npm - @fugood/llama.node - Versions diffs - 0.3.6 → 0.3.8 - Mend

@fugood/llama.node 0.3.6 → 0.3.8

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 (186) hide show

package/README.md +17 -2
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-cuda/arm64/llama-node.node +0 -0
package/bin/linux-cuda/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/lib/binding.ts +3 -1
package/lib/index.js +16 -1
package/lib/index.ts +16 -0
package/package.json +1 -1
package/src/EmbeddingWorker.cpp +4 -3
package/src/LlamaCompletionWorker.cpp +4 -2
package/src/LlamaContext.cpp +61 -6
package/src/LlamaContext.h +1 -0
package/src/common.hpp +6 -11
package/src/llama.cpp/.github/workflows/build.yml +19 -17
package/src/llama.cpp/.github/workflows/docker.yml +77 -30
package/src/llama.cpp/.github/workflows/editorconfig.yml +3 -1
package/src/llama.cpp/.github/workflows/server.yml +22 -3
package/src/llama.cpp/CMakeLists.txt +49 -24
package/src/llama.cpp/common/arg.cpp +82 -26
package/src/llama.cpp/common/arg.h +3 -0
package/src/llama.cpp/common/common.cpp +192 -72
package/src/llama.cpp/common/common.h +51 -18
package/src/llama.cpp/common/ngram-cache.cpp +12 -12
package/src/llama.cpp/common/ngram-cache.h +2 -2
package/src/llama.cpp/common/sampling.cpp +11 -6
package/src/llama.cpp/common/speculative.cpp +18 -15
package/src/llama.cpp/docs/build.md +2 -0
package/src/llama.cpp/examples/batched/batched.cpp +9 -7
package/src/llama.cpp/examples/batched-bench/batched-bench.cpp +3 -3
package/src/llama.cpp/examples/convert-llama2c-to-ggml/convert-llama2c-to-ggml.cpp +10 -8
package/src/llama.cpp/examples/cvector-generator/cvector-generator.cpp +11 -8
package/src/llama.cpp/examples/cvector-generator/mean.hpp +1 -1
package/src/llama.cpp/examples/cvector-generator/pca.hpp +1 -1
package/src/llama.cpp/examples/embedding/embedding.cpp +8 -7
package/src/llama.cpp/examples/eval-callback/eval-callback.cpp +7 -6
package/src/llama.cpp/examples/export-lora/export-lora.cpp +8 -7
package/src/llama.cpp/examples/gguf/gguf.cpp +10 -6
package/src/llama.cpp/examples/gguf-hash/gguf-hash.cpp +1 -0
package/src/llama.cpp/examples/gguf-split/gguf-split.cpp +8 -7
package/src/llama.cpp/examples/gritlm/gritlm.cpp +13 -10
package/src/llama.cpp/examples/imatrix/imatrix.cpp +13 -12
package/src/llama.cpp/examples/infill/infill.cpp +23 -24
package/src/llama.cpp/examples/llama-bench/llama-bench.cpp +44 -13
package/src/llama.cpp/examples/llama.android/llama/src/main/cpp/llama-android.cpp +11 -6
package/src/llama.cpp/examples/llava/clip.cpp +4 -2
package/src/llama.cpp/examples/llava/llava-cli.cpp +9 -6
package/src/llama.cpp/examples/llava/llava.cpp +2 -2
package/src/llama.cpp/examples/llava/minicpmv-cli.cpp +8 -4
package/src/llama.cpp/examples/llava/qwen2vl-cli.cpp +11 -8
package/src/llama.cpp/examples/lookahead/lookahead.cpp +6 -7
package/src/llama.cpp/examples/lookup/lookup-create.cpp +4 -9
package/src/llama.cpp/examples/lookup/lookup-stats.cpp +3 -7
package/src/llama.cpp/examples/lookup/lookup.cpp +5 -6
package/src/llama.cpp/examples/main/main.cpp +51 -29
package/src/llama.cpp/examples/parallel/parallel.cpp +5 -6
package/src/llama.cpp/examples/passkey/passkey.cpp +7 -5
package/src/llama.cpp/examples/perplexity/perplexity.cpp +37 -23
package/src/llama.cpp/examples/quantize-stats/quantize-stats.cpp +12 -14
package/src/llama.cpp/examples/retrieval/retrieval.cpp +8 -8
package/src/llama.cpp/examples/rpc/rpc-server.cpp +12 -0
package/src/llama.cpp/examples/run/CMakeLists.txt +1 -1
package/src/llama.cpp/examples/run/linenoise.cpp/linenoise.cpp +1351 -0
package/src/llama.cpp/examples/run/linenoise.cpp/linenoise.h +114 -0
package/src/llama.cpp/examples/run/run.cpp +175 -61
package/src/llama.cpp/examples/save-load-state/save-load-state.cpp +4 -25
package/src/llama.cpp/examples/server/CMakeLists.txt +1 -0
package/src/llama.cpp/examples/server/httplib.h +1295 -409
package/src/llama.cpp/examples/server/server.cpp +387 -181
package/src/llama.cpp/examples/server/tests/requirements.txt +1 -0
package/src/llama.cpp/examples/server/utils.hpp +170 -58
package/src/llama.cpp/examples/simple/simple.cpp +9 -8
package/src/llama.cpp/examples/simple-chat/simple-chat.cpp +16 -12
package/src/llama.cpp/examples/speculative/speculative.cpp +22 -23
package/src/llama.cpp/examples/speculative-simple/speculative-simple.cpp +8 -12
package/src/llama.cpp/examples/tokenize/tokenize.cpp +17 -5
package/src/llama.cpp/examples/tts/tts.cpp +64 -23
package/src/llama.cpp/ggml/CMakeLists.txt +5 -21
package/src/llama.cpp/ggml/include/ggml-backend.h +2 -0
package/src/llama.cpp/ggml/include/ggml-cpp.h +1 -0
