llama_cpp 0.15.4 → 0.16.1

Sign up to get free protection for your applications and to get access to all the features.
Files changed (161) hide show
  1. checksums.yaml +4 -4
  2. data/CHANGELOG.md +16 -0
  3. data/ext/llama_cpp/extconf.rb +3 -2
  4. data/ext/llama_cpp/llama_cpp.cpp +17 -3
  5. data/lib/llama_cpp/version.rb +2 -2
  6. data/sig/llama_cpp.rbs +15 -1
  7. data/vendor/tmp/llama.cpp/Makefile +166 -82
  8. data/vendor/tmp/llama.cpp/ggml-alloc.c +82 -26
  9. data/vendor/tmp/llama.cpp/ggml-backend-impl.h +20 -8
  10. data/vendor/tmp/llama.cpp/ggml-backend.c +183 -69
  11. data/vendor/tmp/llama.cpp/ggml-backend.h +4 -4
  12. data/vendor/tmp/llama.cpp/ggml-blas.cpp +363 -0
  13. data/vendor/tmp/llama.cpp/ggml-blas.h +23 -0
  14. data/vendor/tmp/llama.cpp/ggml-common.h +6 -0
  15. data/vendor/tmp/llama.cpp/ggml-cuda/acc.cu +47 -0
  16. data/vendor/tmp/llama.cpp/ggml-cuda/arange.cu +34 -0
  17. data/vendor/tmp/llama.cpp/ggml-cuda/argsort.cu +104 -0
  18. data/vendor/tmp/llama.cpp/ggml-cuda/binbcast.cu +280 -0
  19. data/vendor/tmp/llama.cpp/ggml-cuda/clamp.cu +34 -0
  20. data/vendor/tmp/llama.cpp/ggml-cuda/concat.cu +196 -0
  21. data/vendor/tmp/llama.cpp/ggml-cuda/convert.cu +686 -0
  22. data/vendor/tmp/llama.cpp/ggml-cuda/cpy.cu +490 -0
  23. data/vendor/tmp/llama.cpp/ggml-cuda/diagmask.cu +40 -0
  24. data/vendor/tmp/llama.cpp/ggml-cuda/dmmv.cu +674 -0
  25. data/vendor/tmp/llama.cpp/ggml-cuda/fattn-tile-f16.cu +319 -0
  26. data/vendor/tmp/llama.cpp/ggml-cuda/fattn-tile-f32.cu +312 -0
  27. data/vendor/tmp/llama.cpp/ggml-cuda/fattn.cu +345 -0
  28. data/vendor/tmp/llama.cpp/ggml-cuda/getrows.cu +178 -0
  29. data/vendor/tmp/llama.cpp/ggml-cuda/im2col.cu +104 -0
  30. data/vendor/tmp/llama.cpp/ggml-cuda/mmq.cu +88 -0
  31. data/vendor/tmp/llama.cpp/ggml-cuda/mmvq.cu +419 -0
  32. data/vendor/tmp/llama.cpp/ggml-cuda/norm.cu +221 -0
  33. data/vendor/tmp/llama.cpp/ggml-cuda/pad.cu +49 -0
  34. data/vendor/tmp/llama.cpp/ggml-cuda/pool2d.cu +94 -0
  35. data/vendor/tmp/llama.cpp/ggml-cuda/quantize.cu +112 -0
  36. data/vendor/tmp/llama.cpp/ggml-cuda/rope.cu +271 -0
  37. data/vendor/tmp/llama.cpp/ggml-cuda/scale.cu +31 -0
  38. data/vendor/tmp/llama.cpp/ggml-cuda/softmax.cu +206 -0
  39. data/vendor/tmp/llama.cpp/ggml-cuda/sumrows.cu +40 -0
  40. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-f16.cu +5 -0
  41. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q4_0.cu +5 -0
  42. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q4_1.cu +5 -0
  43. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q5_0.cu +5 -0
  44. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q5_1.cu +5 -0
  45. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q8_0.cu +5 -0
  46. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-f16.cu +5 -0
  47. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q4_0.cu +5 -0
  48. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q4_1.cu +5 -0
  49. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q5_0.cu +5 -0
  50. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q5_1.cu +5 -0
  51. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q8_0.cu +5 -0
  52. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-f16.cu +5 -0
  53. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q4_0.cu +5 -0
  54. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q4_1.cu +5 -0
  55. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q5_0.cu +5 -0
  56. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q5_1.cu +5 -0
  57. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q8_0.cu +5 -0
  58. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-f16.cu +5 -0
  59. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q4_0.cu +5 -0
  60. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q4_1.cu +5 -0
  61. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q5_0.cu +5 -0
  62. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q5_1.cu +5 -0
  63. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q8_0.cu +5 -0
  64. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-f16.cu +5 -0
  65. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q4_0.cu +5 -0
  66. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q4_1.cu +5 -0
