@fugood/llama.node 0.3.2 → 0.3.4

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (286) hide show
  1. package/CMakeLists.txt +7 -0
  2. package/bin/darwin/arm64/llama-node.node +0 -0
  3. package/bin/darwin/x64/llama-node.node +0 -0
  4. package/bin/linux/arm64/llama-node.node +0 -0
  5. package/bin/linux/x64/llama-node.node +0 -0
  6. package/bin/linux-vulkan/arm64/llama-node.node +0 -0
  7. package/bin/linux-vulkan/x64/llama-node.node +0 -0
  8. package/bin/win32/arm64/llama-node.node +0 -0
  9. package/bin/win32/arm64/node.lib +0 -0
  10. package/bin/win32/x64/llama-node.node +0 -0
  11. package/bin/win32/x64/node.lib +0 -0
  12. package/bin/win32-vulkan/arm64/llama-node.node +0 -0
  13. package/bin/win32-vulkan/arm64/node.lib +0 -0
  14. package/bin/win32-vulkan/x64/llama-node.node +0 -0
  15. package/bin/win32-vulkan/x64/node.lib +0 -0
  16. package/lib/binding.ts +18 -1
  17. package/package.json +1 -1
  18. package/src/DetokenizeWorker.cpp +1 -1
  19. package/src/EmbeddingWorker.cpp +17 -7
  20. package/src/EmbeddingWorker.h +2 -1
  21. package/src/LlamaCompletionWorker.cpp +8 -8
  22. package/src/LlamaCompletionWorker.h +2 -2
  23. package/src/LlamaContext.cpp +89 -27
  24. package/src/LlamaContext.h +2 -0
  25. package/src/TokenizeWorker.cpp +1 -1
  26. package/src/common.hpp +4 -4
  27. package/src/llama.cpp/.github/workflows/build.yml +240 -168
  28. package/src/llama.cpp/.github/workflows/docker.yml +8 -8
  29. package/src/llama.cpp/.github/workflows/python-lint.yml +8 -1
  30. package/src/llama.cpp/.github/workflows/server.yml +21 -14
  31. package/src/llama.cpp/CMakeLists.txt +14 -6
  32. package/src/llama.cpp/Sources/llama/llama.h +4 -0
  33. package/src/llama.cpp/cmake/arm64-apple-clang.cmake +16 -0
  34. package/src/llama.cpp/cmake/common.cmake +33 -0
  35. package/src/llama.cpp/cmake/x64-windows-llvm.cmake +11 -0
  36. package/src/llama.cpp/common/CMakeLists.txt +6 -4
  37. package/src/llama.cpp/common/arg.cpp +986 -770
  38. package/src/llama.cpp/common/arg.h +22 -22
  39. package/src/llama.cpp/common/common.cpp +212 -351
  40. package/src/llama.cpp/common/common.h +204 -117
  41. package/src/llama.cpp/common/json-schema-to-grammar.cpp +1 -1
  42. package/src/llama.cpp/common/log.cpp +50 -50
  43. package/src/llama.cpp/common/log.h +18 -18
  44. package/src/llama.cpp/common/ngram-cache.cpp +36 -36
  45. package/src/llama.cpp/common/ngram-cache.h +19 -19
  46. package/src/llama.cpp/common/sampling.cpp +163 -121
  47. package/src/llama.cpp/common/sampling.h +41 -20
  48. package/src/llama.cpp/common/speculative.cpp +274 -0
  49. package/src/llama.cpp/common/speculative.h +28 -0
  50. package/src/llama.cpp/docs/build.md +134 -161
  51. package/src/llama.cpp/examples/CMakeLists.txt +33 -14
  52. package/src/llama.cpp/examples/batched/CMakeLists.txt +1 -1
  53. package/src/llama.cpp/examples/batched/batched.cpp +19 -18
  54. package/src/llama.cpp/examples/batched-bench/CMakeLists.txt +1 -1
  55. package/src/llama.cpp/examples/batched-bench/batched-bench.cpp +10 -11
  56. package/src/llama.cpp/examples/convert-llama2c-to-ggml/CMakeLists.txt +1 -1
  57. package/src/llama.cpp/examples/convert-llama2c-to-ggml/convert-llama2c-to-ggml.cpp +1 -1
  58. package/src/llama.cpp/examples/cvector-generator/CMakeLists.txt +1 -1
  59. package/src/llama.cpp/examples/cvector-generator/cvector-generator.cpp +9 -9
  60. package/src/llama.cpp/examples/deprecation-warning/deprecation-warning.cpp +1 -1
  61. package/src/llama.cpp/examples/embedding/CMakeLists.txt +1 -1
  62. package/src/llama.cpp/examples/embedding/embedding.cpp +12 -12
  63. package/src/llama.cpp/examples/eval-callback/CMakeLists.txt +3 -2
  64. package/src/llama.cpp/examples/eval-callback/eval-callback.cpp +8 -8
  65. package/src/llama.cpp/examples/export-lora/CMakeLists.txt +1 -1
  66. package/src/llama.cpp/examples/export-lora/export-lora.cpp +5 -5
  67. package/src/llama.cpp/examples/gbnf-validator/CMakeLists.txt +1 -1
  68. package/src/llama.cpp/examples/gbnf-validator/gbnf-validator.cpp +4 -7
  69. package/src/llama.cpp/examples/gen-docs/CMakeLists.txt +1 -1
  70. package/src/llama.cpp/examples/gen-docs/gen-docs.cpp +7 -7
  71. package/src/llama.cpp/examples/gguf/CMakeLists.txt +1 -1
  72. package/src/llama.cpp/examples/gguf-hash/CMakeLists.txt +8 -1
  73. package/src/llama.cpp/examples/gguf-split/CMakeLists.txt +1 -1
  74. package/src/llama.cpp/examples/gguf-split/gguf-split.cpp +2 -2
  75. package/src/llama.cpp/examples/gritlm/CMakeLists.txt +1 -1
  76. package/src/llama.cpp/examples/gritlm/gritlm.cpp +18 -18
  77. package/src/llama.cpp/examples/imatrix/CMakeLists.txt +1 -1
  78. package/src/llama.cpp/examples/imatrix/imatrix.cpp +31 -13
  79. package/src/llama.cpp/examples/infill/CMakeLists.txt +1 -1
  80. package/src/llama.cpp/examples/infill/infill.cpp +41 -87
  81. package/src/llama.cpp/examples/llama-bench/CMakeLists.txt +1 -1
  82. package/src/llama.cpp/examples/llama-bench/llama-bench.cpp +439 -459
  83. package/src/llama.cpp/examples/llama.android/llama/build.gradle.kts +2 -0
  84. package/src/llama.cpp/examples/llama.android/llama/src/main/cpp/llama-android.cpp +11 -14
