cui-llama.rn 1.4.0 → 1.4.1

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 (73) hide show
  1. package/android/src/main/jni.cpp +9 -9
  2. package/cpp/common.cpp +163 -60
  3. package/cpp/common.h +43 -12
  4. package/cpp/ggml-alloc.c +1042 -1037
  5. package/cpp/ggml-backend-impl.h +255 -256
  6. package/cpp/ggml-backend-reg.cpp +582 -582
  7. package/cpp/ggml-backend.cpp +2002 -2002
  8. package/cpp/ggml-backend.h +354 -352
  9. package/cpp/ggml-common.h +1853 -1853
  10. package/cpp/ggml-cpp.h +39 -39
  11. package/cpp/ggml-cpu-aarch64.cpp +4247 -4247
  12. package/cpp/ggml-cpu-aarch64.h +8 -8
  13. package/cpp/ggml-cpu-impl.h +386 -386
  14. package/cpp/ggml-cpu-quants.c +10920 -10839
  15. package/cpp/ggml-cpu-traits.cpp +36 -36
  16. package/cpp/ggml-cpu-traits.h +38 -38
  17. package/cpp/ggml-cpu.c +329 -60
  18. package/cpp/ggml-cpu.cpp +10 -2
  19. package/cpp/ggml-cpu.h +135 -135
  20. package/cpp/ggml-impl.h +567 -567
  21. package/cpp/ggml-metal-impl.h +17 -17
  22. package/cpp/ggml-metal.m +4884 -4884
  23. package/cpp/ggml-quants.c +5238 -5238
  24. package/cpp/ggml-threading.h +14 -14
  25. package/cpp/ggml.c +6514 -6448
  26. package/cpp/ggml.h +2194 -2163
  27. package/cpp/gguf.cpp +1329 -1325
  28. package/cpp/gguf.h +202 -202
  29. package/cpp/json-schema-to-grammar.cpp +1045 -1045
  30. package/cpp/json-schema-to-grammar.h +8 -8
  31. package/cpp/json.hpp +24766 -24766
  32. package/cpp/llama-adapter.cpp +347 -346
  33. package/cpp/llama-adapter.h +74 -73
  34. package/cpp/llama-arch.cpp +1487 -1434
  35. package/cpp/llama-arch.h +400 -395
  36. package/cpp/llama-batch.cpp +368 -368
  37. package/cpp/llama-batch.h +88 -88
  38. package/cpp/llama-chat.cpp +578 -567
  39. package/cpp/llama-chat.h +52 -51
  40. package/cpp/llama-context.cpp +1775 -1771
  41. package/cpp/llama-context.h +128 -128
  42. package/cpp/llama-cparams.cpp +1 -1
  43. package/cpp/llama-cparams.h +37 -37
  44. package/cpp/llama-cpp.h +30 -30
  45. package/cpp/llama-grammar.cpp +1139 -1139
  46. package/cpp/llama-grammar.h +143 -143
  47. package/cpp/llama-hparams.cpp +71 -71
  48. package/cpp/llama-hparams.h +139 -140
  49. package/cpp/llama-impl.cpp +167 -167
  50. package/cpp/llama-impl.h +61 -61
  51. package/cpp/llama-kv-cache.cpp +718 -718
  52. package/cpp/llama-kv-cache.h +218 -218
  53. package/cpp/llama-mmap.cpp +2 -1
  54. package/cpp/llama-mmap.h +67 -67
  55. package/cpp/llama-model-loader.cpp +1124 -1011
  56. package/cpp/llama-model-loader.h +167 -158
  57. package/cpp/llama-model.cpp +3997 -2202
  58. package/cpp/llama-model.h +370 -391
  59. package/cpp/llama-sampling.cpp +2408 -2406
  60. package/cpp/llama-sampling.h +32 -48
  61. package/cpp/llama-vocab.cpp +3247 -1982
  62. package/cpp/llama-vocab.h +125 -182
  63. package/cpp/llama.cpp +416 -2886
  64. package/cpp/llama.h +1323 -1285
  65. package/cpp/log.cpp +401 -401
  66. package/cpp/log.h +121 -121
  67. package/cpp/rn-llama.hpp +18 -12
  68. package/cpp/sampling.cpp +505 -500
  69. package/cpp/sgemm.cpp +2597 -2597
  70. package/cpp/speculative.cpp +277 -274
  71. package/cpp/speculative.h +28 -28
  72. package/cpp/unicode.cpp +2 -3
  73. package/package.json +1 -1
package/cpp/llama-chat.h CHANGED