  67. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q5_0.cu +5 -0
  68. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q5_1.cu +5 -0
  69. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q8_0.cu +5 -0
  70. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-f16.cu +5 -0
  71. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q4_0.cu +5 -0
  72. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q4_1.cu +5 -0
  73. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q5_0.cu +5 -0
  74. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q5_1.cu +5 -0
  75. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q8_0.cu +5 -0
  76. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs256-f16-f16.cu +5 -0
  77. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-f16.cu +5 -0
  78. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q4_0.cu +5 -0
  79. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q4_1.cu +5 -0
  80. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q5_0.cu +5 -0
  81. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q5_1.cu +5 -0
  82. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q8_0.cu +5 -0
  83. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-f16.cu +5 -0
  84. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q4_0.cu +5 -0
  85. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q4_1.cu +5 -0
  86. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q5_0.cu +5 -0
  87. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q5_1.cu +5 -0
  88. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q8_0.cu +5 -0
  89. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-f16.cu +5 -0
  90. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q4_0.cu +5 -0
  91. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q4_1.cu +5 -0
  92. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q5_0.cu +5 -0
  93. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q5_1.cu +5 -0
  94. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q8_0.cu +5 -0
  95. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-f16.cu +5 -0
  96. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q4_0.cu +5 -0
  97. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q4_1.cu +5 -0
  98. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q5_0.cu +5 -0
  99. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q5_1.cu +5 -0
  100. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q8_0.cu +5 -0
  101. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-f16.cu +5 -0
  102. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q4_0.cu +5 -0
  103. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q4_1.cu +5 -0
  104. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q5_0.cu +5 -0
  105. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q5_1.cu +5 -0
  106. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q8_0.cu +5 -0
  107. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-f16.cu +5 -0
  108. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q4_0.cu +5 -0
  109. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q4_1.cu +5 -0
  110. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q5_0.cu +5 -0
  111. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q5_1.cu +5 -0
  112. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q8_0.cu +5 -0
  113. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-f16.cu +5 -0
  114. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q4_0.cu +5 -0
  115. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q4_1.cu +5 -0
  116. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q5_0.cu +5 -0
  117. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q5_1.cu +5 -0
  118. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q8_0.cu +5 -0
  119. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs256-f16-f16.cu +5 -0
  120. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-f16.cu +5 -0
  121. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q4_0.cu +5 -0
  122. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q4_1.cu +5 -0
  123. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q5_0.cu +5 -0
  124. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q5_1.cu +5 -0