  85. package/src/llama.cpp/examples/llava/CMakeLists.txt +10 -3
  86. package/src/llama.cpp/examples/llava/clip.cpp +263 -66
  87. package/src/llama.cpp/examples/llava/clip.h +8 -2
  88. package/src/llama.cpp/examples/llava/llava-cli.cpp +23 -23
  89. package/src/llama.cpp/examples/llava/llava.cpp +83 -22
  90. package/src/llama.cpp/examples/llava/minicpmv-cli.cpp +21 -21
  91. package/src/llama.cpp/examples/llava/qwen2vl-cli.cpp +581 -0
  92. package/src/llama.cpp/examples/lookahead/CMakeLists.txt +1 -1
  93. package/src/llama.cpp/examples/lookahead/lookahead.cpp +26 -26
  94. package/src/llama.cpp/examples/lookup/CMakeLists.txt +4 -4
  95. package/src/llama.cpp/examples/lookup/lookup-create.cpp +7 -7
  96. package/src/llama.cpp/examples/lookup/lookup-merge.cpp +4 -4
  97. package/src/llama.cpp/examples/lookup/lookup-stats.cpp +16 -15
  98. package/src/llama.cpp/examples/lookup/lookup.cpp +30 -30
  99. package/src/llama.cpp/examples/main/CMakeLists.txt +1 -1
  100. package/src/llama.cpp/examples/main/main.cpp +73 -114
  101. package/src/llama.cpp/examples/main-cmake-pkg/CMakeLists.txt +1 -1
  102. package/src/llama.cpp/examples/parallel/CMakeLists.txt +1 -1
  103. package/src/llama.cpp/examples/parallel/parallel.cpp +18 -19
  104. package/src/llama.cpp/examples/passkey/CMakeLists.txt +1 -1
  105. package/src/llama.cpp/examples/passkey/passkey.cpp +14 -14
  106. package/src/llama.cpp/examples/perplexity/CMakeLists.txt +1 -1
  107. package/src/llama.cpp/examples/perplexity/perplexity.cpp +99 -120
  108. package/src/llama.cpp/examples/quantize/CMakeLists.txt +1 -1
  109. package/src/llama.cpp/examples/quantize/quantize.cpp +0 -3
  110. package/src/llama.cpp/examples/quantize-stats/CMakeLists.txt +1 -1
  111. package/src/llama.cpp/examples/quantize-stats/quantize-stats.cpp +10 -9
  112. package/src/llama.cpp/examples/retrieval/CMakeLists.txt +1 -1
  113. package/src/llama.cpp/examples/retrieval/retrieval.cpp +16 -16
  114. package/src/llama.cpp/examples/rpc/rpc-server.cpp +3 -1
  115. package/src/llama.cpp/examples/run/CMakeLists.txt +5 -0
  116. package/src/llama.cpp/examples/run/run.cpp +911 -0
  117. package/src/llama.cpp/examples/save-load-state/CMakeLists.txt +1 -1
  118. package/src/llama.cpp/examples/save-load-state/save-load-state.cpp +38 -21
  119. package/src/llama.cpp/examples/server/CMakeLists.txt +3 -16
  120. package/src/llama.cpp/examples/server/server.cpp +2073 -1339
  121. package/src/llama.cpp/examples/server/tests/requirements.txt +2 -2
  122. package/src/llama.cpp/examples/server/utils.hpp +354 -277
  123. package/src/llama.cpp/examples/simple/CMakeLists.txt +2 -2
  124. package/src/llama.cpp/examples/simple/simple.cpp +130 -94
  125. package/src/llama.cpp/examples/simple-chat/CMakeLists.txt +5 -0
  126. package/src/llama.cpp/examples/simple-chat/simple-chat.cpp +200 -0
  127. package/src/llama.cpp/examples/speculative/CMakeLists.txt +1 -1
  128. package/src/llama.cpp/examples/speculative/speculative.cpp +68 -64
  129. package/src/llama.cpp/examples/speculative-simple/CMakeLists.txt +5 -0
  130. package/src/llama.cpp/examples/speculative-simple/speculative-simple.cpp +265 -0
  131. package/src/llama.cpp/examples/tokenize/CMakeLists.txt +1 -1
  132. package/src/llama.cpp/examples/tokenize/tokenize.cpp +3 -3
  133. package/src/llama.cpp/examples/tts/CMakeLists.txt +5 -0
  134. package/src/llama.cpp/examples/tts/tts.cpp +932 -0
  135. package/src/llama.cpp/ggml/CMakeLists.txt +54 -36
  136. package/src/llama.cpp/ggml/include/ggml-backend.h +63 -34
  137. package/src/llama.cpp/ggml/include/ggml-blas.h +5 -3
  138. package/src/llama.cpp/ggml/include/ggml-cann.h +9 -7
  139. package/src/llama.cpp/ggml/include/ggml-cpp.h +38 -0
  140. package/src/llama.cpp/ggml/include/ggml-cpu.h +135 -0
  141. package/src/llama.cpp/ggml/include/ggml-cuda.h +12 -12
  142. package/src/llama.cpp/ggml/include/ggml-kompute.h +7 -3
  143. package/src/llama.cpp/ggml/include/ggml-metal.h +11 -7
  144. package/src/llama.cpp/ggml/include/ggml-opencl.h +26 -0
  145. package/src/llama.cpp/ggml/include/ggml-opt.h +216 -0
  146. package/src/llama.cpp/ggml/include/ggml-rpc.h +9 -5
  147. package/src/llama.cpp/ggml/include/ggml-sycl.h +18 -11
  148. package/src/llama.cpp/ggml/include/ggml-vulkan.h +10 -8
  149. package/src/llama.cpp/ggml/include/ggml.h +159 -417
  150. package/src/llama.cpp/ggml/src/CMakeLists.txt +121 -1155
  151. package/src/llama.cpp/ggml/src/ggml-alloc.c +23 -28
  152. package/src/llama.cpp/ggml/src/ggml-backend-impl.h +57 -36
  153. package/src/llama.cpp/ggml/src/ggml-backend-reg.cpp +552 -0
  154. package/src/llama.cpp/ggml/src/ggml-backend.cpp +306 -867
  155. package/src/llama.cpp/ggml/src/ggml-blas/CMakeLists.txt +87 -0
  156. package/src/llama.cpp/ggml/src/{ggml-blas.cpp → ggml-blas/ggml-blas.cpp} +216 -65
  157. package/src/llama.cpp/ggml/src/ggml-cann/CMakeLists.txt +76 -0
  158. package/src/llama.cpp/ggml/src/ggml-cann/aclnn_ops.cpp +456 -111
  159. package/src/llama.cpp/ggml/src/ggml-cann/common.h +6 -3
  160. package/src/llama.cpp/ggml/src/{ggml-cann.cpp → ggml-cann/ggml-cann.cpp} +343 -177
