@novastera-oss/llamarn 0.4.1 → 0.4.3-beta4

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 (976) hide show
  1. package/RNLlamaCpp.podspec +3 -0
  2. package/android/CMakeLists.txt +2 -0
  3. package/android/src/main/cpp/include/llama.h +44 -21
  4. package/android/src/main/jniLibs/arm64-v8a/libggml-base.so +0 -0
  5. package/android/src/main/jniLibs/arm64-v8a/libggml-cpu.so +0 -0
  6. package/android/src/main/jniLibs/arm64-v8a/libggml.so +0 -0
  7. package/android/src/main/jniLibs/arm64-v8a/libllama.so +0 -0
  8. package/android/src/main/jniLibs/armeabi-v7a/libggml-base.so +0 -0
  9. package/android/src/main/jniLibs/armeabi-v7a/libggml-cpu.so +0 -0
  10. package/android/src/main/jniLibs/armeabi-v7a/libggml.so +0 -0
  11. package/android/src/main/jniLibs/armeabi-v7a/libllama.so +0 -0
  12. package/android/src/main/jniLibs/x86/libggml-base.so +0 -0
  13. package/android/src/main/jniLibs/x86/libggml-cpu.so +0 -0
  14. package/android/src/main/jniLibs/x86/libggml.so +0 -0
  15. package/android/src/main/jniLibs/x86/libllama.so +0 -0
  16. package/android/src/main/jniLibs/x86_64/libggml-base.so +0 -0
  17. package/android/src/main/jniLibs/x86_64/libggml-cpu.so +0 -0
  18. package/android/src/main/jniLibs/x86_64/libggml.so +0 -0
  19. package/android/src/main/jniLibs/x86_64/libllama.so +0 -0
  20. package/cpp/build-info.cpp +2 -2
  21. package/cpp/llama.cpp/CMakeLists.txt +12 -0
  22. package/cpp/llama.cpp/CODEOWNERS +116 -10
  23. package/cpp/llama.cpp/CONTRIBUTING.md +30 -3
  24. package/cpp/llama.cpp/README.md +13 -5
  25. package/cpp/llama.cpp/build-xcframework.sh +5 -0
  26. package/cpp/llama.cpp/cmake/riscv64-spacemit-linux-gnu-gcc.cmake +29 -0
  27. package/cpp/llama.cpp/common/CMakeLists.txt +12 -2
  28. package/cpp/llama.cpp/common/arg.cpp +303 -795
  29. package/cpp/llama.cpp/common/arg.h +2 -3
  30. package/cpp/llama.cpp/common/chat-parser-xml-toolcall.cpp +861 -0
  31. package/cpp/llama.cpp/common/chat-parser-xml-toolcall.h +45 -0
  32. package/cpp/llama.cpp/common/chat-parser.cpp +156 -15
  33. package/cpp/llama.cpp/common/chat-parser.h +13 -0
  34. package/cpp/llama.cpp/common/chat.cpp +1147 -88
  35. package/cpp/llama.cpp/common/chat.h +16 -3
  36. package/cpp/llama.cpp/common/common.cpp +70 -15
  37. package/cpp/llama.cpp/common/common.h +57 -19
  38. package/cpp/llama.cpp/common/download.cpp +1072 -0
  39. package/cpp/llama.cpp/common/download.h +55 -0
  40. package/cpp/llama.cpp/common/http.h +73 -0
  41. package/cpp/llama.cpp/common/json-partial.cpp +70 -2
  42. package/cpp/llama.cpp/common/json-schema-to-grammar.cpp +61 -22
  43. package/cpp/llama.cpp/common/json-schema-to-grammar.h +2 -0
  44. package/cpp/llama.cpp/common/log.cpp +59 -2
  45. package/cpp/llama.cpp/common/log.h +12 -4
  46. package/cpp/llama.cpp/common/sampling.cpp +84 -8
  47. package/cpp/llama.cpp/common/sampling.h +3 -1
  48. package/cpp/llama.cpp/common/speculative.cpp +1 -1
  49. package/cpp/llama.cpp/convert_hf_to_gguf.py +1608 -233
  50. package/cpp/llama.cpp/convert_hf_to_gguf_update.py +6 -1
  51. package/cpp/llama.cpp/convert_lora_to_gguf.py +37 -5
  52. package/cpp/llama.cpp/ggml/CMakeLists.txt +47 -28
  53. package/cpp/llama.cpp/ggml/include/ggml-backend.h +19 -1
  54. package/cpp/llama.cpp/ggml/include/ggml-cpu.h +1 -1
  55. package/cpp/llama.cpp/ggml/include/ggml-hexagon.h +19 -0
  56. package/cpp/llama.cpp/ggml/include/ggml-metal.h +1 -6
  57. package/cpp/llama.cpp/ggml/include/ggml-rpc.h +7 -9
  58. package/cpp/llama.cpp/ggml/include/ggml-zdnn.h +2 -1
  59. package/cpp/llama.cpp/ggml/include/ggml.h +199 -6
  60. package/cpp/llama.cpp/ggml/src/CMakeLists.txt +38 -0
  61. package/cpp/llama.cpp/ggml/src/ggml-alloc.c +299 -130
  62. package/cpp/llama.cpp/ggml/src/ggml-backend-impl.h +4 -4
  63. package/cpp/llama.cpp/ggml/src/ggml-backend-reg.cpp +21 -5
  64. package/cpp/llama.cpp/ggml/src/ggml-backend.cpp +99 -2
  65. package/cpp/llama.cpp/ggml/src/ggml-blas/CMakeLists.txt +1 -1
  66. package/cpp/llama.cpp/ggml/src/ggml-blas/ggml-blas.cpp +1 -0
  67. package/cpp/llama.cpp/ggml/src/ggml-cann/acl_tensor.cpp +57 -45
  68. package/cpp/llama.cpp/ggml/src/ggml-cann/acl_tensor.h +138 -47
  69. package/cpp/llama.cpp/ggml/src/ggml-cann/aclnn_ops.cpp +1584 -1773
  70. package/cpp/llama.cpp/ggml/src/ggml-cann/aclnn_ops.h +201 -317
  71. package/cpp/llama.cpp/ggml/src/ggml-cann/common.h +146 -187
  72. package/cpp/llama.cpp/ggml/src/ggml-cann/ggml-cann.cpp +771 -713
  73. package/cpp/llama.cpp/ggml/src/ggml-cpu/CMakeLists.txt +135 -77
  74. package/cpp/llama.cpp/ggml/src/ggml-cpu/amx/amx.cpp +5 -2
  75. package/cpp/llama.cpp/ggml/src/ggml-cpu/arch/arm/quants.c +428 -26
  76. package/cpp/llama.cpp/ggml/src/ggml-cpu/arch/loongarch/quants.c +16 -17
  77. package/cpp/llama.cpp/ggml/src/ggml-cpu/arch/riscv/quants.c +318 -145
  78. package/cpp/llama.cpp/ggml/src/ggml-cpu/arch/s390/cpu-feats.cpp +50 -0
  79. package/cpp/llama.cpp/ggml/src/ggml-cpu/arch/s390/quants.c +155 -60
  80. package/cpp/llama.cpp/ggml/src/ggml-cpu/arch/x86/repack.cpp +8 -8
  81. package/cpp/llama.cpp/ggml/src/ggml-cpu/arch-fallback.h +0 -1
  82. package/cpp/llama.cpp/ggml/src/ggml-cpu/common.h +14 -0
  83. package/cpp/llama.cpp/ggml/src/ggml-cpu/ggml-cpu-impl.h +10 -9
  84. package/cpp/llama.cpp/ggml/src/ggml-cpu/ggml-cpu.c +108 -64
  85. package/cpp/llama.cpp/ggml/src/ggml-cpu/ggml-cpu.cpp +14 -4
  86. package/cpp/llama.cpp/ggml/src/ggml-cpu/kleidiai/kernels.cpp +530 -87
  87. package/cpp/llama.cpp/ggml/src/ggml-cpu/kleidiai/kernels.h +37 -45
  88. package/cpp/llama.cpp/ggml/src/ggml-cpu/kleidiai/kleidiai.cpp +349 -127
  89. package/cpp/llama.cpp/ggml/src/ggml-cpu/ops.cpp +947 -1218
  90. package/cpp/llama.cpp/ggml/src/ggml-cpu/ops.h +5 -4
  91. package/cpp/llama.cpp/ggml/src/ggml-cpu/repack.cpp +143 -29
  92. package/cpp/llama.cpp/ggml/src/ggml-cpu/simd-mappings.h +82 -76
  93. package/cpp/llama.cpp/ggml/src/ggml-cpu/spacemit/ime.cpp +1025 -0
  94. package/cpp/llama.cpp/ggml/src/ggml-cpu/spacemit/ime.h +13 -0
  95. package/cpp/llama.cpp/ggml/src/ggml-cpu/spacemit/ime1_kernels.cpp +3196 -0
  96. package/cpp/llama.cpp/ggml/src/ggml-cpu/spacemit/ime_kernels.h +26 -0
  97. package/cpp/llama.cpp/ggml/src/ggml-cpu/unary-ops.cpp +151 -0
  98. package/cpp/llama.cpp/ggml/src/ggml-cpu/unary-ops.h +7 -0
  99. package/cpp/llama.cpp/ggml/src/ggml-cpu/vec.cpp +233 -28
  100. package/cpp/llama.cpp/ggml/src/ggml-cpu/vec.h +326 -66
  101. package/cpp/llama.cpp/ggml/src/ggml-cuda/CMakeLists.txt +12 -3
  102. package/cpp/llama.cpp/ggml/src/ggml-cuda/argsort.cu +102 -6
  103. package/cpp/llama.cpp/ggml/src/ggml-cuda/binbcast.cu +110 -76
  104. package/cpp/llama.cpp/ggml/src/ggml-cuda/common.cuh +167 -38
  105. package/cpp/llama.cpp/ggml/src/ggml-cuda/conv2d.cu +6 -11
  106. package/cpp/llama.cpp/ggml/src/ggml-cuda/convert.cuh +12 -0
  107. package/cpp/llama.cpp/ggml/src/ggml-cuda/cpy-utils.cuh +1 -1
  108. package/cpp/llama.cpp/ggml/src/ggml-cuda/cpy.cu +245 -151
  109. package/cpp/llama.cpp/ggml/src/ggml-cuda/cpy.cuh +1 -5
  110. package/cpp/llama.cpp/ggml/src/ggml-cuda/fattn-common.cuh +341 -289
  111. package/cpp/llama.cpp/ggml/src/ggml-cuda/fattn-tile.cu +49 -0
  112. package/cpp/llama.cpp/ggml/src/ggml-cuda/fattn-tile.cuh +1233 -0
  113. package/cpp/llama.cpp/ggml/src/ggml-cuda/fattn-vec.cuh +586 -0
  114. package/cpp/llama.cpp/ggml/src/ggml-cuda/fattn-wmma-f16.cu +6 -6
  115. package/cpp/llama.cpp/ggml/src/ggml-cuda/fattn-wmma-f16.cuh +48 -0
  116. package/cpp/llama.cpp/ggml/src/ggml-cuda/fattn.cu +123 -220
  117. package/cpp/llama.cpp/ggml/src/ggml-cuda/getrows.cu +41 -39
  118. package/cpp/llama.cpp/ggml/src/ggml-cuda/ggml-cuda.cu +715 -45
  119. package/cpp/llama.cpp/ggml/src/ggml-cuda/im2col.cu +150 -0
  120. package/cpp/llama.cpp/ggml/src/ggml-cuda/im2col.cuh +1 -0
  121. package/cpp/llama.cpp/ggml/src/ggml-cuda/mma.cuh +321 -24
  122. package/cpp/llama.cpp/ggml/src/ggml-cuda/mmf.cu +93 -351
  123. package/cpp/llama.cpp/ggml/src/ggml-cuda/mmf.cuh +828 -1
  124. package/cpp/llama.cpp/ggml/src/ggml-cuda/mmid.cu +164 -0
  125. package/cpp/llama.cpp/ggml/src/ggml-cuda/mmid.cuh +5 -0
  126. package/cpp/llama.cpp/ggml/src/ggml-cuda/mmq.cu +3 -166
  127. package/cpp/llama.cpp/ggml/src/ggml-cuda/mmq.cuh +1 -1
  128. package/cpp/llama.cpp/ggml/src/ggml-cuda/mmvf.cu +371 -78
  129. package/cpp/llama.cpp/ggml/src/ggml-cuda/mmvf.cuh +3 -2
  130. package/cpp/llama.cpp/ggml/src/ggml-cuda/mmvq.cu +279 -147
  131. package/cpp/llama.cpp/ggml/src/ggml-cuda/mmvq.cuh +1 -1
  132. package/cpp/llama.cpp/ggml/src/ggml-cuda/norm.cu +97 -85
  133. package/cpp/llama.cpp/ggml/src/ggml-cuda/pad.cu +46 -23
  134. package/cpp/llama.cpp/ggml/src/ggml-cuda/pad_reflect_1d.cu +63 -54
  135. package/cpp/llama.cpp/ggml/src/ggml-cuda/quantize.cu +12 -10
  136. package/cpp/llama.cpp/ggml/src/ggml-cuda/rope.cu +192 -77
  137. package/cpp/llama.cpp/ggml/src/ggml-cuda/rope.cuh +2 -0
  138. package/cpp/llama.cpp/ggml/src/ggml-cuda/scale.cu +10 -9
  139. package/cpp/llama.cpp/ggml/src/ggml-cuda/set-rows.cu +137 -75
  140. package/cpp/llama.cpp/ggml/src/ggml-cuda/set.cu +39 -0
  141. package/cpp/llama.cpp/ggml/src/ggml-cuda/set.cuh +7 -0
  142. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq112-dv112.cu +5 -0
  143. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq128-dv128.cu +5 -0
  144. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq256-dv256.cu +5 -0
  145. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq40-dv40.cu +5 -0
  146. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq576-dv512.cu +5 -0
  147. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq64-dv64.cu +5 -0
  148. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq72-dv72.cu +5 -0
  149. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq80-dv80.cu +5 -0
  150. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-tile-instance-dkq96-dv96.cu +5 -0
  151. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-f16.cu +7 -0
  152. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-q4_0.cu +7 -0
  153. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-q4_1.cu +7 -0
  154. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-q5_0.cu +7 -0
  155. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-q5_1.cu +7 -0
  156. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-q8_0.cu +7 -0
  157. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-f16.cu +7 -0
  158. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-q4_0.cu +7 -0
  159. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-q4_1.cu +7 -0
  160. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-q5_0.cu +7 -0
  161. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-q5_1.cu +7 -0
  162. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-q8_0.cu +7 -0
  163. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-f16.cu +7 -0
  164. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-q4_0.cu +7 -0
  165. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-q4_1.cu +7 -0
  166. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-q5_0.cu +7 -0
  167. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-q5_1.cu +7 -0
  168. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-q8_0.cu +7 -0
  169. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-f16.cu +7 -0
  170. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-q4_0.cu +7 -0
  171. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-q4_1.cu +7 -0
  172. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-q5_0.cu +7 -0
  173. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-q5_1.cu +7 -0
  174. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-q8_0.cu +7 -0
  175. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-f16.cu +7 -0
  176. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-q4_0.cu +7 -0
  177. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-q4_1.cu +7 -0
  178. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-q5_0.cu +7 -0
  179. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-q5_1.cu +7 -0
  180. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-q8_0.cu +7 -0
  181. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-f16.cu +7 -0
  182. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-q4_0.cu +7 -0
  183. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-q4_1.cu +7 -0
  184. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-q5_0.cu +7 -0
  185. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-q5_1.cu +7 -0
  186. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-q8_0.cu +7 -0
  187. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/generate_cu_files.py +40 -19
  188. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_1.cu +5 -0
  189. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_10.cu +5 -0
  190. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_11.cu +5 -0
  191. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_12.cu +5 -0
  192. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_13.cu +5 -0
  193. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_14.cu +5 -0
  194. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_15.cu +5 -0
  195. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_16.cu +5 -0
  196. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_2.cu +5 -0
  197. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_3.cu +5 -0
  198. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_4.cu +5 -0
  199. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_5.cu +5 -0
  200. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_6.cu +5 -0
  201. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_7.cu +5 -0
  202. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_8.cu +5 -0
  203. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_9.cu +5 -0
  204. package/cpp/llama.cpp/ggml/src/ggml-cuda/topk-moe.cu +336 -0
  205. package/cpp/llama.cpp/ggml/src/ggml-cuda/topk-moe.cuh +16 -0
  206. package/cpp/llama.cpp/ggml/src/ggml-cuda/tsembd.cu +3 -3
  207. package/cpp/llama.cpp/ggml/src/ggml-cuda/unary.cu +105 -11
  208. package/cpp/llama.cpp/ggml/src/ggml-cuda/unary.cuh +36 -0
  209. package/cpp/llama.cpp/ggml/src/ggml-cuda/upscale.cu +87 -6
  210. package/cpp/llama.cpp/ggml/src/ggml-cuda/vendors/hip.h +28 -12
  211. package/cpp/llama.cpp/ggml/src/ggml-hexagon/CMakeLists.txt +68 -0
  212. package/cpp/llama.cpp/ggml/src/ggml-hexagon/ggml-hexagon.cpp +3807 -0
  213. package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp/CMakeLists.txt +40 -0
  214. package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp/act-ops.c +442 -0
  215. package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp/binary-ops.c +360 -0
  216. package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp/cmake-toolchain.cmake +157 -0
  217. package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp/htp-ctx.h +40 -0
  218. package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp/htp-dma.c +69 -0
  219. package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp/htp-dma.h +119 -0
  220. package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp/htp-msg.h +156 -0
  221. package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp/htp-ops.h +64 -0
  222. package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp/htp_iface.idl +16 -0
  223. package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp/hvx-exp.c +93 -0
  224. package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp/hvx-inverse.c +60 -0
  225. package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp/hvx-sigmoid.c +49 -0
  226. package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp/hvx-utils.c +960 -0
  227. package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp/hvx-utils.h +1032 -0
  228. package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp/main.c +829 -0
  229. package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp/matmul-ops.c +2223 -0
  230. package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp/ops-utils.h +149 -0
  231. package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp/rope-ops.c +418 -0
  232. package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp/softmax-ops.c +402 -0
  233. package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp/unary-ops.c +255 -0
  234. package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp/worker-pool.c +297 -0
  235. package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp/worker-pool.h +57 -0
  236. package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp-utils.c +448 -0
  237. package/cpp/llama.cpp/ggml/src/ggml-hexagon/htp-utils.h +220 -0
  238. package/cpp/llama.cpp/ggml/src/ggml-hip/CMakeLists.txt +8 -13
  239. package/cpp/llama.cpp/ggml/src/ggml-impl.h +110 -12
  240. package/cpp/llama.cpp/ggml/src/ggml-metal/CMakeLists.txt +6 -5
  241. package/cpp/llama.cpp/ggml/src/ggml-metal/ggml-metal-common.cpp +446 -0
  242. package/cpp/llama.cpp/ggml/src/ggml-metal/ggml-metal-common.h +52 -0
  243. package/cpp/llama.cpp/ggml/src/ggml-metal/ggml-metal-context.h +33 -0
  244. package/cpp/llama.cpp/ggml/src/ggml-metal/ggml-metal-context.m +599 -0
  245. package/cpp/llama.cpp/ggml/src/ggml-metal/ggml-metal-device.cpp +1662 -0
  246. package/cpp/llama.cpp/ggml/src/ggml-metal/ggml-metal-device.h +251 -0
  247. package/cpp/llama.cpp/ggml/src/ggml-metal/ggml-metal-device.m +1527 -0
  248. package/cpp/llama.cpp/ggml/src/ggml-metal/ggml-metal-impl.h +244 -39
  249. package/cpp/llama.cpp/ggml/src/ggml-metal/ggml-metal-ops.cpp +3844 -0
  250. package/cpp/llama.cpp/ggml/src/ggml-metal/ggml-metal-ops.h +90 -0
  251. package/cpp/llama.cpp/ggml/src/ggml-metal/ggml-metal.cpp +723 -0
  252. package/cpp/llama.cpp/ggml/src/ggml-metal/ggml-metal.metal +3453 -1907
  253. package/cpp/llama.cpp/ggml/src/ggml-musa/CMakeLists.txt +3 -1
  254. package/cpp/llama.cpp/ggml/src/ggml-opencl/CMakeLists.txt +10 -0
  255. package/cpp/llama.cpp/ggml/src/ggml-opencl/ggml-opencl.cpp +1331 -109
  256. package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/cvt.cl +126 -0
  257. package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/flash_attn_f16.cl +31 -4
  258. package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/flash_attn_f32.cl +35 -7
  259. package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/flash_attn_f32_f16.cl +31 -4
  260. package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/gemm_moe_mxfp4_f32.cl +162 -0
  261. package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/gemv_moe_mxfp4_f32.cl +156 -0
  262. package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/get_rows.cl +36 -12
  263. package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/mul_mm_f16_f32_kq_kqv.cl +273 -0
  264. package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/mul_mm_f16_f32_l4_lm.cl +24 -10
