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
  1. package/README.md +3 -2
  2. package/android/src/main/CMakeLists.txt +20 -5
  3. package/android/src/main/java/com/rnllama/LlamaContext.java +115 -27
  4. package/android/src/main/java/com/rnllama/RNLlama.java +40 -7
  5. package/android/src/main/jni.cpp +222 -34
  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 -2114
  9. package/cpp/common.h +600 -613
  10. package/cpp/ggml-aarch64.c +129 -3478
  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 -216
  15. package/cpp/ggml-backend-reg.cpp +195 -0
  16. package/cpp/ggml-backend.cpp +1997 -2661
  17. package/cpp/ggml-backend.h +328 -314
  18. package/cpp/ggml-common.h +1853 -1853
  19. package/cpp/ggml-cpp.h +38 -38
  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 -13720
  26. package/cpp/ggml-cpu.cpp +663 -0
  27. package/cpp/ggml-cpu.h +177 -150
  28. package/cpp/ggml-impl.h +550 -296
  29. package/cpp/ggml-metal.h +66 -66
  30. package/cpp/ggml-metal.m +4294 -3933
  31. package/cpp/ggml-quants.c +5247 -15739
  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 -8390
  36. package/cpp/ggml.h +2411 -2441
  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 -2345
  41. package/cpp/llama-sampling.h +48 -48
  42. package/cpp/llama-vocab.cpp +1984 -1984
  43. package/cpp/llama-vocab.h +170 -170
  44. package/cpp/llama.cpp +22132 -22046
  45. package/cpp/llama.h +1253 -1255
  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 -466
  50. package/cpp/sgemm.cpp +1884 -1276
  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 -5
  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 -6
  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-cpu-quants.h ADDED
@@ -0,0 +1,63 @@
1
+ #pragma once
2
+
3
+ #define LM_GGML_COMMON_DECL_C
4
+ #include "ggml-common.h"
5
+
6
+ #include "ggml.h"
7
+
8
+ // GGML CPU internal header
9
+
10
+ #ifdef __cplusplus
11
+ extern "C" {
12
+ #endif
13
+
14
+ // Quantization
15
+ void quantize_row_q4_0(const float * LM_GGML_RESTRICT x, void * LM_GGML_RESTRICT y, int64_t k);
16
+ void quantize_row_q4_1(const float * LM_GGML_RESTRICT x, void * LM_GGML_RESTRICT y, int64_t k);
17
+ void quantize_row_q5_0(const float * LM_GGML_RESTRICT x, void * LM_GGML_RESTRICT y, int64_t k);
18
+ void quantize_row_q5_1(const float * LM_GGML_RESTRICT x, void * LM_GGML_RESTRICT y, int64_t k);
19
+ void quantize_row_q8_0(const float * LM_GGML_RESTRICT x, void * LM_GGML_RESTRICT y, int64_t k);
20
+ void quantize_row_q8_1(const float * LM_GGML_RESTRICT x, void * LM_GGML_RESTRICT y, int64_t k);
21
+
22
+ void quantize_row_q2_K(const float * LM_GGML_RESTRICT x, void * LM_GGML_RESTRICT y, int64_t k);
23
+ void quantize_row_q3_K(const float * LM_GGML_RESTRICT x, void * LM_GGML_RESTRICT y, int64_t k);
24
+ void quantize_row_q4_K(const float * LM_GGML_RESTRICT x, void * LM_GGML_RESTRICT y, int64_t k);
25
+ void quantize_row_q5_K(const float * LM_GGML_RESTRICT x, void * LM_GGML_RESTRICT y, int64_t k);
26
+ void quantize_row_q6_K(const float * LM_GGML_RESTRICT x, void * LM_GGML_RESTRICT y, int64_t k);
27
+ void quantize_row_q8_K(const float * LM_GGML_RESTRICT x, void * LM_GGML_RESTRICT y, int64_t k);
28
+
29
+ void quantize_row_tq1_0(const float * LM_GGML_RESTRICT x, void * LM_GGML_RESTRICT y, int64_t k);
30
+ void quantize_row_tq2_0(const float * LM_GGML_RESTRICT x, void * LM_GGML_RESTRICT y, int64_t k);
31
+
32
+ void quantize_row_iq4_nl (const float * LM_GGML_RESTRICT x, void * LM_GGML_RESTRICT y, int64_t k);
33
+ void quantize_row_iq4_xs (const float * LM_GGML_RESTRICT x, void * LM_GGML_RESTRICT y, int64_t k);
34
+
35
+ // Dot product
36
+ void lm_ggml_vec_dot_q4_0_q8_0(int n, float * LM_GGML_RESTRICT s, size_t bs, const void * LM_GGML_RESTRICT vx, size_t bx, const void * LM_GGML_RESTRICT vy, size_t by, int nrc);
