llama_cpp 0.15.3 → 0.16.0

Sign up to get free protection for your applications and to get access to all the features.
Files changed (149) hide show
  1. checksums.yaml +4 -4
  2. data/CHANGELOG.md +16 -0
  3. data/ext/llama_cpp/extconf.rb +1 -2
  4. data/ext/llama_cpp/llama_cpp.cpp +27 -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 +66 -36
  8. data/vendor/tmp/llama.cpp/ggml-alloc.c +4 -4
  9. data/vendor/tmp/llama.cpp/ggml-backend.c +5 -5
  10. data/vendor/tmp/llama.cpp/ggml-backend.h +1 -1
  11. data/vendor/tmp/llama.cpp/ggml-cuda/acc.cu +47 -0
  12. data/vendor/tmp/llama.cpp/ggml-cuda/arange.cu +34 -0
  13. data/vendor/tmp/llama.cpp/ggml-cuda/argsort.cu +103 -0
  14. data/vendor/tmp/llama.cpp/ggml-cuda/binbcast.cu +280 -0
  15. data/vendor/tmp/llama.cpp/ggml-cuda/clamp.cu +34 -0
  16. data/vendor/tmp/llama.cpp/ggml-cuda/concat.cu +196 -0
  17. data/vendor/tmp/llama.cpp/ggml-cuda/convert.cu +686 -0
  18. data/vendor/tmp/llama.cpp/ggml-cuda/cpy.cu +490 -0
  19. data/vendor/tmp/llama.cpp/ggml-cuda/diagmask.cu +40 -0
  20. data/vendor/tmp/llama.cpp/ggml-cuda/dmmv.cu +662 -0
  21. data/vendor/tmp/llama.cpp/ggml-cuda/fattn-tile-f16.cu +319 -0
  22. data/vendor/tmp/llama.cpp/ggml-cuda/fattn-tile-f32.cu +312 -0
  23. data/vendor/tmp/llama.cpp/ggml-cuda/fattn.cu +345 -0
  24. data/vendor/tmp/llama.cpp/ggml-cuda/getrows.cu +178 -0
  25. data/vendor/tmp/llama.cpp/ggml-cuda/im2col.cu +104 -0
  26. data/vendor/tmp/llama.cpp/ggml-cuda/mmq.cu +1564 -0
  27. data/vendor/tmp/llama.cpp/ggml-cuda/mmvq.cu +404 -0
  28. data/vendor/tmp/llama.cpp/ggml-cuda/norm.cu +221 -0
  29. data/vendor/tmp/llama.cpp/ggml-cuda/pad.cu +49 -0
  30. data/vendor/tmp/llama.cpp/ggml-cuda/pool2d.cu +94 -0
  31. data/vendor/tmp/llama.cpp/ggml-cuda/quantize.cu +45 -0
  32. data/vendor/tmp/llama.cpp/ggml-cuda/rope.cu +271 -0
  33. data/vendor/tmp/llama.cpp/ggml-cuda/scale.cu +31 -0
  34. data/vendor/tmp/llama.cpp/ggml-cuda/softmax.cu +205 -0
  35. data/vendor/tmp/llama.cpp/ggml-cuda/sumrows.cu +40 -0
  36. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-f16.cu +5 -0
  37. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q4_0.cu +5 -0
  38. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q4_1.cu +5 -0
  39. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q5_0.cu +5 -0
  40. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q5_1.cu +5 -0
  41. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q8_0.cu +5 -0
  42. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-f16.cu +5 -0
  43. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q4_0.cu +5 -0
  44. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q4_1.cu +5 -0
  45. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q5_0.cu +5 -0
  46. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q5_1.cu +5 -0
  47. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q8_0.cu +5 -0
  48. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-f16.cu +5 -0
  49. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q4_0.cu +5 -0
  50. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q4_1.cu +5 -0
  51. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q5_0.cu +5 -0
  52. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q5_1.cu +5 -0
  53. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q8_0.cu +5 -0
  54. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-f16.cu +5 -0
  55. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q4_0.cu +5 -0
  56. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q4_1.cu +5 -0
  57. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q5_0.cu +5 -0
  58. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q5_1.cu +5 -0
  59. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q8_0.cu +5 -0
  60. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-f16.cu +5 -0
  61. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q4_0.cu +5 -0
  62. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q4_1.cu +5 -0
  63. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q5_0.cu +5 -0
  64. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q5_1.cu +5 -0
  65. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q8_0.cu +5 -0
  66. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-f16.cu +5 -0
  67. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q4_0.cu +5 -0
  68. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q4_1.cu +5 -0
  69. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q5_0.cu +5 -0
  70. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q5_1.cu +5 -0
  71. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q8_0.cu +5 -0
  72. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs256-f16-f16.cu +5 -0
  73. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-f16.cu +5 -0
  74. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q4_0.cu +5 -0
  75. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q4_1.cu +5 -0
  76. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q5_0.cu +5 -0
  77. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q5_1.cu +5 -0
  78. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q8_0.cu +5 -0
