRA_LIBS ${GGML_EXTRA_LIBS} -fsycl OpenCL mkl_core pthread m dl mkl_sycl_blas mkl_intel_ilp64 mkl_tbb_thread)
+            list(APPEND GGML_EXTRA_LIBS_PRIVATE sycl OpenCL mkl_core pthread m dl mkl_sycl_blas mkl_intel_ilp64 mkl_tbb_thread)
         elseif (GGML_SYCL_TARGET STREQUAL "NVIDIA")
             set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fsycl-targets=nvptx64-nvidia-cuda")
-            set(GGML_EXTRA_LIBS ${GGML_EXTRA_LIBS} -fsycl pthread m dl onemkl)
+            list(APPEND GGML_EXTRA_LIBS_PRIVATE sycl pthread m dl onemkl)
+        elseif (GGML_SYCL_TARGET STREQUAL "AMD")
+            if (GGML_SYCL_HIP_TARGET STREQUAL "")
+                message(ERROR "Can't enable SYCL hip backend, GGML_SYCL_HIP_TARGET has not been set.")
+            endif()
+            set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fsycl-targets=amdgcn-amd-amdhsa -Xsycl-target-backend --offload-arch=${GGML_SYCL_HIP_TARGET}")
+            list(APPEND GGML_EXTRA_LIBS_PRIVATE sycl pthread m dl onemkl)
         endif()
     endif()
 endif()
@@ -569,7 +601,7 @@ if (GGML_RPC)
     list(APPEND GGML_CDEF_PUBLIC GGML_USE_RPC)
     if (WIN32)
-        set(GGML_EXTRA_LIBS ${GGML_EXTRA_LIBS} ws2_32)
+        list(APPEND GGML_EXTRA_LIBS_PRIVATE ws2_32)
     endif()
     set(GGML_HEADERS_RPC ../include/ggml-rpc.h)
@@ -602,6 +634,14 @@ if (GGML_VULKAN)
             add_compile_definitions(GGML_VULKAN_MEMORY_DEBUG)
         endif()
+        if (GGML_VULKAN_SHADER_DEBUG_INFO)
+            add_compile_definitions(GGML_VULKAN_SHADER_DEBUG_INFO)
+        endif()
+        if (GGML_VULKAN_PERF)
+            add_compile_definitions(GGML_VULKAN_PERF)
+        endif()
         if (GGML_VULKAN_VALIDATE)
             add_compile_definitions(GGML_VULKAN_VALIDATE)
         endif()
@@ -639,8 +679,8 @@ if (GGML_VULKAN)
         set(GGML_HEADERS_VULKAN ${CMAKE_CURRENT_SOURCE_DIR}/../include/ggml-vulkan.h ${_ggml_vk_header})
         set(GGML_SOURCES_VULKAN ggml-vulkan.cpp ${_ggml_vk_source})
-        set(GGML_EXTRA_LIBS     ${GGML_EXTRA_LIBS} Vulkan::Vulkan)
-        set(GGML_EXTRA_INCLUDES ${GGML_EXTRA_INCLUDES} ${CMAKE_CURRENT_BINARY_DIR})
+        list(APPEND GGML_EXTRA_LIBS_PRIVATE Vulkan::Vulkan)
+        list(APPEND GGML_EXTRA_INCLUDES     ${CMAKE_CURRENT_BINARY_DIR})
     else()
         message(WARNING "Vulkan not found")
     endif()
@@ -799,8 +839,8 @@ if (GGML_KOMPUTE)
         list(APPEND GGML_CDEF_PUBLIC GGML_USE_KOMPUTE)
-        set(GGML_EXTRA_LIBS     ${GGML_EXTRA_LIBS}     kompute)
-        set(GGML_EXTRA_INCLUDES ${GGML_EXTRA_INCLUDES} ${CMAKE_CURRENT_BINARY_DIR})
+        list(APPEND GGML_EXTRA_LIBS_PRIVATE kompute)
+        list(APPEND GGML_EXTRA_INCLUDES     ${CMAKE_CURRENT_BINARY_DIR})
     else()
         message(WARNING "Kompute not found")
     endif()
@@ -849,11 +889,6 @@ if (GGML_CANN)
                 ${CANN_INSTALL_DIR}/acllib/include
             )
-            # TODO: find libs
-            link_directories(
-                ${CANN_INSTALL_DIR}/lib64
-            )