  161. package/src/llama.cpp/ggml/src/ggml-cann/kernels/CMakeLists.txt +2 -5
  162. package/src/llama.cpp/ggml/src/ggml-cann/kernels/dup.cpp +22 -9
  163. package/src/llama.cpp/ggml/src/ggml-cann/kernels/get_row_f16.cpp +24 -13
  164. package/src/llama.cpp/ggml/src/ggml-cann/kernels/get_row_f32.cpp +23 -13
  165. package/src/llama.cpp/ggml/src/ggml-cann/kernels/get_row_q4_0.cpp +11 -0
  166. package/src/llama.cpp/ggml/src/ggml-cann/kernels/quantize_f16_q8_0.cpp +10 -0
  167. package/src/llama.cpp/ggml/src/ggml-cann/kernels/quantize_f32_q8_0.cpp +10 -0
  168. package/src/llama.cpp/ggml/src/ggml-cann/kernels/quantize_float_to_q4_0.cpp +17 -0
  169. package/src/llama.cpp/ggml/src/ggml-common.h +42 -42
  170. package/src/llama.cpp/ggml/src/ggml-cpu/CMakeLists.txt +336 -0
  171. package/src/llama.cpp/ggml/src/ggml-cpu/amx/amx.cpp +220 -0
  172. package/src/llama.cpp/ggml/src/ggml-cpu/amx/amx.h +8 -0
  173. package/src/llama.cpp/ggml/src/ggml-cpu/amx/common.h +91 -0
  174. package/src/llama.cpp/ggml/src/ggml-cpu/amx/mmq.cpp +2511 -0
  175. package/src/llama.cpp/ggml/src/ggml-cpu/amx/mmq.h +10 -0
  176. package/src/llama.cpp/ggml/src/ggml-cpu/cpu-feats-x86.cpp +323 -0
  177. package/src/llama.cpp/ggml/src/{ggml-aarch64.c → ggml-cpu/ggml-cpu-aarch64.cpp} +1299 -246
  178. package/src/llama.cpp/ggml/src/ggml-cpu/ggml-cpu-aarch64.h +8 -0
  179. package/src/llama.cpp/ggml/src/ggml-cpu/ggml-cpu-hbm.cpp +55 -0
  180. package/src/llama.cpp/ggml/src/ggml-cpu/ggml-cpu-hbm.h +8 -0
  181. package/src/llama.cpp/ggml/src/{ggml-cpu-impl.h → ggml-cpu/ggml-cpu-impl.h} +14 -242
  182. package/src/llama.cpp/ggml/src/ggml-cpu/ggml-cpu-quants.c +10835 -0
  183. package/src/llama.cpp/ggml/src/ggml-cpu/ggml-cpu-quants.h +63 -0
  184. package/src/llama.cpp/ggml/src/ggml-cpu/ggml-cpu-traits.cpp +36 -0
  185. package/src/llama.cpp/ggml/src/ggml-cpu/ggml-cpu-traits.h +38 -0
  186. package/src/llama.cpp/ggml/src/ggml-cpu/ggml-cpu.c +14123 -0
  187. package/src/llama.cpp/ggml/src/ggml-cpu/ggml-cpu.cpp +628 -0
  188. package/src/llama.cpp/ggml/src/{llamafile → ggml-cpu/llamafile}/sgemm.cpp +666 -0
  189. package/src/llama.cpp/ggml/src/ggml-cuda/CMakeLists.txt +152 -0
  190. package/src/llama.cpp/ggml/src/ggml-cuda/vendors/hip.h +8 -0
  191. package/src/llama.cpp/ggml/src/ggml-hip/CMakeLists.txt +104 -0
  192. package/src/llama.cpp/ggml/src/ggml-impl.h +393 -22
  193. package/src/llama.cpp/ggml/src/ggml-kompute/CMakeLists.txt +166 -0
  194. package/src/llama.cpp/ggml/src/{ggml-kompute.cpp → ggml-kompute/ggml-kompute.cpp} +360 -127
  195. package/src/llama.cpp/ggml/src/ggml-metal/CMakeLists.txt +105 -0
  196. package/src/llama.cpp/ggml/src/ggml-metal/ggml-metal-impl.h +288 -0
  197. package/src/llama.cpp/ggml/src/ggml-musa/CMakeLists.txt +107 -0
  198. package/src/llama.cpp/ggml/src/ggml-opencl/CMakeLists.txt +147 -0
  199. package/src/llama.cpp/ggml/src/ggml-opencl/ggml-opencl.cpp +4004 -0
  200. package/src/llama.cpp/ggml/src/ggml-opt.cpp +854 -0
  201. package/src/llama.cpp/ggml/src/ggml-quants.c +188 -10702
  202. package/src/llama.cpp/ggml/src/ggml-quants.h +78 -125
  203. package/src/llama.cpp/ggml/src/ggml-rpc/CMakeLists.txt +9 -0
  204. package/src/llama.cpp/ggml/src/{ggml-rpc.cpp → ggml-rpc/ggml-rpc.cpp} +478 -300
  205. package/src/llama.cpp/ggml/src/ggml-sycl/CMakeLists.txt +84 -0
  206. package/src/llama.cpp/ggml/src/ggml-sycl/backend.hpp +3 -0
  207. package/src/llama.cpp/ggml/src/ggml-sycl/common.cpp +36 -5
  208. package/src/llama.cpp/ggml/src/ggml-sycl/common.hpp +259 -0
  209. package/src/llama.cpp/ggml/src/ggml-sycl/concat.cpp +3 -2
  210. package/src/llama.cpp/ggml/src/ggml-sycl/convert.cpp +1 -1
  211. package/src/llama.cpp/ggml/src/ggml-sycl/dmmv.cpp +5 -5
  212. package/src/llama.cpp/ggml/src/ggml-sycl/dpct/helper.hpp +34 -35
  213. package/src/llama.cpp/ggml/src/ggml-sycl/element_wise.cpp +1030 -0
  214. package/src/llama.cpp/ggml/src/ggml-sycl/element_wise.hpp +76 -0
  215. package/src/llama.cpp/ggml/src/ggml-sycl/gemm.hpp +4 -4
  216. package/src/llama.cpp/ggml/src/{ggml-sycl.cpp → ggml-sycl/ggml-sycl.cpp} +3638 -4151
  217. package/src/llama.cpp/ggml/src/ggml-sycl/im2col.cpp +3 -2
  218. package/src/llama.cpp/ggml/src/ggml-sycl/mmq.cpp +6 -6
  219. package/src/llama.cpp/ggml/src/ggml-sycl/mmvq.cpp +75 -87
  220. package/src/llama.cpp/ggml/src/ggml-sycl/norm.cpp +7 -6
  221. package/src/llama.cpp/ggml/src/ggml-sycl/outprod.cpp +56 -0
  222. package/src/llama.cpp/ggml/src/ggml-sycl/outprod.hpp +11 -0
  223. package/src/llama.cpp/ggml/src/ggml-sycl/presets.hpp +6 -0
  224. package/src/llama.cpp/ggml/src/ggml-sycl/rope.cpp +4 -3
  225. package/src/llama.cpp/ggml/src/ggml-sycl/softmax.cpp +7 -7
  226. package/src/llama.cpp/ggml/src/ggml-sycl/tsembd.cpp +1 -0
  227. package/src/llama.cpp/ggml/src/ggml-sycl/vecdotq.hpp +4 -4
  228. package/src/llama.cpp/ggml/src/ggml-sycl/wkv6.cpp +141 -0
  229. package/src/llama.cpp/ggml/src/ggml-sycl/wkv6.hpp +10 -0
  230. package/src/llama.cpp/ggml/src/ggml-threading.cpp +12 -0
  231. package/src/llama.cpp/ggml/src/ggml-threading.h +14 -0