package/src/llama.cpp/ggml/include/ggml.h +36 -145
package/src/llama.cpp/ggml/include/gguf.h +202 -0
package/src/llama.cpp/ggml/src/CMakeLists.txt +6 -3
package/src/llama.cpp/ggml/src/ggml-alloc.c +5 -0
package/src/llama.cpp/ggml/src/ggml-backend-impl.h +0 -1
package/src/llama.cpp/ggml/src/ggml-backend-reg.cpp +79 -49
package/src/llama.cpp/ggml/src/ggml-backend.cpp +5 -2
package/src/llama.cpp/ggml/src/ggml-cpu/CMakeLists.txt +33 -23
package/src/llama.cpp/ggml/src/ggml-cpu/ggml-cpu-aarch64.cpp +57 -72
package/src/llama.cpp/ggml/src/ggml-cpu/ggml-cpu-quants.c +87 -2
package/src/llama.cpp/ggml/src/ggml-cpu/ggml-cpu.c +335 -66
package/src/llama.cpp/ggml/src/ggml-cpu/ggml-cpu.cpp +10 -2
package/src/llama.cpp/ggml/src/ggml-cpu/llamafile/sgemm.cpp +1090 -378
package/src/llama.cpp/ggml/src/ggml-cpu/llamafile/sgemm.h +2 -2
package/src/llama.cpp/ggml/src/ggml-cuda/vendors/cuda.h +1 -0
package/src/llama.cpp/ggml/src/ggml-cuda/vendors/hip.h +3 -0
package/src/llama.cpp/ggml/src/ggml-cuda/vendors/musa.h +3 -0
package/src/llama.cpp/ggml/src/ggml-hip/CMakeLists.txt +3 -1
package/src/llama.cpp/ggml/src/ggml-impl.h +11 -16
package/src/llama.cpp/ggml/src/ggml-metal/CMakeLists.txt +16 -0
package/src/llama.cpp/ggml/src/ggml-opencl/ggml-opencl.cpp +6 -6
package/src/llama.cpp/ggml/src/ggml-rpc/ggml-rpc.cpp +154 -35
package/src/llama.cpp/ggml/src/ggml-sycl/backend.hpp +1 -0
package/src/llama.cpp/ggml/src/ggml-sycl/common.cpp +9 -3
package/src/llama.cpp/ggml/src/ggml-sycl/common.hpp +18 -0
package/src/llama.cpp/ggml/src/ggml-sycl/concat.cpp +3 -2
package/src/llama.cpp/ggml/src/ggml-sycl/concat.hpp +1 -2
package/src/llama.cpp/ggml/src/ggml-sycl/conv.cpp +3 -2
package/src/llama.cpp/ggml/src/ggml-sycl/conv.hpp +1 -2
package/src/llama.cpp/ggml/src/ggml-sycl/dpct/helper.hpp +40 -95
package/src/llama.cpp/ggml/src/ggml-sycl/element_wise.cpp +48 -48
package/src/llama.cpp/ggml/src/ggml-sycl/element_wise.hpp +24 -24
package/src/llama.cpp/ggml/src/ggml-sycl/ggml-sycl.cpp +238 -164
package/src/llama.cpp/ggml/src/ggml-sycl/gla.cpp +105 -0
package/src/llama.cpp/ggml/src/ggml-sycl/gla.hpp +8 -0
package/src/llama.cpp/ggml/src/ggml-sycl/outprod.cpp +3 -3
package/src/llama.cpp/ggml/src/ggml-sycl/outprod.hpp +1 -2
package/src/llama.cpp/ggml/src/ggml-sycl/tsembd.cpp +3 -2
package/src/llama.cpp/ggml/src/ggml-sycl/tsembd.hpp +1 -2
package/src/llama.cpp/ggml/src/ggml-sycl/wkv6.cpp +7 -5
package/src/llama.cpp/ggml/src/ggml-sycl/wkv6.hpp +1 -2
package/src/llama.cpp/ggml/src/ggml-vulkan/CMakeLists.txt +74 -4
package/src/llama.cpp/ggml/src/ggml-vulkan/ggml-vulkan.cpp +314 -116
package/src/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/CMakeLists.txt +4 -2
package/src/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/vulkan-shaders-gen.cpp +9 -3
package/src/llama.cpp/ggml/src/ggml.c +117 -1327
package/src/llama.cpp/ggml/src/gguf.cpp +1329 -0
package/src/llama.cpp/include/llama-cpp.h +6 -1
package/src/llama.cpp/include/llama.h +138 -75
package/src/llama.cpp/src/CMakeLists.txt +13 -1
package/src/llama.cpp/src/llama-adapter.cpp +347 -0
package/src/llama.cpp/src/llama-adapter.h +74 -0
package/src/llama.cpp/src/llama-arch.cpp +1487 -0
package/src/llama.cpp/src/llama-arch.h +400 -0
package/src/llama.cpp/src/llama-batch.cpp +368 -0
package/src/llama.cpp/src/llama-batch.h +88 -0
package/src/llama.cpp/src/llama-chat.cpp +578 -0
package/src/llama.cpp/src/llama-chat.h +52 -0
package/src/llama.cpp/src/llama-context.cpp +1775 -0
package/src/llama.cpp/src/llama-context.h +128 -0
package/src/llama.cpp/src/llama-cparams.cpp +1 -0
package/src/llama.cpp/src/llama-cparams.h +37 -0
package/src/llama.cpp/src/llama-grammar.cpp +5 -4
package/src/llama.cpp/src/llama-grammar.h +3 -1
package/src/llama.cpp/src/llama-hparams.cpp +71 -0
package/src/llama.cpp/src/llama-hparams.h +139 -0
package/src/llama.cpp/src/llama-impl.cpp +167 -0
package/src/llama.cpp/src/llama-impl.h +16 -136
package/src/llama.cpp/src/llama-kv-cache.cpp +718 -0
package/src/llama.cpp/src/llama-kv-cache.h +218 -0
package/src/llama.cpp/src/llama-mmap.cpp +589 -0
package/src/llama.cpp/src/llama-mmap.h +67 -0
package/src/llama.cpp/src/llama-model-loader.cpp +1124 -0
package/src/llama.cpp/src/llama-model-loader.h +167 -0
package/src/llama.cpp/src/llama-model.cpp +3953 -0
package/src/llama.cpp/src/llama-model.h +370 -0
package/src/llama.cpp/src/llama-quant.cpp +934 -0
package/src/llama.cpp/src/llama-quant.h +1 -0
package/src/llama.cpp/src/llama-sampling.cpp +147 -32
package/src/llama.cpp/src/llama-sampling.h +3 -19
package/src/llama.cpp/src/llama-vocab.cpp +1832 -575
package/src/llama.cpp/src/llama-vocab.h +97 -142
