cui-llama.rn 1.4.3 → 1.4.6

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 (134) hide show
  1. package/README.md +93 -114
  2. package/android/src/main/CMakeLists.txt +5 -0
  3. package/android/src/main/java/com/rnllama/LlamaContext.java +91 -17
  4. package/android/src/main/java/com/rnllama/RNLlama.java +37 -4
  5. package/android/src/main/jni-utils.h +6 -0
  6. package/android/src/main/jni.cpp +289 -31
  7. package/android/src/main/jniLibs/arm64-v8a/librnllama.so +0 -0
  8. package/android/src/main/jniLibs/arm64-v8a/librnllama_v8.so +0 -0
  9. package/android/src/main/jniLibs/arm64-v8a/librnllama_v8_2.so +0 -0
  10. package/android/src/main/jniLibs/arm64-v8a/librnllama_v8_2_dotprod.so +0 -0
  11. package/android/src/main/jniLibs/arm64-v8a/librnllama_v8_2_dotprod_i8mm.so +0 -0
  12. package/android/src/main/jniLibs/arm64-v8a/librnllama_v8_2_i8mm.so +0 -0
  13. package/android/src/main/jniLibs/x86_64/librnllama.so +0 -0
  14. package/android/src/main/jniLibs/x86_64/librnllama_x86_64.so +0 -0
  15. package/android/src/newarch/java/com/rnllama/RNLlamaModule.java +7 -2
  16. package/android/src/oldarch/java/com/rnllama/RNLlamaModule.java +7 -2
  17. package/cpp/chat-template.hpp +529 -0
  18. package/cpp/chat.cpp +1779 -0
  19. package/cpp/chat.h +135 -0
  20. package/cpp/common.cpp +2064 -1873
  21. package/cpp/common.h +700 -699
  22. package/cpp/ggml-alloc.c +1039 -1042
  23. package/cpp/ggml-alloc.h +1 -1
  24. package/cpp/ggml-backend-impl.h +255 -255
  25. package/cpp/ggml-backend-reg.cpp +586 -582
  26. package/cpp/ggml-backend.cpp +2004 -2002
  27. package/cpp/ggml-backend.h +354 -354
  28. package/cpp/ggml-common.h +1851 -1853
  29. package/cpp/ggml-cpp.h +39 -39
  30. package/cpp/ggml-cpu-aarch64.cpp +4248 -4247
  31. package/cpp/ggml-cpu-aarch64.h +8 -8
  32. package/cpp/ggml-cpu-impl.h +531 -386
  33. package/cpp/ggml-cpu-quants.c +12527 -10920
  34. package/cpp/ggml-cpu-traits.cpp +36 -36
  35. package/cpp/ggml-cpu-traits.h +38 -38
  36. package/cpp/ggml-cpu.c +15766 -14391
  37. package/cpp/ggml-cpu.cpp +655 -635
  38. package/cpp/ggml-cpu.h +138 -135
  39. package/cpp/ggml-impl.h +567 -567
  40. package/cpp/ggml-metal-impl.h +235 -0
  41. package/cpp/ggml-metal.h +1 -1
  42. package/cpp/ggml-metal.m +5146 -4884
  43. package/cpp/ggml-opt.cpp +854 -854
  44. package/cpp/ggml-opt.h +216 -216
  45. package/cpp/ggml-quants.c +5238 -5238
  46. package/cpp/ggml-threading.h +14 -14
  47. package/cpp/ggml.c +6529 -6514
  48. package/cpp/ggml.h +2198 -2194
  49. package/cpp/gguf.cpp +1329 -1329
  50. package/cpp/gguf.h +202 -202
  51. package/cpp/json-schema-to-grammar.cpp +1024 -1045
  52. package/cpp/json-schema-to-grammar.h +21 -8
  53. package/cpp/json.hpp +24766 -24766
  54. package/cpp/llama-adapter.cpp +347 -347
  55. package/cpp/llama-adapter.h +74 -74
  56. package/cpp/llama-arch.cpp +1513 -1487
  57. package/cpp/llama-arch.h +403 -400
  58. package/cpp/llama-batch.cpp +368 -368
  59. package/cpp/llama-batch.h +88 -88
  60. package/cpp/llama-chat.cpp +588 -578
  61. package/cpp/llama-chat.h +53 -52
  62. package/cpp/llama-context.cpp +1775 -1775
  63. package/cpp/llama-context.h +128 -128
  64. package/cpp/llama-cparams.cpp +1 -1
  65. package/cpp/llama-cparams.h +37 -37
  66. package/cpp/llama-cpp.h +30 -30
  67. package/cpp/llama-grammar.cpp +1219 -1139
  68. package/cpp/llama-grammar.h +173 -143
  69. package/cpp/llama-hparams.cpp +71 -71
