llama_cpp 0.14.2 → 0.14.4

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 (33) hide show
  1. checksums.yaml +4 -4
  2. data/CHANGELOG.md +14 -0
  3. data/ext/llama_cpp/llama_cpp.cpp +64 -0
  4. data/lib/llama_cpp/version.rb +2 -2
  5. data/sig/llama_cpp.rbs +6 -0
  6. data/vendor/tmp/llama.cpp/Makefile +91 -21
  7. data/vendor/tmp/llama.cpp/ggml-alloc.c +14 -5
  8. data/vendor/tmp/llama.cpp/ggml-backend-impl.h +5 -0
  9. data/vendor/tmp/llama.cpp/ggml-backend.c +155 -125
  10. data/vendor/tmp/llama.cpp/ggml-backend.h +4 -4
  11. data/vendor/tmp/llama.cpp/ggml-common.h +25 -2
  12. data/vendor/tmp/llama.cpp/ggml-cuda.cu +1779 -10762
  13. data/vendor/tmp/llama.cpp/ggml-cuda.h +6 -15
  14. data/vendor/tmp/llama.cpp/ggml-kompute.cpp +5 -0
  15. data/vendor/tmp/llama.cpp/ggml-metal.m +167 -124
  16. data/vendor/tmp/llama.cpp/ggml-metal.metal +603 -303
  17. data/vendor/tmp/llama.cpp/ggml-opencl.cpp +5 -0
  18. data/vendor/tmp/llama.cpp/ggml-quants.c +663 -56
  19. data/vendor/tmp/llama.cpp/ggml-quants.h +3 -0
  20. data/vendor/tmp/llama.cpp/ggml-sycl.cpp +341 -469
  21. data/vendor/tmp/llama.cpp/ggml-sycl.h +19 -4
  22. data/vendor/tmp/llama.cpp/ggml-vulkan-shaders.hpp +37199 -14939
  23. data/vendor/tmp/llama.cpp/ggml-vulkan.cpp +335 -307
  24. data/vendor/tmp/llama.cpp/ggml-vulkan.h +0 -11
  25. data/vendor/tmp/llama.cpp/ggml.c +229 -107
  26. data/vendor/tmp/llama.cpp/ggml.h +11 -5
  27. data/vendor/tmp/llama.cpp/llama.cpp +2136 -464
  28. data/vendor/tmp/llama.cpp/llama.h +86 -23
  29. data/vendor/tmp/llama.cpp/unicode-data.cpp +1651 -0
  30. data/vendor/tmp/llama.cpp/unicode-data.h +16 -0
  31. data/vendor/tmp/llama.cpp/unicode.cpp +8 -1403
  32. data/vendor/tmp/llama.cpp/unicode.h +2 -0
  33. metadata +5 -3
data/vendor/tmp/llama.cpp/ggml-quants.h CHANGED
@@ -72,6 +72,7 @@ void dequantize_row_iq2_xs (const block_iq2_xs * GGML_RESTRICT x, float * GGML_
72
72
  void dequantize_row_iq2_s (const block_iq2_s * GGML_RESTRICT x, float * GGML_RESTRICT y, int k);
73
73
  void dequantize_row_iq3_xxs(const block_iq3_xxs * GGML_RESTRICT x, float * GGML_RESTRICT y, int k);
74
74
  void dequantize_row_iq1_s (const block_iq1_s * GGML_RESTRICT x, float * GGML_RESTRICT y, int k);
75
+ void dequantize_row_iq1_m (const block_iq1_m * GGML_RESTRICT x, float * GGML_RESTRICT y, int k);
75
76
  void dequantize_row_iq4_nl (const block_iq4_nl * GGML_RESTRICT x, float * GGML_RESTRICT y, int k);
76
77
  void dequantize_row_iq4_xs (const block_iq4_xs * GGML_RESTRICT x, float * GGML_RESTRICT y, int k);
77
78
  void dequantize_row_iq3_s (const block_iq3_s * GGML_RESTRICT x, float * GGML_RESTRICT y, int k);
@@ -94,6 +95,7 @@ void ggml_vec_dot_iq2_xs_q8_K (int n, float * GGML_RESTRICT s, size_t bs, const
94
95
  void ggml_vec_dot_iq2_s_q8_K (int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
95
96
  void ggml_vec_dot_iq3_xxs_q8_K(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
96
97
  void ggml_vec_dot_iq1_s_q8_K (int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
98
+ void ggml_vec_dot_iq1_m_q8_K (int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
97
99
  void ggml_vec_dot_iq4_nl_q8_0 (int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
98
100
  void ggml_vec_dot_iq4_xs_q8_K (int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
99
101
  void ggml_vec_dot_iq3_s_q8_K (int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
@@ -104,6 +106,7 @@ size_t quantize_iq2_xs (const float * GGML_RESTRICT src, void * GGML_RESTRICT ds
104
106
  size_t quantize_iq2_s (const float * GGML_RESTRICT src, void * GGML_RESTRICT dst, int nrows, int n_per_row, const float * imatrix);
105
107
  size_t quantize_iq3_xxs(const float * GGML_RESTRICT src, void * GGML_RESTRICT dst, int nrows, int n_per_row, const float * imatrix);
106
108
  size_t quantize_iq1_s (const float * GGML_RESTRICT src, void * GGML_RESTRICT dst, int nrows, int n_per_row, const float * imatrix);
109
+ size_t quantize_iq1_m (const float * GGML_RESTRICT src, void * GGML_RESTRICT dst, int nrows, int n_per_row, const float * imatrix);
107
110
  size_t quantize_iq4_nl (const float * GGML_RESTRICT src, void * GGML_RESTRICT dst, int nrows, int n_per_row, const float * imatrix);
108
111
  size_t quantize_iq4_xs (const float * GGML_RESTRICT src, void * GGML_RESTRICT dst, int nrows, int n_per_row, const float * imatrix);
109
112
  size_t quantize_iq3_s (const float * GGML_RESTRICT src, void * GGML_RESTRICT dst, int nrows, int n_per_row, const float * imatrix);