package/src/llama.cpp/src/llama.cpp +7160 -20314
package/src/llama.cpp/src/unicode.cpp +8 -3
package/src/llama.cpp/tests/CMakeLists.txt +2 -0
package/src/llama.cpp/tests/test-autorelease.cpp +3 -3
package/src/llama.cpp/tests/test-backend-ops.cpp +370 -59
package/src/llama.cpp/tests/test-chat-template.cpp +162 -125
package/src/llama.cpp/tests/test-gguf.cpp +222 -187
package/src/llama.cpp/tests/test-model-load-cancel.cpp +1 -1
package/src/llama.cpp/tests/test-sampling.cpp +0 -1
package/src/llama.cpp/tests/test-tokenizer-0.cpp +4 -4
package/src/llama.cpp/tests/test-tokenizer-1-bpe.cpp +9 -7
package/src/llama.cpp/tests/test-tokenizer-1-spm.cpp +8 -6

package/src/llama.cpp/ggml/src/ggml-sycl/dpct/helper.hpp CHANGED Viewed

@@ -82,6 +82,14 @@ inline std::string get_device_backend_and_type(const sycl::device &device) {
     return device_type.str();
 }
+template <typename Ts> struct matrix_info_t {
+    oneapi::mkl::transpose transpose_info[2];
+    Ts                     value_info[2];
+    std::int64_t           size_info[3];
+    std::int64_t           ld_info[3];
+    std::int64_t           groupsize_info;
+};
 namespace dpct
 {
     typedef sycl::queue *queue_ptr;
@@ -1727,26 +1735,13 @@ namespace dpct
         };
         template <class Ta, class Tb, class Tc, class Ts>
-        inline void gemm_batch_impl(sycl::queue &q, oneapi::mkl::transpose a_trans,
-                                    oneapi::mkl::transpose b_trans, int m, int n, int k,
-                                    const void *alpha, const void **a, int lda,
-                                    const void **b, int ldb, const void *beta, void **c,
-                                    int ldc, int batch_size)
-        {
-            struct matrix_info_t
-            {
-                oneapi::mkl::transpose transpose_info[2];
-                Ts value_info[2];
-                std::int64_t size_info[3];
-                std::int64_t ld_info[3];
-                std::int64_t groupsize_info;
-            };
+        inline void gemm_batch_impl(sycl::queue & q, oneapi::mkl::transpose a_trans, oneapi::mkl::transpose b_trans,
+                                    int m, int n, int k, const void * alpha, const void ** a, int lda, const void ** b,
+                                    int ldb, const void * beta, void ** c, int ldc, int batch_size,
+                                    matrix_info_t<float> * matrix_info) {
             Ts alpha_value = dpct::get_value(reinterpret_cast<const Ts *>(alpha), q);
             Ts beta_value = dpct::get_value(reinterpret_cast<const Ts *>(beta), q);
-            matrix_info_t *matrix_info =
-                (matrix_info_t *)std::malloc(sizeof(matrix_info_t));
             matrix_info->transpose_info[0] = a_trans;
             matrix_info->transpose_info[1] = b_trans;
             matrix_info->value_info[0] = alpha_value;
@@ -1763,23 +1758,18 @@ namespace dpct
             sycl::event e = oneapi::mkl::blas::column_major::gemm_batch(
                 oneapi::mkl::backend_selector<oneapi::mkl::backend::cublas>{ q }, matrix_info->transpose_info,
                 matrix_info->transpose_info + 1, matrix_info->size_info, matrix_info->size_info + 1,
-                matrix_info->size_info + 2, matrix_info->value_info, reinterpret_cast<const Ta **>(a),
-                matrix_info->ld_info, reinterpret_cast<const Tb **>(b), matrix_info->ld_info + 1,
-                matrix_info->value_info + 1, reinterpret_cast<Tc **>(c), matrix_info->ld_info + 2, 1,
-                &(matrix_info->groupsize_info));
+                matrix_info->size_info + 2, reinterpret_cast<Ts *>(matrix_info->value_info),
+                reinterpret_cast<const Ta **>(a), matrix_info->ld_info, reinterpret_cast<const Tb **>(b),
+                matrix_info->ld_info + 1, reinterpret_cast<Ts *>(matrix_info->value_info + 1),
+                reinterpret_cast<Tc **>(c), matrix_info->ld_info + 2, 1, &(matrix_info->groupsize_info));
 #else
             sycl::event e = oneapi::mkl::blas::column_major::gemm_batch(
                 q, matrix_info->transpose_info, matrix_info->transpose_info + 1, matrix_info->size_info,
-                matrix_info->size_info + 1, matrix_info->size_info + 2, matrix_info->value_info,
+                matrix_info->size_info + 1, matrix_info->size_info + 2, reinterpret_cast<Ts *>(matrix_info->value_info),
                 reinterpret_cast<const Ta **>(a), matrix_info->ld_info, reinterpret_cast<const Tb **>(b),
-                matrix_info->ld_info + 1, matrix_info->value_info + 1, reinterpret_cast<Tc **>(c),
