npm - llama-cpp-capacitor - Versions diffs - 0.0.5 → 0.0.7 - Mend

llama-cpp-capacitor 0.0.5 → 0.0.7

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 (149) hide show

package/cpp/LICENSE +21 -0
package/cpp/README.md +4 -0
package/cpp/anyascii.c +22223 -0
package/cpp/anyascii.h +42 -0
package/cpp/chat-parser.cpp +393 -0
package/cpp/chat-parser.h +120 -0
package/cpp/chat.cpp +2315 -0
package/cpp/chat.h +221 -0
package/cpp/common.cpp +1619 -0
package/cpp/common.h +744 -0
package/cpp/ggml-alloc.c +1028 -0
package/cpp/ggml-alloc.h +76 -0
package/cpp/ggml-backend-impl.h +255 -0
package/cpp/ggml-backend-reg.cpp +600 -0
package/cpp/ggml-backend.cpp +2118 -0
package/cpp/ggml-backend.h +354 -0
package/cpp/ggml-common.h +1878 -0
package/cpp/ggml-cpp.h +39 -0
package/cpp/ggml-cpu/amx/amx.cpp +221 -0
package/cpp/ggml-cpu/amx/amx.h +8 -0
package/cpp/ggml-cpu/amx/common.h +91 -0
package/cpp/ggml-cpu/amx/mmq.cpp +2512 -0
package/cpp/ggml-cpu/amx/mmq.h +10 -0
package/cpp/ggml-cpu/arch/arm/cpu-feats.cpp +94 -0
package/cpp/ggml-cpu/arch/arm/quants.c +3650 -0
package/cpp/ggml-cpu/arch/arm/repack.cpp +1891 -0
package/cpp/ggml-cpu/arch/x86/cpu-feats.cpp +327 -0
package/cpp/ggml-cpu/arch/x86/quants.c +3820 -0
package/cpp/ggml-cpu/arch/x86/repack.cpp +6307 -0
package/cpp/ggml-cpu/arch-fallback.h +215 -0
package/cpp/ggml-cpu/binary-ops.cpp +158 -0
package/cpp/ggml-cpu/binary-ops.h +16 -0
package/cpp/ggml-cpu/common.h +73 -0
package/cpp/ggml-cpu/ggml-cpu-impl.h +525 -0
package/cpp/ggml-cpu/ggml-cpu.c +3578 -0
package/cpp/ggml-cpu/ggml-cpu.cpp +672 -0
package/cpp/ggml-cpu/ops.cpp +10587 -0
package/cpp/ggml-cpu/ops.h +114 -0
package/cpp/ggml-cpu/quants.c +1193 -0
package/cpp/ggml-cpu/quants.h +97 -0
package/cpp/ggml-cpu/repack.cpp +1982 -0
package/cpp/ggml-cpu/repack.h +120 -0
package/cpp/ggml-cpu/simd-mappings.h +1184 -0
package/cpp/ggml-cpu/traits.cpp +36 -0
package/cpp/ggml-cpu/traits.h +38 -0
package/cpp/ggml-cpu/unary-ops.cpp +186 -0
package/cpp/ggml-cpu/unary-ops.h +28 -0
package/cpp/ggml-cpu/vec.cpp +348 -0
package/cpp/ggml-cpu/vec.h +1121 -0
package/cpp/ggml-cpu.h +145 -0
package/cpp/ggml-impl.h +622 -0
package/cpp/ggml-metal-impl.h +688 -0
package/cpp/ggml-metal.h +66 -0
package/cpp/ggml-metal.m +6833 -0
package/cpp/ggml-opt.cpp +1093 -0
package/cpp/ggml-opt.h +256 -0
package/cpp/ggml-quants.c +5324 -0
package/cpp/ggml-quants.h +106 -0
package/cpp/ggml-threading.cpp +12 -0
package/cpp/ggml-threading.h +14 -0
package/cpp/ggml.c +7108 -0
package/cpp/ggml.h +2492 -0
package/cpp/gguf.cpp +1358 -0
package/cpp/gguf.h +202 -0
package/cpp/json-partial.cpp +256 -0
package/cpp/json-partial.h +38 -0
package/cpp/json-schema-to-grammar.cpp +985 -0
package/cpp/json-schema-to-grammar.h +21 -0
package/cpp/llama-adapter.cpp +388 -0
package/cpp/llama-adapter.h +76 -0
package/cpp/llama-arch.cpp +2355 -0
package/cpp/llama-arch.h +499 -0
package/cpp/llama-batch.cpp +875 -0
package/cpp/llama-batch.h +160 -0
package/cpp/llama-chat.cpp +783 -0
package/cpp/llama-chat.h +65 -0
package/cpp/llama-context.cpp +2748 -0
package/cpp/llama-context.h +306 -0
package/cpp/llama-cparams.cpp +5 -0
package/cpp/llama-cparams.h +41 -0
package/cpp/llama-cpp.h +30 -0
package/cpp/llama-grammar.cpp +1229 -0
package/cpp/llama-grammar.h +173 -0
package/cpp/llama-graph.cpp +1891 -0
package/cpp/llama-graph.h +810 -0
package/cpp/llama-hparams.cpp +180 -0
package/cpp/llama-hparams.h +233 -0
package/cpp/llama-impl.cpp +167 -0
package/cpp/llama-impl.h +61 -0
package/cpp/llama-io.cpp +15 -0
package/cpp/llama-io.h +35 -0
package/cpp/llama-kv-cache-iswa.cpp +318 -0
package/cpp/llama-kv-cache-iswa.h +135 -0
package/cpp/llama-kv-cache.cpp +2059 -0
package/cpp/llama-kv-cache.h +374 -0
package/cpp/llama-kv-cells.h +491 -0
package/cpp/llama-memory-hybrid.cpp +258 -0
package/cpp/llama-memory-hybrid.h +137 -0
package/cpp/llama-memory-recurrent.cpp +1146 -0
package/cpp/llama-memory-recurrent.h +179 -0
package/cpp/llama-memory.cpp +59 -0
package/cpp/llama-memory.h +119 -0
package/cpp/llama-mmap.cpp +600 -0
package/cpp/llama-mmap.h +68 -0
package/cpp/llama-model-loader.cpp +1164 -0
package/cpp/llama-model-loader.h +170 -0
package/cpp/llama-model-saver.cpp +282 -0
package/cpp/llama-model-saver.h +37 -0
package/cpp/llama-model.cpp +19042 -0
package/cpp/llama-model.h +491 -0
package/cpp/llama-sampling.cpp +2575 -0
package/cpp/llama-sampling.h +32 -0
package/cpp/llama-vocab.cpp +3792 -0
package/cpp/llama-vocab.h +176 -0
package/cpp/llama.cpp +358 -0
package/cpp/llama.h +1373 -0
package/cpp/log.cpp +427 -0
package/cpp/log.h +103 -0
package/cpp/minja/chat-template.hpp +550 -0
package/cpp/minja/minja.hpp +3009 -0
package/cpp/nlohmann/json.hpp +25526 -0
package/cpp/nlohmann/json_fwd.hpp +187 -0
package/cpp/regex-partial.cpp +204 -0
package/cpp/regex-partial.h +56 -0
package/cpp/rn-completion.cpp +681 -0
package/cpp/rn-completion.h +116 -0
package/cpp/rn-llama.cpp +345 -0
package/cpp/rn-llama.h +149 -0
package/cpp/rn-mtmd.hpp +602 -0
package/cpp/rn-tts.cpp +591 -0
package/cpp/rn-tts.h +59 -0
package/cpp/sampling.cpp +579 -0
package/cpp/sampling.h +107 -0
package/cpp/tools/mtmd/clip-impl.h +473 -0
package/cpp/tools/mtmd/clip.cpp +4322 -0
package/cpp/tools/mtmd/clip.h +106 -0
package/cpp/tools/mtmd/miniaudio/miniaudio.h +93468 -0
package/cpp/tools/mtmd/mtmd-audio.cpp +769 -0
package/cpp/tools/mtmd/mtmd-audio.h +47 -0
package/cpp/tools/mtmd/mtmd-helper.cpp +460 -0
package/cpp/tools/mtmd/mtmd-helper.h +91 -0
package/cpp/tools/mtmd/mtmd.cpp +1066 -0
package/cpp/tools/mtmd/mtmd.h +298 -0
package/cpp/tools/mtmd/stb/stb_image.h +7988 -0
package/cpp/unicode-data.cpp +7034 -0
package/cpp/unicode-data.h +20 -0
package/cpp/unicode.cpp +1061 -0
package/cpp/unicode.h +68 -0
package/package.json +2 -1

