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/ggml-cpu-traits.cpp CHANGED
@@ -1,36 +1,36 @@
1
- #include "ggml-cpu-traits.h"
2
-
3
- #include "ggml-backend-impl.h"
4
- #include "ggml-backend.h"
5
-
6
- namespace ggml::cpu {
7
- tensor_traits::~tensor_traits() {}
8
-
9
- extra_buffer_type::~extra_buffer_type() {}
10
- } // namespace ggml::cpu
11
-
12
- bool lm_ggml_cpu_extra_compute_forward(struct lm_ggml_compute_params * params, struct lm_ggml_tensor * op) {
13
- for (auto extra : lm_ggml_backend_cpu_get_extra_buffers_type()) {
14
- if (extra && extra->context) {
15
- auto buf_extra = (ggml::cpu::extra_buffer_type *) extra->context;
16
- auto tensor_traits = buf_extra->get_tensor_traits(op);
17
- if (tensor_traits && tensor_traits->compute_forward(params, op)) {
18
- return true;
19
- }
20
- }
21
- }
22
- return false;
23
- }
24
-
25
- bool lm_ggml_cpu_extra_work_size(int n_threads, const struct lm_ggml_tensor * op, size_t * size) {
26
- for (auto extra : lm_ggml_backend_cpu_get_extra_buffers_type()) {
27
- if (extra && extra->context) {
28
- auto buf_extra = (ggml::cpu::extra_buffer_type *) extra->context;
29
- auto tensor_traits = buf_extra->get_tensor_traits(op);
30
- if (tensor_traits && tensor_traits->work_size(n_threads, op, *size)) {
31
- return true;
32
- }
33
- }
34
- }
35
- return false;
36
- }
1
+ #include "ggml-cpu-traits.h"
2
+
3
+ #include "ggml-backend-impl.h"
4
+ #include "ggml-backend.h"
5
+
6
+ namespace ggml::cpu {
7
+ tensor_traits::~tensor_traits() {}
8
+
9
+ extra_buffer_type::~extra_buffer_type() {}
10
+ } // namespace ggml::cpu
11
+
12
+ bool lm_ggml_cpu_extra_compute_forward(struct lm_ggml_compute_params * params, struct lm_ggml_tensor * op) {
13
+ for (auto extra : lm_ggml_backend_cpu_get_extra_buffers_type()) {
14
+ if (extra && extra->context) {
15
+ auto buf_extra = (ggml::cpu::extra_buffer_type *) extra->context;
16
+ auto tensor_traits = buf_extra->get_tensor_traits(op);
17
+ if (tensor_traits && tensor_traits->compute_forward(params, op)) {
18
+ return true;
19
+ }
20
+ }
21
+ }
22
+ return false;
23
+ }
24
+
25
+ bool lm_ggml_cpu_extra_work_size(int n_threads, const struct lm_ggml_tensor * op, size_t * size) {
26
+ for (auto extra : lm_ggml_backend_cpu_get_extra_buffers_type()) {
27
+ if (extra && extra->context) {
28
+ auto buf_extra = (ggml::cpu::extra_buffer_type *) extra->context;
29
+ auto tensor_traits = buf_extra->get_tensor_traits(op);
30
+ if (tensor_traits && tensor_traits->work_size(n_threads, op, *size)) {
31
+ return true;
32
+ }
33
+ }
34
+ }
35
+ return false;
36
+ }
package/cpp/ggml-cpu-traits.h CHANGED
@@ -1,38 +1,38 @@
1
- #pragma once
2
- #include "ggml-backend-impl.h"
3
- #include "ggml-cpu-impl.h"
4
- #include "ggml.h"
5
-
6
- #ifdef __cplusplus
7
- # include <vector>
8
- extern "C" {
9
- #endif
10
-
11
- // return true if op part of extra "accelerator"
12
- bool lm_ggml_cpu_extra_compute_forward(struct lm_ggml_compute_params * params, struct lm_ggml_tensor * op);
13
- bool lm_ggml_cpu_extra_work_size(int n_threads, const struct lm_ggml_tensor * op, size_t * size);
14
-
15
- #ifdef __cplusplus
16
- }
17
-
18
- namespace ggml::cpu {
19
- // register in tensor->extra
20
- class tensor_traits {
21
- public:
22
- virtual ~tensor_traits();
23
- virtual bool work_size(int n_threads, const struct lm_ggml_tensor * op, size_t & size) = 0;
24
- virtual bool compute_forward(struct lm_ggml_compute_params * params, struct lm_ggml_tensor * op) = 0;
25
- };
26
-
27
- class extra_buffer_type {
28
- public:
29
- virtual ~extra_buffer_type();
30
- virtual bool supports_op(lm_ggml_backend_dev_t dev, const struct lm_ggml_tensor * op) = 0;
31
- virtual tensor_traits * get_tensor_traits(const struct lm_ggml_tensor * op) = 0;
32
- };
33
- } // namespace ggml::cpu
34
-
35
- // implemented in ggml-cpu.cpp.
36
- std::vector<lm_ggml_backend_buffer_type_t> & lm_ggml_backend_cpu_get_extra_buffers_type();
37
-
38
- #endif
1
+ #pragma once
2
+ #include "ggml-backend-impl.h"
3
+ #include "ggml-cpu-impl.h"
4
+ #include "ggml.h"
5
+
6
+ #ifdef __cplusplus
7
+ # include <vector>
8
+ extern "C" {
9
+ #endif
10
+
11
+ // return true if op part of extra "accelerator"
12
+ bool lm_ggml_cpu_extra_compute_forward(struct lm_ggml_compute_params * params, struct lm_ggml_tensor * op);
13
+ bool lm_ggml_cpu_extra_work_size(int n_threads, const struct lm_ggml_tensor * op, size_t * size);
14
+
15
+ #ifdef __cplusplus
16
+ }
17
+
18
+ namespace ggml::cpu {
19
+ // register in tensor->extra
20
+ class tensor_traits {
21
+ public:
22
+ virtual ~tensor_traits();
23
+ virtual bool work_size(int n_threads, const struct lm_ggml_tensor * op, size_t & size) = 0;
24
+ virtual bool compute_forward(struct lm_ggml_compute_params * params, struct lm_ggml_tensor * op) = 0;
25
+ };
26
+
27
+ class extra_buffer_type {
28
+ public:
29
+ virtual ~extra_buffer_type();
30
+ virtual bool supports_op(lm_ggml_backend_dev_t dev, const struct lm_ggml_tensor * op) = 0;
31
+ virtual tensor_traits * get_tensor_traits(const struct lm_ggml_tensor * op) = 0;
32
+ };
33
+ } // namespace ggml::cpu
34
+
35
+ // implemented in ggml-cpu.cpp.
36
+ std::vector<lm_ggml_backend_buffer_type_t> & lm_ggml_backend_cpu_get_extra_buffers_type();
37
+
38
+ #endif