-                matrix_info->ld_info + 2, 1, &(matrix_info->groupsize_info));
+                matrix_info->ld_info + 1, reinterpret_cast<Ts *>(matrix_info->value_info + 1),
+                reinterpret_cast<Tc **>(c), matrix_info->ld_info + 2, 1, &(matrix_info->groupsize_info));
 #endif
-            q.submit([&](sycl::handler &cgh)
-                     {
-    cgh.depends_on(e);
-    cgh.host_task([=] { std::free(matrix_info); }); });
         }
         template <class Ta, class Tb, class Tc, class Ts>
@@ -2422,25 +2412,11 @@ namespace dpct
     /// \param [in] ldc Leading dimension of C.
     /// \param [in] batch_size Specifies the number of matrix multiply operations to perform.
     /// \param [in] scaling_type Data type of the scaling factors.
-    inline void gemm_batch(sycl::queue &q, oneapi::mkl::transpose a_trans,
-                           oneapi::mkl::transpose b_trans, int m, int n, int k,
-                           const void *alpha, const void *a[],
-                           library_data_t a_type, int lda, const void *b[],
-                           library_data_t b_type, int ldb, const void *beta,
-                           void *c[], library_data_t c_type, int ldc,
-                           int batch_size, library_data_t scaling_type)
-    {
-        if (scaling_type == library_data_t::real_float &&
-            c_type == library_data_t::complex_float)
-        {
-            scaling_type = library_data_t::complex_float;
-        }
-        else if (scaling_type == library_data_t::real_double &&
-                 c_type == library_data_t::complex_double)
-        {
-            scaling_type = library_data_t::complex_double;
-        }
+    inline void gemm_batch(sycl::queue & q, oneapi::mkl::transpose a_trans, oneapi::mkl::transpose b_trans, int m,
+                           int n, int k, const void * alpha, const void * a[], library_data_t a_type, int lda,
+                           const void * b[], library_data_t b_type, int ldb, const void * beta, void * c[],
+                           library_data_t c_type, int ldc, int batch_size, library_data_t scaling_type,
+                           matrix_info_t<float> * matrix_info) {
         std::uint64_t key =
             detail::get_type_combination_id(a_type, b_type, c_type, scaling_type);
         switch (key)
@@ -2449,48 +2425,24 @@ namespace dpct
             library_data_t::real_float, library_data_t::real_float,
             library_data_t::real_float, library_data_t::real_float):
         {
-            detail::gemm_batch_impl<float, float, float, float>(
-                q, a_trans, b_trans, m, n, k, alpha, a, lda, b, ldb, beta, c, ldc,
-                batch_size);
+            detail::gemm_batch_impl<float, float, float, float>(q, a_trans, b_trans, m, n, k, alpha, a, lda, b, ldb,
+                                                                beta, c, ldc, batch_size, matrix_info);
             break;
         }
         case detail::get_type_combination_id(
             library_data_t::real_double, library_data_t::real_double,
             library_data_t::real_double, library_data_t::real_double):
         {
-            detail::gemm_batch_impl<double, double, double, double>(
-                q, a_trans, b_trans, m, n, k, alpha, a, lda, b, ldb, beta, c, ldc,
-                batch_size);
-            break;
-        }
-        case detail::get_type_combination_id(
-            library_data_t::complex_float, library_data_t::complex_float,
-            library_data_t::complex_float, library_data_t::complex_float):
-        {
-            detail::gemm_batch_impl<std::complex<float>, std::complex<float>,
-                                    std::complex<float>, std::complex<float>>(
-                q, a_trans, b_trans, m, n, k, alpha, a, lda, b, ldb, beta, c, ldc,
-                batch_size);
-            break;
-        }
-        case detail::get_type_combination_id(
-            library_data_t::complex_double, library_data_t::complex_double,
-            library_data_t::complex_double, library_data_t::complex_double):
-        {
-            detail::gemm_batch_impl<std::complex<double>, std::complex<double>,
-                                    std::complex<double>, std::complex<double>>(
-                q, a_trans, b_trans, m, n, k, alpha, a, lda, b, ldb, beta, c, ldc,
-                batch_size);
+            detail::gemm_batch_impl<double, double, double, double>(q, a_trans, b_trans, m, n, k, alpha, a, lda, b, ldb,
+                                                                    beta, c, ldc, batch_size, matrix_info);
             break;