  79. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-f16.cu +5 -0
  80. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q4_0.cu +5 -0
  81. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q4_1.cu +5 -0
  82. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q5_0.cu +5 -0
  83. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q5_1.cu +5 -0
  84. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q8_0.cu +5 -0
  85. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-f16.cu +5 -0
  86. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q4_0.cu +5 -0
  87. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q4_1.cu +5 -0
  88. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q5_0.cu +5 -0
  89. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q5_1.cu +5 -0
  90. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q8_0.cu +5 -0
  91. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-f16.cu +5 -0
  92. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q4_0.cu +5 -0
  93. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q4_1.cu +5 -0
  94. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q5_0.cu +5 -0
  95. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q5_1.cu +5 -0
  96. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q8_0.cu +5 -0
  97. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-f16.cu +5 -0
  98. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q4_0.cu +5 -0
  99. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q4_1.cu +5 -0
  100. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q5_0.cu +5 -0
  101. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q5_1.cu +5 -0
  102. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q8_0.cu +5 -0
  103. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-f16.cu +5 -0
  104. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q4_0.cu +5 -0
  105. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q4_1.cu +5 -0
  106. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q5_0.cu +5 -0
  107. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q5_1.cu +5 -0
  108. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q8_0.cu +5 -0
  109. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-f16.cu +5 -0
  110. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q4_0.cu +5 -0
  111. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q4_1.cu +5 -0
  112. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q5_0.cu +5 -0
  113. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q5_1.cu +5 -0
  114. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q8_0.cu +5 -0
  115. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs256-f16-f16.cu +5 -0
  116. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-f16.cu +5 -0
  117. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q4_0.cu +5 -0
  118. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q4_1.cu +5 -0
  119. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q5_0.cu +5 -0
  120. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q5_1.cu +5 -0
  121. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q8_0.cu +5 -0
  122. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-wmma-f16-instance-kqfloat-cpb16.cu +10 -0
  123. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-wmma-f16-instance-kqfloat-cpb32.cu +9 -0
  124. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-wmma-f16-instance-kqhalf-cpb16.cu +10 -0
  125. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-wmma-f16-instance-kqhalf-cpb32.cu +10 -0
  126. data/vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-wmma-f16-instance-kqhalf-cpb8.cu +8 -0
  127. data/vendor/tmp/llama.cpp/ggml-cuda/tsembd.cu +47 -0
  128. data/vendor/tmp/llama.cpp/ggml-cuda/unary.cu +266 -0
  129. data/vendor/tmp/llama.cpp/ggml-cuda/upscale.cu +51 -0
  130. data/vendor/tmp/llama.cpp/ggml-cuda.cu +35 -16
  131. data/vendor/tmp/llama.cpp/ggml-impl.h +4 -0
  132. data/vendor/tmp/llama.cpp/ggml-kompute.cpp +21 -7
  133. data/vendor/tmp/llama.cpp/ggml-metal.h +1 -1
  134. data/vendor/tmp/llama.cpp/ggml-metal.m +99 -35
  135. data/vendor/tmp/llama.cpp/ggml-metal.metal +146 -80
  136. data/vendor/tmp/llama.cpp/ggml-quants.c +101 -11
  137. data/vendor/tmp/llama.cpp/ggml-rpc.cpp +75 -58
  138. data/vendor/tmp/llama.cpp/ggml-sycl.cpp +345 -227
  139. data/vendor/tmp/llama.cpp/ggml-vulkan-shaders.hpp +99301 -39793
  140. data/vendor/tmp/llama.cpp/ggml-vulkan.cpp +458 -329
  141. data/vendor/tmp/llama.cpp/ggml.c +301 -409
  142. data/vendor/tmp/llama.cpp/ggml.h +19 -23
  143. data/vendor/tmp/llama.cpp/llama.cpp +855 -651
  144. data/vendor/tmp/llama.cpp/llama.h +28 -48
  145. metadata +121 -6
  146. data/vendor/tmp/llama.cpp/ggml-mpi.c +0 -216
  147. data/vendor/tmp/llama.cpp/ggml-mpi.h +0 -39
  148. data/vendor/tmp/llama.cpp/ggml-opencl.cpp +0 -2305
  149. data/vendor/tmp/llama.cpp/ggml-opencl.h +0 -36
data/vendor/tmp/llama.cpp/ggml.h CHANGED
@@ -756,7 +756,6 @@ extern "C" {
756
756
  GGML_API enum ggml_type ggml_ftype_to_ggml_type(enum ggml_ftype ftype);
757
757
 
