cui-llama.rn 1.3.0 → 1.3.3

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 (76) hide show
  1. package/android/src/main/CMakeLists.txt +6 -1
  2. package/android/src/main/jni.cpp +6 -6
  3. package/cpp/amx/amx.cpp +196 -0
  4. package/cpp/amx/amx.h +20 -0
  5. package/cpp/amx/common.h +101 -0
  6. package/cpp/amx/mmq.cpp +2524 -0
  7. package/cpp/amx/mmq.h +16 -0
  8. package/cpp/common.cpp +1981 -1682
  9. package/cpp/common.h +636 -600
  10. package/cpp/ggml-aarch64.c +129 -129
  11. package/cpp/ggml-aarch64.h +19 -19
  12. package/cpp/ggml-alloc.c +1038 -1040
  13. package/cpp/ggml-alloc.h +76 -76
  14. package/cpp/ggml-backend-impl.h +238 -216
  15. package/cpp/ggml-backend-reg.cpp +423 -195
  16. package/cpp/ggml-backend.cpp +1999 -1997
  17. package/cpp/ggml-backend.h +351 -328
  18. package/cpp/ggml-common.h +1859 -1853
  19. package/cpp/ggml-cpp.h +38 -38
  20. package/cpp/ggml-cpu-aarch64.c +3823 -3560
  21. package/cpp/ggml-cpu-aarch64.h +32 -30
  22. package/cpp/ggml-cpu-impl.h +386 -371
  23. package/cpp/ggml-cpu-quants.c +10835 -10822
  24. package/cpp/ggml-cpu-quants.h +63 -63
  25. package/cpp/ggml-cpu.c +99 -103
  26. package/cpp/ggml-cpu.cpp +69 -17
  27. package/cpp/ggml-cpu.h +152 -177
  28. package/cpp/ggml-impl.h +556 -550
  29. package/cpp/ggml-metal.h +66 -66
  30. package/cpp/ggml-metal.m +4426 -4294
  31. package/cpp/ggml-quants.c +5247 -5247
  32. package/cpp/ggml-quants.h +100 -100
  33. package/cpp/ggml-threading.cpp +12 -12
  34. package/cpp/ggml-threading.h +12 -12
  35. package/cpp/ggml.c +7618 -8180
  36. package/cpp/ggml.h +2255 -2411
  37. package/cpp/json-schema-to-grammar.cpp +1045 -0
  38. package/cpp/json-schema-to-grammar.h +8 -0
  39. package/cpp/json.hpp +24766 -0
  40. package/cpp/llama-grammar.cpp +1138 -1138
  41. package/cpp/llama-grammar.h +144 -144
  42. package/cpp/llama-impl.h +181 -181
  43. package/cpp/llama-sampling.cpp +2348 -2348
  44. package/cpp/llama-sampling.h +48 -48
  45. package/cpp/llama-vocab.cpp +1984 -1984
  46. package/cpp/llama-vocab.h +170 -170
  47. package/cpp/llama.cpp +22332 -22132
  48. package/cpp/llama.h +1259 -1253
  49. package/cpp/log.cpp +401 -401
  50. package/cpp/log.h +121 -121
  51. package/cpp/rn-llama.hpp +6 -6
  52. package/cpp/sampling.cpp +505 -466
  53. package/cpp/sampling.h +22 -1
  54. package/cpp/sgemm.cpp +1884 -1884
  55. package/cpp/speculative.cpp +270 -0
  56. package/cpp/speculative.h +28 -0
  57. package/cpp/unicode.cpp +11 -0
  58. package/ios/RNLlamaContext.mm +13 -0
  59. package/lib/commonjs/NativeRNLlama.js.map +1 -1
  60. package/lib/commonjs/grammar.js +4 -2
  61. package/lib/commonjs/grammar.js.map +1 -1
  62. package/lib/commonjs/index.js.map +1 -1
  63. package/lib/module/NativeRNLlama.js.map +1 -1
  64. package/lib/module/grammar.js +2 -1
  65. package/lib/module/grammar.js.map +1 -1
  66. package/lib/module/index.js.map +1 -1
  67. package/lib/typescript/NativeRNLlama.d.ts +94 -4
  68. package/lib/typescript/NativeRNLlama.d.ts.map +1 -1
  69. package/lib/typescript/grammar.d.ts +5 -6
  70. package/lib/typescript/grammar.d.ts.map +1 -1
  71. package/lib/typescript/index.d.ts +4 -2
  72. package/lib/typescript/index.d.ts.map +1 -1
  73. package/package.json +2 -1
  74. package/src/NativeRNLlama.ts +97 -10
  75. package/src/grammar.ts +10 -8
  76. package/src/index.ts +22 -1
package/cpp/llama-sampling.h CHANGED
@@ -1,48 +1,48 @@
1
- #pragma once
2
-
3
- // TODO: rename llama-sampling.h/.cpp to llama-sampler.h/.cpp ?