         }
         case detail::get_type_combination_id(
             library_data_t::real_half, library_data_t::real_half,
             library_data_t::real_half, library_data_t::real_half):
         {
-            detail::gemm_batch_impl<sycl::half, sycl::half, sycl::half,
-                                    sycl::half>(q, a_trans, b_trans, m, n, k, alpha,
-                                                a, lda, b, ldb, beta, c, ldc,
-                                                batch_size);
+            detail::gemm_batch_impl<sycl::half, sycl::half, sycl::half, sycl::half>(
+                q, a_trans, b_trans, m, n, k, alpha, a, lda, b, ldb, beta, c, ldc, batch_size, matrix_info);
             break;
         }
 #ifdef __INTEL_MKL__
@@ -2498,19 +2450,16 @@ namespace dpct
             library_data_t::real_bfloat16, library_data_t::real_bfloat16,
             library_data_t::real_bfloat16, library_data_t::real_float):
         {
-            detail::gemm_batch_impl<oneapi::mkl::bfloat16, oneapi::mkl::bfloat16,
-                                    oneapi::mkl::bfloat16, float>(
-                q, a_trans, b_trans, m, n, k, alpha, a, lda, b, ldb, beta, c, ldc,
-                batch_size);
+            detail::gemm_batch_impl<oneapi::mkl::bfloat16, oneapi::mkl::bfloat16, oneapi::mkl::bfloat16, float>(
+                q, a_trans, b_trans, m, n, k, alpha, a, lda, b, ldb, beta, c, ldc, batch_size, matrix_info);
             break;
         }
         case detail::get_type_combination_id(
             library_data_t::real_bfloat16, library_data_t::real_bfloat16,
             library_data_t::real_float, library_data_t::real_float):
         {
-            detail::gemm_batch_impl<oneapi::mkl::bfloat16, oneapi::mkl::bfloat16, float,
-                                    float>(q, a_trans, b_trans, m, n, k, alpha, a, lda,
-                                           b, ldb, beta, c, ldc, batch_size);
+            detail::gemm_batch_impl<oneapi::mkl::bfloat16, oneapi::mkl::bfloat16, float, float>(
+                q, a_trans, b_trans, m, n, k, alpha, a, lda, b, ldb, beta, c, ldc, batch_size, matrix_info);
             break;
         }
 #endif
@@ -2522,10 +2471,9 @@ namespace dpct
                 dpct::get_value(reinterpret_cast<const std::int32_t *>(alpha), q);
             float beta_float =
                 dpct::get_value(reinterpret_cast<const std::int32_t *>(beta), q);
-            detail::gemm_batch_impl<std::int8_t, std::int8_t, std::int32_t,
-                                    float>(q, a_trans, b_trans, m, n, k, &alpha_float,
-                                           a, lda, b, ldb, &beta_float, c, ldc,
-                                           batch_size);
+            detail::gemm_batch_impl<std::int8_t, std::int8_t, std::int32_t, float>(
+                q, a_trans, b_trans, m, n, k, &alpha_float, a, lda, b, ldb, &beta_float, c, ldc, batch_size,
+                matrix_info);
             break;
         }
         case detail::get_type_combination_id(
@@ -2533,8 +2481,7 @@ namespace dpct
             library_data_t::real_float, library_data_t::real_float):
         {
             detail::gemm_batch_impl<std::int8_t, std::int8_t, float, float>(
-                q, a_trans, b_trans, m, n, k, alpha, a, lda, b, ldb, beta, c, ldc,
-                batch_size);
+                q, a_trans, b_trans, m, n, k, alpha, a, lda, b, ldb, beta, c, ldc, batch_size, matrix_info);
             break;
         }
         case detail::get_type_combination_id(
@@ -2542,8 +2489,7 @@ namespace dpct
             library_data_t::real_float, library_data_t::real_float):
         {
             detail::gemm_batch_impl<sycl::half, sycl::half, float, float>(
-                q, a_trans, b_trans, m, n, k, alpha, a, lda, b, ldb, beta, c, ldc,
-                batch_size);
+                q, a_trans, b_trans, m, n, k, alpha, a, lda, b, ldb, beta, c, ldc, batch_size, matrix_info);
             break;
         }
         case detail::get_type_combination_id(
@@ -2557,8 +2503,7 @@ namespace dpct
             sycl::half alpha_half(alpha_value);
             sycl::half beta_half(beta_value);
             detail::gemm_batch_impl<sycl::half, sycl::half, sycl::half, sycl::half>(
-                q, a_trans, b_trans, m, n, k, &alpha_half, a, lda, b, ldb, &beta_half, c, ldc,
-                batch_size);
+                q, a_trans, b_trans, m, n, k, &alpha_half, a, lda, b, ldb, &beta_half, c, ldc, batch_size, matrix_info);
             break;
         }
         default:

package/src/llama.cpp/ggml/src/ggml-sycl/element_wise.cpp CHANGED Viewed