@@ -1,51 +1,52 @@
1
- #pragma once
2
-
3
- #include <string>
4
- #include <vector>
5
- #include <cstdint>
6
-
7
- enum llm_chat_template {
8
- LLM_CHAT_TEMPLATE_CHATML,
9
- LLM_CHAT_TEMPLATE_LLAMA_2,
10
- LLM_CHAT_TEMPLATE_LLAMA_2_SYS,
11
- LLM_CHAT_TEMPLATE_LLAMA_2_SYS_BOS,
12
- LLM_CHAT_TEMPLATE_LLAMA_2_SYS_STRIP,
13
- LLM_CHAT_TEMPLATE_MISTRAL_V1,
14
- LLM_CHAT_TEMPLATE_MISTRAL_V3,
15
- LLM_CHAT_TEMPLATE_MISTRAL_V3_TEKKEN,
16
- LLM_CHAT_TEMPLATE_MISTRAL_V7,
17
- LLM_CHAT_TEMPLATE_PHI_3,
18
- LLM_CHAT_TEMPLATE_FALCON_3,
19
- LLM_CHAT_TEMPLATE_ZEPHYR,
20
- LLM_CHAT_TEMPLATE_MONARCH,
21
- LLM_CHAT_TEMPLATE_GEMMA,
22
- LLM_CHAT_TEMPLATE_ORION,
23
- LLM_CHAT_TEMPLATE_OPENCHAT,
24
- LLM_CHAT_TEMPLATE_VICUNA,
25
- LLM_CHAT_TEMPLATE_VICUNA_ORCA,
26
- LLM_CHAT_TEMPLATE_DEEPSEEK,
27
- LLM_CHAT_TEMPLATE_DEEPSEEK_2,
28
- LLM_CHAT_TEMPLATE_DEEPSEEK_3,
29
- LLM_CHAT_TEMPLATE_COMMAND_R,
30
- LLM_CHAT_TEMPLATE_LLAMA_3,
31
- LLM_CHAT_TEMPLATE_CHATGML_3,
32
- LLM_CHAT_TEMPLATE_CHATGML_4,
33
- LLM_CHAT_TEMPLATE_MINICPM,
34
- LLM_CHAT_TEMPLATE_EXAONE_3,
35
- LLM_CHAT_TEMPLATE_RWKV_WORLD,
36
- LLM_CHAT_TEMPLATE_GRANITE,
37
- LLM_CHAT_TEMPLATE_GIGACHAT,
38
- LLM_CHAT_TEMPLATE_MEGREZ,
39
- LLM_CHAT_TEMPLATE_UNKNOWN,
40
- };
41
-
42
- struct llama_chat_message;
43
-
44
- llm_chat_template llm_chat_template_from_str(const std::string & name);
45
-
46
- llm_chat_template llm_chat_detect_template(const std::string & tmpl);
47
-
48
- int32_t llm_chat_apply_template(
49
- llm_chat_template tmpl,
50
- const std::vector<const llama_chat_message *> & chat,
51
- std::string & dest, bool add_ass);
1
+ #pragma once
2
+
3
+ #include <string>
4
+ #include <vector>
5
+ #include <cstdint>
6
+
7
+ enum llm_chat_template {
8
+ LLM_CHAT_TEMPLATE_CHATML,
9
+ LLM_CHAT_TEMPLATE_LLAMA_2,
10
+ LLM_CHAT_TEMPLATE_LLAMA_2_SYS,
11
+ LLM_CHAT_TEMPLATE_LLAMA_2_SYS_BOS,
12
+ LLM_CHAT_TEMPLATE_LLAMA_2_SYS_STRIP,
13
+ LLM_CHAT_TEMPLATE_MISTRAL_V1,
14
+ LLM_CHAT_TEMPLATE_MISTRAL_V3,
15
+ LLM_CHAT_TEMPLATE_MISTRAL_V3_TEKKEN,
16
+ LLM_CHAT_TEMPLATE_MISTRAL_V7,
17
+ LLM_CHAT_TEMPLATE_PHI_3,
18
+ LLM_CHAT_TEMPLATE_PHI_4,
19
+ LLM_CHAT_TEMPLATE_FALCON_3,
20
+ LLM_CHAT_TEMPLATE_ZEPHYR,
21
+ LLM_CHAT_TEMPLATE_MONARCH,
22
+ LLM_CHAT_TEMPLATE_GEMMA,
23
+ LLM_CHAT_TEMPLATE_ORION,
24
+ LLM_CHAT_TEMPLATE_OPENCHAT,
25
+ LLM_CHAT_TEMPLATE_VICUNA,
26
+ LLM_CHAT_TEMPLATE_VICUNA_ORCA,
27
+ LLM_CHAT_TEMPLATE_DEEPSEEK,
28
+ LLM_CHAT_TEMPLATE_DEEPSEEK_2,
29
+ LLM_CHAT_TEMPLATE_DEEPSEEK_3,
30
+ LLM_CHAT_TEMPLATE_COMMAND_R,
31
+ LLM_CHAT_TEMPLATE_LLAMA_3,
32
+ LLM_CHAT_TEMPLATE_CHATGML_3,
33
+ LLM_CHAT_TEMPLATE_CHATGML_4,
34
+ LLM_CHAT_TEMPLATE_MINICPM,
35
+ LLM_CHAT_TEMPLATE_EXAONE_3,
36
+ LLM_CHAT_TEMPLATE_RWKV_WORLD,
37
+ LLM_CHAT_TEMPLATE_GRANITE,
38
+ LLM_CHAT_TEMPLATE_GIGACHAT,
39
+ LLM_CHAT_TEMPLATE_MEGREZ,
40
+ LLM_CHAT_TEMPLATE_UNKNOWN,
41
+ };
42
+
43
+ struct llama_chat_message;
44
+
45
+ llm_chat_template llm_chat_template_from_str(const std::string & name);
46
+
47
+ llm_chat_template llm_chat_detect_template(const std::string & tmpl);
48
+
49
+ int32_t llm_chat_apply_template(
50
+ llm_chat_template tmpl,
51
+ const std::vector<const llama_chat_message *> & chat,
52
+ std::string & dest, bool add_ass);