npm - cui-llama.rn - Versions diffs - 1.2.6 → 1.3.0 - Mend

cui-llama.rn 1.2.6 → 1.3.0

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 (70) hide show

package/README.md +3 -2
package/android/src/main/CMakeLists.txt +20 -5
package/android/src/main/java/com/rnllama/LlamaContext.java +115 -27
package/android/src/main/java/com/rnllama/RNLlama.java +40 -7
package/android/src/main/jni.cpp +222 -34
package/android/src/newarch/java/com/rnllama/RNLlamaModule.java +9 -4
package/android/src/oldarch/java/com/rnllama/RNLlamaModule.java +9 -4
package/cpp/common.cpp +1682 -2114
package/cpp/common.h +600 -613
package/cpp/ggml-aarch64.c +129 -3478
package/cpp/ggml-aarch64.h +19 -39
package/cpp/ggml-alloc.c +1040 -1040
package/cpp/ggml-alloc.h +76 -76
package/cpp/ggml-backend-impl.h +216 -216
package/cpp/ggml-backend-reg.cpp +195 -0
package/cpp/ggml-backend.cpp +1997 -2661
package/cpp/ggml-backend.h +328 -314
package/cpp/ggml-common.h +1853 -1853
package/cpp/ggml-cpp.h +38 -38
package/cpp/ggml-cpu-aarch64.c +3560 -0
package/cpp/ggml-cpu-aarch64.h +30 -0
package/cpp/ggml-cpu-impl.h +371 -614
package/cpp/ggml-cpu-quants.c +10822 -0
package/cpp/ggml-cpu-quants.h +63 -0
package/cpp/ggml-cpu.c +13975 -13720
package/cpp/ggml-cpu.cpp +663 -0
package/cpp/ggml-cpu.h +177 -150
package/cpp/ggml-impl.h +550 -296
package/cpp/ggml-metal.h +66 -66
package/cpp/ggml-metal.m +4294 -3933
package/cpp/ggml-quants.c +5247 -15739
package/cpp/ggml-quants.h +100 -147
package/cpp/ggml-threading.cpp +12 -0
package/cpp/ggml-threading.h +12 -0
package/cpp/ggml.c +8180 -8390
package/cpp/ggml.h +2411 -2441
package/cpp/llama-grammar.cpp +1138 -1138
package/cpp/llama-grammar.h +144 -144
package/cpp/llama-impl.h +181 -181
package/cpp/llama-sampling.cpp +2348 -2345
package/cpp/llama-sampling.h +48 -48
package/cpp/llama-vocab.cpp +1984 -1984
package/cpp/llama-vocab.h +170 -170
package/cpp/llama.cpp +22132 -22046
package/cpp/llama.h +1253 -1255
package/cpp/log.cpp +401 -401
package/cpp/log.h +121 -121
package/cpp/rn-llama.hpp +83 -19
package/cpp/sampling.cpp +466 -466
package/cpp/sgemm.cpp +1884 -1276
package/ios/RNLlama.mm +43 -20
package/ios/RNLlamaContext.h +9 -3
package/ios/RNLlamaContext.mm +133 -33
package/jest/mock.js +0 -1
package/lib/commonjs/NativeRNLlama.js.map +1 -1
package/lib/commonjs/index.js +52 -15
package/lib/commonjs/index.js.map +1 -1
package/lib/module/NativeRNLlama.js.map +1 -1
package/lib/module/index.js +51 -15
package/lib/module/index.js.map +1 -1
package/lib/typescript/NativeRNLlama.d.ts +29 -5
package/lib/typescript/NativeRNLlama.d.ts.map +1 -1
package/lib/typescript/index.d.ts +12 -5
package/lib/typescript/index.d.ts.map +1 -1
package/package.json +1 -1
package/src/NativeRNLlama.ts +41 -6
package/src/index.ts +82 -27
package/cpp/json-schema-to-grammar.cpp +0 -1045
package/cpp/json-schema-to-grammar.h +0 -8
package/cpp/json.hpp +0 -24766

package/cpp/ggml-cpu-aarch64.h ADDED Viewed

@@ -0,0 +1,30 @@
+#pragma once
+#include "ggml.h"
+// GGML internal header
+#ifdef __cplusplus
+extern "C" {
+#endif
+// Quantization
+void quantize_mat_q8_0(const float * LM_GGML_RESTRICT x, void * LM_GGML_RESTRICT y, int64_t nrows, int64_t n_per_row, int64_t blck_size_interleave);
+// GEMV
+void lm_ggml_gemv_q4_0_4x4_q8_0(int n, float * LM_GGML_RESTRICT s, size_t bs, const void * LM_GGML_RESTRICT vx, const void * LM_GGML_RESTRICT vy, int nr, int nc);
+void lm_ggml_gemv_q4_0_4x8_q8_0(int n, float * LM_GGML_RESTRICT s, size_t bs, const void * LM_GGML_RESTRICT vx, const void * LM_GGML_RESTRICT vy, int nr, int nc);
+void lm_ggml_gemv_q4_0_8x8_q8_0(int n, float * LM_GGML_RESTRICT s, size_t bs, const void * LM_GGML_RESTRICT vx, const void * LM_GGML_RESTRICT vy, int nr, int nc);
+// GEMM
+void lm_ggml_gemm_q4_0_4x4_q8_0(int n, float * LM_GGML_RESTRICT s, size_t bs, const void * LM_GGML_RESTRICT vx, const void * LM_GGML_RESTRICT vy, int nr, int nc);
+void lm_ggml_gemm_q4_0_4x8_q8_0(int n, float * LM_GGML_RESTRICT s, size_t bs, const void * LM_GGML_RESTRICT vx, const void * LM_GGML_RESTRICT vy, int nr, int nc);
+void lm_ggml_gemm_q4_0_8x8_q8_0(int n, float * LM_GGML_RESTRICT s, size_t bs, const void * LM_GGML_RESTRICT vx, const void * LM_GGML_RESTRICT vy, int nr, int nc);
+void           lm_ggml_aarch64_repack_tensor(struct lm_ggml_tensor * cur, enum lm_ggml_type repack_type, const void * data, size_t data_size);
+enum lm_ggml_type lm_ggml_aarch64_get_optimal_repack_type(const struct lm_ggml_tensor * cur);
+#ifdef __cplusplus
+}
+#endif