  232. package/src/llama.cpp/ggml/src/ggml-vulkan/CMakeLists.txt +92 -0
  233. package/src/llama.cpp/ggml/src/{ggml-vulkan.cpp → ggml-vulkan/ggml-vulkan.cpp} +2138 -887
  234. package/src/llama.cpp/ggml/src/{vulkan-shaders → ggml-vulkan/vulkan-shaders}/CMakeLists.txt +3 -1
  235. package/src/llama.cpp/ggml/src/ggml-vulkan/vulkan-shaders/vulkan-shaders-gen.cpp +593 -0
  236. package/src/llama.cpp/ggml/src/ggml.c +4427 -20125
  237. package/src/llama.cpp/include/llama-cpp.h +25 -0
  238. package/src/llama.cpp/include/llama.h +93 -52
  239. package/src/llama.cpp/models/ggml-vocab-roberta-bpe.gguf.inp +112 -0
  240. package/src/llama.cpp/models/ggml-vocab-roberta-bpe.gguf.out +46 -0
  241. package/src/llama.cpp/pocs/CMakeLists.txt +3 -1
  242. package/src/llama.cpp/pocs/vdot/CMakeLists.txt +2 -2
  243. package/src/llama.cpp/pocs/vdot/q8dot.cpp +4 -3
  244. package/src/llama.cpp/pocs/vdot/vdot.cpp +8 -7
  245. package/src/llama.cpp/src/CMakeLists.txt +4 -8
  246. package/src/llama.cpp/src/llama-grammar.cpp +15 -15
  247. package/src/llama.cpp/src/llama-grammar.h +2 -5
  248. package/src/llama.cpp/src/llama-sampling.cpp +779 -194
  249. package/src/llama.cpp/src/llama-sampling.h +21 -2
  250. package/src/llama.cpp/src/llama-vocab.cpp +55 -10
  251. package/src/llama.cpp/src/llama-vocab.h +35 -11
  252. package/src/llama.cpp/src/llama.cpp +4317 -2979
  253. package/src/llama.cpp/src/unicode-data.cpp +2 -2
  254. package/src/llama.cpp/src/unicode.cpp +62 -51
  255. package/src/llama.cpp/src/unicode.h +9 -10
  256. package/src/llama.cpp/tests/CMakeLists.txt +48 -38
  257. package/src/llama.cpp/tests/test-arg-parser.cpp +15 -15
  258. package/src/llama.cpp/tests/test-backend-ops.cpp +324 -80
  259. package/src/llama.cpp/tests/test-barrier.cpp +1 -0
  260. package/src/llama.cpp/tests/test-chat-template.cpp +59 -9
  261. package/src/llama.cpp/tests/test-gguf.cpp +1303 -0
  262. package/src/llama.cpp/tests/test-grammar-integration.cpp +3 -6
  263. package/src/llama.cpp/tests/test-json-schema-to-grammar.cpp +17 -4
  264. package/src/llama.cpp/tests/test-llama-grammar.cpp +2 -4
  265. package/src/llama.cpp/tests/test-log.cpp +2 -2
  266. package/src/llama.cpp/tests/test-opt.cpp +853 -142
  267. package/src/llama.cpp/tests/test-quantize-fns.cpp +24 -21
  268. package/src/llama.cpp/tests/test-quantize-perf.cpp +16 -14
  269. package/src/llama.cpp/tests/test-rope.cpp +62 -20
  270. package/src/llama.cpp/tests/test-sampling.cpp +163 -138
  271. package/src/llama.cpp/tests/test-tokenizer-0.cpp +7 -7
  272. package/src/llama.cpp/tests/test-tokenizer-1-bpe.cpp +5 -5
  273. package/src/llama.cpp/tests/test-tokenizer-1-spm.cpp +5 -5
  274. package/src/llama.cpp/.github/workflows/nix-ci-aarch64.yml +0 -72
  275. package/src/llama.cpp/.github/workflows/nix-ci.yml +0 -79
  276. package/src/llama.cpp/.github/workflows/nix-flake-update.yml +0 -22
  277. package/src/llama.cpp/.github/workflows/nix-publish-flake.yml +0 -36
  278. package/src/llama.cpp/common/train.cpp +0 -1515
  279. package/src/llama.cpp/common/train.h +0 -233
  280. package/src/llama.cpp/examples/baby-llama/CMakeLists.txt +0 -5
  281. package/src/llama.cpp/examples/baby-llama/baby-llama.cpp +0 -1639
  282. package/src/llama.cpp/ggml/src/ggml-aarch64.h +0 -39
  283. package/src/llama.cpp/ggml/src/vulkan-shaders/vulkan-shaders-gen.cpp +0 -600
  284. package/src/llama.cpp/tests/test-grad0.cpp +0 -1683
  285. /package/src/llama.cpp/ggml/{cmake → src/ggml-cpu/cmake}/FindSIMD.cmake +0 -0
  286. /package/src/llama.cpp/ggml/src/{llamafile → ggml-cpu/llamafile}/sgemm.h +0 -0
@@ -1,4 +1,5 @@
1
1
  #include "ggml.h"
2
+ #include "ggml-cpu.h"
2
3
  #include "ggml-backend.h"
3
4
 
4
5
  #include <chrono>
@@ -21,7 +21,7 @@ int main(void) {
21
21
  std::vector<std::string> templates = {
22
22
  // teknium/OpenHermes-2.5-Mistral-7B
23
23
  "{% for message in messages %}{{'<|im_start|>' + message['role'] + '\\n' + message['content'] + '<|im_end|>' + '\\n'}}{% endfor %}{% if add_generation_prompt %}{{ '<|im_start|>assistant\\n' }}{% endif %}",
24
- // mistralai/Mistral-7B-Instruct-v0.2
24
+ // mistralai/Mistral-7B-Instruct-v0.2 (NOTE: Old pre-v1 without a system prompt)
25
25
  "{{ bos_token }}{% for message in messages %}{% if (message['role'] == 'user') != (loop.index0 % 2 == 0) %}{{ raise_exception('Conversation roles must alternate user/assistant/user/assistant/...') }}{% endif %}{% if message['role'] == 'user' %}{{ '[INST] ' + message['content'] + ' [/INST]' }}{% elif message['role'] == 'assistant' %}{{ message['content'] + eos_token}}{% else %}{{ raise_exception('Only user and assistant roles are supported!') }}{% endif %}{% endfor %}",
26
26
  // TheBloke/FusionNet_34Bx2_MoE-AWQ
27
27
  "{%- for idx in range(0, messages|length) -%}\\n{%- if messages[idx]['role'] == 'user' -%}\\n{%- if idx > 1 -%}\\n{{- bos_token + '[INST] ' + messages[idx]['content'] + ' [/INST]' -}}\\n{%- else -%}\\n{{- messages[idx]['content'] + ' [/INST]' -}}\\n{%- endif -%}\\n{% elif messages[idx]['role'] == 'system' %}\\n{{- '[INST] <<SYS>>\\\\n' + messages[idx]['content'] + '\\\\n<</SYS>>\\\\n\\\\n' -}}\\n{%- elif messages[idx]['role'] == 'assistant' -%}\\n{{- ' ' + messages[idx]['content'] + ' ' + eos_token -}}\\n{% endif %}\\n{% endfor %}",