package/cpp/ggml-cpu/ops.h ADDED Viewed

@@ -0,0 +1,114 @@
+#pragma once
+#include "ggml.h"
+//
+// cache line
+//
+#if defined(__cpp_lib_hardware_interference_size)
+#define CACHE_LINE_SIZE std::hardware_destructive_interference_size
+#else
+#if defined(__POWER9_VECTOR__)
+#define CACHE_LINE_SIZE 128
+#elif defined(__VXE__) || defined(__VXE2__)
+#define CACHE_LINE_SIZE 256
+#else
+#define CACHE_LINE_SIZE 64
+#endif
+#endif
+static const size_t CACHE_LINE_SIZE_F32 = CACHE_LINE_SIZE/sizeof(float);
+// Work buffer size for im2col operations in CONV2D
+#define LM_GGML_IM2COL_WORK_SIZE (16 * 1024 * 1024)
+#ifdef __cplusplus
+extern "C" {
+#endif
+void lm_ggml_compute_forward_dup(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_add(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_add_id(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_add1(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_acc(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_sum(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_sum_rows(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_mean(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_argmax(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_count_equal(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_repeat(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_repeat_back(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_concat(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_silu_back(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_norm(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_rms_norm(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_rms_norm_back(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_group_norm(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_l2_norm(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_out_prod(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_scale(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_set(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_cpy(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_cont(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_reshape(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_view(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_permute(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_transpose(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_get_rows(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_get_rows_back(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_set_rows(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_diag(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_diag_mask_inf(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_diag_mask_zero(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_soft_max(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_soft_max_ext_back(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_rope(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_rope_back(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_clamp(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_conv_transpose_1d(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_im2col(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_im2col_back_f32(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_conv_2d(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_conv_3d(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_conv_transpose_2d(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_conv_2d_dw(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_pool_1d(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_pool_2d(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_pool_2d_back(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_upscale(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_pad(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_pad_reflect_1d(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_roll(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_arange(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_timestep_embedding(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_argsort(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_leaky_relu(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_flash_attn_ext(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_flash_attn_back(
+        const struct lm_ggml_compute_params * params,
+        const bool masked,
+        struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_ssm_conv(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_ssm_scan(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_win_part(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_win_unpart(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_unary(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_glu(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_get_rel_pos(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_add_rel_pos(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_rwkv_wkv6(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_rwkv_wkv7(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_gla(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_map_custom1(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_map_custom2(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_map_custom3(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_custom(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_cross_entropy_loss(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_cross_entropy_loss_back(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_opt_step_adamw(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_mul_mat(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+void lm_ggml_compute_forward_opt_step_sgd(const struct lm_ggml_compute_params * params, struct lm_ggml_tensor * dst);
+#ifdef __cplusplus
+}
+#endif