  265. package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/mul_mm_f32_f32_l4_lm.cl +24 -10
  266. package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/mul_mm_q8_0_f32_l4_lm.cl +154 -0
  267. package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/mul_mv_id_mxfp4_f32_flat.cl +176 -0
  268. package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/mul_mv_id_q8_0_f32.cl +140 -0
  269. package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/mul_mv_id_q8_0_f32_flat.cl +222 -0
  270. package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/mul_mv_mxfp4_f32_flat.cl +167 -0
  271. package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/mul_mv_q8_0_f32.cl +125 -0
  272. package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/mul_mv_q8_0_f32_flat.cl +202 -0
  273. package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/pad.cl +29 -20
  274. package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/rms_norm.cl +25 -10
  275. package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/rope.cl +50 -24
  276. package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/set_rows.cl +123 -10
  277. package/cpp/llama.cpp/ggml/src/ggml-opencl/kernels/tsembd.cl +2 -2
  278. package/cpp/llama.cpp/ggml/src/ggml-quants.c +1 -0
  279. package/cpp/llama.cpp/ggml/src/ggml-rpc/ggml-rpc.cpp +341 -161
  280. package/cpp/llama.cpp/ggml/src/ggml-sycl/backend.hpp +6 -0
  281. package/cpp/llama.cpp/ggml/src/ggml-sycl/binbcast.cpp +6 -5
  282. package/cpp/llama.cpp/ggml/src/ggml-sycl/common.hpp +74 -15
  283. package/cpp/llama.cpp/ggml/src/ggml-sycl/concat.cpp +50 -30
  284. package/cpp/llama.cpp/ggml/src/ggml-sycl/conv.cpp +10 -4
  285. package/cpp/llama.cpp/ggml/src/ggml-sycl/convert.cpp +166 -99
  286. package/cpp/llama.cpp/ggml/src/ggml-sycl/count-equal.cpp +79 -0
  287. package/cpp/llama.cpp/ggml/src/ggml-sycl/count-equal.hpp +9 -0
  288. package/cpp/llama.cpp/ggml/src/ggml-sycl/cpy.cpp +72 -94
  289. package/cpp/llama.cpp/ggml/src/ggml-sycl/dmmv.cpp +67 -49
  290. package/cpp/llama.cpp/ggml/src/ggml-sycl/dpct/helper.hpp +21 -31
  291. package/cpp/llama.cpp/ggml/src/ggml-sycl/element_wise.cpp +252 -316
  292. package/cpp/llama.cpp/ggml/src/ggml-sycl/element_wise.hpp +6 -2
  293. package/cpp/llama.cpp/ggml/src/ggml-sycl/getrows.cpp +9 -6
  294. package/cpp/llama.cpp/ggml/src/ggml-sycl/ggml-sycl.cpp +359 -142
  295. package/cpp/llama.cpp/ggml/src/ggml-sycl/gla.cpp +2 -2
  296. package/cpp/llama.cpp/ggml/src/ggml-sycl/im2col.cpp +1 -1
  297. package/cpp/llama.cpp/ggml/src/ggml-sycl/mmq.cpp +80 -60
  298. package/cpp/llama.cpp/ggml/src/ggml-sycl/mmvq.cpp +201 -132
  299. package/cpp/llama.cpp/ggml/src/ggml-sycl/norm.cpp +230 -55
  300. package/cpp/llama.cpp/ggml/src/ggml-sycl/norm.hpp +2 -0
  301. package/cpp/llama.cpp/ggml/src/ggml-sycl/pad.cpp +97 -0
  302. package/cpp/llama.cpp/ggml/src/ggml-sycl/pad.hpp +24 -0
  303. package/cpp/llama.cpp/ggml/src/ggml-sycl/pad_reflect_1d.cpp +72 -0
  304. package/cpp/llama.cpp/ggml/src/ggml-sycl/pad_reflect_1d.hpp +8 -0
  305. package/cpp/llama.cpp/ggml/src/ggml-sycl/presets.hpp +2 -0
  306. package/cpp/llama.cpp/ggml/src/ggml-sycl/repeat_back.cpp +76 -0
  307. package/cpp/llama.cpp/ggml/src/ggml-sycl/repeat_back.hpp +8 -0
  308. package/cpp/llama.cpp/ggml/src/ggml-sycl/roll.cpp +122 -0
  309. package/cpp/llama.cpp/ggml/src/ggml-sycl/roll.hpp +20 -0
  310. package/cpp/llama.cpp/ggml/src/ggml-sycl/rope.cpp +50 -41
  311. package/cpp/llama.cpp/ggml/src/ggml-sycl/set.cpp +73 -0
  312. package/cpp/llama.cpp/ggml/src/ggml-sycl/set.hpp +5 -0
  313. package/cpp/llama.cpp/ggml/src/ggml-sycl/set_rows.cpp +45 -36
  314. package/cpp/llama.cpp/ggml/src/ggml-sycl/softmax.cpp +330 -165
  315. package/cpp/llama.cpp/ggml/src/ggml-sycl/softmax.hpp +4 -0
  316. package/cpp/llama.cpp/ggml/src/ggml-sycl/ssm_conv.cpp +127 -0
  317. package/cpp/llama.cpp/ggml/src/ggml-sycl/ssm_conv.hpp +5 -0
  318. package/cpp/llama.cpp/ggml/src/ggml-sycl/tsembd.cpp +12 -6
  319. package/cpp/llama.cpp/ggml/src/ggml-sycl/wkv.cpp +16 -12
  320. package/cpp/llama.cpp/ggml/src/ggml-vulkan/CMakeLists.txt +38 -18
  321. package/cpp/llama.cpp/ggml/src/ggml-vulkan/ggml-vulkan.cpp +4184 -2159
  322. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/abs.comp +21 -0
  323. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/acc.comp +2 -2
  324. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/add.comp +2 -2
  325. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/add1.comp +28 -0
  326. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/add_id.comp +1 -1
  327. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/arange.comp +20 -0
  328. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/argmax.comp +2 -2
  329. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/argsort.comp +33 -26
  330. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/argsort_large.comp +114 -0
  331. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/ceil.comp +22 -0
  332. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/clamp.comp +2 -2
  333. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/concat.comp +2 -2
  334. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/contig_copy.comp +2 -2
  335. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/conv2d_dw.comp +1 -1
  336. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/conv2d_mm.comp +53 -30
  337. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/conv_transpose_1d.comp +1 -1
  338. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/copy.comp +2 -2
  339. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/copy_from_quant.comp +3 -3
  340. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/copy_to_quant.comp +13 -6
  341. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/copy_transpose.comp +67 -0
  342. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/cos.comp +2 -2
  343. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/count_equal.comp +2 -2
  344. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_f32.comp +1 -1
  345. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/{dequant_funcs.comp → dequant_funcs.glsl} +138 -2
  346. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/{dequant_funcs_cm2.comp → dequant_funcs_cm2.glsl} +18 -4
  347. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/{dequant_head.comp → dequant_head.glsl} +1 -1
  348. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq1_m.comp +1 -1
  349. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq1_s.comp +1 -1
  350. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq2_s.comp +2 -2
  351. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq2_xs.comp +1 -1
  352. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq2_xxs.comp +3 -2
  353. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq3_s.comp +7 -6
  354. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq3_xxs.comp +5 -3
  355. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq4_nl.comp +1 -1
  356. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq4_xs.comp +1 -1
  357. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_mxfp4.comp +3 -3
  358. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q2_k.comp +3 -3
  359. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q3_k.comp +1 -1
  360. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q4_0.comp +1 -1
  361. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q4_1.comp +1 -1
  362. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q4_k.comp +3 -3
  363. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q5_0.comp +1 -1
  364. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q5_1.comp +1 -1
  365. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q5_k.comp +3 -3
  366. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q6_k.comp +1 -1
  367. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q8_0.comp +1 -1
  368. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/diag_mask_inf.comp +1 -1
  369. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/div.comp +2 -2
  370. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/exp.comp +3 -2
  371. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/fill.comp +19 -0
  372. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/flash_attn.comp +52 -14
  373. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/{flash_attn_base.comp → flash_attn_base.glsl} +50 -12
  374. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_cm1.comp +61 -12
  375. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_cm2.comp +54 -12
  376. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_split_k_reduce.comp +5 -1
  377. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/floor.comp +22 -0
  378. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/geglu.comp +2 -2
  379. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/geglu_erf.comp +2 -2
  380. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/geglu_quick.comp +2 -2
  381. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/gelu.comp +2 -2
  382. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/gelu_erf.comp +2 -2
  383. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/gelu_quick.comp +2 -2
  384. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/{generic_binary_head.comp → generic_binary_head.glsl} +10 -2
  385. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/get_rows.comp +21 -12
  386. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/get_rows_quant.comp +28 -18
  387. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/{glu_head.comp → glu_head.glsl} +1 -1
  388. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/group_norm.comp +2 -2
  389. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/hardsigmoid.comp +22 -0
  390. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/hardswish.comp +22 -0
  391. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/im2col.comp +15 -7
  392. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/im2col_3d.comp +125 -0
  393. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/l2_norm.comp +2 -2
  394. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/leaky_relu.comp +2 -2
  395. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/log.comp +18 -0
  396. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul.comp +2 -2
  397. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec.comp +1 -1
  398. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_base.glsl +229 -0
  399. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iface.glsl +33 -0
  400. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iq1_m.comp +1 -1
  401. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iq1_s.comp +1 -1
  402. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iq2_s.comp +1 -1
  403. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iq2_xs.comp +1 -1
  404. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iq2_xxs.comp +1 -1
  405. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iq3_s.comp +1 -1
  406. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iq3_xxs.comp +1 -1
  407. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_nc.comp +9 -7
  408. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_p021.comp +9 -7
  409. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_q2_k.comp +3 -5
  410. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_q3_k.comp +1 -1
  411. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_q4_k.comp +3 -5
  412. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_q5_k.comp +3 -5
  413. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_q6_k.comp +1 -1
  414. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vecq.comp +140 -0
  415. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mm.comp +106 -634
  416. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mm_cm2.comp +118 -9
  417. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mm_funcs.glsl +556 -0
  418. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mm_id_funcs.glsl +70 -0
  419. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mmq.comp +77 -214
  420. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mmq_funcs.glsl +589 -0
  421. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mmq_shmem_types.glsl +78 -0
  422. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/multi_add.comp +97 -13
  423. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/neg.comp +20 -0
  424. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/norm.comp +2 -2
  425. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/opt_step_adamw.comp +2 -2
  426. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/opt_step_sgd.comp +1 -1
  427. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/pad.comp +25 -4
  428. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/pool2d.comp +1 -1
  429. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/quantize_q8_1.comp +55 -5
  430. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/reglu.comp +2 -2
  431. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/relu.comp +2 -2
  432. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/repeat.comp +2 -2
  433. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/repeat_back.comp +2 -2
  434. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/rms_norm.comp +45 -3
  435. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/rms_norm_back.comp +2 -2
  436. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/rms_norm_partials.comp +2 -2
  437. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/roll.comp +2 -2
  438. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/rope_funcs.glsl +227 -0
  439. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/rope_head.glsl +20 -0
  440. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/rope_multi.comp +5 -52
  441. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/rope_neox.comp +5 -35
  442. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/rope_norm.comp +5 -35
  443. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/rope_params.glsl +27 -0
  444. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/rope_vision.comp +5 -41
  445. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/round.comp +29 -0
  446. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/scale.comp +2 -2
  447. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/sigmoid.comp +2 -2
  448. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/silu.comp +2 -2
  449. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/silu_back.comp +2 -2
  450. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/sin.comp +2 -2
  451. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/soft_max.comp +1 -1
  452. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/soft_max_back.comp +6 -2
  453. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/softplus.comp +23 -0
  454. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/sqrt.comp +2 -2
  455. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/square.comp +2 -2
  456. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/ssm_conv.comp +44 -0
  457. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/ssm_scan.comp +140 -0
  458. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/step.comp +22 -0
  459. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/sub.comp +2 -2
  460. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/sum_rows.comp +1 -1
  461. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/swiglu.comp +2 -2
  462. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/swiglu_oai.comp +2 -2
  463. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/tanh.comp +2 -2
  464. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/timestep_embedding.comp +5 -4
  465. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/topk_moe.comp +171 -0
  466. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/trunc.comp +22 -0
  467. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/{types.comp → types.glsl} +79 -29
  468. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/upscale.comp +36 -12
  469. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/vulkan-shaders-gen.cpp +471 -196
  470. package/cpp/llama.cpp/ggml/src/ggml-webgpu/CMakeLists.txt +8 -0
  471. package/cpp/llama.cpp/ggml/src/ggml-webgpu/ggml-webgpu.cpp +1690 -383
  472. package/cpp/llama.cpp/ggml/src/ggml-webgpu/wgsl-shaders/bin_op.tmpl.wgsl +188 -0
  473. package/cpp/llama.cpp/ggml/src/ggml-webgpu/wgsl-shaders/binary_head.tmpl +45 -0
  474. package/cpp/llama.cpp/ggml/src/ggml-webgpu/wgsl-shaders/common_decls.tmpl +930 -0
  475. package/cpp/llama.cpp/ggml/src/ggml-webgpu/wgsl-shaders/cpy.tmpl.wgsl +101 -0
  476. package/cpp/llama.cpp/ggml/src/ggml-webgpu/wgsl-shaders/embed_wgsl.py +57 -10
  477. package/cpp/llama.cpp/ggml/src/ggml-webgpu/wgsl-shaders/get_rows.tmpl.wgsl +874 -0
  478. package/cpp/llama.cpp/ggml/src/ggml-webgpu/wgsl-shaders/glu.tmpl.wgsl +323 -0
  479. package/cpp/llama.cpp/ggml/src/ggml-webgpu/wgsl-shaders/mul_mat.tmpl.wgsl +25 -912
  480. package/cpp/llama.cpp/ggml/src/ggml-webgpu/wgsl-shaders/mul_mat_decls.tmpl +97 -0
  481. package/cpp/llama.cpp/ggml/src/ggml-webgpu/wgsl-shaders/mul_mat_reg_tile.tmpl.wgsl +247 -0
  482. package/cpp/llama.cpp/ggml/src/ggml-webgpu/wgsl-shaders/mul_mat_subgroup_matrix.tmpl.wgsl +302 -0
  483. package/cpp/llama.cpp/ggml/src/ggml-webgpu/wgsl-shaders/mul_mat_vec.tmpl.wgsl +267 -0
  484. package/cpp/llama.cpp/ggml/src/ggml-webgpu/wgsl-shaders/rms_norm.wgsl +123 -0
  485. package/cpp/llama.cpp/ggml/src/ggml-webgpu/wgsl-shaders/rope.tmpl.wgsl +295 -0
  486. package/cpp/llama.cpp/ggml/src/ggml-webgpu/wgsl-shaders/scale.tmpl.wgsl +90 -0
  487. package/cpp/llama.cpp/ggml/src/ggml-webgpu/wgsl-shaders/{set_rows.wgsl → set_rows.tmpl.wgsl} +38 -8
  488. package/cpp/llama.cpp/ggml/src/ggml-webgpu/wgsl-shaders/soft_max.tmpl.wgsl +345 -0
  489. package/cpp/llama.cpp/ggml/src/ggml-zdnn/common.hpp +59 -0
  490. package/cpp/llama.cpp/ggml/src/ggml-zdnn/ggml-zdnn.cpp +96 -314
  491. package/cpp/llama.cpp/ggml/src/ggml-zdnn/mmf.cpp +80 -0
  492. package/cpp/llama.cpp/ggml/src/ggml-zdnn/mmf.hpp +12 -0
  493. package/cpp/llama.cpp/ggml/src/ggml-zdnn/utils.cpp +79 -0
  494. package/cpp/llama.cpp/ggml/src/ggml-zdnn/utils.hpp +19 -0
  495. package/cpp/llama.cpp/ggml/src/ggml.c +440 -17
  496. package/cpp/llama.cpp/ggml/src/gguf.cpp +104 -29
  497. package/cpp/llama.cpp/gguf-py/gguf/constants.py +363 -13
  498. package/cpp/llama.cpp/gguf-py/gguf/gguf_writer.py +64 -0
  499. package/cpp/llama.cpp/gguf-py/gguf/lazy.py +8 -3
  500. package/cpp/llama.cpp/gguf-py/gguf/scripts/gguf_convert_endian.py +6 -0
  501. package/cpp/llama.cpp/gguf-py/gguf/tensor_mapping.py +156 -18
  502. package/cpp/llama.cpp/gguf-py/gguf/utility.py +80 -0