4
-
5
- #include "llama-grammar.h"
6
-
7
- struct llama_vocab;
8
- struct llama_grammar;
9
-
10
- // sampler chain
11
-
12
- struct llama_sampler_chain {
13
- llama_sampler_chain_params params;
14
-
15
- std::vector<struct llama_sampler *> samplers;
16
-
17
- // timing
18
-
19
- mutable int64_t t_sample_us;
20
-
21
- mutable int32_t n_sample;
22
- };
23
-
24
- struct llama_sampler * llama_sampler_init_grammar_impl(
25
- const struct llama_vocab & vocab,
26
- const char * grammar_str,
27
- const char * grammar_root);
28
-
29
- struct llama_sampler * llama_sampler_init_infill_impl(
30
- const struct llama_vocab & vocab);
31
-
32
- struct llama_sampler * llama_sampler_init_dry_impl(
33
- const struct llama_vocab & vocab,
34
- int32_t context_size,
35
- float dry_multiplier,
36
- float dry_base,
37
- int32_t dry_allowed_length,
38
- int32_t dry_penalty_last_n,
39
- const char ** seq_breakers,
40
- size_t num_breakers);
41
-
42
- struct llama_sampler * llama_sampler_init_dry_testing(
43
- int32_t context_size,
44
- float dry_multiplier,
45
- float dry_base,
46
- int32_t dry_allowed_length,
47
- int32_t dry_penalty_last_n,
48
- const std::vector<std::vector<llama_token>>& seq_breakers);
1
+ #pragma once
2
+
3
+ // TODO: rename llama-sampling.h/.cpp to llama-sampler.h/.cpp ?
4
+
5
+ #include "llama-grammar.h"
6
+
7
+ struct llama_vocab;
8
+ struct llama_grammar;
9
+
10
+ // sampler chain
11
+
12
+ struct llama_sampler_chain {
13
+ llama_sampler_chain_params params;
14
+
15
+ std::vector<struct llama_sampler *> samplers;
16
+
17
+ // timing
18
+
19
+ mutable int64_t t_sample_us;
20
+
21
+ mutable int32_t n_sample;
22
+ };
23
+
24
+ struct llama_sampler * llama_sampler_init_grammar_impl(
25
+ const struct llama_vocab & vocab,
26
+ const char * grammar_str,
27
+ const char * grammar_root);
28
+
29
+ struct llama_sampler * llama_sampler_init_infill_impl(
30
+ const struct llama_vocab & vocab);
31
+
32
+ struct llama_sampler * llama_sampler_init_dry_impl(
33
+ const struct llama_vocab & vocab,
34
+ int32_t context_size,
35
+ float dry_multiplier,
36
+ float dry_base,
37
+ int32_t dry_allowed_length,
38
+ int32_t dry_penalty_last_n,
39
+ const char ** seq_breakers,
40
+ size_t num_breakers);
41
+
42
+ struct llama_sampler * llama_sampler_init_dry_testing(
43
+ int32_t context_size,
44
+ float dry_multiplier,
45
+ float dry_base,
46
+ int32_t dry_allowed_length,
47
+ int32_t dry_penalty_last_n,
48
+ const std::vector<std::vector<llama_token>>& seq_breakers);