cui-llama.rn 1.2.4 → 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
  1. package/README.md +3 -4
  2. package/android/src/main/CMakeLists.txt +21 -5
  3. package/android/src/main/java/com/rnllama/LlamaContext.java +115 -30
  4. package/android/src/main/java/com/rnllama/RNLlama.java +40 -7
  5. package/android/src/main/jni.cpp +222 -36
  6. package/android/src/newarch/java/com/rnllama/RNLlamaModule.java +9 -4
  7. package/android/src/oldarch/java/com/rnllama/RNLlamaModule.java +9 -4
  8. package/cpp/common.cpp +1682 -2122
  9. package/cpp/common.h +600 -594
  10. package/cpp/ggml-aarch64.c +129 -3209
  11. package/cpp/ggml-aarch64.h +19 -39
  12. package/cpp/ggml-alloc.c +1040 -1040
  13. package/cpp/ggml-alloc.h +76 -76
  14. package/cpp/ggml-backend-impl.h +216 -227
  15. package/cpp/ggml-backend-reg.cpp +195 -0
  16. package/cpp/ggml-backend.cpp +1997 -2625
  17. package/cpp/ggml-backend.h +328 -326
  18. package/cpp/ggml-common.h +1853 -1853
  19. package/cpp/ggml-cpp.h +38 -0
  20. package/cpp/ggml-cpu-aarch64.c +3560 -0
  21. package/cpp/ggml-cpu-aarch64.h +30 -0
  22. package/cpp/ggml-cpu-impl.h +371 -614
  23. package/cpp/ggml-cpu-quants.c +10822 -0
  24. package/cpp/ggml-cpu-quants.h +63 -0
  25. package/cpp/ggml-cpu.c +13975 -0
  26. package/cpp/ggml-cpu.cpp +663 -0
  27. package/cpp/ggml-cpu.h +177 -0
  28. package/cpp/ggml-impl.h +550 -209
  29. package/cpp/ggml-metal.h +66 -66
  30. package/cpp/ggml-metal.m +4294 -3819
  31. package/cpp/ggml-quants.c +5247 -15752
  32. package/cpp/ggml-quants.h +100 -147
  33. package/cpp/ggml-threading.cpp +12 -0
  34. package/cpp/ggml-threading.h +12 -0
  35. package/cpp/ggml.c +8180 -23464
  36. package/cpp/ggml.h +2411 -2562
  37. package/cpp/llama-grammar.cpp +1138 -1138
  38. package/cpp/llama-grammar.h +144 -144
  39. package/cpp/llama-impl.h +181 -181
  40. package/cpp/llama-sampling.cpp +2348 -2194
  41. package/cpp/llama-sampling.h +48 -30
  42. package/cpp/llama-vocab.cpp +1984 -1968
  43. package/cpp/llama-vocab.h +170 -165
  44. package/cpp/llama.cpp +22132 -21969
  45. package/cpp/llama.h +1253 -1253
  46. package/cpp/log.cpp +401 -401
  47. package/cpp/log.h +121 -121
  48. package/cpp/rn-llama.hpp +83 -19
  49. package/cpp/sampling.cpp +466 -458
  50. package/cpp/sgemm.cpp +1884 -1219
  51. package/ios/RNLlama.mm +43 -20
  52. package/ios/RNLlamaContext.h +9 -3
  53. package/ios/RNLlamaContext.mm +133 -33
  54. package/jest/mock.js +0 -1
  55. package/lib/commonjs/NativeRNLlama.js.map +1 -1
  56. package/lib/commonjs/index.js +52 -15
  57. package/lib/commonjs/index.js.map +1 -1
  58. package/lib/module/NativeRNLlama.js.map +1 -1
  59. package/lib/module/index.js +51 -15
  60. package/lib/module/index.js.map +1 -1
  61. package/lib/typescript/NativeRNLlama.d.ts +29 -6
  62. package/lib/typescript/NativeRNLlama.d.ts.map +1 -1
  63. package/lib/typescript/index.d.ts +12 -5
  64. package/lib/typescript/index.d.ts.map +1 -1
  65. package/package.json +1 -1
  66. package/src/NativeRNLlama.ts +41 -7
  67. package/src/index.ts +82 -27
  68. package/cpp/json-schema-to-grammar.cpp +0 -1045
  69. package/cpp/json-schema-to-grammar.h +0 -8
  70. package/cpp/json.hpp +0 -24766
package/cpp/ggml-aarch64.h CHANGED
@@ -1,39 +1,19 @@
1
- // SPDX-FileCopyrightText: Copyright 2024 Arm Ltd.
2
- #pragma once
3
-
4
- #define LM_GGML_COMMON_DECL_C
5
- #include "ggml-common.h"
6
-
7
- #include "ggml.h"
8
-
9
- // GGML internal header
10
-
11
- #ifdef __cplusplus
12
- extern "C" {
13
- #endif
14
-
15
- // Quantization
16
- void quantize_q8_0_4x4(const float * LM_GGML_RESTRICT x, void * LM_GGML_RESTRICT y, int64_t k);
17
- void quantize_q8_0_4x8(const float * LM_GGML_RESTRICT x, void * LM_GGML_RESTRICT y, int64_t k);
18
-
19
- 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);
20
-
21
- // Quantization utilizing an importance matrix (a.k.a. "Activation aWare Quantization")
22
- size_t quantize_q4_0_4x4(const float * LM_GGML_RESTRICT src, void * LM_GGML_RESTRICT dst, int64_t nrows, int64_t n_per_row, const float * imatrix);
23
- size_t quantize_q4_0_4x8(const float * LM_GGML_RESTRICT src, void * LM_GGML_RESTRICT dst, int64_t nrows, int64_t n_per_row, const float * imatrix);
24
- size_t quantize_q4_0_8x8(const float * LM_GGML_RESTRICT src, void * LM_GGML_RESTRICT dst, int64_t nrows, int64_t n_per_row, const float * imatrix);
25
-
26
- // GEMV
27
- 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);
28
- 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);
29
- 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);
30
-
31
- // GEMM
32
- 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);
33
- 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);
34
- 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);
35
-
36
- #ifdef __cplusplus
37
- }
38
- #endif
39
-
1
+ #pragma once
2
+
3
+ #include "ggml.h"
4
+
5
+ // GGML internal header
6
+
7
+ #ifdef __cplusplus
8
+ extern "C" {
9
+ #endif
10
+
11
+ // Quantization utilizing an importance matrix (a.k.a. "Activation aWare Quantization")
12
+ size_t quantize_q4_0_4x4(const float * LM_GGML_RESTRICT src, void * LM_GGML_RESTRICT dst, int64_t nrows, int64_t n_per_row, const float * imatrix);
13
+ size_t quantize_q4_0_4x8(const float * LM_GGML_RESTRICT src, void * LM_GGML_RESTRICT dst, int64_t nrows, int64_t n_per_row, const float * imatrix);
14
+ size_t quantize_q4_0_8x8(const float * LM_GGML_RESTRICT src, void * LM_GGML_RESTRICT dst, int64_t nrows, int64_t n_per_row, const float * imatrix);
15
+
16
+ #ifdef __cplusplus
17
+ }
18
+ #endif
19
+