  503. package/cpp/llama.cpp/gguf-py/gguf/vocab.py +4 -4
  504. package/cpp/llama.cpp/include/llama.h +44 -21
  505. package/cpp/llama.cpp/media/llama1-icon-transparent.png +0 -0
  506. package/cpp/llama.cpp/media/llama1-icon-transparent.svg +77 -0
  507. package/cpp/llama.cpp/media/llama1-icon.png +0 -0
  508. package/cpp/llama.cpp/media/llama1-icon.svg +87 -0
  509. package/cpp/llama.cpp/requirements/requirements-all.txt +2 -0
  510. package/cpp/llama.cpp/requirements/requirements-convert_hf_to_gguf.txt +3 -3
  511. package/cpp/llama.cpp/requirements/requirements-convert_legacy_llama.txt +3 -1
  512. package/cpp/llama.cpp/requirements/requirements-tool_bench.txt +1 -1
  513. package/cpp/llama.cpp/src/CMakeLists.txt +101 -0
  514. package/cpp/llama.cpp/src/llama-adapter.cpp +33 -0
  515. package/cpp/llama.cpp/src/llama-adapter.h +3 -0
  516. package/cpp/llama.cpp/src/llama-arch.cpp +344 -14
  517. package/cpp/llama.cpp/src/llama-arch.h +50 -0
  518. package/cpp/llama.cpp/src/llama-batch.cpp +63 -31
  519. package/cpp/llama.cpp/src/llama-batch.h +13 -2
  520. package/cpp/llama.cpp/src/llama-chat.cpp +85 -3
  521. package/cpp/llama.cpp/src/llama-chat.h +4 -0
  522. package/cpp/llama.cpp/src/llama-context.cpp +300 -45
  523. package/cpp/llama.cpp/src/llama-context.h +16 -6
  524. package/cpp/llama.cpp/src/llama-cparams.h +2 -1
  525. package/cpp/llama.cpp/src/llama-grammar.cpp +17 -9
  526. package/cpp/llama.cpp/src/llama-graph.cpp +226 -64
  527. package/cpp/llama.cpp/src/llama-graph.h +27 -5
  528. package/cpp/llama.cpp/src/llama-hparams.cpp +53 -2
  529. package/cpp/llama.cpp/src/llama-hparams.h +48 -8
  530. package/cpp/llama.cpp/src/llama-impl.cpp +3 -3
  531. package/cpp/llama.cpp/src/llama-impl.h +2 -0
  532. package/cpp/llama.cpp/src/llama-kv-cache-iswa.cpp +13 -3
  533. package/cpp/llama.cpp/src/llama-kv-cache-iswa.h +2 -0
  534. package/cpp/llama.cpp/src/llama-kv-cache.cpp +120 -62
  535. package/cpp/llama.cpp/src/llama-kv-cache.h +13 -4
  536. package/cpp/llama.cpp/src/llama-kv-cells.h +44 -2
  537. package/cpp/llama.cpp/src/llama-memory-hybrid.cpp +19 -9
  538. package/cpp/llama.cpp/src/llama-memory-hybrid.h +2 -0
  539. package/cpp/llama.cpp/src/llama-memory-recurrent.cpp +38 -17
  540. package/cpp/llama.cpp/src/llama-memory-recurrent.h +5 -2
  541. package/cpp/llama.cpp/src/llama-memory.h +3 -0
  542. package/cpp/llama.cpp/src/llama-model-loader.cpp +2 -0
  543. package/cpp/llama.cpp/src/llama-model.cpp +1070 -12614
  544. package/cpp/llama.cpp/src/llama-model.h +40 -4
  545. package/cpp/llama.cpp/src/llama-quant.cpp +14 -6
  546. package/cpp/llama.cpp/src/llama-sampling.cpp +243 -136
  547. package/cpp/llama.cpp/src/llama-vocab.cpp +43 -3
  548. package/cpp/llama.cpp/src/llama-vocab.h +43 -39
  549. package/cpp/llama.cpp/src/llama.cpp +69 -10
  550. package/cpp/llama.cpp/src/models/afmoe.cpp +187 -0
  551. package/cpp/llama.cpp/src/models/apertus.cpp +125 -0
  552. package/cpp/llama.cpp/src/models/arcee.cpp +135 -0
  553. package/cpp/llama.cpp/src/models/arctic.cpp +138 -0
  554. package/cpp/llama.cpp/src/models/arwkv7.cpp +86 -0
  555. package/cpp/llama.cpp/src/models/baichuan.cpp +122 -0
  556. package/cpp/llama.cpp/src/models/bailingmoe.cpp +144 -0
  557. package/cpp/llama.cpp/src/models/bailingmoe2.cpp +135 -0
  558. package/cpp/llama.cpp/src/models/bert.cpp +176 -0
  559. package/cpp/llama.cpp/src/models/bitnet.cpp +160 -0
  560. package/cpp/llama.cpp/src/models/bloom.cpp +101 -0
  561. package/cpp/llama.cpp/src/models/chameleon.cpp +178 -0
  562. package/cpp/llama.cpp/src/models/chatglm.cpp +132 -0
  563. package/cpp/llama.cpp/src/models/codeshell.cpp +111 -0
  564. package/cpp/llama.cpp/src/models/cogvlm.cpp +100 -0
  565. package/cpp/llama.cpp/src/models/cohere2-iswa.cpp +131 -0
  566. package/cpp/llama.cpp/src/models/command-r.cpp +122 -0
  567. package/cpp/llama.cpp/src/models/dbrx.cpp +123 -0
  568. package/cpp/llama.cpp/src/models/deci.cpp +135 -0
  569. package/cpp/llama.cpp/src/models/deepseek.cpp +144 -0
  570. package/cpp/llama.cpp/src/models/deepseek2.cpp +237 -0
  571. package/cpp/llama.cpp/src/models/dots1.cpp +134 -0
  572. package/cpp/llama.cpp/src/models/dream.cpp +105 -0
  573. package/cpp/llama.cpp/src/models/ernie4-5-moe.cpp +150 -0
  574. package/cpp/llama.cpp/src/models/ernie4-5.cpp +110 -0
  575. package/cpp/llama.cpp/src/models/exaone.cpp +114 -0
  576. package/cpp/llama.cpp/src/models/exaone4.cpp +123 -0
  577. package/cpp/llama.cpp/src/models/falcon-h1.cpp +113 -0
  578. package/cpp/llama.cpp/src/models/falcon.cpp +120 -0
  579. package/cpp/llama.cpp/src/models/gemma-embedding.cpp +120 -0
  580. package/cpp/llama.cpp/src/models/gemma.cpp +112 -0
  581. package/cpp/llama.cpp/src/models/gemma2-iswa.cpp +125 -0
  582. package/cpp/llama.cpp/src/models/gemma3-iswa.cpp +131 -0
  583. package/cpp/llama.cpp/src/models/gemma3n-iswa.cpp +377 -0
  584. package/cpp/llama.cpp/src/models/glm4-moe.cpp +153 -0
  585. package/cpp/llama.cpp/src/models/glm4.cpp +127 -0
  586. package/cpp/llama.cpp/src/models/gpt2.cpp +105 -0
  587. package/cpp/llama.cpp/src/models/gptneox.cpp +144 -0
  588. package/cpp/llama.cpp/src/models/granite-hybrid.cpp +196 -0
  589. package/cpp/llama.cpp/src/models/granite.cpp +211 -0
  590. package/cpp/llama.cpp/src/models/graph-context-mamba.cpp +283 -0
  591. package/cpp/llama.cpp/src/models/grok.cpp +159 -0
  592. package/cpp/llama.cpp/src/models/grovemoe.cpp +141 -0
  593. package/cpp/llama.cpp/src/models/hunyuan-dense.cpp +132 -0
  594. package/cpp/llama.cpp/src/models/hunyuan-moe.cpp +154 -0
  595. package/cpp/llama.cpp/src/models/internlm2.cpp +120 -0
  596. package/cpp/llama.cpp/src/models/jais.cpp +86 -0
  597. package/cpp/llama.cpp/src/models/jamba.cpp +106 -0
  598. package/cpp/llama.cpp/src/models/lfm2.cpp +173 -0
  599. package/cpp/llama.cpp/src/models/llada-moe.cpp +122 -0
  600. package/cpp/llama.cpp/src/models/llada.cpp +99 -0
  601. package/cpp/llama.cpp/src/models/llama-iswa.cpp +174 -0
  602. package/cpp/llama.cpp/src/models/llama.cpp +155 -0
  603. package/cpp/llama.cpp/src/models/mamba.cpp +55 -0
  604. package/cpp/llama.cpp/src/models/minicpm3.cpp +199 -0
  605. package/cpp/llama.cpp/src/models/minimax-m2.cpp +124 -0
  606. package/cpp/llama.cpp/src/models/models.h +485 -0
  607. package/cpp/llama.cpp/src/models/mpt.cpp +126 -0
  608. package/cpp/llama.cpp/src/models/nemotron-h.cpp +121 -0
  609. package/cpp/llama.cpp/src/models/nemotron.cpp +122 -0
  610. package/cpp/llama.cpp/src/models/neo-bert.cpp +104 -0
  611. package/cpp/llama.cpp/src/models/olmo.cpp +121 -0
  612. package/cpp/llama.cpp/src/models/olmo2.cpp +150 -0
  613. package/cpp/llama.cpp/src/models/olmoe.cpp +124 -0
  614. package/cpp/llama.cpp/src/models/openai-moe-iswa.cpp +124 -0
  615. package/cpp/llama.cpp/src/models/openelm.cpp +124 -0
  616. package/cpp/llama.cpp/src/models/orion.cpp +123 -0
  617. package/cpp/llama.cpp/src/models/pangu-embedded.cpp +121 -0
  618. package/cpp/llama.cpp/src/models/phi2.cpp +121 -0
  619. package/cpp/llama.cpp/src/models/phi3.cpp +152 -0
  620. package/cpp/llama.cpp/src/models/plamo.cpp +110 -0
  621. package/cpp/llama.cpp/src/models/plamo2.cpp +316 -0
  622. package/cpp/llama.cpp/src/models/plm.cpp +168 -0
  623. package/cpp/llama.cpp/src/models/qwen.cpp +108 -0
  624. package/cpp/llama.cpp/src/models/qwen2.cpp +117 -0
  625. package/cpp/llama.cpp/src/models/qwen2moe.cpp +151 -0
  626. package/cpp/llama.cpp/src/models/qwen2vl.cpp +117 -0
  627. package/cpp/llama.cpp/src/models/qwen3.cpp +117 -0
  628. package/cpp/llama.cpp/src/models/qwen3moe.cpp +124 -0
  629. package/cpp/llama.cpp/src/models/qwen3vl-moe.cpp +149 -0
  630. package/cpp/llama.cpp/src/models/qwen3vl.cpp +141 -0
  631. package/cpp/llama.cpp/src/models/refact.cpp +94 -0
  632. package/cpp/llama.cpp/src/models/rwkv6-base.cpp +162 -0
  633. package/cpp/llama.cpp/src/models/rwkv6.cpp +94 -0
  634. package/cpp/llama.cpp/src/models/rwkv6qwen2.cpp +86 -0
  635. package/cpp/llama.cpp/src/models/rwkv7-base.cpp +135 -0
  636. package/cpp/llama.cpp/src/models/rwkv7.cpp +90 -0
  637. package/cpp/llama.cpp/src/models/seed-oss.cpp +124 -0
  638. package/cpp/llama.cpp/src/models/smallthinker.cpp +120 -0
  639. package/cpp/llama.cpp/src/models/smollm3.cpp +128 -0
  640. package/cpp/llama.cpp/src/models/stablelm.cpp +146 -0
  641. package/cpp/llama.cpp/src/models/starcoder.cpp +100 -0
  642. package/cpp/llama.cpp/src/models/starcoder2.cpp +121 -0
  643. package/cpp/llama.cpp/src/models/t5-dec.cpp +166 -0
  644. package/cpp/llama.cpp/src/models/t5-enc.cpp +96 -0
  645. package/cpp/llama.cpp/src/models/wavtokenizer-dec.cpp +149 -0
  646. package/cpp/llama.cpp/src/models/xverse.cpp +108 -0
  647. package/cpp/llama.cpp/src/unicode.cpp +77 -0
  648. package/cpp/llama.cpp/src/unicode.h +43 -0
  649. package/cpp/llama.cpp/vendor/cpp-httplib/CMakeLists.txt +94 -0
  650. package/cpp/llama.cpp/vendor/cpp-httplib/httplib.cpp +9339 -0
  651. package/cpp/llama.cpp/vendor/cpp-httplib/httplib.h +433 -8222
  652. package/cpp/llama.cpp/vendor/cpp-httplib/patch-boringssl.cmake +6 -0
  653. package/cpp/llama.cpp/vendor/miniaudio/miniaudio.h +4179 -1900
  654. package/cpp/llama.cpp/vendor/minja/chat-template.hpp +9 -2
  655. package/cpp/llama.cpp/vendor/minja/minja.hpp +101 -22
  656. package/ios/include/chat.h +16 -3
  657. package/ios/include/common/minja/chat-template.hpp +9 -2
  658. package/ios/include/common/minja/minja.hpp +101 -22
  659. package/ios/include/common.h +57 -19
  660. package/ios/include/json-schema-to-grammar.h +2 -0
  661. package/ios/include/llama.h +44 -21
  662. package/ios/include/log.h +12 -4
  663. package/ios/include/sampling.h +3 -1
  664. package/ios/libs/llama.xcframework/Info.plist +20 -20
  665. package/ios/libs/llama.xcframework/ios-arm64/dSYMs/llama.dSYM/Contents/Resources/DWARF/llama +0 -0
  666. package/ios/libs/llama.xcframework/ios-arm64/dSYMs/llama.dSYM/Contents/Resources/Relocations/aarch64/llama.yml +6399 -5557
  667. package/ios/libs/llama.xcframework/ios-arm64/llama.framework/Headers/ggml-backend.h +19 -1
  668. package/ios/libs/llama.xcframework/ios-arm64/llama.framework/Headers/ggml-cpu.h +1 -1
  669. package/ios/libs/llama.xcframework/ios-arm64/llama.framework/Headers/ggml-metal.h +1 -6
  670. package/ios/libs/llama.xcframework/ios-arm64/llama.framework/Headers/ggml.h +199 -6
  671. package/ios/libs/llama.xcframework/ios-arm64/llama.framework/Headers/llama.h +44 -21
  672. package/ios/libs/llama.xcframework/ios-arm64/llama.framework/llama +0 -0
  673. package/ios/libs/llama.xcframework/ios-arm64_x86_64-simulator/dSYMs/llama.dSYM/Contents/Resources/DWARF/llama +0 -0
  674. package/ios/libs/llama.xcframework/ios-arm64_x86_64-simulator/dSYMs/llama.dSYM/Contents/Resources/Relocations/aarch64/llama.yml +6362 -5520
  675. package/ios/libs/llama.xcframework/ios-arm64_x86_64-simulator/dSYMs/llama.dSYM/Contents/Resources/Relocations/x86_64/llama.yml +4813 -4241
  676. package/ios/libs/llama.xcframework/ios-arm64_x86_64-simulator/llama.framework/Headers/ggml-backend.h +19 -1
  677. package/ios/libs/llama.xcframework/ios-arm64_x86_64-simulator/llama.framework/Headers/ggml-cpu.h +1 -1
  678. package/ios/libs/llama.xcframework/ios-arm64_x86_64-simulator/llama.framework/Headers/ggml-metal.h +1 -6
  679. package/ios/libs/llama.xcframework/ios-arm64_x86_64-simulator/llama.framework/Headers/ggml.h +199 -6
  680. package/ios/libs/llama.xcframework/ios-arm64_x86_64-simulator/llama.framework/Headers/llama.h +44 -21
  681. package/ios/libs/llama.xcframework/ios-arm64_x86_64-simulator/llama.framework/llama +0 -0
  682. package/package.json +10 -4
  683. package/cpp/llama.cpp/ggml/src/ggml-cann/Doxyfile +0 -2579
  684. package/cpp/llama.cpp/ggml/src/ggml-cuda/fattn-tile-f16.cu +0 -371
  685. package/cpp/llama.cpp/ggml/src/ggml-cuda/fattn-tile-f16.cuh +0 -3
  686. package/cpp/llama.cpp/ggml/src/ggml-cuda/fattn-tile-f32.cu +0 -379
  687. package/cpp/llama.cpp/ggml/src/ggml-cuda/fattn-tile-f32.cuh +0 -3
  688. package/cpp/llama.cpp/ggml/src/ggml-cuda/fattn-vec-f16.cuh +0 -495
  689. package/cpp/llama.cpp/ggml/src/ggml-cuda/fattn-vec-f32.cuh +0 -486
  690. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-f16.cu +0 -5
  691. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q4_0.cu +0 -5
  692. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q4_1.cu +0 -5
  693. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q5_0.cu +0 -5
  694. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q5_1.cu +0 -5
  695. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q8_0.cu +0 -5
  696. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-f16.cu +0 -5
  697. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q4_0.cu +0 -5
  698. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q4_1.cu +0 -5
  699. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q5_0.cu +0 -5
  700. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q5_1.cu +0 -5
  701. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q8_0.cu +0 -5
  702. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-f16.cu +0 -5
  703. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q4_0.cu +0 -5
  704. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q4_1.cu +0 -5
  705. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q5_0.cu +0 -5
  706. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q5_1.cu +0 -5
  707. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q8_0.cu +0 -5
  708. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-f16.cu +0 -5
  709. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q4_0.cu +0 -5
  710. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q4_1.cu +0 -5
  711. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q5_0.cu +0 -5
  712. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q5_1.cu +0 -5
  713. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q8_0.cu +0 -5
  714. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-f16.cu +0 -5
  715. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q4_0.cu +0 -5
  716. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q4_1.cu +0 -5
  717. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q5_0.cu +0 -5
  718. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q5_1.cu +0 -5
  719. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q8_0.cu +0 -5
  720. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-f16.cu +0 -5
  721. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q4_0.cu +0 -5
  722. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q4_1.cu +0 -5
  723. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q5_0.cu +0 -5
  724. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q5_1.cu +0 -5
  725. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q8_0.cu +0 -5
  726. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs256-f16-f16.cu +0 -5
  727. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-f16.cu +0 -5
  728. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q4_0.cu +0 -5
  729. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q4_1.cu +0 -5
  730. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q5_0.cu +0 -5
  731. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q5_1.cu +0 -5
  732. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q8_0.cu +0 -5
  733. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-f16.cu +0 -5
  734. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q4_0.cu +0 -5
  735. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q4_1.cu +0 -5
  736. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q5_0.cu +0 -5
  737. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q5_1.cu +0 -5
  738. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q8_0.cu +0 -5
  739. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-f16.cu +0 -5
  740. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q4_0.cu +0 -5
  741. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q4_1.cu +0 -5
  742. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q5_0.cu +0 -5
  743. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q5_1.cu +0 -5
  744. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q8_0.cu +0 -5
  745. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-f16.cu +0 -5
  746. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q4_0.cu +0 -5
  747. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q4_1.cu +0 -5
  748. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q5_0.cu +0 -5
  749. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q5_1.cu +0 -5
  750. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q8_0.cu +0 -5
  751. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-f16.cu +0 -5
  752. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q4_0.cu +0 -5
  753. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q4_1.cu +0 -5
  754. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q5_0.cu +0 -5
  755. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q5_1.cu +0 -5
  756. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q8_0.cu +0 -5
  757. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-f16.cu +0 -5
  758. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q4_0.cu +0 -5
  759. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q4_1.cu +0 -5
  760. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q5_0.cu +0 -5
  761. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q5_1.cu +0 -5
  762. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q8_0.cu +0 -5
  763. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-f16.cu +0 -5
  764. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q4_0.cu +0 -5
  765. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q4_1.cu +0 -5
  766. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q5_0.cu +0 -5
  767. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q5_1.cu +0 -5
  768. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q8_0.cu +0 -5
  769. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs256-f16-f16.cu +0 -5
  770. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-f16.cu +0 -5
  771. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q4_0.cu +0 -5
  772. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q4_1.cu +0 -5
  773. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q5_0.cu +0 -5
  774. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q5_1.cu +0 -5
  775. package/cpp/llama.cpp/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q8_0.cu +0 -5
  776. package/cpp/llama.cpp/ggml/src/ggml-metal/ggml-metal.m +0 -6886
  777. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_base.comp +0 -154
  778. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/mul_mmq_funcs.comp +0 -105
  779. package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/rope_head.comp +0 -55