@@ -882,149 +882,149 @@ inline void ggml_sycl_op_div(ggml_backend_sycl_context & ctx, const ggml_tensor
 }
-void ggml_sycl_sqrt(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst) {
+void ggml_sycl_sqrt(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
     GGML_SYCL_DEBUG("call %s\n", __func__);
-    ggml_sycl_op_flatten(ctx, src0, src1, dst, ggml_sycl_op_sqrt);
+    ggml_sycl_op_flatten(ctx, dst->src[0], dst->src[1], dst, ggml_sycl_op_sqrt);
     GGML_SYCL_DEBUG("call %s done\n", __func__);
 }
-void ggml_sycl_sin(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst) {
+void ggml_sycl_sin(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
     GGML_SYCL_DEBUG("call %s\n", __func__);
-    ggml_sycl_op_flatten(ctx, src0, src1, dst, ggml_sycl_op_sin);
+    ggml_sycl_op_flatten(ctx, dst->src[0], dst->src[1], dst, ggml_sycl_op_sin);
     GGML_SYCL_DEBUG("call %s done\n", __func__);
 }
-void ggml_sycl_cos(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst) {
+void ggml_sycl_cos(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
     GGML_SYCL_DEBUG("call %s\n", __func__);
-    ggml_sycl_op_flatten(ctx, src0, src1, dst, ggml_sycl_op_cos);
+    ggml_sycl_op_flatten(ctx, dst->src[0], dst->src[1], dst, ggml_sycl_op_cos);
     GGML_SYCL_DEBUG("call %s done\n", __func__);
 }
-void ggml_sycl_acc(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst) {
+void ggml_sycl_acc(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
     GGML_SYCL_DEBUG("call %s\n", __func__);
-    ggml_sycl_op_flatten(ctx, src0, src1, dst, ggml_sycl_op_acc);
+    ggml_sycl_op_flatten(ctx, dst->src[0], dst->src[1], dst, ggml_sycl_op_acc);
     GGML_SYCL_DEBUG("call %s done\n", __func__);
 }
-void ggml_sycl_gelu(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst) {
+void ggml_sycl_gelu(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
     GGML_SYCL_DEBUG("call %s\n", __func__);
-    ggml_sycl_op_flatten(ctx, src0, src1, dst, ggml_sycl_op_gelu);
+    ggml_sycl_op_flatten(ctx, dst->src[0], dst->src[1], dst, ggml_sycl_op_gelu);
     GGML_SYCL_DEBUG("call %s done\n", __func__);
 }
-void ggml_sycl_silu(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst) {
+void ggml_sycl_silu(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
     GGML_SYCL_DEBUG("call %s\n", __func__);
-    ggml_sycl_op_flatten(ctx, src0, src1, dst, ggml_sycl_op_silu);
+    ggml_sycl_op_flatten(ctx, dst->src[0], dst->src[1], dst, ggml_sycl_op_silu);
     GGML_SYCL_DEBUG("call %s done\n", __func__);
 }
-void ggml_sycl_gelu_quick(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst) {
+void ggml_sycl_gelu_quick(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
     GGML_SYCL_DEBUG("call %s\n", __func__);
-    ggml_sycl_op_flatten(ctx, src0, src1, dst, ggml_sycl_op_gelu_quick);
+    ggml_sycl_op_flatten(ctx, dst->src[0], dst->src[1], dst, ggml_sycl_op_gelu_quick);
     GGML_SYCL_DEBUG("call %s done\n", __func__);
 }
-void ggml_sycl_tanh(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst) {
+void ggml_sycl_tanh(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
     GGML_SYCL_DEBUG("call %s\n", __func__);
-    ggml_sycl_op_flatten(ctx, src0, src1, dst, ggml_sycl_op_tanh);
+    ggml_sycl_op_flatten(ctx, dst->src[0], dst->src[1], dst, ggml_sycl_op_tanh);
     GGML_SYCL_DEBUG("call %s done\n", __func__);
 }
-void ggml_sycl_relu(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst) {
+void ggml_sycl_relu(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
     GGML_SYCL_DEBUG("call %s\n", __func__);
-    ggml_sycl_op_flatten(ctx, src0, src1, dst, ggml_sycl_op_relu);
+    ggml_sycl_op_flatten(ctx, dst->src[0], dst->src[1], dst, ggml_sycl_op_relu);
     GGML_SYCL_DEBUG("call %s done\n", __func__);
 }
-void ggml_sycl_sigmoid(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst) {
+void ggml_sycl_sigmoid(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
     GGML_SYCL_DEBUG("call %s\n", __func__);
-    ggml_sycl_op_flatten(ctx, src0, src1, dst, ggml_sycl_op_sigmoid);
+    ggml_sycl_op_flatten(ctx, dst->src[0], dst->src[1], dst, ggml_sycl_op_sigmoid);
     GGML_SYCL_DEBUG("call %s done\n", __func__);
 }
-void ggml_sycl_hardsigmoid(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst) {
+void ggml_sycl_hardsigmoid(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
     GGML_SYCL_DEBUG("call %s\n", __func__);
-    ggml_sycl_op_flatten(ctx, src0, src1, dst, ggml_sycl_op_hardsigmoid);
+    ggml_sycl_op_flatten(ctx, dst->src[0], dst->src[1], dst, ggml_sycl_op_hardsigmoid);
     GGML_SYCL_DEBUG("call %s done\n", __func__);