  70. package/cpp/llama-hparams.h +139 -139
  71. package/cpp/llama-impl.cpp +167 -167
  72. package/cpp/llama-impl.h +61 -61
  73. package/cpp/llama-kv-cache.cpp +718 -718
  74. package/cpp/llama-kv-cache.h +219 -218
  75. package/cpp/llama-mmap.cpp +600 -590
  76. package/cpp/llama-mmap.h +68 -67
  77. package/cpp/llama-model-loader.cpp +1124 -1124
  78. package/cpp/llama-model-loader.h +167 -167
  79. package/cpp/llama-model.cpp +4087 -3997
  80. package/cpp/llama-model.h +370 -370
  81. package/cpp/llama-sampling.cpp +2558 -2408
  82. package/cpp/llama-sampling.h +32 -32
  83. package/cpp/llama-vocab.cpp +3264 -3247
  84. package/cpp/llama-vocab.h +125 -125
  85. package/cpp/llama.cpp +10284 -10077
  86. package/cpp/llama.h +1354 -1323
  87. package/cpp/log.cpp +393 -401
  88. package/cpp/log.h +132 -121
  89. package/cpp/minja/chat-template.hpp +529 -0
  90. package/cpp/minja/minja.hpp +2915 -0
  91. package/cpp/minja.hpp +2915 -0
  92. package/cpp/rn-llama.cpp +66 -6
  93. package/cpp/rn-llama.h +26 -1
  94. package/cpp/sampling.cpp +570 -505
  95. package/cpp/sampling.h +3 -0
  96. package/cpp/sgemm.cpp +2598 -2597
  97. package/cpp/sgemm.h +14 -14
  98. package/cpp/speculative.cpp +278 -277
  99. package/cpp/speculative.h +28 -28
  100. package/cpp/unicode.cpp +9 -2
  101. package/ios/CMakeLists.txt +6 -0
  102. package/ios/RNLlama.h +0 -8
  103. package/ios/RNLlama.mm +27 -3
  104. package/ios/RNLlamaContext.h +10 -1
  105. package/ios/RNLlamaContext.mm +269 -57
  106. package/jest/mock.js +21 -2
  107. package/lib/commonjs/NativeRNLlama.js.map +1 -1
  108. package/lib/commonjs/grammar.js +3 -0
  109. package/lib/commonjs/grammar.js.map +1 -1
  110. package/lib/commonjs/index.js +87 -13
  111. package/lib/commonjs/index.js.map +1 -1
  112. package/lib/module/NativeRNLlama.js.map +1 -1
  113. package/lib/module/grammar.js +3 -0
  114. package/lib/module/grammar.js.map +1 -1
  115. package/lib/module/index.js +86 -13
  116. package/lib/module/index.js.map +1 -1
  117. package/lib/typescript/NativeRNLlama.d.ts +107 -2
  118. package/lib/typescript/NativeRNLlama.d.ts.map +1 -1
  119. package/lib/typescript/grammar.d.ts.map +1 -1
  120. package/lib/typescript/index.d.ts +32 -7
  121. package/lib/typescript/index.d.ts.map +1 -1
  122. package/llama-rn.podspec +1 -1
  123. package/package.json +3 -2
  124. package/src/NativeRNLlama.ts +115 -3
  125. package/src/grammar.ts +3 -0
  126. package/src/index.ts +138 -21
  127. package/android/src/main/build-arm64/CMakeFiles/3.31.4/CMakeCCompiler.cmake +0 -81
  128. package/android/src/main/build-arm64/CMakeFiles/3.31.4/CMakeSystem.cmake +0 -15
  129. package/android/src/main/build-arm64/CMakeFiles/3.31.4/CompilerIdC/CMakeCCompilerId.c +0 -904
  130. package/android/src/main/build-arm64/CMakeFiles/3.31.4/CompilerIdC/CMakeCCompilerId.o +0 -0
  131. package/android/src/main/build-arm64/CMakeFiles/3.31.4/CompilerIdCXX/CMakeCXXCompilerId.cpp +0 -919
  132. package/android/src/main/build-arm64/CMakeFiles/3.31.4/CompilerIdCXX/CMakeCXXCompilerId.o +0 -0
  133. package/android/src/main/build-arm64/CMakeFiles/CMakeConfigureLog.yaml +0 -55
  134. package/cpp/rn-llama.hpp +0 -913
package/cpp/llama-vocab.h CHANGED
@@ -1,125 +1,125 @@
1
- #pragma once
2
-
3
- #include "llama.h"
4
-
5
- #include <string>
6
- #include <vector>
7
- #include <memory>
8
-
9
- struct LLM_KV;
10
- struct llama_model_loader;
11
-
12
- struct llama_vocab {
13