  780. package/cpp/llama.cpp/ggml/src/ggml-webgpu/wgsl-shaders/cpy.wgsl +0 -60
  781. package/cpp/llama.cpp/ggml/src/ggml-zdnn/ggml-zdnn-impl.h +0 -97
  782. package/cpp/llama.cpp/models/ggml-vocab-aquila.gguf +0 -0
  783. package/cpp/llama.cpp/models/ggml-vocab-baichuan.gguf +0 -0
  784. package/cpp/llama.cpp/models/ggml-vocab-bert-bge.gguf +0 -0
  785. package/cpp/llama.cpp/models/ggml-vocab-bert-bge.gguf.inp +0 -112
  786. package/cpp/llama.cpp/models/ggml-vocab-bert-bge.gguf.out +0 -46
  787. package/cpp/llama.cpp/models/ggml-vocab-command-r.gguf +0 -0
  788. package/cpp/llama.cpp/models/ggml-vocab-command-r.gguf.inp +0 -112
  789. package/cpp/llama.cpp/models/ggml-vocab-command-r.gguf.out +0 -46
  790. package/cpp/llama.cpp/models/ggml-vocab-deepseek-coder.gguf +0 -0
  791. package/cpp/llama.cpp/models/ggml-vocab-deepseek-coder.gguf.inp +0 -112
  792. package/cpp/llama.cpp/models/ggml-vocab-deepseek-coder.gguf.out +0 -46
  793. package/cpp/llama.cpp/models/ggml-vocab-deepseek-llm.gguf +0 -0
  794. package/cpp/llama.cpp/models/ggml-vocab-deepseek-llm.gguf.inp +0 -112
  795. package/cpp/llama.cpp/models/ggml-vocab-deepseek-llm.gguf.out +0 -46
  796. package/cpp/llama.cpp/models/ggml-vocab-falcon.gguf +0 -0
  797. package/cpp/llama.cpp/models/ggml-vocab-falcon.gguf.inp +0 -112
  798. package/cpp/llama.cpp/models/ggml-vocab-falcon.gguf.out +0 -46
  799. package/cpp/llama.cpp/models/ggml-vocab-gpt-2.gguf +0 -0
  800. package/cpp/llama.cpp/models/ggml-vocab-gpt-2.gguf.inp +0 -112
  801. package/cpp/llama.cpp/models/ggml-vocab-gpt-2.gguf.out +0 -46
  802. package/cpp/llama.cpp/models/ggml-vocab-gpt-neox.gguf +0 -0
  803. package/cpp/llama.cpp/models/ggml-vocab-llama-bpe.gguf +0 -0
  804. package/cpp/llama.cpp/models/ggml-vocab-llama-bpe.gguf.inp +0 -112
  805. package/cpp/llama.cpp/models/ggml-vocab-llama-bpe.gguf.out +0 -46
  806. package/cpp/llama.cpp/models/ggml-vocab-llama-spm.gguf +0 -0
  807. package/cpp/llama.cpp/models/ggml-vocab-llama-spm.gguf.inp +0 -112
  808. package/cpp/llama.cpp/models/ggml-vocab-llama-spm.gguf.out +0 -46
  809. package/cpp/llama.cpp/models/ggml-vocab-mpt.gguf +0 -0
  810. package/cpp/llama.cpp/models/ggml-vocab-mpt.gguf.inp +0 -112
  811. package/cpp/llama.cpp/models/ggml-vocab-mpt.gguf.out +0 -46
  812. package/cpp/llama.cpp/models/ggml-vocab-nomic-bert-moe.gguf +0 -0
  813. package/cpp/llama.cpp/models/ggml-vocab-phi-3.gguf +0 -0
  814. package/cpp/llama.cpp/models/ggml-vocab-phi-3.gguf.inp +0 -112
  815. package/cpp/llama.cpp/models/ggml-vocab-phi-3.gguf.out +0 -46
  816. package/cpp/llama.cpp/models/ggml-vocab-qwen2.gguf +0 -0
  817. package/cpp/llama.cpp/models/ggml-vocab-qwen2.gguf.inp +0 -112
  818. package/cpp/llama.cpp/models/ggml-vocab-qwen2.gguf.out +0 -46
  819. package/cpp/llama.cpp/models/ggml-vocab-refact.gguf +0 -0
  820. package/cpp/llama.cpp/models/ggml-vocab-refact.gguf.inp +0 -112
  821. package/cpp/llama.cpp/models/ggml-vocab-refact.gguf.out +0 -46
  822. package/cpp/llama.cpp/models/ggml-vocab-starcoder.gguf +0 -0
  823. package/cpp/llama.cpp/models/ggml-vocab-starcoder.gguf.inp +0 -112
  824. package/cpp/llama.cpp/models/ggml-vocab-starcoder.gguf.out +0 -46
  825. package/cpp/llama.cpp/models/templates/ByteDance-Seed-OSS.jinja +0 -171
  826. package/cpp/llama.cpp/models/templates/CohereForAI-c4ai-command-r-plus-tool_use.jinja +0 -202
  827. package/cpp/llama.cpp/models/templates/CohereForAI-c4ai-command-r7b-12-2024-tool_use.jinja +0 -156
  828. package/cpp/llama.cpp/models/templates/Mistral-Small-3.2-24B-Instruct-2506.jinja +0 -124
  829. package/cpp/llama.cpp/models/templates/NousResearch-Hermes-2-Pro-Llama-3-8B-tool_use.jinja +0 -152
  830. package/cpp/llama.cpp/models/templates/NousResearch-Hermes-3-Llama-3.1-8B-tool_use.jinja +0 -152
  831. package/cpp/llama.cpp/models/templates/Qwen-QwQ-32B.jinja +0 -62
  832. package/cpp/llama.cpp/models/templates/Qwen-Qwen2.5-7B-Instruct.jinja +0 -54
  833. package/cpp/llama.cpp/models/templates/Qwen-Qwen3-0.6B.jinja +0 -85
  834. package/cpp/llama.cpp/models/templates/README.md +0 -25
  835. package/cpp/llama.cpp/models/templates/deepseek-ai-DeepSeek-R1-Distill-Llama-8B.jinja +0 -1
  836. package/cpp/llama.cpp/models/templates/deepseek-ai-DeepSeek-R1-Distill-Qwen-32B.jinja +0 -1
  837. package/cpp/llama.cpp/models/templates/fireworks-ai-llama-3-firefunction-v2.jinja +0 -57
  838. package/cpp/llama.cpp/models/templates/google-gemma-2-2b-it.jinja +0 -4
  839. package/cpp/llama.cpp/models/templates/ibm-granite-granite-3.3-2B-Instruct.jinja +0 -59
  840. package/cpp/llama.cpp/models/templates/llama-cpp-deepseek-r1.jinja +0 -76
  841. package/cpp/llama.cpp/models/templates/llama-cpp-rwkv-world.jinja +0 -34
  842. package/cpp/llama.cpp/models/templates/meetkai-functionary-medium-v3.1.jinja +0 -58
  843. package/cpp/llama.cpp/models/templates/meetkai-functionary-medium-v3.2.jinja +0 -287
  844. package/cpp/llama.cpp/models/templates/meta-llama-Llama-3.1-8B-Instruct.jinja +0 -109
  845. package/cpp/llama.cpp/models/templates/meta-llama-Llama-3.2-3B-Instruct.jinja +0 -93
  846. package/cpp/llama.cpp/models/templates/meta-llama-Llama-3.3-70B-Instruct.jinja +0 -109
  847. package/cpp/llama.cpp/models/templates/microsoft-Phi-3.5-mini-instruct.jinja +0 -8
  848. package/cpp/llama.cpp/models/templates/mistralai-Mistral-Nemo-Instruct-2407.jinja +0 -87
  849. package/cpp/llama.cpp/models/templates/moonshotai-Kimi-K2.jinja +0 -43
  850. package/cpp/llama.cpp/models/templates/openai-gpt-oss-120b.jinja +0 -331
  851. package/cpp/llama.cpp/models/templates/unsloth-mistral-Devstral-Small-2507.jinja +0 -105
  852. package/cpp/llama.cpp/prompts/LLM-questions.txt +0 -49
  853. package/cpp/llama.cpp/prompts/alpaca.txt +0 -1
  854. package/cpp/llama.cpp/prompts/assistant.txt +0 -31
  855. package/cpp/llama.cpp/prompts/chat-with-baichuan.txt +0 -4
  856. package/cpp/llama.cpp/prompts/chat-with-bob.txt +0 -7
  857. package/cpp/llama.cpp/prompts/chat-with-qwen.txt +0 -1
  858. package/cpp/llama.cpp/prompts/chat-with-vicuna-v0.txt +0 -7
  859. package/cpp/llama.cpp/prompts/chat-with-vicuna-v1.txt +0 -7
  860. package/cpp/llama.cpp/prompts/chat.txt +0 -28
  861. package/cpp/llama.cpp/prompts/dan-modified.txt +0 -1
  862. package/cpp/llama.cpp/prompts/dan.txt +0 -1
  863. package/cpp/llama.cpp/prompts/mnemonics.txt +0 -93
  864. package/cpp/llama.cpp/prompts/parallel-questions.txt +0 -43
  865. package/cpp/llama.cpp/prompts/reason-act.txt +0 -18
  866. package/ios/libs/llama.xcframework/macos-arm64_x86_64/dSYMs/llama.dSYM/Contents/Info.plist +0 -20
  867. package/ios/libs/llama.xcframework/macos-arm64_x86_64/dSYMs/llama.dSYM/Contents/Resources/DWARF/llama +0 -0
  868. package/ios/libs/llama.xcframework/macos-arm64_x86_64/dSYMs/llama.dSYM/Contents/Resources/Relocations/aarch64/llama.yml +0 -5524
  869. package/ios/libs/llama.xcframework/macos-arm64_x86_64/dSYMs/llama.dSYM/Contents/Resources/Relocations/x86_64/llama.yml +0 -4247
  870. package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Headers/ggml-alloc.h +0 -76
  871. package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Headers/ggml-backend.h +0 -354
  872. package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Headers/ggml-blas.h +0 -25
  873. package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Headers/ggml-cpu.h +0 -145
  874. package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Headers/ggml-metal.h +0 -66
  875. package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Headers/ggml-opt.h +0 -256
  876. package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Headers/ggml.h +0 -2492
  877. package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Headers/gguf.h +0 -202
  878. package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Headers/llama.h +0 -1391
  879. package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Modules/module.modulemap +0 -17
  880. package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Resources/Info.plist +0 -32
  881. package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/A/Headers/ggml-alloc.h +0 -76
  882. package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/A/Headers/ggml-backend.h +0 -354
  883. package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/A/Headers/ggml-blas.h +0 -25
  884. package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/A/Headers/ggml-cpu.h +0 -145
  885. package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/A/Headers/ggml-metal.h +0 -66
  886. package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/A/Headers/ggml-opt.h +0 -256
  887. package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/A/Headers/ggml.h +0 -2492
  888. package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/A/Headers/gguf.h +0 -202
  889. package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/A/Headers/llama.h +0 -1391
  890. package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/A/Modules/module.modulemap +0 -17
  891. package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/A/Resources/Info.plist +0 -32
  892. package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/A/llama +0 -0
  893. package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/Current/Headers/ggml-alloc.h +0 -76
  894. package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/Current/Headers/ggml-backend.h +0 -354
  895. package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/Current/Headers/ggml-blas.h +0 -25
  896. package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/Current/Headers/ggml-cpu.h +0 -145
  897. package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/Current/Headers/ggml-metal.h +0 -66
  898. package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/Current/Headers/ggml-opt.h +0 -256
  899. package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/Current/Headers/ggml.h +0 -2492
  900. package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/Current/Headers/gguf.h +0 -202
  901. package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/Current/Headers/llama.h +0 -1391
  902. package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/Current/Modules/module.modulemap +0 -17
  903. package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/Current/Resources/Info.plist +0 -32
  904. package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/Versions/Current/llama +0 -0
  905. package/ios/libs/llama.xcframework/macos-arm64_x86_64/llama.framework/llama +0 -0
  906. package/ios/libs/llama.xcframework/tvos-arm64/dSYMs/llama.dSYM/Contents/Info.plist +0 -20
  907. package/ios/libs/llama.xcframework/tvos-arm64/dSYMs/llama.dSYM/Contents/Resources/DWARF/llama +0 -0
  908. package/ios/libs/llama.xcframework/tvos-arm64/dSYMs/llama.dSYM/Contents/Resources/Relocations/aarch64/llama.yml +0 -5561
  909. package/ios/libs/llama.xcframework/tvos-arm64/llama.framework/Headers/ggml-alloc.h +0 -76
  910. package/ios/libs/llama.xcframework/tvos-arm64/llama.framework/Headers/ggml-backend.h +0 -354
  911. package/ios/libs/llama.xcframework/tvos-arm64/llama.framework/Headers/ggml-blas.h +0 -25
  912. package/ios/libs/llama.xcframework/tvos-arm64/llama.framework/Headers/ggml-cpu.h +0 -145
  913. package/ios/libs/llama.xcframework/tvos-arm64/llama.framework/Headers/ggml-metal.h +0 -66
  914. package/ios/libs/llama.xcframework/tvos-arm64/llama.framework/Headers/ggml-opt.h +0 -256
  915. package/ios/libs/llama.xcframework/tvos-arm64/llama.framework/Headers/ggml.h +0 -2492
  916. package/ios/libs/llama.xcframework/tvos-arm64/llama.framework/Headers/gguf.h +0 -202
  917. package/ios/libs/llama.xcframework/tvos-arm64/llama.framework/Headers/llama.h +0 -1391
  918. package/ios/libs/llama.xcframework/tvos-arm64/llama.framework/Info.plist +0 -35
  919. package/ios/libs/llama.xcframework/tvos-arm64/llama.framework/Modules/module.modulemap +0 -17
  920. package/ios/libs/llama.xcframework/tvos-arm64/llama.framework/llama +0 -0
  921. package/ios/libs/llama.xcframework/tvos-arm64_x86_64-simulator/dSYMs/llama.dSYM/Contents/Info.plist +0 -20
  922. package/ios/libs/llama.xcframework/tvos-arm64_x86_64-simulator/dSYMs/llama.dSYM/Contents/Resources/DWARF/llama +0 -0
  923. package/ios/libs/llama.xcframework/tvos-arm64_x86_64-simulator/dSYMs/llama.dSYM/Contents/Resources/Relocations/aarch64/llama.yml +0 -5524
  924. package/ios/libs/llama.xcframework/tvos-arm64_x86_64-simulator/dSYMs/llama.dSYM/Contents/Resources/Relocations/x86_64/llama.yml +0 -4246
  925. package/ios/libs/llama.xcframework/tvos-arm64_x86_64-simulator/llama.framework/Headers/ggml-alloc.h +0 -76
  926. package/ios/libs/llama.xcframework/tvos-arm64_x86_64-simulator/llama.framework/Headers/ggml-backend.h +0 -354
  927. package/ios/libs/llama.xcframework/tvos-arm64_x86_64-simulator/llama.framework/Headers/ggml-blas.h +0 -25
  928. package/ios/libs/llama.xcframework/tvos-arm64_x86_64-simulator/llama.framework/Headers/ggml-cpu.h +0 -145
  929. package/ios/libs/llama.xcframework/tvos-arm64_x86_64-simulator/llama.framework/Headers/ggml-metal.h +0 -66
  930. package/ios/libs/llama.xcframework/tvos-arm64_x86_64-simulator/llama.framework/Headers/ggml-opt.h +0 -256
  931. package/ios/libs/llama.xcframework/tvos-arm64_x86_64-simulator/llama.framework/Headers/ggml.h +0 -2492
  932. package/ios/libs/llama.xcframework/tvos-arm64_x86_64-simulator/llama.framework/Headers/gguf.h +0 -202
  933. package/ios/libs/llama.xcframework/tvos-arm64_x86_64-simulator/llama.framework/Headers/llama.h +0 -1391
  934. package/ios/libs/llama.xcframework/tvos-arm64_x86_64-simulator/llama.framework/Info.plist +0 -35
  935. package/ios/libs/llama.xcframework/tvos-arm64_x86_64-simulator/llama.framework/Modules/module.modulemap +0 -17
  936. package/ios/libs/llama.xcframework/tvos-arm64_x86_64-simulator/llama.framework/llama +0 -0
  937. package/ios/libs/llama.xcframework/xros-arm64/dSYMs/llama.dSYM/Contents/Info.plist +0 -20
  938. package/ios/libs/llama.xcframework/xros-arm64/dSYMs/llama.dSYM/Contents/Resources/DWARF/llama +0 -0
  939. package/ios/libs/llama.xcframework/xros-arm64/dSYMs/llama.dSYM/Contents/Resources/Relocations/aarch64/llama.yml +0 -5558
  940. package/ios/libs/llama.xcframework/xros-arm64/llama.framework/Headers/ggml-alloc.h +0 -76
  941. package/ios/libs/llama.xcframework/xros-arm64/llama.framework/Headers/ggml-backend.h +0 -354
  942. package/ios/libs/llama.xcframework/xros-arm64/llama.framework/Headers/ggml-blas.h +0 -25
  943. package/ios/libs/llama.xcframework/xros-arm64/llama.framework/Headers/ggml-cpu.h +0 -145
  944. package/ios/libs/llama.xcframework/xros-arm64/llama.framework/Headers/ggml-metal.h +0 -66
  945. package/ios/libs/llama.xcframework/xros-arm64/llama.framework/Headers/ggml-opt.h +0 -256
  946. package/ios/libs/llama.xcframework/xros-arm64/llama.framework/Headers/ggml.h +0 -2492
  947. package/ios/libs/llama.xcframework/xros-arm64/llama.framework/Headers/gguf.h +0 -202
  948. package/ios/libs/llama.xcframework/xros-arm64/llama.framework/Headers/llama.h +0 -1391
  949. package/ios/libs/llama.xcframework/xros-arm64/llama.framework/Info.plist +0 -32
  950. package/ios/libs/llama.xcframework/xros-arm64/llama.framework/Modules/module.modulemap +0 -17
  951. package/ios/libs/llama.xcframework/xros-arm64/llama.framework/llama +0 -0
  952. package/ios/libs/llama.xcframework/xros-arm64_x86_64-simulator/dSYMs/llama.dSYM/Contents/Info.plist +0 -20
  953. package/ios/libs/llama.xcframework/xros-arm64_x86_64-simulator/dSYMs/llama.dSYM/Contents/Resources/DWARF/llama +0 -0
  954. package/ios/libs/llama.xcframework/xros-arm64_x86_64-simulator/dSYMs/llama.dSYM/Contents/Resources/Relocations/aarch64/llama.yml +0 -5520
  955. package/ios/libs/llama.xcframework/xros-arm64_x86_64-simulator/dSYMs/llama.dSYM/Contents/Resources/Relocations/x86_64/llama.yml +0 -4243
  956. package/ios/libs/llama.xcframework/xros-arm64_x86_64-simulator/llama.framework/Headers/ggml-alloc.h +0 -76
  957. package/ios/libs/llama.xcframework/xros-arm64_x86_64-simulator/llama.framework/Headers/ggml-backend.h +0 -354
  958. package/ios/libs/llama.xcframework/xros-arm64_x86_64-simulator/llama.framework/Headers/ggml-blas.h +0 -25
  959. package/ios/libs/llama.xcframework/xros-arm64_x86_64-simulator/llama.framework/Headers/ggml-cpu.h +0 -145
  960. package/ios/libs/llama.xcframework/xros-arm64_x86_64-simulator/llama.framework/Headers/ggml-metal.h +0 -66
  961. package/ios/libs/llama.xcframework/xros-arm64_x86_64-simulator/llama.framework/Headers/ggml-opt.h +0 -256
  962. package/ios/libs/llama.xcframework/xros-arm64_x86_64-simulator/llama.framework/Headers/ggml.h +0 -2492
  963. package/ios/libs/llama.xcframework/xros-arm64_x86_64-simulator/llama.framework/Headers/gguf.h +0 -202
  964. package/ios/libs/llama.xcframework/xros-arm64_x86_64-simulator/llama.framework/Headers/llama.h +0 -1391
  965. package/ios/libs/llama.xcframework/xros-arm64_x86_64-simulator/llama.framework/Info.plist +0 -32
  966. package/ios/libs/llama.xcframework/xros-arm64_x86_64-simulator/llama.framework/Modules/module.modulemap +0 -17
  967. package/ios/libs/llama.xcframework/xros-arm64_x86_64-simulator/llama.framework/llama +0 -0
  968. /package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/{test_bfloat16_support.comp → feature-tests/bfloat16.comp} +0 -0
  969. /package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/{test_coopmat_support.comp → feature-tests/coopmat.comp} +0 -0
  970. /package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/{test_coopmat2_support.comp → feature-tests/coopmat2.comp} +0 -0
  971. /package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/{test_integer_dot_support.comp → feature-tests/integer_dot.comp} +0 -0
  972. /package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/{generic_head.comp → generic_head.glsl} +0 -0
  973. /package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/{generic_unary_head.comp → generic_unary_head.glsl} +0 -0
  974. /package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/{glu_main.comp → glu_main.glsl} +0 -0
  975. /package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/{rte.comp → rte.glsl} +0 -0
  976. /package/cpp/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/{utils.comp → utils.glsl} +0 -0
@@ -192,18 +192,25 @@ class chat_template {
192
192
  };
193
193
  };
194
194
  const json dummy_args_obj {{"argument_needle", "print('Hello, World!')"}};
195
+ const auto contains_arg_needle = [&](const std::string & out_str) {
196
+ return contains(out_str, "<parameter=argument_needle>")
197
+ || contains(out_str, "\"argument_needle\":")
198
+ || contains(out_str, "'argument_needle':")
199
+ || contains(out_str, ">argument_needle<")
200
+ || contains(out_str, "<parameter name=\"argument_needle\">");
201
+ };
195
202
 