@@ -65,16 +65,28 @@ int main(void) {
65
65
  u8"{% for message in messages %}{% if message['role'] == 'user' %}{{'<用户>' + message['content'].strip() + '<AI>'}}{% else %}{{message['content'].strip()}}{% endif %}{% endfor %}",
66
66
  // DeepSeek-V2
67
67
  "{% if not add_generation_prompt is defined %}{% set add_generation_prompt = false %}{% endif %}{{ bos_token }}{% for message in messages %}{% if message['role'] == 'user' %}{{ 'User: ' + message['content'] + '\n\n' }}{% elif message['role'] == 'assistant' %}{{ 'Assistant: ' + message['content'] + eos_token }}{% elif message['role'] == 'system' %}{{ message['content'] + '\n\n' }}{% endif %}{% endfor %}{% if add_generation_prompt %}{{ 'Assistant:' }}{% endif %}",
68
+ // ibm-granite/granite-3.0-8b-instruct
69
+ "{%- if tools %}\n {{- '<|start_of_role|>available_tools<|end_of_role|>\n' }}\n {%- for tool in tools %}\n {{- tool | tojson(indent=4) }}\n {%- if not loop.last %}\n {{- '\n\n' }}\n {%- endif %}\n {%- endfor %}\n {{- '<|end_of_text|>\n' }}\n{%- endif %}\n{%- for message in messages %}\n {%- if message['role'] == 'system' %}\n {{- '<|start_of_role|>system<|end_of_role|>' + message['content'] + '<|end_of_text|>\n' }}\n {%- elif message['role'] == 'user' %}\n {{- '<|start_of_role|>user<|end_of_role|>' + message['content'] + '<|end_of_text|>\n' }}\n {%- elif message['role'] == 'assistant' %}\n {{- '<|start_of_role|>assistant<|end_of_role|>' + message['content'] + '<|end_of_text|>\n' }}\n {%- elif message['role'] == 'assistant_tool_call' %}\n {{- '<|start_of_role|>assistant<|end_of_role|><|tool_call|>' + message['content'] + '<|end_of_text|>\n' }}\n {%- elif message['role'] == 'tool_response' %}\n {{- '<|start_of_role|>tool_response<|end_of_role|>' + message['content'] + '<|end_of_text|>\n' }}\n {%- endif %}\n {%- if loop.last and add_generation_prompt %}\n {{- '<|start_of_role|>assistant<|end_of_role|>' }}\n {%- endif %}\n{%- endfor %}",
70
+ // mistralai/Mistral-7B-Instruct-v0.2 (mistralai 'v1' template with a system prompt)
71
+ "{%- if messages[0]['role'] == 'system' %}\n {%- set system_message = messages[0]['content'] %}\n {%- set loop_messages = messages[1:] %}\n{%- else %}\n {%- set loop_messages = messages %}\n{%- endif %}\n\n{{- bos_token }}\n{%- for message in loop_messages %}\n {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) %}\n {{- raise_exception('After the optional system message, conversation roles must alternate user/assistant/user/assistant/...') }}\n {%- endif %}\n {%- if message['role'] == 'user' %}\n {%- if loop.first and system_message is defined %}\n {{- ' [INST] ' + system_message + '\\n\\n' + message['content'] + ' [/INST]' }}\n {%- else %}\n {{- ' [INST] ' + message['content'] + ' [/INST]' }}\n {%- endif %}\n {%- elif message['role'] == 'assistant' %}\n {{- ' ' + message['content'] + eos_token}}\n {%- else %}\n {{- raise_exception('Only user and assistant roles are supported, with the exception of an initial optional system message!') }}\n {%- endif %}\n{%- endfor %}\n",
72
+ // Mistral-Large-Instruct-2407 (mistralai 'v3' template)
73
+ "{%- if messages[0][\"role\"] == \"system\" %}\n {%- set system_message = messages[0][\"content\"] %}\n {%- set loop_messages = messages[1:] %}\n{%- else %}\n {%- set loop_messages = messages %}\n{%- endif %}\n{%- if not tools is defined %}\n {%- set tools = none %}\n{%- endif %}\n{%- set user_messages = loop_messages | selectattr(\"role\", \"equalto\", \"user\") | list %}\n\n{#- This block checks for alternating user/assistant messages, skipping tool calling messages #}\n{%- set ns = namespace() %}\n{%- set ns.index = 0 %}\n{%- for message in loop_messages %}\n {%- if not (message.role == \"tool\" or message.role == \"tool_results\" or (message.tool_calls is defined and message.tool_calls is not none)) %}\n {%- if (message[\"role\"] == \"user\") != (ns.index % 2 == 0) %}\n {{- raise_exception(\"After the optional system message, conversation roles must alternate user/assistant/user/assistant/...\") }}\n {%- endif %}\n {%- set ns.index = ns.index + 1 %}\n {%- endif %}\n{%- endfor %}\n\n{{- bos_token }}\n{%- for message in loop_messages %}\n {%- if message[\"role\"] == \"user\" %}\n {%- if tools is not none and (message == user_messages[-1]) %}\n {{- \"[AVAILABLE_TOOLS] [\" }}\n {%- for tool in tools %}\n {%- set tool = tool.function %}\n {{- '{\"type\": \"function\", \"function\": {' }}\n {%- for key, val in tool.items() if key != \"return\" %}\n {%- if val is string %}\n {{- '\"' + key + '\": \"' + val + '\"' }}\n {%- else %}\n {{- '\"' + key + '\": ' + val|tojson }}\n {%- endif %}\n {%- if not loop.last %}\n {{- \", \" }}\n {%- endif %}\n {%- endfor %}\n {{- \"}}\" }}\n {%- if not loop.last %}\n {{- \", \" }}\n {%- else %}\n {{- \"]\" }}\n {%- endif %}\n {%- endfor %}\n {{- \"[/AVAILABLE_TOOLS]\" }}\n {%- endif %}\n {%- if loop.last and system_message is defined %}\n {{- \"[INST] \" + system_message + \"\\n\\n\" + message[\"content\"] + \"[/INST]\" }}\n {%- else %}\n {{- \"[INST] \" + message[\"content\"] + \"[/INST]\" }}\n {%- endif %}\n {%- elif message.tool_calls is defined and message.tool_calls is not none %}\n {{- \"[TOOL_CALLS] [\" }}\n {%- for tool_call in message.tool_calls %}\n {%- set out = tool_call.function|tojson %}\n {{- out[:-1] }}\n {%- if not tool_call.id is defined or tool_call.id|length != 9 %}\n {{- raise_exception(\"Tool call IDs should be alphanumeric strings with length 9!