 }
-void ggml_sycl_hardswish(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst) {
+void ggml_sycl_hardswish(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
     GGML_SYCL_DEBUG("call %s\n", __func__);
-    ggml_sycl_op_flatten(ctx, src0, src1, dst, ggml_sycl_op_hardswish);
+    ggml_sycl_op_flatten(ctx, dst->src[0], dst->src[1], dst, ggml_sycl_op_hardswish);
     GGML_SYCL_DEBUG("call %s done\n", __func__);
 }
-void ggml_sycl_exp(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst) {
+void ggml_sycl_exp(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
     GGML_SYCL_DEBUG("call %s\n", __func__);
-    ggml_sycl_op_flatten(ctx, src0, src1, dst, ggml_sycl_op_exp);
+    ggml_sycl_op_flatten(ctx, dst->src[0], dst->src[1], dst, ggml_sycl_op_exp);
     GGML_SYCL_DEBUG("call %s done\n", __func__);
 }
-void ggml_sycl_log(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst) {
+void ggml_sycl_log(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
     GGML_SYCL_DEBUG("call %s\n", __func__);
-    ggml_sycl_op_flatten(ctx, src0, src1, dst, ggml_sycl_op_log);
+    ggml_sycl_op_flatten(ctx, dst->src[0], dst->src[1], dst, ggml_sycl_op_log);
     GGML_SYCL_DEBUG("call %s done\n", __func__);
 }
-void ggml_sycl_neg(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst) {
+void ggml_sycl_neg(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
     GGML_SYCL_DEBUG("call %s\n", __func__);
-    ggml_sycl_op_flatten(ctx, src0, src1, dst, ggml_sycl_op_neg);
+    ggml_sycl_op_flatten(ctx, dst->src[0], dst->src[1], dst, ggml_sycl_op_neg);
     GGML_SYCL_DEBUG("call %s done\n", __func__);
 }
-void ggml_sycl_step(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst) {
+void ggml_sycl_step(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
     GGML_SYCL_DEBUG("call %s\n", __func__);
-    ggml_sycl_op_flatten(ctx, src0, src1, dst, ggml_sycl_op_step);
+    ggml_sycl_op_flatten(ctx, dst->src[0], dst->src[1], dst, ggml_sycl_op_step);
     GGML_SYCL_DEBUG("call %s done\n", __func__);
 }
-void ggml_sycl_leaky_relu(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst) {
+void ggml_sycl_leaky_relu(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
     GGML_SYCL_DEBUG("call %s\n", __func__);
-    ggml_sycl_op_flatten(ctx, src0, src1, dst, ggml_sycl_op_leaky_relu);
+    ggml_sycl_op_flatten(ctx, dst->src[0], dst->src[1], dst, ggml_sycl_op_leaky_relu);
     GGML_SYCL_DEBUG("call %s done\n", __func__);
 }
-void ggml_sycl_sqr(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst) {
+void ggml_sycl_sqr(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
     GGML_SYCL_DEBUG("call %s\n", __func__);
-    ggml_sycl_op_flatten(ctx, src0, src1, dst, ggml_sycl_op_sqr);
+    ggml_sycl_op_flatten(ctx, dst->src[0], dst->src[1], dst, ggml_sycl_op_sqr);
     GGML_SYCL_DEBUG("call %s done\n", __func__);
 }
-void ggml_sycl_upscale(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst) {
+void ggml_sycl_upscale(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
     GGML_SYCL_DEBUG("call %s\n", __func__);
-    ggml_sycl_op_flatten(ctx, src0, src1, dst, ggml_sycl_op_upscale);
+    ggml_sycl_op_flatten(ctx, dst->src[0], dst->src[1], dst, ggml_sycl_op_upscale);
     GGML_SYCL_DEBUG("call %s done\n", __func__);
 }
-void ggml_sycl_pad(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst) {
+void ggml_sycl_pad(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
     GGML_SYCL_DEBUG("call %s\n", __func__);
-    ggml_sycl_op_flatten(ctx, src0, src1, dst, ggml_sycl_op_pad);
+    ggml_sycl_op_flatten(ctx, dst->src[0], dst->src[1], dst, ggml_sycl_op_pad);
     GGML_SYCL_DEBUG("call %s done\n", __func__);
 }
-void ggml_sycl_add(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst) {
+void ggml_sycl_add(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
     GGML_SYCL_DEBUG("call %s\n", __func__);
-    ggml_sycl_op_flatten(ctx, src0, src1, dst, ggml_sycl_op_add);
+    ggml_sycl_op_flatten(ctx, dst->src[0], dst->src[1], dst, ggml_sycl_op_add);
     GGML_SYCL_DEBUG("call %s done\n", __func__);
 }
-void ggml_sycl_sub(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst) {
+void ggml_sycl_sub(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
     GGML_SYCL_DEBUG("call %s\n", __func__);
-    ggml_sycl_op_flatten(ctx, src0, src1, dst, ggml_sycl_op_sub);
+    ggml_sycl_op_flatten(ctx, dst->src[0], dst->src[1], dst, ggml_sycl_op_sub);
     GGML_SYCL_DEBUG("call %s done\n", __func__);
 }
-void ggml_sycl_mul(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst) {