196
203
  // Note: the arguments are rendered in both cases, but may be double-escaped, which we don't want.
197
204
  out = try_raw_render(json::array({
198
205
  dummy_user_msg,
199
206
  make_tool_calls_msg(json::array({make_tool_call("ipython", dummy_args_obj.dump())})),
200
207
  }), {}, false);
201
- auto tool_call_renders_str_arguments = contains(out, "<parameter=argument_needle>") || contains(out, "\"argument_needle\":") || contains(out, "'argument_needle':");
208
+ auto tool_call_renders_str_arguments = contains_arg_needle(out);
202
209
  out = try_raw_render(json::array({
203
210
  dummy_user_msg,
204
211
  make_tool_calls_msg(json::array({make_tool_call("ipython", dummy_args_obj)})),
205
212
  }), {}, false);
206
- auto tool_call_renders_obj_arguments = contains(out, "<parameter=argument_needle>") || contains(out, "\"argument_needle\":") || contains(out, "'argument_needle':");
213
+ auto tool_call_renders_obj_arguments = contains_arg_needle(out);
207
214
 
208
215
  caps_.supports_tool_calls = tool_call_renders_str_arguments || tool_call_renders_obj_arguments;
209
216
  caps_.requires_object_arguments = !tool_call_renders_str_arguments && tool_call_renders_obj_arguments;
@@ -55,7 +55,7 @@ inline std::string normalize_newlines(const std::string & s) {
55
55
  }
56
56
 