758
758
  GGML_API GGML_CALL bool ggml_is_transposed(const struct ggml_tensor * tensor);
759
- GGML_API GGML_CALL bool ggml_is_contiguous(const struct ggml_tensor * tensor);
760
759
  GGML_API GGML_CALL bool ggml_is_permuted (const struct ggml_tensor * tensor);
761
760
  GGML_API GGML_CALL bool ggml_is_empty (const struct ggml_tensor * tensor);
762
761
  GGML_API bool ggml_is_scalar (const struct ggml_tensor * tensor);
@@ -765,6 +764,11 @@ extern "C" {
765
764
  GGML_API bool ggml_is_3d (const struct ggml_tensor * tensor);
766
765
  GGML_API int ggml_n_dims (const struct ggml_tensor * tensor); // returns 1 for scalars
767
766
 
767
+ GGML_API GGML_CALL bool ggml_is_contiguous (const struct ggml_tensor * tensor);
768
+ GGML_API GGML_CALL bool ggml_is_contiguous_0(const struct ggml_tensor * tensor); // same as ggml_is_contiguous()
769
+ GGML_API GGML_CALL bool ggml_is_contiguous_1(const struct ggml_tensor * tensor); // contiguous for dims >= 1
770
+ GGML_API GGML_CALL bool ggml_is_contiguous_2(const struct ggml_tensor * tensor); // contiguous for dims >= 2
771
+
768
772
  GGML_API bool ggml_are_same_shape (const struct ggml_tensor * t0, const struct ggml_tensor * t1);
769
773
  GGML_API bool ggml_are_same_stride(const struct ggml_tensor * t0, const struct ggml_tensor * t1);
770
774
 
@@ -1007,12 +1011,13 @@ extern "C" {
1007
1011
  struct ggml_tensor * a,
1008
1012
  struct ggml_tensor * b);
1009
1013
 
1010
- // concat a and b on dim 2
1014
+ // concat a and b along dim
1011
1015
  // used in stable-diffusion
1012
1016
  GGML_API struct ggml_tensor * ggml_concat(
1013
1017
  struct ggml_context * ctx,
1014
1018
  struct ggml_tensor * a,
1015
- struct ggml_tensor * b);
1019
+ struct ggml_tensor * b,
1020
+ int dim);
1016
1021
 
1017
1022
  GGML_API struct ggml_tensor * ggml_abs(
1018
1023
  struct ggml_context * ctx,
@@ -1460,7 +1465,6 @@ extern "C" {
1460
1465
  // rotary position embedding
1461
1466
  // if mode & 1 == 1, skip n_past elements (NOT SUPPORTED)
1462
1467
  // if mode & 2 == 1, GPT-NeoX style
1463
- // if mode & 4 == 1, ChatGLM style
1464
1468
  //
1465
1469
  // b is an int32 vector with size a->ne[2], it contains the positions
1466
1470
  // c is freq factors (e.g. phi3-128k), (optional)
@@ -1469,8 +1473,7 @@ extern "C" {
1469
1473
  struct ggml_tensor * a,
1470
1474
  struct ggml_tensor * b,
1471
1475
  int n_dims,
1472
- int mode,
1473
- int n_ctx);
1476
+ int mode);
1474
1477
 
1475
1478
  // in-place, returns view(a)
1476
1479
  GGML_API struct ggml_tensor * ggml_rope_inplace(
@@ -1478,8 +1481,7 @@ extern "C" {
1478
1481
  struct ggml_tensor * a,
1479
1482
  struct ggml_tensor * b,
1480
1483
  int n_dims,
1481
- int mode,
1482
- int n_ctx);
1484
+ int mode);
1483
1485
 