- struct token_data {
14
- std::string text;
15
- float score;
16
- llama_token_attr attr;
17
- };
18
-
19
- llama_vocab();
20
- ~llama_vocab();
21
-
22
- void load(llama_model_loader & ml, const LLM_KV & kv);
23
-
24
- enum llama_vocab_type get_type() const;
25
- enum llama_vocab_pre_type get_pre_type() const;
26
-
27
- uint32_t n_tokens() const;
28
- uint32_t n_token_types() const;
29
-
30
- std::string type_name() const;
31
-
32
- bool is_normal (llama_token id) const;
33
- bool is_unknown (llama_token id) const;
34
- bool is_control (llama_token id) const;
35
- bool is_byte (llama_token id) const;
36
- bool is_user_defined(llama_token id) const;
37
- bool is_unused (llama_token id) const;
38
- bool is_eog (llama_token id) const;
39
-
40
- uint8_t token_to_byte(llama_token id) const;
41
- llama_token byte_to_token(uint8_t ch) const;
42
-
43
- llama_token text_to_token(const std::string & text) const;
44
-
45
- const token_data & get_token_data(llama_token id) const;
46
-
47
- const char * token_get_text (llama_token id) const;
48
- float token_get_score(llama_token id) const;
49
- llama_token_attr token_get_attr (llama_token id) const;
50
-
51
- llama_token token_bos() const;
52
- llama_token token_eos() const;
53
- llama_token token_eot() const;
54
- llama_token token_eom() const;
55
- llama_token token_unk() const;
56
- llama_token token_sep() const;
57
- llama_token token_nl () const;
58
- llama_token token_pad() const;
59
-
60
- llama_token token_prefix() const;
61
- llama_token token_middle() const;
62
- llama_token token_suffix() const;
63
-
64
- llama_token token_fim_pre() const;
65
- llama_token token_fim_suf() const;
66
- llama_token token_fim_mid() const;
67
- llama_token token_fim_pad() const;
68
- llama_token token_fim_rep() const;
69
- llama_token token_fim_sep() const;
70
-
71
- bool get_add_space_prefix () const;
72
- bool get_add_bos () const;
73
- bool get_add_eos () const;
74
- bool get_ignore_merges () const;
75
- bool get_clean_spaces () const;
76
- bool get_remove_extra_whitespaces () const;
77
- bool get_escape_whitespaces () const;
78
- bool get_treat_whitespace_as_suffix() const;
79
-
80
- int max_token_len() const;
81
-
82
- int find_bpe_rank(const std::string & token_left, const std::string & token_right) const;
83
-
84
- int32_t tokenize(
85
- const char * text,
86
- int32_t text_len,
87
- llama_token * tokens,
88
- int32_t n_tokens_max,
89
- bool add_special,
90
- bool parse_special) const;
91
-
92
- std::vector<llama_token> tokenize(
93
- const std::string & raw_text,
94
- bool add_special,
95
- bool parse_special = false) const;
96
-
97
- // does not write null-terminator to buf
98
- int32_t token_to_piece(
99
- llama_token token,
100
- char * buf,
101
- int32_t length,
102
- int32_t lstrip,
103
- bool special) const;
104
-
105
- // use cached data
106
- const std::string & token_to_piece(llama_token token) const;
107
-
108
- int32_t detokenize(
109
- const llama_token * tokens,
110
- int32_t n_tokens,
111
- char * text,
112
- int32_t text_len_max,
113
- bool remove_special,
114
- bool unparse_special) const;
115
-
116
- std::string detokenize(
117
- const std::vector<llama_token> & tokens,
118
- bool special) const;
119
-
120
- void print_info() const;
121
-
122
- private:
123
- struct impl;
124
- std::unique_ptr<impl> pimpl;
125
- };
1
+ #pragma once
2
+
3
+ #include "llama.h"
4
+
5
+ #include <string>
6
+ #include <vector>
7