  125. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q8_0.cu +5 -0
  126. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-wmma-f16-instance-kqfloat-cpb16.cu +10 -0
  127. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-wmma-f16-instance-kqfloat-cpb32.cu +9 -0
  128. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-wmma-f16-instance-kqhalf-cpb16.cu +10 -0
  129. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-wmma-f16-instance-kqhalf-cpb32.cu +10 -0
  130. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-wmma-f16-instance-kqhalf-cpb8.cu +8 -0
  131. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/mmq-instance-q2_k.cu +5 -0
  132. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/mmq-instance-q3_k.cu +5 -0
  133. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/mmq-instance-q4_0.cu +5 -0
  134. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/mmq-instance-q4_1.cu +5 -0
  135. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/mmq-instance-q4_k.cu +5 -0
  136. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/mmq-instance-q5_0.cu +5 -0
  137. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/mmq-instance-q5_1.cu +5 -0
  138. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/mmq-instance-q5_k.cu +5 -0
  139. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/mmq-instance-q6_k.cu +5 -0
  140. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/mmq-instance-q8_0.cu +5 -0
  141. data/vendor/tmp/llama.cpp/ggml-cuda/tsembd.cu +47 -0
  142. data/vendor/tmp/llama.cpp/ggml-cuda/unary.cu +286 -0
  143. data/vendor/tmp/llama.cpp/ggml-cuda/upscale.cu +51 -0
  144. data/vendor/tmp/llama.cpp/ggml-cuda.cu +103 -135
  145. data/vendor/tmp/llama.cpp/ggml-kompute.cpp +29 -13
  146. data/vendor/tmp/llama.cpp/ggml-metal.h +1 -1
  147. data/vendor/tmp/llama.cpp/ggml-metal.m +45 -33
  148. data/vendor/tmp/llama.cpp/ggml-metal.metal +83 -59
  149. data/vendor/tmp/llama.cpp/ggml-rpc.cpp +15 -14
  150. data/vendor/tmp/llama.cpp/ggml-sycl.cpp +26 -90
  151. data/vendor/tmp/llama.cpp/ggml-vulkan-shaders.hpp +74522 -14913
  152. data/vendor/tmp/llama.cpp/ggml-vulkan.cpp +631 -471
  153. data/vendor/tmp/llama.cpp/ggml.c +278 -603
  154. data/vendor/tmp/llama.cpp/ggml.h +9 -28
  155. data/vendor/tmp/llama.cpp/llama.cpp +345 -473
  156. data/vendor/tmp/llama.cpp/llama.h +21 -43
  157. metadata +134 -7
  158. data/vendor/tmp/llama.cpp/ggml-mpi.c +0 -216
  159. data/vendor/tmp/llama.cpp/ggml-mpi.h +0 -39
  160. data/vendor/tmp/llama.cpp/ggml-opencl.cpp +0 -2305
  161. data/vendor/tmp/llama.cpp/ggml-opencl.h +0 -36
data/vendor/tmp/llama.cpp/ggml-opencl.h DELETED
@@ -1,36 +0,0 @@
1
- #pragma once
2
-
3
- #include "ggml.h"
4
- #include "ggml-backend.h"
5
-
6
- #ifdef __cplusplus
7
- extern "C" {
8
- #endif
9
-
10
- GGML_API void ggml_cl_init(void);
11
-
12
- GGML_API void ggml_cl_mul(const struct ggml_tensor * src0, const struct ggml_tensor * src1, struct ggml_tensor * dst);
13
- GGML_API void ggml_cl_add(const struct ggml_tensor * src0, const struct ggml_tensor * src1, struct ggml_tensor * dst);
14
- GGML_API bool ggml_cl_can_mul_mat(const struct ggml_tensor * src0, const struct ggml_tensor * src1, const struct ggml_tensor * dst);
15
- GGML_API size_t ggml_cl_mul_mat_get_wsize(const struct ggml_tensor * src0, const struct ggml_tensor * src1, struct ggml_tensor * dst);
16
- GGML_API void ggml_cl_mul_mat(const struct ggml_tensor * src0, const struct ggml_tensor * src1, struct ggml_tensor * dst, void * wdata, size_t wsize);
17
-
18
- // GGML_API void * ggml_cl_host_malloc(size_t size);
19
- // GGML_API void ggml_cl_host_free(void * ptr);
20
-
21
- GGML_API void ggml_cl_free_data(const struct ggml_tensor* tensor);
22
-
23
- GGML_API void ggml_cl_transform_tensor(void * data, struct ggml_tensor * tensor);
24
-
25
- // backend API
26
-
27
- // GGML_API ggml_backend_t ggml_backend_opencl_init(void);
28
-
29
- // GGML_API bool ggml_backend_is_opencl(ggml_backend_t backend);
30
-
31
- GGML_API ggml_backend_buffer_type_t ggml_backend_opencl_buffer_type(void);
32
- // GGML_API ggml_backend_buffer_type_t ggml_backend_opencl_host_buffer_type(void);
33
-
34
- #ifdef __cplusplus
35
- }
36
- #endif