\") }}\n {%- endif %}\n {{- ', \"id\": \"' + tool_call.id + '\"}' }}\n {%- if not loop.last %}\n {{- \", \" }}\n {%- else %}\n {{- \"]\" + eos_token }}\n {%- endif %}\n {%- endfor %}\n {%- elif message[\"role\"] == \"assistant\" %}\n {{- \" \" + message[\"content\"]|trim + eos_token}}\n {%- elif message[\"role\"] == \"tool_results\" or message[\"role\"] == \"tool\" %}\n {%- if message.content is defined and message.content.content is defined %}\n {%- set content = message.content.content %}\n {%- else %}\n {%- set content = message.content %}\n {%- endif %}\n {{- '[TOOL_RESULTS] {\"content\": ' + content|string + \", \" }}\n {%- if not message.tool_call_id is defined or message.tool_call_id|length != 9 %}\n {{- raise_exception(\"Tool call IDs should be alphanumeric strings with length 9!\") }}\n {%- endif %}\n {{- '\"call_id\": \"' + message.tool_call_id + '\"}[/TOOL_RESULTS]' }}\n {%- else %}\n {{- raise_exception(\"Only user and assistant roles are supported, with the exception of an initial optional system message!\") }}\n {%- endif %}\n{%- endfor %}\n",
74
+ // Mistral-Nemo-Instruct-2407 (mistralai 'v3-tekken' template)
75
+ "{%- if messages[0][\"role\"] == \"system\" %}\n {%- set system_message = messages[0][\"content\"] %}\n {%- set loop_messages = messages[1:] %}\n{%- else %}\n {%- set loop_messages = messages %}\n{%- endif %}\n{%- if not tools is defined %}\n {%- set tools = none %}\n{%- endif %}\n{%- set user_messages = loop_messages | selectattr(\"role\", \"equalto\", \"user\") | list %}\n\n{#- This block checks for alternating user/assistant messages, skipping tool calling messages #}\n{%- set ns = namespace() %}\n{%- set ns.index = 0 %}\n{%- for message in loop_messages %}\n {%- if not (message.role == \"tool\" or message.role == \"tool_results\" or (message.tool_calls is defined and message.tool_calls is not none)) %}\n {%- if (message[\"role\"] == \"user\") != (ns.index % 2 == 0) %}\n {{- raise_exception(\"After the optional system message, conversation roles must alternate user/assistant/user/assistant/...\") }}\n {%- endif %}\n {%- set ns.index = ns.index + 1 %}\n {%- endif %}\n{%- endfor %}\n\n{{- bos_token }}\n{%- for message in loop_messages %}\n {%- if message[\"role\"] == \"user\" %}\n {%- if tools is not none and (message == user_messages[-1]) %}\n {{- \"[AVAILABLE_TOOLS][\" }}\n {%- for tool in tools %}\n {%- set tool = tool.function %}\n {{- '{\"type\": \"function\", \"function\": {' }}\n {%- for key, val in tool.items() if key != \"return\" %}\n {%- if val is string %}\n {{- '\"' + key + '\": \"' + val + '\"' }}\n {%- else %}\n {{- '\"' + key + '\": ' + val|tojson }}\n {%- endif %}\n {%- if not loop.last %}\n {{- \", \" }}\n {%- endif %}\n {%- endfor %}\n {{- \"}}\" }}\n {%- if not loop.last %}\n {{- \", \" }}\n {%- else %}\n {{- \"]\" }}\n {%- endif %}\n {%- endfor %}\n {{- \"[/AVAILABLE_TOOLS]\" }}\n {%- endif %}\n {%- if loop.last and system_message is defined %}\n {{- \"[INST]\" + system_message + \"\\n\\n\" + message[\"content\"] + \"[/INST]\" }}\n {%- else %}\n {{- \"[INST]\" + message[\"content\"] + \"[/INST]\" }}\n {%- endif %}\n {%- elif (message.tool_calls is defined and message.tool_calls is not none) %}\n {{- \"[TOOL_CALLS][\" }}\n {%- for tool_call in message.tool_calls %}\n {%- set out = tool_call.function|tojson %}\n {{- out[:-1] }}\n {%- if not tool_call.id is defined or tool_call.id|length != 9 %}\n {{- raise_exception(\"Tool call IDs should be alphanumeric strings with length 9!\") }}\n {%- endif %}\n {{- ', \"id\": \"' + tool_call.id + '\"}' }}\n {%- if not loop.last %}\n {{- \", \" }}\n {%- else %}\n {{- \"]\" + eos_token }}\n {%- endif %}\n {%- endfor %}\n {%- elif message[\"role\"] == \"assistant\" %}\n {{- message[\"content\"] + eos_token}}\n {%- elif message[\"role\"] == \"tool_results\" or message[\"role\"] == \"tool\" %}\n {%- if message.content is defined and message.content.content is defined %}\n {%- set content = message.content.content %}\n {%- else %}\n {%- set content = message.content %}\n {%- endif %}\n {{- '[TOOL_RESULTS]{\"content\": ' + content|string + \", \" }}\n {%- if not message.tool_call_id is defined or message.tool_call_id|length != 9 %}\n {{- raise_exception(\"Tool call IDs should be alphanumeric strings with length 9!\") }}\n {%- endif %}\n {{- '\"call_id\": \"' + message.tool_call_id + '\"}[/TOOL_RESULTS]' }}\n {%- else %}\n {{- raise_exception(\"Only user and assistant roles are supported, with the exception of an initial optional system message!\") }}\n {%- endif %}\n{%- endfor %}\n",
76
+ // mistralai/Mistral-Large-Instruct-2411 (mistralai 'v7' template)
77