57
57
  /* Values that behave roughly like in Python. */
58
- class Value : public std::enable_shared_from_this<Value> {
58
+ class Value {
59
59
  public:
60
60
  using CallableType = std::function<Value(const std::shared_ptr<Context> &, ArgumentsValue &)>;
61
61
  using FilterType = std::function<Value(const std::shared_ptr<Context> &, ArgumentsValue &)>;
@@ -158,12 +158,14 @@ public:
158
158
  Value(const json & v) {
159
159
  if (v.is_object()) {
160
160
  auto object = std::make_shared<ObjectType>();
161
+ object->reserve(v.size());
161
162
  for (auto it = v.begin(); it != v.end(); ++it) {
162
- (*object)[it.key()] = it.value();
163
+ object->emplace_back(it.key(), Value(it.value()));
163
164
  }
164
165
  object_ = std::move(object);
165
166
  } else if (v.is_array()) {
166
167
  auto array = std::make_shared<ArrayType>();
168
+ array->reserve(v.size());
167
169
  for (const auto& item : v) {
168
170
  array->push_back(Value(item));
169
171
  }
@@ -610,7 +612,7 @@ static std::string error_location_suffix(const std::string & source, size_t pos)
610
612
  return out.str();
611
613
  }
612
614
 
613
- class Context : public std::enable_shared_from_this<Context> {
615
+ class Context {
614
616
  protected:
615
617
  Value values_;
616
618
  std::shared_ptr<Context> parent_;
@@ -706,7 +708,7 @@ enum SpaceHandling { Keep, Strip, StripSpaces, StripNewline };
706
708
 
707
709
  class TemplateToken {
708
710
  public:
709
- enum class Type { Text, Expression, If, Else, Elif, EndIf, For, EndFor, Generation, EndGeneration, Set, EndSet, Comment, Macro, EndMacro, Filter, EndFilter, Break, Continue };
711
+ enum class Type { Text, Expression, If, Else, Elif, EndIf, For, EndFor, Generation, EndGeneration, Set, EndSet, Comment, Macro, EndMacro, Filter, EndFilter, Break, Continue, Call, EndCall };
710
712
 
711
713
  static std::string typeToString(Type t) {
712
714
  switch (t) {
@@ -729,6 +731,8 @@ public:
729
731
  case Type::EndGeneration: return "endgeneration";
730
732
  case Type::Break: return "break";
731
733
  case Type::Continue: return "continue";
734
+ case Type::Call: return "call";
735
+ case Type::EndCall: return "endcall";
732
736
  }
733
737
  return "Unknown";
734
738
  }
@@ -846,6 +850,17 @@ struct LoopControlTemplateToken : public TemplateToken {
846
850
  LoopControlTemplateToken(const Location & loc, SpaceHandling pre, SpaceHandling post, LoopControlType control_type) : TemplateToken(Type::Break, loc, pre, post), control_type(control_type) {}
847
851
  };
848
852
 
853
+ struct CallTemplateToken : public TemplateToken {
854
+ std::shared_ptr<Expression> expr;
855
+ CallTemplateToken(const Location & loc, SpaceHandling pre, SpaceHandling post, std::shared_ptr<Expression> && e)
856
+ : TemplateToken(Type::Call, loc, pre, post), expr(std::move(e)) {}
857
+ };
858
+
859
+ struct EndCallTemplateToken : public TemplateToken {
860
+ EndCallTemplateToken(const Location & loc, SpaceHandling pre, SpaceHandling post)
861
+ : TemplateToken(Type::EndCall, loc, pre, post) {}
862
+ };
863
+
849
864
  class TemplateNode {
850
865
  Location location_;
851
866
  protected:
@@ -1047,36 +1062,48 @@ public:
1047
1062
  }
1048
1063
  }
1049
1064
  }
1050
- void do_render(std::ostringstream &, const std::shared_ptr<Context> & macro_context) const override {
1065
+ void do_render(std::ostringstream &, const std::shared_ptr<Context> & context) const override {
1051
1066
  if (!name) throw std::runtime_error("MacroNode.name is null");
1052
1067
  if (!body) throw std::runtime_error("MacroNode.body is null");
1053
- auto callable = Value::callable([&](const std::shared_ptr<Context> & context, ArgumentsValue & args) {
1054
- auto call_context = macro_context;
1068
+
1069
+ // Use init-capture to avoid dangling 'this' pointer and circular references
1070
+ auto callable = Value::callable([weak_context = std::weak_ptr<Context>(context),
1071
+ name = name, params = params, body = body,
1072
+ named_param_positions = named_param_positions]
1073
+ (const std::shared_ptr<Context> & call_context, ArgumentsValue & args) {
1074
+ auto context_locked = weak_context.lock();
1075
+ if (!context_locked) throw std::runtime_error("Macro context no longer valid");
1076
+ auto execution_context = Context::make(Value::object(), context_locked);
1077
+
1078
+ if (call_context->contains("caller")) {
1079
+ execution_context->set("caller", call_context->get("caller"));
1080
+ }
1081
+
1055
1082
  std::vector<bool> param_set(params.size(), false);
1056
1083
  for (size_t i = 0, n = args.args.size(); i < n; i++) {
1057
1084
  auto & arg = args.args[i];
1058
1085
  if (i >= params.size()) throw std::runtime_error("Too many positional arguments for macro " + name->get_name());
1059
1086
  param_set[i] = true;
1060
- auto & param_name = params[i].first;
1061
- call_context->set(param_name, arg);
1087
+ const auto & param_name = params[i].first;
1088
+ execution_context->set(param_name, arg);
1062
1089
  }
1063
1090
  for (auto & [arg_name, value] : args.kwargs) {
1064
1091
  auto it = named_param_positions.find(arg_name);
1065
1092
  if (it == named_param_positions.end()) throw std::runtime_error("Unknown parameter name for macro " + name->get_name() + ": " + arg_name);
1066
1093
 
1067
- call_context->set(arg_name, value);
1094
+ execution_context->set(arg_name, value);
1068
1095
  param_set[it->second] = true;
1069
1096
  }
1070
1097
  // Set default values for parameters that were not passed
1071
1098
  for (size_t i = 0, n = params.size(); i < n; i++) {
1072
1099
  if (!param_set[i] && params[i].second != nullptr) {
1073
- auto val = params[i].second->evaluate(context);
1074
- call_context->set(params[i].first, val);
1100
+ auto val = params[i].second->evaluate(call_context);
1101
+ execution_context->set(params[i].first, val);
1075
1102
  }
1076
1103
  }
1077
- return body->render(call_context);
1104
+ return body->render(execution_context);
1078
1105
  });
1079
- macro_context->set(name->get_name(), callable);
1106
+ context->set(name->get_name(), callable);
1080
1107
  }
1081
1108
  };
1082
1109
 