             add_subdirectory(ggml-cann/kernels)
             list(APPEND CANN_LIBRARIES
                 ascendcl
@@ -870,8 +905,10 @@ if (GGML_CANN)
             message(STATUS "CANN: CANN_INCLUDE_DIRS =  ${CANN_INCLUDE_DIRS}")
             message(STATUS "CANN: CANN_LIBRARIES =  ${CANN_LIBRARIES}")
-            set(GGML_EXTRA_LIBS     ${GGML_EXTRA_LIBS}     ${CANN_LIBRARIES} )
-            set(GGML_EXTRA_INCLUDES ${GGML_EXTRA_INCLUDES} ${CANN_INCLUDE_DIRS})
+            list(APPEND GGML_EXTRA_LIBS_PRIVATE ${CANN_LIBRARIES} )
+            list(APPEND GGML_EXTRA_INCLUDES     ${CANN_INCLUDE_DIRS})
+            list(APPEND GGML_EXTRA_LIBDIRS      ${CANN_INSTALL_DIR}/lib64)
             list(APPEND GGML_CDEF_PUBLIC GGML_USE_CANN)
         endif()
     else()
@@ -1164,6 +1201,7 @@ elseif (CMAKE_OSX_ARCHITECTURES STREQUAL "x86_64" OR CMAKE_GENERATOR_PLATFORM_LW
         endif()
         if (GGML_AVX512)
             list(APPEND ARCH_FLAGS -mavx512f)
+            list(APPEND ARCH_FLAGS -mavx512dq)
             list(APPEND ARCH_FLAGS -mavx512bw)
         endif()
         if (GGML_AVX512_VBMI)
@@ -1237,7 +1275,7 @@ endif()
 # Data types, macros and functions related to controlling CPU affinity and
 # some memory allocation are available on Linux through GNU extensions in libc
-if (CMAKE_SYSTEM_NAME MATCHES "Linux")
+if (CMAKE_SYSTEM_NAME MATCHES "Linux" OR CMAKE_SYSTEM_NAME MATCHES "Android")
     add_compile_definitions(_GNU_SOURCE)
 endif()
@@ -1287,7 +1325,7 @@ add_library(ggml
             ../include/ggml-backend.h
             ggml.c
             ggml-alloc.c
-            ggml-backend.c
+            ggml-backend.cpp
             ggml-quants.c
             ggml-quants.h
             ${GGML_SOURCES_CUDA}      ${GGML_HEADERS_CUDA}
@@ -1308,21 +1346,25 @@ if (EMSCRIPTEN)
     set_target_properties(ggml PROPERTIES COMPILE_FLAGS "-msimd128")
 endif()
-target_compile_definitions(ggml PUBLIC  ${GGML_CDEF_PUBLIC})
-target_include_directories(ggml PUBLIC ../include)
+target_compile_definitions(ggml PUBLIC    ${GGML_CDEF_PUBLIC})
+target_include_directories(ggml PUBLIC  ../include)
 target_include_directories(ggml PRIVATE . ${GGML_EXTRA_INCLUDES})
-target_link_directories(ggml PRIVATE ${GGML_EXTRA_LIBDIRS})
+target_link_directories   (ggml PRIVATE   ${GGML_EXTRA_LIBDIRS})
 target_compile_features   (ggml PRIVATE c_std_11) # don't bump
-target_link_libraries(ggml PRIVATE Threads::Threads ${GGML_EXTRA_LIBS})
+list(APPEND GGML_EXTRA_LIBS_PRIVATE Threads::Threads)
 find_library(MATH_LIBRARY m)
 if (MATH_LIBRARY)
     if (NOT WIN32 OR NOT GGML_SYCL)
-        target_link_libraries(ggml PRIVATE ${MATH_LIBRARY})
+        list(APPEND GGML_EXTRA_LIBS_PRIVATE m)
     endif()
 endif()
+list(REMOVE_DUPLICATES GGML_EXTRA_LIBS_PRIVATE)
+list(REMOVE_DUPLICATES GGML_EXTRA_LIBS_PUBLIC)
+target_link_libraries(ggml PRIVATE ${GGML_EXTRA_LIBS_PRIVATE} PUBLIC ${GGML_EXTRA_LIBS_PUBLIC})
 if (BUILD_SHARED_LIBS)
     set_target_properties(ggml PROPERTIES POSITION_INDEPENDENT_CODE ON)
     target_compile_definitions(ggml PRIVATE GGML_SHARED GGML_BUILD)