+ "{{ bos_token }}{% for message in messages %}{% if message['role'] == 'user' %}{{ '[INST] ' + message['content'] + '[/INST]' }}{% elif message['role'] == 'system' %}{{ '[SYSTEM_PROMPT] ' + message['content'] + '[/SYSTEM_PROMPT]' }}{% elif message['role'] == 'assistant' %}{{ ' ' + message['content'] + eos_token }}{% else %}{{ raise_exception('Only user, system and assistant roles are supported!') }}{% endif %}{% endfor %}",
78
+ // ai-sage/GigaChat-20B-A3B-instruct
79
+ "{% if messages[0]['role'] == 'system' -%}\n {%- set loop_messages = messages[1:] -%}\n {%- set system_message = bos_token + messages[0]['content'] + additional_special_tokens[1] -%}\n{%- else -%}\n {%- set loop_messages = messages -%}\n {%- set system_message = bos_token + '' -%}\n{%- endif -%}\n{%- for message in loop_messages %}\n {% if (message['role'] == 'user') != (loop.index0 % 2 == 0) %}\n {{ raise_exception('Conversation roles must alternate user/assistant/user/assistant/...') }}\n {% endif %}\n \n {%- if loop.index0 == 0 -%}\n {{ system_message -}}\n {%- endif -%}\n {%- if message['role'] == 'user' -%}\n {{ message['role'] + additional_special_tokens[0] + message['content'] + additional_special_tokens[1] -}}\n {{ 'available functions' + additional_special_tokens[0] + additional_special_tokens[2] + additional_special_tokens[3] + additional_special_tokens[1] -}}\n {%- endif -%}\n {%- if message['role'] == 'assistant' -%}\n {{ message['role'] + additional_special_tokens[0] + message['content'] + additional_special_tokens[1] -}}\n {%- endif -%}\n {%- if loop.last and add_generation_prompt -%}\n {{ 'assistant' + additional_special_tokens[0] -}}\n {%- endif -%}\n{%- endfor %}",
68
80
  };
69
81
  std::vector<std::string> expected_output = {
70
82
  // teknium/OpenHermes-2.5-Mistral-7B
71
83
  "<|im_start|>system\nYou are a helpful assistant<|im_end|>\n<|im_start|>user\nHello<|im_end|>\n<|im_start|>assistant\nHi there<|im_end|>\n<|im_start|>user\nWho are you<|im_end|>\n<|im_start|>assistant\n I am an assistant <|im_end|>\n<|im_start|>user\nAnother question<|im_end|>\n<|im_start|>assistant\n",
72
- // mistralai/Mistral-7B-Instruct-v0.2
84
+ // mistralai/Mistral-7B-Instruct-v0.2 (NOTE: Old pre-v1 without a system prompt)
73
85
  "[INST] You are a helpful assistant\nHello [/INST]Hi there</s>[INST] Who are you [/INST] I am an assistant </s>[INST] Another question [/INST]",
74
86
  // TheBloke/FusionNet_34Bx2_MoE-AWQ
75
- "[INST] <<SYS>>\nYou are a helpful assistant\n<</SYS>>\n\nHello [/INST] Hi there </s><s>[INST] Who are you [/INST] I am an assistant </s><s>[INST] Another question [/INST]",
87
+ "[INST] <<SYS>>\nYou are a helpful assistant\n<</SYS>>\n\nHello [/INST]Hi there</s><s>[INST] Who are you [/INST] I am an assistant </s><s>[INST] Another question [/INST]",
76
88
  // bofenghuang/vigogne-2-70b-chat
77
- "[INST] <<SYS>>\nYou are a helpful assistant\n<</SYS>>\n\nHello [/INST] Hi there </s>[INST] Who are you [/INST] I am an assistant </s>[INST] Another question [/INST]",
89
+ "[INST] <<SYS>>\nYou are a helpful assistant\n<</SYS>>\n\nHello [/INST]Hi there</s>[INST] Who are you [/INST]I am an assistant</s>[INST] Another question [/INST]",
78
90
  // mlabonne/AlphaMonarch-7B
79
91
  "system\nYou are a helpful assistant</s>\n<s>user\nHello</s>\n<s>assistant\nHi there</s>\n<s>user\nWho are you</s>\n<s>assistant\n I am an assistant </s>\n<s>user\nAnother question</s>\n<s>assistant\n",
80
92
  // google/gemma-7b-it
@@ -109,10 +121,33 @@ int main(void) {
109
121
  u8"You are a helpful assistant<用户>Hello<AI>Hi there<用户>Who are you<AI>I am an assistant<用户>Another question<AI>",
110
122
  // DeepSeek-V2
111
123
  u8"You are a helpful assistant\n\nUser: Hello\n\nAssistant: Hi there<|end▁of▁sentence|>User: Who are you\n\nAssistant: I am an assistant <|end▁of▁sentence|>User: Another question\n\nAssistant:",
124
+ // ibm-granite/granite-3.0-8b-instruct
125
+ "<|start_of_role|>system<|end_of_role|>You are a helpful assistant<|end_of_text|>\n<|start_of_role|>user<|end_of_role|>Hello<|end_of_text|>\n<|start_of_role|>assistant<|end_of_role|>Hi there<|end_of_text|>\n<|start_of_role|>user<|end_of_role|>Who are you<|end_of_text|>\n<|start_of_role|>assistant<|end_of_role|> I am an assistant <|end_of_text|>\n<|start_of_role|>user<|end_of_role|>Another question<|end_of_text|>\n<|start_of_role|>assistant<|end_of_role|>\n",
126
+ // mistralai/Mistral-7B-Instruct-v0.2 (mistralai 'v1' template with a system prompt)
127
+ " [INST] You are a helpful assistant\n\nHello [/INST] Hi there</s> [INST] Who are you [/INST] I am an assistant </s> [INST] Another question [/INST]",
128
+ // Mistral-Large-Instruct-2407 (mistralai 'v3' template; modified to have system prompt at start)
129
+ "[INST] You are a helpful assistant\n\nHello[/INST] Hi there</s>[INST] Who are you[/INST] I am an assistant</s>[INST] Another question[/INST]",
130
+ // Mistral-Nemo-Instruct-2407 (mistralai 'v3-tekken' template; modified to have system prompt at start)
131
+ "[INST]You are a helpful assistant\n\nHello[/INST]Hi there</s>[INST]Who are you[/INST] I am an assistant </s>[INST]Another question[/INST]",
132
+ // mistralai/Mistral-Large-Instruct-2411 (mistralai 'v7' template)
133