@@ -1611,6 +1638,44 @@ public:
1611
1638
  }
1612
1639
  };
1613
1640
 
1641
+ class CallNode : public TemplateNode {
1642
+ std::shared_ptr<Expression> expr;
1643
+ std::shared_ptr<TemplateNode> body;
1644
+
1645
+ public:
1646
+ CallNode(const Location & loc, std::shared_ptr<Expression> && e, std::shared_ptr<TemplateNode> && b)
1647
+ : TemplateNode(loc), expr(std::move(e)), body(std::move(b)) {}
1648
+
1649
+ void do_render(std::ostringstream & out, const std::shared_ptr<Context> & context) const override {
1650
+ if (!expr) throw std::runtime_error("CallNode.expr is null");
1651
+ if (!body) throw std::runtime_error("CallNode.body is null");
1652
+
1653
+ // Use init-capture to avoid dangling 'this' pointer and circular references
1654
+ auto caller = Value::callable([weak_context = std::weak_ptr<Context>(context), body=body]
1655
+ (const std::shared_ptr<Context> &, ArgumentsValue &) -> Value {
1656
+ auto context_locked = weak_context.lock();
1657
+ if (!context_locked) throw std::runtime_error("Caller context no longer valid");
1658
+ return Value(body->render(context_locked));
1659
+ });
1660
+
1661
+ context->set("caller", caller);
1662
+
1663
+ auto call_expr = dynamic_cast<CallExpr*>(expr.get());
1664
+ if (!call_expr) {
1665
+ throw std::runtime_error("Invalid call block syntax - expected function call");
1666
+ }
1667
+
1668
+ Value function = call_expr->object->evaluate(context);
1669
+ if (!function.is_callable()) {
1670
+ throw std::runtime_error("Call target must be callable: " + function.dump());
1671
+ }
1672
+ ArgumentsValue args = call_expr->args.evaluate(context);
1673
+
1674
+ Value result = function.call(context, args);
1675
+ out << result.to_str();
1676
+ }
1677
+ };
1678
+
1614
1679
  class FilterExpr : public Expression {
1615
1680
  std::vector<std::shared_ptr<Expression>> parts;
1616
1681
  public:
@@ -2140,7 +2205,7 @@ private:
2140
2205
 
2141
2206
  auto value = parseValue();
2142
2207
 
2143
- while (it != end && consumeSpaces() && peekSymbols({ "[", "." })) {
2208
+ while (it != end && consumeSpaces() && peekSymbols({ "[", ".", "(" })) {
2144
2209
  if (!consumeToken("[").empty()) {
2145
2210
  std::shared_ptr<Expression> index;
2146
2211
  auto slice_loc = get_location();
@@ -2185,15 +2250,13 @@ private:
2185
2250
  auto key = std::make_shared<LiteralExpr>(identifier->location, Value(identifier->get_name()));
2186
2251
  value = std::make_shared<SubscriptExpr>(identifier->location, std::move(value), std::move(key));
2187
2252
  }
2253
+ } else if (peekSymbols({ "(" })) {
2254
+ auto callParams = parseCallArgs();
2255
+ value = std::make_shared<CallExpr>(get_location(), std::move(value), std::move(callParams));
2188
2256
  }
2189
2257
  consumeSpaces();
2190
2258
  }
2191
2259
 
2192
- if (peekSymbols({ "(" })) {
2193
- auto location = get_location();
2194
- auto callParams = parseCallArgs();
2195
- value = std::make_shared<CallExpr>(location, std::move(value), std::move(callParams));
2196
- }
2197
2260
  return value;
2198
2261
  }
2199
2262
 
@@ -2320,7 +2383,7 @@ private:
2320
2383
  static std::regex comment_tok(R"(\{#([-~]?)([\s\S]*?)([-~]?)#\})");
2321
2384
  static std::regex expr_open_regex(R"(\{\{([-~])?)");
2322
2385
  static std::regex block_open_regex(R"(^\{%([-~])?\s*)");
2323
- static std::regex block_keyword_tok(R"((if|else|elif|endif|for|endfor|generation|endgeneration|set|endset|block|endblock|macro|endmacro|filter|endfilter|break|continue)\b)");
2386
+ static std::regex block_keyword_tok(R"((if|else|elif|endif|for|endfor|generation|endgeneration|set|endset|block|endblock|macro|endmacro|filter|endfilter|break|continue|call|endcall)\b)");
2324
2387
  static std::regex non_text_open_regex(R"(\{\{|\{%|\{#)");
2325
2388
  static std::regex expr_close_regex(R"(\s*([-~])?\}\})");
2326
2389
  static std::regex block_close_regex(R"(\s*([-~])?%\})");
@@ -2443,6 +2506,15 @@ private:
2443
2506
  } else if (keyword == "endmacro") {
2444
2507
  auto post_space = parseBlockClose();
2445
2508
  tokens.push_back(std::make_unique<EndMacroTemplateToken>(location, pre_space, post_space));
2509
+ } else if (keyword == "call") {
2510
+ auto expr = parseExpression();
2511
+ if (!expr) throw std::runtime_error("Expected expression in call block");
2512
+
2513
+ auto post_space = parseBlockClose();
2514
+ tokens.push_back(std::make_unique<CallTemplateToken>(location, pre_space, post_space, std::move(expr)));
2515
+ } else if (keyword == "endcall") {
2516
+ auto post_space = parseBlockClose();
2517
+ tokens.push_back(std::make_unique<EndCallTemplateToken>(location, pre_space, post_space));
2446
2518
  } else if (keyword == "filter") {
2447
2519
  auto filter = parseExpression();
2448
2520
  if (!filter) throw std::runtime_error("Expected expression in filter block");
@@ -2575,6 +2647,12 @@ private:
2575
2647
  throw unterminated(**start);
2576
2648
  }
2577
2649
  children.emplace_back(std::make_shared<MacroNode>(token->location, std::move(macro_token->name), std::move(macro_token->params), std::move(body)));
2650
+ } else if (auto call_token = dynamic_cast<CallTemplateToken*>(token.get())) {
2651
+ auto body = parseTemplate(begin, it, end);
2652
+ if (it == end || (*(it++))->type != TemplateToken::Type::EndCall) {
2653
+ throw unterminated(**start);
2654
+ }
2655
+ children.emplace_back(std::make_shared<CallNode>(token->location, std::move(call_token->expr), std::move(body)));
2578
2656
  } else if (auto filter_token = dynamic_cast<FilterTemplateToken*>(token.get())) {
2579
2657
  auto body = parseTemplate(begin, it, end);
2580
2658
  if (it == end || (*(it++))->type != TemplateToken::Type::EndFilter) {
@@ -2588,6 +2666,7 @@ private:
2588
2666
  } else if (dynamic_cast<EndForTemplateToken*>(token.get())
2589
2667
  || dynamic_cast<EndSetTemplateToken*>(token.get())
2590
2668
  || dynamic_cast<EndMacroTemplateToken*>(token.get())
2669
+ || dynamic_cast<EndCallTemplateToken*>(token.get())
2591
2670
  || dynamic_cast<EndFilterTemplateToken*>(token.get())
2592
2671
  || dynamic_cast<EndIfTemplateToken*>(token.get())
2593
2672
  || dynamic_cast<ElseTemplateToken*>(token.get())
@@ -2657,7 +2736,7 @@ inline std::shared_ptr<Context> Context::builtins() {
2657
2736
  globals.set("raise_exception", simple_function("raise_exception", { "message" }, [](const std::shared_ptr<Context> &, Value & args) -> Value {
2658
2737
  throw std::runtime_error(args.at("message").get<std::string>());
2659
2738
  }));
2660
- globals.set("tojson", simple_function("tojson", { "value", "indent" }, [](const std::shared_ptr<Context> &, Value & args) {
2739
+ globals.set("tojson", simple_function("tojson", { "value", "indent", "ensure_ascii" }, [](const std::shared_ptr<Context> &, Value & args) {
2661
2740
  return Value(args.at("value").dump(args.get<int64_t>("indent", -1), /* to_json= */ true));
2662
2741
  }));
2663
2742
  globals.set("items", simple_function("items", { "object" }, [](const std::shared_ptr<Context> &, Value & args) {
@@ -33,8 +33,8 @@ struct common_chat_msg_content_part {
33
33
  struct common_chat_msg {
34
34
  std::string role;
35
35
  std::string content;
36
- std::vector<common_chat_msg_content_part> content_parts = {};
37
- std::vector<common_chat_tool_call> tool_calls = {};
36
+ std::vector<common_chat_msg_content_part> content_parts;
37
+ std::vector<common_chat_tool_call> tool_calls;
38
38
  std::string reasoning_content;
39
39
  std::string tool_name;
40
40
  std::string tool_call_id;
@@ -44,7 +44,7 @@ struct common_chat_msg {
44
44
  bool empty() const {
45
45
  return content.empty() && content_parts.empty() && tool_calls.empty() && reasoning_content.empty() && tool_name.empty() && tool_call_id.empty();
46
46
  }
47
- void ensure_tool_call_ids_set(std::vector<std::string> & ids_cache, const std::function<std::string()> & gen_tool_call_id) {
47
+ void set_tool_call_ids(std::vector<std::string> & ids_cache, const std::function<std::string()> & gen_tool_call_id) {
48
48
  for (auto i = 0u; i < tool_calls.size(); i++) {
49
49
  if (ids_cache.size() <= i) {
50
50
  auto id = tool_calls[i].id;
@@ -101,17 +101,28 @@ enum common_chat_format {
101
101
  COMMON_CHAT_FORMAT_CONTENT_ONLY,
102
102
  COMMON_CHAT_FORMAT_GENERIC,
103
103
  COMMON_CHAT_FORMAT_MISTRAL_NEMO,
104
+ COMMON_CHAT_FORMAT_MAGISTRAL,
104
105
  COMMON_CHAT_FORMAT_LLAMA_3_X,
105
106
  COMMON_CHAT_FORMAT_LLAMA_3_X_WITH_BUILTIN_TOOLS,
106
107
  COMMON_CHAT_FORMAT_DEEPSEEK_R1,
107
108
  COMMON_CHAT_FORMAT_FIREFUNCTION_V2,
108
109
  COMMON_CHAT_FORMAT_FUNCTIONARY_V3_2,
109
110
  COMMON_CHAT_FORMAT_FUNCTIONARY_V3_1_LLAMA_3_1,
111
+ COMMON_CHAT_FORMAT_DEEPSEEK_V3_1,
110
112
  COMMON_CHAT_FORMAT_HERMES_2_PRO,
111
113
  COMMON_CHAT_FORMAT_COMMAND_R7B,
112
114
  COMMON_CHAT_FORMAT_GRANITE,
113
115
  COMMON_CHAT_FORMAT_GPT_OSS,
114
116
  COMMON_CHAT_FORMAT_SEED_OSS,
117
+ COMMON_CHAT_FORMAT_NEMOTRON_V2,
118
+ COMMON_CHAT_FORMAT_APERTUS,
119
+ COMMON_CHAT_FORMAT_LFM2_WITH_JSON_TOOLS,
120
+ COMMON_CHAT_FORMAT_GLM_4_5,
121
+ COMMON_CHAT_FORMAT_MINIMAX_M2,
122
+ COMMON_CHAT_FORMAT_KIMI_K2,
123
+ COMMON_CHAT_FORMAT_QWEN3_CODER_XML,
124
+ COMMON_CHAT_FORMAT_APRIEL_1_5,
125
+ COMMON_CHAT_FORMAT_XIAOMI_MIMO,
115
126
 
116
127
  COMMON_CHAT_FORMAT_COUNT, // Not a format, just the # formats
117
128
  };
@@ -198,6 +209,8 @@ common_chat_msg common_chat_parse(const std::string & input, bool is_p
198
209
 
199
210
  common_chat_tool_choice common_chat_tool_choice_parse_oaicompat(const std::string & tool_choice);
200
211
 
212
+ bool common_chat_templates_support_enable_thinking(const common_chat_templates * chat_templates);
213
+
201
214
  // Parses a JSON array of messages in OpenAI's chat completion API format.
202
215
  // T can be std::string containing JSON or nlohmann::ordered_json
203
216
  template <class T> std::vector<common_chat_msg> common_chat_msgs_parse_oaicompat(const T & messages);
@@ -192,18 +192,25 @@ class chat_template {
192
192
  };
193
193
  };
194
194
  const json dummy_args_obj {{"argument_needle", "print('Hello, World!')"}};
195
+ const auto contains_arg_needle = [&](const std::string & out_str) {
196
+ return contains(out_str, "<parameter=argument_needle>")
197
+ || contains(out_str, "\"argument_needle\":")
198
+ || contains(out_str, "'argument_needle':")
199
+ || contains(out_str, ">argument_needle<")
200
+ || contains(out_str, "<parameter name=\"argument_needle\">");
201
+ };
195
202
 
196
203
  // Note: the arguments are rendered in both cases, but may be double-escaped, which we don't want.
197
204
  out = try_raw_render(json::array({
198
205
  dummy_user_msg,
199
206
  make_tool_calls_msg(json::array({make_tool_call("ipython", dummy_args_obj.dump())})),
200
207
  }), {}, false);
201
- auto tool_call_renders_str_arguments = contains(out, "<parameter=argument_needle>") || contains(out, "\"argument_needle\":") || contains(out, "'argument_needle':");
208
+ auto tool_call_renders_str_arguments = contains_arg_needle(out);
202
209
  out = try_raw_render(json::array({
203
210
  dummy_user_msg,
204
211
  make_tool_calls_msg(json::array({make_tool_call("ipython", dummy_args_obj)})),
205
212
  }), {}, false);
206
- auto tool_call_renders_obj_arguments = contains(out, "<parameter=argument_needle>") || contains(out, "\"argument_needle\":") || contains(out, "'argument_needle':");
213
+ auto tool_call_renders_obj_arguments = contains_arg_needle(out);
207
214
 
208
215
  caps_.supports_tool_calls = tool_call_renders_str_arguments || tool_call_renders_obj_arguments;
209
216
  caps_.requires_object_arguments = !tool_call_renders_str_arguments && tool_call_renders_obj_arguments;
@@ -55,7 +55,7 @@ inline std::string normalize_newlines(const std::string & s) {
55
55
  }
56
56
 