+void ggml_sycl_mul(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
     GGML_SYCL_DEBUG("call %s\n", __func__);
-    ggml_sycl_op_flatten(ctx, src0, src1, dst, ggml_sycl_op_mul);
+    ggml_sycl_op_flatten(ctx, dst->src[0], dst->src[1], dst, ggml_sycl_op_mul);
     GGML_SYCL_DEBUG("call %s done\n", __func__);
 }
-void ggml_sycl_div(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst) {
+void ggml_sycl_div(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
     GGML_SYCL_DEBUG("call %s\n", __func__);
-    ggml_sycl_op_flatten(ctx, src0, src1, dst, ggml_sycl_op_div);
+    ggml_sycl_op_flatten(ctx, dst->src[0], dst->src[1], dst, ggml_sycl_op_div);
     GGML_SYCL_DEBUG("call %s done\n", __func__);
 }

package/src/llama.cpp/ggml/src/ggml-sycl/element_wise.hpp CHANGED Viewed

@@ -25,52 +25,52 @@ static __dpct_inline__ float op_div(const float a, const float b) {
 }
-void ggml_sycl_sqrt(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst);
+void ggml_sycl_sqrt(ggml_backend_sycl_context & ctx, ggml_tensor * dst);
-void ggml_sycl_sin(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst);
+void ggml_sycl_sin(ggml_backend_sycl_context & ctx, ggml_tensor * dst);
-void ggml_sycl_cos(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst);
+void ggml_sycl_cos(ggml_backend_sycl_context & ctx, ggml_tensor * dst);
-void ggml_sycl_acc(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst);
+void ggml_sycl_acc(ggml_backend_sycl_context & ctx, ggml_tensor * dst);
-void ggml_sycl_gelu(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst);
+void ggml_sycl_gelu(ggml_backend_sycl_context & ctx, ggml_tensor * dst);
-void ggml_sycl_silu(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst);
+void ggml_sycl_silu(ggml_backend_sycl_context & ctx, ggml_tensor * dst);
-void ggml_sycl_gelu_quick(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst);
+void ggml_sycl_gelu_quick(ggml_backend_sycl_context & ctx, ggml_tensor * dst);
-void ggml_sycl_tanh(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst);
+void ggml_sycl_tanh(ggml_backend_sycl_context & ctx, ggml_tensor * dst);
-void ggml_sycl_relu(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst);
+void ggml_sycl_relu(ggml_backend_sycl_context & ctx, ggml_tensor * dst);
-void ggml_sycl_sigmoid(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst);
+void ggml_sycl_sigmoid(ggml_backend_sycl_context & ctx, ggml_tensor * dst);
-void ggml_sycl_hardsigmoid(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst);
+void ggml_sycl_hardsigmoid(ggml_backend_sycl_context & ctx, ggml_tensor * dst);
-void ggml_sycl_hardswish(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst);
+void ggml_sycl_hardswish(ggml_backend_sycl_context & ctx, ggml_tensor * dst);
-void ggml_sycl_exp(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst);
+void ggml_sycl_exp(ggml_backend_sycl_context & ctx, ggml_tensor * dst);
-void ggml_sycl_log(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst);
+void ggml_sycl_log(ggml_backend_sycl_context & ctx, ggml_tensor * dst);
-void ggml_sycl_neg(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst);
+void ggml_sycl_neg(ggml_backend_sycl_context & ctx, ggml_tensor * dst);
-void ggml_sycl_step(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst);
+void ggml_sycl_step(ggml_backend_sycl_context & ctx, ggml_tensor * dst);
-void ggml_sycl_leaky_relu(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst);
+void ggml_sycl_leaky_relu(ggml_backend_sycl_context & ctx, ggml_tensor * dst);
-void ggml_sycl_sqr(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst);
+void ggml_sycl_sqr(ggml_backend_sycl_context & ctx, ggml_tensor * dst);
-void ggml_sycl_upscale(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst);
+void ggml_sycl_upscale(ggml_backend_sycl_context & ctx, ggml_tensor * dst);
-void ggml_sycl_pad(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst);
+void ggml_sycl_pad(ggml_backend_sycl_context & ctx, ggml_tensor * dst);
-void ggml_sycl_add(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst);
+void ggml_sycl_add(ggml_backend_sycl_context & ctx, ggml_tensor * dst);
-void ggml_sycl_sub(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst);
+void ggml_sycl_sub(ggml_backend_sycl_context & ctx, ggml_tensor * dst);
-void ggml_sycl_mul(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst);
+void ggml_sycl_mul(ggml_backend_sycl_context & ctx, ggml_tensor * dst);
-void ggml_sycl_div(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst);
+void ggml_sycl_div(ggml_backend_sycl_context & ctx, ggml_tensor * dst);
 #endif // GGML_SYCL_ELEMENTWISE_HPP