37
+ void lm_ggml_vec_dot_q4_1_q8_1(int n, float * LM_GGML_RESTRICT s, size_t bs, const void * LM_GGML_RESTRICT vx, size_t bx, const void * LM_GGML_RESTRICT vy, size_t by, int nrc);
38
+ void lm_ggml_vec_dot_q5_0_q8_0(int n, float * LM_GGML_RESTRICT s, size_t bs, const void * LM_GGML_RESTRICT vx, size_t bx, const void * LM_GGML_RESTRICT vy, size_t by, int nrc);
39
+ void lm_ggml_vec_dot_q5_1_q8_1(int n, float * LM_GGML_RESTRICT s, size_t bs, const void * LM_GGML_RESTRICT vx, size_t bx, const void * LM_GGML_RESTRICT vy, size_t by, int nrc);
40
+ void lm_ggml_vec_dot_q8_0_q8_0(int n, float * LM_GGML_RESTRICT s, size_t bs, const void * LM_GGML_RESTRICT vx, size_t bx, const void * LM_GGML_RESTRICT vy, size_t by, int nrc);
41
+
42
+ void lm_ggml_vec_dot_q2_K_q8_K(int n, float * LM_GGML_RESTRICT s, size_t bs, const void * LM_GGML_RESTRICT vx, size_t bx, const void * LM_GGML_RESTRICT vy, size_t by, int nrc);
43
+ void lm_ggml_vec_dot_q3_K_q8_K(int n, float * LM_GGML_RESTRICT s, size_t bs, const void * LM_GGML_RESTRICT vx, size_t bx, const void * LM_GGML_RESTRICT vy, size_t by, int nrc);
44
+ void lm_ggml_vec_dot_q4_K_q8_K(int n, float * LM_GGML_RESTRICT s, size_t bs, const void * LM_GGML_RESTRICT vx, size_t bx, const void * LM_GGML_RESTRICT vy, size_t by, int nrc);
45
+ void lm_ggml_vec_dot_q5_K_q8_K(int n, float * LM_GGML_RESTRICT s, size_t bs, const void * LM_GGML_RESTRICT vx, size_t bx, const void * LM_GGML_RESTRICT vy, size_t by, int nrc);
46
+ void lm_ggml_vec_dot_q6_K_q8_K(int n, float * LM_GGML_RESTRICT s, size_t bs, const void * LM_GGML_RESTRICT vx, size_t bx, const void * LM_GGML_RESTRICT vy, size_t by, int nrc);
47
+
48
+ void lm_ggml_vec_dot_tq1_0_q8_K(int n, float * LM_GGML_RESTRICT s, size_t bs, const void * LM_GGML_RESTRICT vx, size_t bx, const void * LM_GGML_RESTRICT vy, size_t by, int nrc);
49
+ void lm_ggml_vec_dot_tq2_0_q8_K(int n, float * LM_GGML_RESTRICT s, size_t bs, const void * LM_GGML_RESTRICT vx, size_t bx, const void * LM_GGML_RESTRICT vy, size_t by, int nrc);
50
+
51
+ void lm_ggml_vec_dot_iq2_xxs_q8_K(int n, float * LM_GGML_RESTRICT s, size_t bs, const void * LM_GGML_RESTRICT vx, size_t bx, const void * LM_GGML_RESTRICT vy, size_t by, int nrc);
52
+ void lm_ggml_vec_dot_iq2_xs_q8_K (int n, float * LM_GGML_RESTRICT s, size_t bs, const void * LM_GGML_RESTRICT vx, size_t bx, const void * LM_GGML_RESTRICT vy, size_t by, int nrc);
53
+ void lm_ggml_vec_dot_iq2_s_q8_K (int n, float * LM_GGML_RESTRICT s, size_t bs, const void * LM_GGML_RESTRICT vx, size_t bx, const void * LM_GGML_RESTRICT vy, size_t by, int nrc);
54
+ void lm_ggml_vec_dot_iq3_xxs_q8_K(int n, float * LM_GGML_RESTRICT s, size_t bs, const void * LM_GGML_RESTRICT vx, size_t bx, const void * LM_GGML_RESTRICT vy, size_t by, int nrc);
55
+ void lm_ggml_vec_dot_iq1_s_q8_K (int n, float * LM_GGML_RESTRICT s, size_t bs, const void * LM_GGML_RESTRICT vx, size_t bx, const void * LM_GGML_RESTRICT vy, size_t by, int nrc);
56
+ void lm_ggml_vec_dot_iq1_m_q8_K (int n, float * LM_GGML_RESTRICT s, size_t bs, const void * LM_GGML_RESTRICT vx, size_t bx, const void * LM_GGML_RESTRICT vy, size_t by, int nrc);
57
+ void lm_ggml_vec_dot_iq4_nl_q8_0 (int n, float * LM_GGML_RESTRICT s, size_t bs, const void * LM_GGML_RESTRICT vx, size_t bx, const void * LM_GGML_RESTRICT vy, size_t by, int nrc);
58
+ void lm_ggml_vec_dot_iq4_xs_q8_K (int n, float * LM_GGML_RESTRICT s, size_t bs, const void * LM_GGML_RESTRICT vx, size_t bx, const void * LM_GGML_RESTRICT vy, size_t by, int nrc);
59
+ void lm_ggml_vec_dot_iq3_s_q8_K (int n, float * LM_GGML_RESTRICT s, size_t bs, const void * LM_GGML_RESTRICT vx, size_t bx, const void * LM_GGML_RESTRICT vy, size_t by, int nrc);
60
+
61
+ #ifdef __cplusplus
62
+ }
63
+ #endif