+ #include <memory>
8
+
9
+ struct LLM_KV;
10
+ struct llama_model_loader;
11
+
12
+ struct llama_vocab {
13
+ struct token_data {
14
+ std::string text;
15
+ float score;
16
+ llama_token_attr attr;
17
+ };
18
+
19
+ llama_vocab();
20
+ ~llama_vocab();
21
+
22
+ void load(llama_model_loader & ml, const LLM_KV & kv);
23
+
24
+ enum llama_vocab_type get_type() const;
25
+ enum llama_vocab_pre_type get_pre_type() const;
26
+
27
+ uint32_t n_tokens() const;
28
+ uint32_t n_token_types() const;
29
+
30
+ std::string type_name() const;
31
+
32
+ bool is_normal (llama_token id) const;
33
+ bool is_unknown (llama_token id) const;
34
+ bool is_control (llama_token id) const;
35
+ bool is_byte (llama_token id) const;
36
+ bool is_user_defined(llama_token id) const;
37
+ bool is_unused (llama_token id) const;
38
+ bool is_eog (llama_token id) const;
39
+
40
+ uint8_t token_to_byte(llama_token id) const;
41
+ llama_token byte_to_token(uint8_t ch) const;
42
+
43
+ llama_token text_to_token(const std::string & text) const;
44
+
45
+ const token_data & get_token_data(llama_token id) const;
46
+
47
+ const char * token_get_text (llama_token id) const;
48
+ float token_get_score(llama_token id) const;
49
+ llama_token_attr token_get_attr (llama_token id) const;
50
+
51
+ llama_token token_bos() const;
52
+ llama_token token_eos() const;
53
+ llama_token token_eot() const;
54
+ llama_token token_eom() const;
55
+ llama_token token_unk() const;
56
+ llama_token token_sep() const;
57
+ llama_token token_nl () const;
58
+ llama_token token_pad() const;
59
+
60
+ llama_token token_prefix() const;
61
+ llama_token token_middle() const;
62
+ llama_token token_suffix() const;
63
+
64
+ llama_token token_fim_pre() const;
65
+ llama_token token_fim_suf() const;
66
+ llama_token token_fim_mid() const;
67
+ llama_token token_fim_pad() const;
68
+ llama_token token_fim_rep() const;
69
+ llama_token token_fim_sep() const;
70
+
71
+ bool get_add_space_prefix () const;
72
+ bool get_add_bos () const;
73
+ bool get_add_eos () const;
74
+ bool get_ignore_merges () const;
75
+ bool get_clean_spaces () const;
76
+ bool get_remove_extra_whitespaces () const;
77
+ bool get_escape_whitespaces () const;
78
+ bool get_treat_whitespace_as_suffix() const;
79
+
80
+ int max_token_len() const;
81
+
82
+ int find_bpe_rank(const std::string & token_left, const std::string & token_right) const;
83
+
84
+ int32_t tokenize(
85
+ const char * text,
86
+ int32_t text_len,
87
+ llama_token * tokens,
88
+ int32_t n_tokens_max,
89
+ bool add_special,
90
+ bool parse_special) const;
91
+
92
+ std::vector<llama_token> tokenize(
93
+ const std::string & raw_text,
94
+ bool add_special,
95
+ bool parse_special = false) const;
96
+
97
+ // does not write null-terminator to buf
98
+ int32_t token_to_piece(
99
+ llama_token token,
100
+ char * buf,
101
+ int32_t length,
102
+ int32_t lstrip,
103
+ bool special) const;
104
+
105
+ // use cached data
106
+ const std::string & token_to_piece(llama_token token) const;
107
+
108
+ int32_t detokenize(
109
+ const llama_token * tokens,
110
+ int32_t n_tokens,
111
+ char * text,
112
+ int32_t text_len_max,
113
+ bool remove_special,
114
+ bool unparse_special) const;
115
+
116
+ std::string detokenize(
117
+ const std::vector<llama_token> & tokens,
118
+ bool special) const;
119
+
120
+ void print_info() const;
121
+
122
+ private:
123
+ struct impl;
124
+ std::unique_ptr<impl> pimpl;
125
+ };