57
57
  /* Values that behave roughly like in Python. */
58
- class Value : public std::enable_shared_from_this<Value> {
58
+ class Value {
59
59
  public:
60
60
  using CallableType = std::function<Value(const std::shared_ptr<Context> &, ArgumentsValue &)>;
61
61
  using FilterType = std::function<Value(const std::shared_ptr<Context> &, ArgumentsValue &)>;
@@ -158,12 +158,14 @@ public:
158
158
  Value(const json & v) {
159
159
  if (v.is_object()) {
160
160
  auto object = std::make_shared<ObjectType>();
161
+ object->reserve(v.size());
161
162
  for (auto it = v.begin(); it != v.end(); ++it) {
162
- (*object)[it.key()] = it.value();
163
+ object->emplace_back(it.key(), Value(it.value()));
163
164
  }
164
165
  object_ = std::move(object);
165
166
  } else if (v.is_array()) {
166
167
  auto array = std::make_shared<ArrayType>();
168
+ array->reserve(v.size());
167
169
  for (const auto& item : v) {
168
170
  array->push_back(Value(item));
169
171
  }
@@ -610,7 +612,7 @@ static std::string error_location_suffix(const std::string & source, size_t pos)
610
612
  return out.str();
611
613
  }
612
614
 
613
- class Context : public std::enable_shared_from_this<Context> {
615
+ class Context {
614
616
  protected:
615
617
  Value values_;
616
618
  std::shared_ptr<Context> parent_;
@@ -706,7 +708,7 @@ enum SpaceHandling { Keep, Strip, StripSpaces, StripNewline };
706
708
 
707
709
  class TemplateToken {
708
710
  public:
709
- enum class Type { Text, Expression, If, Else, Elif, EndIf, For, EndFor, Generation, EndGeneration, Set, EndSet, Comment, Macro, EndMacro, Filter, EndFilter, Break, Continue };
711
+ enum class Type { Text, Expression, If, Else, Elif, EndIf, For, EndFor, Generation, EndGeneration, Set, EndSet, Comment, Macro, EndMacro, Filter, EndFilter, Break, Continue, Call, EndCall };
710
712
 
711
713
  static std::string typeToString(Type t) {
712
714
  switch (t) {
@@ -729,6 +731,8 @@ public:
729
731
  case Type::EndGeneration: return "endgeneration";
730
732
  case Type::Break: return "break";
731
733
  case Type::Continue: return "continue";
734
+ case Type::Call: return "call";
735
+ case Type::EndCall: return "endcall";
732
736
  }
733
737
  return "Unknown";
734
738
  }
@@ -846,6 +850,17 @@ struct LoopControlTemplateToken : public TemplateToken {
846
850
  LoopControlTemplateToken(const Location & loc, SpaceHandling pre, SpaceHandling post, LoopControlType control_type) : TemplateToken(Type::Break, loc, pre, post), control_type(control_type) {}
847
851
  };
848
852
 
853
+ struct CallTemplateToken : public TemplateToken {
854
+ std::shared_ptr<Expression> expr;
855
+ CallTemplateToken(const Location & loc, SpaceHandling pre, SpaceHandling post, std::shared_ptr<Expression> && e)
856
+ : TemplateToken(Type::Call, loc, pre, post), expr(std::move(e)) {}
857
+ };
858
+
859
+ struct EndCallTemplateToken : public TemplateToken {
860
+ EndCallTemplateToken(const Location & loc, SpaceHandling pre, SpaceHandling post)
861
+ : TemplateToken(Type::EndCall, loc, pre, post) {}
862
+ };
863
+
849
864
  class TemplateNode {
850
865
  Location location_;
851
866
  protected:
@@ -1047,36 +1062,48 @@ public:
1047
1062
  }
1048
1063
  }
1049
1064
  }
1050
- void do_render(std::ostringstream &, const std::shared_ptr<Context> & macro_context) const override {
1065
+ void do_render(std::ostringstream &, const std::shared_ptr<Context> & context) const override {
1051
1066
  if (!name) throw std::runtime_error("MacroNode.name is null");
1052
1067
  if (!body) throw std::runtime_error("MacroNode.body is null");
1053
- auto callable = Value::callable([&](const std::shared_ptr<Context> & context, ArgumentsValue & args) {
1054
- auto call_context = macro_context;
1068
+
1069
+ // Use init-capture to avoid dangling 'this' pointer and circular references
1070
+ auto callable = Value::callable([weak_context = std::weak_ptr<Context>(context),
1071
+ name = name, params = params, body = body,
1072
+ named_param_positions = named_param_positions]
1073
+ (const std::shared_ptr<Context> & call_context, ArgumentsValue & args) {
1074
+ auto context_locked = weak_context.lock();
1075
+ if (!context_locked) throw std::runtime_error("Macro context no longer valid");
1076
+ auto execution_context = Context::make(Value::object(), context_locked);
1077
+
1078
+ if (call_context->contains("caller")) {
1079
+ execution_context->set("caller", call_context->get("caller"));
1080
+ }
1081
+
1055
1082
  std::vector<bool> param_set(params.size(), false);
1056
1083
  for (size_t i = 0, n = args.args.size(); i < n; i++) {
1057
1084
  auto & arg = args.args[i];
1058
1085
  if (i >= params.size()) throw std::runtime_error("Too many positional arguments for macro " + name->get_name());
1059
1086
  param_set[i] = true;
1060
- auto & param_name = params[i].first;
1061
- call_context->set(param_name, arg);
1087
+ const auto & param_name = params[i].first;
1088
+ execution_context->set(param_name, arg);
1062
1089
  }
1063
1090
  for (auto & [arg_name, value] : args.kwargs) {
1064
1091
  auto it = named_param_positions.find(arg_name);
1065
1092
  if (it == named_param_positions.end()) throw std::runtime_error("Unknown parameter name for macro " + name->get_name() + ": " + arg_name);
1066
1093
 
1067
- call_context->set(arg_name, value);
1094
+ execution_context->set(arg_name, value);
1068
1095
  param_set[it->second] = true;
1069
1096
  }
1070
1097
  // Set default values for parameters that were not passed
1071
1098
  for (size_t i = 0, n = params.size(); i < n; i++) {
1072
1099
  if (!param_set[i] && params[i].second != nullptr) {
1073
- auto val = params[i].second->evaluate(context);
1074
- call_context->set(params[i].first, val);
1100
+ auto val = params[i].second->evaluate(call_context);
1101
+ execution_context->set(params[i].first, val);
1075
1102
  }
1076
1103
  }
1077
- return body->render(call_context);
1104
+ return body->render(execution_context);
1078
1105
  });
1079
- macro_context->set(name->get_name(), callable);
1106
+ context->set(name->get_name(), callable);
1080
1107
  }
1081
1108
  };
1082
1109
 
@@ -1611,6 +1638,44 @@ public:
1611
1638
  }
1612
1639
  };
1613
1640
 
1641
+ class CallNode : public TemplateNode {
1642
+ std::shared_ptr<Expression> expr;
1643
+ std::shared_ptr<TemplateNode> body;
1644
+
1645
+ public:
1646
+ CallNode(const Location & loc, std::shared_ptr<Expression> && e, std::shared_ptr<TemplateNode> && b)
1647
+ : TemplateNode(loc), expr(std::move(e)), body(std::move(b)) {}
1648
+
1649
+ void do_render(std::ostringstream & out, const std::shared_ptr<Context> & context) const override {
1650
+ if (!expr) throw std::runtime_error("CallNode.expr is null");
1651
+ if (!body) throw std::runtime_error("CallNode.body is null");
1652
+
1653
+ // Use init-capture to avoid dangling 'this' pointer and circular references
1654
+ auto caller = Value::callable([weak_context = std::weak_ptr<Context>(context), body=body]
1655
+ (const std::shared_ptr<Context> &, ArgumentsValue &) -> Value {
1656
+ auto context_locked = weak_context.lock();
1657
+ if (!context_locked) throw std::runtime_error("Caller context no longer valid");
1658
+ return Value(body->render(context_locked));
1659
+ });
1660
+
1661
+ context->set("caller", caller);
1662
+
1663
+ auto call_expr = dynamic_cast<CallExpr*>(expr.get());
1664
+ if (!call_expr) {
1665
+ throw std::runtime_error("Invalid call block syntax - expected function call");
1666
+ }
1667
+
1668
+ Value function = call_expr->object->evaluate(context);
1669
+ if (!function.is_callable()) {
1670
+ throw std::runtime_error("Call target must be callable: " + function.dump());
1671
+ }
1672
+ ArgumentsValue args = call_expr->args.evaluate(context);
1673
+
1674
+ Value result = function.call(context, args);
1675
+ out << result.to_str();
1676
+ }
1677
+ };
1678
+
1614
1679
  class FilterExpr : public Expression {
1615
1680
  std::vector<std::shared_ptr<Expression>> parts;
1616
1681
  public:
@@ -2140,7 +2205,7 @@ private:
2140
2205
 
2141
2206
  auto value = parseValue();
2142
2207
 
2143
- while (it != end && consumeSpaces() && peekSymbols({ "[", "." })) {
2208
+ while (it != end && consumeSpaces() && peekSymbols({ "[", ".", "(" })) {
2144
2209
  if (!consumeToken("[").empty()) {
2145
2210
  std::shared_ptr<Expression> index;
2146
2211
  auto slice_loc = get_location();
@@ -2185,15 +2250,13 @@ private:
2185
2250
  auto key = std::make_shared<LiteralExpr>(identifier->location, Value(identifier->get_name()));
2186
2251
  value = std::make_shared<SubscriptExpr>(identifier->location, std::move(value), std::move(key));
2187
2252
  }
2253
+ } else if (peekSymbols({ "(" })) {
2254
+ auto callParams = parseCallArgs();
2255
+ value = std::make_shared<CallExpr>(get_location(), std::move(value), std::move(callParams));
2188
2256
  }
2189
2257
  consumeSpaces();
2190
2258
  }
2191
2259
 
2192
- if (peekSymbols({ "(" })) {
2193
- auto location = get_location();
2194
- auto callParams = parseCallArgs();
2195
- value = std::make_shared<CallExpr>(location, std::move(value), std::move(callParams));
2196
- }
2197
2260
  return value;
2198
2261
  }
2199
2262
 
@@ -2320,7 +2383,7 @@ private:
2320
2383
  static std::regex comment_tok(R"(\{#([-~]?)([\s\S]*?)([-~]?)#\})");
2321
2384
  static std::regex expr_open_regex(R"(\{\{([-~])?)");
2322
2385
  static std::regex block_open_regex(R"(^\{%([-~])?\s*)");
2323
- static std::regex block_keyword_tok(R"((if|else|elif|endif|for|endfor|generation|endgeneration|set|endset|block|endblock|macro|endmacro|filter|endfilter|break|continue)\b)");
2386
+ static std::regex block_keyword_tok(R"((if|else|elif|endif|for|endfor|generation|endgeneration|set|endset|block|endblock|macro|endmacro|filter|endfilter|break|continue|call|endcall)\b)");
2324
2387
  static std::regex non_text_open_regex(R"(\{\{|\{%|\{#)");
2325
2388
  static std::regex expr_close_regex(R"(\s*([-~])?\}\})");
2326
2389
  static std::regex block_close_regex(R"(\s*([-~])?%\})");
@@ -2443,6 +2506,15 @@ private:
2443
2506
  } else if (keyword == "endmacro") {
2444
2507
  auto post_space = parseBlockClose();
2445
2508
  tokens.push_back(std::make_unique<EndMacroTemplateToken>(location, pre_space, post_space));
2509
+ } else if (keyword == "call") {
2510
+ auto expr = parseExpression();
2511
+ if (!expr) throw std::runtime_error("Expected expression in call block");
2512
+
2513
+ auto post_space = parseBlockClose();
2514
+ tokens.push_back(std::make_unique<CallTemplateToken>(location, pre_space, post_space, std::move(expr)));
2515
+ } else if (keyword == "endcall") {
2516
+ auto post_space = parseBlockClose();
2517
+ tokens.push_back(std::make_unique<EndCallTemplateToken>(location, pre_space, post_space));
2446
2518
  } else if (keyword == "filter") {
2447
2519
  auto filter = parseExpression();
2448
2520
  if (!filter) throw std::runtime_error("Expected expression in filter block");
@@ -2575,6 +2647,12 @@ private:
2575
2647
  throw unterminated(**start);
2576
2648
  }
2577
2649
  children.emplace_back(std::make_shared<MacroNode>(token->location, std::move(macro_token->name), std::move(macro_token->params), std::move(body)));
2650
+ } else if (auto call_token = dynamic_cast<CallTemplateToken*>(token.get())) {
2651
+ auto body = parseTemplate(begin, it, end);
2652
+ if (it == end || (*(it++))->type != TemplateToken::Type::EndCall) {
2653
+ throw unterminated(**start);
2654
+ }
2655
+ children.emplace_back(std::make_shared<CallNode>(token->location, std::move(call_token->expr), std::move(body)));
2578
2656
  } else if (auto filter_token = dynamic_cast<FilterTemplateToken*>(token.get())) {
2579
2657
  auto body = parseTemplate(begin, it, end);
2580
2658
  if (it == end || (*(it++))->type != TemplateToken::Type::EndFilter) {
@@ -2588,6 +2666,7 @@ private:
2588
2666
  } else if (dynamic_cast<EndForTemplateToken*>(token.get())
2589
2667
  || dynamic_cast<EndSetTemplateToken*>(token.get())
2590
2668
  || dynamic_cast<EndMacroTemplateToken*>(token.get())
2669
+ || dynamic_cast<EndCallTemplateToken*>(token.get())
2591
2670
  || dynamic_cast<EndFilterTemplateToken*>(token.get())
2592
2671
  || dynamic_cast<EndIfTemplateToken*>(token.get())
2593
2672
  || dynamic_cast<ElseTemplateToken*>(token.get())
@@ -2657,7 +2736,7 @@ inline std::shared_ptr<Context> Context::builtins() {
2657
2736
  globals.set("raise_exception", simple_function("raise_exception", { "message" }, [](const std::shared_ptr<Context> &, Value & args) -> Value {
2658
2737
  throw std::runtime_error(args.at("message").get<std::string>());
2659
2738
  }));
2660
- globals.set("tojson", simple_function("tojson", { "value", "indent" }, [](const std::shared_ptr<Context> &, Value & args) {
2739
+ globals.set("tojson", simple_function("tojson", { "value", "indent", "ensure_ascii" }, [](const std::shared_ptr<Context> &, Value & args) {
2661
2740
  return Value(args.at("value").dump(args.get<int64_t>("indent", -1), /* to_json= */ true));
2662
2741
  }));
2663
2742
  globals.set("items", simple_function("items", { "object" }, [](const std::shared_ptr<Context> &, Value & args) {