1484
1486
  // custom RoPE
1485
1487
  GGML_API struct ggml_tensor * ggml_rope_ext(
@@ -1489,8 +1491,7 @@ extern "C" {
1489
1491
  struct ggml_tensor * c,
1490
1492
  int n_dims,
1491
1493
  int mode,
1492
- int n_ctx,
1493
- int n_orig_ctx,
1494
+ int n_ctx_orig,
1494
1495
  float freq_base,
1495
1496
  float freq_scale,
1496
1497
  float ext_factor,
@@ -1506,8 +1507,7 @@ extern "C" {
1506
1507
  struct ggml_tensor * c,
1507
1508
  int n_dims,
1508
1509
  int mode,
1509
- int n_ctx,
1510
- int n_orig_ctx,
1510
+ int n_ctx_orig,
1511
1511
  float freq_base,
1512
1512
  float freq_scale,
1513
1513
  float ext_factor,
@@ -1521,8 +1521,7 @@ extern "C" {
1521
1521
  struct ggml_tensor * b,
1522
1522
  int n_dims,
1523
1523
  int mode,
1524
- int n_ctx,
1525
- int n_orig_ctx,
1524
+ int n_ctx_orig,
1526
1525
  float freq_base,
1527
1526
  float freq_scale,
1528
1527
  float ext_factor,
@@ -1537,8 +1536,7 @@ extern "C" {
1537
1536
  struct ggml_tensor * b,
1538
1537
  int n_dims,
1539
1538
  int mode,
1540
- int n_ctx,
1541
- int n_orig_ctx,
1539
+ int n_ctx_orig,
1542
1540
  float freq_base,
1543
1541
  float freq_scale,
1544
1542
  float ext_factor,
@@ -1549,7 +1547,7 @@ extern "C" {
1549
1547
 
1550
1548
  // compute correction dims for YaRN RoPE scaling
1551
1549
  GGML_CALL void ggml_rope_yarn_corr_dims(
1552
- int n_dims, int n_orig_ctx, float freq_base, float beta_fast, float beta_slow, float dims[2]);
1550
+ int n_dims, int n_ctx_orig, float freq_base, float beta_fast, float beta_slow, float dims[2]);
1553
1551
 
1554
1552
  // rotary position embedding backward, i.e compute dx from dy
1555
1553
  // a - dy
@@ -1560,16 +1558,13 @@ extern "C" {
1560
1558
  struct ggml_tensor * c,
1561
1559
  int n_dims,
1562
1560
  int mode,
1563
- int n_ctx,
1564
- int n_orig_ctx,
1561
+ int n_ctx_orig,
1565
1562
  float freq_base,
1566
1563
  float freq_scale,
1567
1564
  float ext_factor,
1568
1565
  float attn_factor,
1569
1566
  float beta_fast,
1570
- float beta_slow,
1571
- float xpos_base,
1572
- bool xpos_down);
1567
+ float beta_slow);
1573
1568
 
1574
1569
  // clamp
1575
1570
  // in-place, returns view(a)
@@ -2404,6 +2399,7 @@ extern "C" {
2404
2399
  GGML_API int ggml_cpu_has_avx512_bf16(void);
2405
2400
  GGML_API int ggml_cpu_has_fma (void);
2406
2401
  GGML_API int ggml_cpu_has_neon (void);
2402
+ GGML_API int ggml_cpu_has_sve (void);
2407
2403
  GGML_API int ggml_cpu_has_arm_fma (void);
2408
2404
  GGML_API int ggml_cpu_has_metal (void);
2409
2405
  GGML_API int ggml_cpu_has_f16c (void);
@@ -2411,13 +2407,13 @@ extern "C" {
2411
2407
  GGML_API int ggml_cpu_has_wasm_simd (void);
2412
2408
  GGML_API int ggml_cpu_has_blas (void);
2413
2409
  GGML_API int ggml_cpu_has_cuda (void);
2414
- GGML_API int ggml_cpu_has_clblast (void);
2415
2410
  GGML_API int ggml_cpu_has_vulkan (void);
2416
2411
  GGML_API int ggml_cpu_has_kompute (void);
2417
2412
  GGML_API int ggml_cpu_has_gpublas (void);
2418
2413
  GGML_API int ggml_cpu_has_sse3 (void);
2419
2414
  GGML_API int ggml_cpu_has_ssse3 (void);
2420
2415
  GGML_API int ggml_cpu_has_sycl (void);
2416
+ GGML_API int ggml_cpu_has_rpc (void);
2421
2417
  GGML_API int ggml_cpu_has_vsx (void);
2422
2418
  GGML_API int ggml_cpu_has_matmul_int8(void);
2423
2419