+ "[SYSTEM_PROMPT] You are a helpful assistant[/SYSTEM_PROMPT][INST] Hello[/INST] Hi there</s>[INST] Who are you[/INST] I am an assistant </s>[INST] Another question[/INST]",
134
+ // ai-sage/GigaChat-20B-A3B-instruct
135
+ "<s>You are a helpful assistant<|message_sep|>user<|role_sep|>Hello<|message_sep|>available functions<|role_sep|>[]<|message_sep|>assistant<|role_sep|>Hi there<|message_sep|>user<|role_sep|>Who are you<|message_sep|>available functions<|role_sep|>[]<|message_sep|>assistant<|role_sep|> I am an assistant <|message_sep|>user<|role_sep|>Another question<|message_sep|>available functions<|role_sep|>[]<|message_sep|>assistant<|role_sep|>",
112
136
  };
113
137
  std::vector<char> formatted_chat(1024);
114
138
  int32_t res;
115
139
 
140
+ // list all supported templates
141
+ std::vector<const char *> supported_tmpl;
142
+ res = llama_chat_builtin_templates(nullptr, 0);
143
+ assert(res > 0);
144
+ supported_tmpl.resize(res);
145
+ res = llama_chat_builtin_templates(supported_tmpl.data(), supported_tmpl.size());
146
+ printf("Built-in chat templates:\n");
147
+ for (auto tmpl : supported_tmpl) {
148
+ printf(" %s\n", tmpl);
149
+ }
150
+
116
151
  // test invalid chat template
117
152
  res = llama_chat_apply_template(nullptr, "INVALID TEMPLATE", conversation, message_count, true, formatted_chat.data(), formatted_chat.size());
118
153
  assert(res < 0);
@@ -140,19 +175,26 @@ int main(void) {
140
175
 
141
176
  // test llama_chat_format_single for system message
142
177
  printf("\n\n=== llama_chat_format_single (system message) ===\n\n");
143
- std::vector<llama_chat_msg> chat2;
144
- llama_chat_msg sys_msg{"system", "You are a helpful assistant"};
178
+ std::vector<common_chat_msg> chat2;
179
+ common_chat_msg sys_msg{"system", "You are a helpful assistant"};
145
180
 
146
181
  auto fmt_sys = [&](std::string tmpl) {
147
- auto output = llama_chat_format_single(nullptr, tmpl, chat2, sys_msg, false);
182
+ auto output = common_chat_format_single(nullptr, tmpl, chat2, sys_msg, false);
148
183
  printf("fmt_sys(%s) : %s\n", tmpl.c_str(), output.c_str());
149
184
  printf("-------------------------\n");
150
185
  return output;
151
186
  };
152
187
  assert(fmt_sys("chatml") == "<|im_start|>system\nYou are a helpful assistant<|im_end|>\n");
188
+ assert(fmt_sys("mistral-v1") == " [INST] You are a helpful assistant\n\n");
189
+ assert(fmt_sys("mistral-v3") == "[INST] You are a helpful assistant\n\n");
190
+ assert(fmt_sys("mistral-v3-tekken") == "[INST]You are a helpful assistant\n\n");
191
+ assert(fmt_sys("mistral-v7") == "[SYSTEM_PROMPT] You are a helpful assistant[/SYSTEM_PROMPT]");
153
192
  assert(fmt_sys("llama2") == "[INST] You are a helpful assistant\n");
193
+ assert(fmt_sys("llama2-sys") == "[INST] <<SYS>>\nYou are a helpful assistant\n<</SYS>>\n\n");
194
+ assert(fmt_sys("mistral") == "[INST] You are a helpful assistant\n"); // for old pre-v1 templates
154
195
  assert(fmt_sys("gemma") == ""); // for gemma, system message is merged with user message
155
196
  assert(fmt_sys("llama3") == "<|start_header_id|>system<|end_header_id|>\n\nYou are a helpful assistant<|eot_id|>");
197
+ assert(fmt_sys("gigachat") == "<s>You are a helpful assistant<|message_sep|>");
156
198
 
157
199
 
158
200
  // test llama_chat_format_single for user message
@@ -160,18 +202,26 @@ int main(void) {
160
202
  chat2.push_back({"system", "You are a helpful assistant"});
161
203
  chat2.push_back({"user", "Hello"});
162
204
  chat2.push_back({"assistant", "I am assistant"});
163
- llama_chat_msg new_msg{"user", "How are you"};
205
+ common_chat_msg new_msg{"user", "How are you"};
164
206
 
165
207
  auto fmt_single = [&](std::string tmpl) {
166
- auto output = llama_chat_format_single(nullptr, tmpl, chat2, new_msg, true);
208
+ auto output = common_chat_format_single(nullptr, tmpl, chat2, new_msg, true);
167
209
  printf("fmt_single(%s) : %s\n", tmpl.c_str(), output.c_str());
168
210
  printf("-------------------------\n");
169
211
  return output;
170
212
  };
171
213
  assert(fmt_single("chatml") == "\n<|im_start|>user\nHow are you<|im_end|>\n<|im_start|>assistant\n");
214
+ assert(fmt_single("mistral-v1") == " [INST] How are you [/INST]");
215
+ assert(fmt_single("mistral-v3") == "[INST] How are you[/INST]");
216
+ assert(fmt_single("mistral-v3-tekken") == "[INST]How are you[/INST]");
217
+ assert(fmt_single("mistral-v7") == "[INST] How are you[/INST]");
172
218
  assert(fmt_single("llama2") == "[INST] How are you [/INST]");
219
+ assert(fmt_single("mistral") == "[INST] How are you [/INST]"); // for old pre-v1 templates
173
220
  assert(fmt_single("gemma") == "\n<start_of_turn>user\nHow are you<end_of_turn>\n<start_of_turn>model\n");
174
221
  assert(fmt_single("llama3") == "<|start_header_id|>user<|end_header_id|>\n\nHow are you<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n");
222
+ assert(fmt_single("gigachat") == "user<|role_sep|>How are you<|message_sep|>available functions<|role_sep|>[]<|message_sep|>assistant<|role_sep|>");
223
+
224
+ printf("Test chat templates: OK\n");
175
225
 
176
226
  return 0;
177
227
  }