whispercpp 1.3.1 → 1.3.3

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 (857) hide show
  1. checksums.yaml +4 -4
  2. data/.gitignore +7 -3
  3. data/README.md +161 -43
  4. data/Rakefile +45 -13
  5. data/ext/.gitignore +4 -8
  6. data/ext/dependencies.rb +73 -0
  7. data/ext/extconf.rb +21 -198
  8. data/ext/options.rb +85 -0
  9. data/ext/ruby_whisper.c +177 -0
  10. data/ext/ruby_whisper.h +17 -2
  11. data/ext/ruby_whisper_context.c +672 -0
  12. data/ext/ruby_whisper_error.c +52 -0
  13. data/ext/ruby_whisper_model.c +232 -0
  14. data/ext/ruby_whisper_params.c +1303 -0
  15. data/ext/ruby_whisper_segment.c +220 -0
  16. data/ext/ruby_whisper_transcribe.cpp +93 -0
  17. data/ext/ruby_whisper_vad_params.c +288 -0
  18. data/ext/sources/CMakeGraphVizOptions.cmake +8 -0
  19. data/ext/sources/CMakeLists.txt +255 -0
  20. data/ext/sources/bindings/javascript/CMakeLists.txt +41 -0
  21. data/ext/sources/bindings/javascript/emscripten.cpp +93 -0
  22. data/ext/sources/bindings/javascript/libwhisper.worker.js +1 -0
  23. data/ext/sources/bindings/javascript/package-tmpl.json +26 -0
  24. data/ext/sources/bindings/javascript/package.json +26 -0
  25. data/ext/sources/bindings/javascript/whisper.js +19 -0
  26. data/ext/sources/build-xcframework.sh +547 -0
  27. data/ext/sources/cmake/DefaultTargetOptions.cmake +16 -0
  28. data/ext/sources/cmake/FindFFmpeg.cmake +163 -0
  29. data/ext/sources/cmake/build-info.cmake +60 -0
  30. data/ext/sources/cmake/git-vars.cmake +22 -0
  31. data/ext/sources/cmake/whisper-config.cmake.in +65 -0
  32. data/ext/sources/cmake/whisper.pc.in +10 -0
  33. data/ext/sources/examples/CMakeLists.txt +124 -0
  34. data/ext/sources/examples/addon.node/CMakeLists.txt +31 -0
  35. data/ext/sources/examples/addon.node/__test__/whisper.spec.js +133 -0
  36. data/ext/sources/examples/addon.node/addon.cpp +557 -0
  37. data/ext/sources/examples/addon.node/index.js +57 -0
  38. data/ext/sources/examples/addon.node/package.json +16 -0
  39. data/ext/sources/examples/addon.node/vad-example.js +132 -0
  40. data/ext/sources/examples/bench/CMakeLists.txt +8 -0
  41. data/ext/sources/examples/bench/bench.cpp +176 -0
  42. data/ext/sources/examples/bench.wasm/CMakeLists.txt +49 -0
  43. data/ext/sources/examples/bench.wasm/emscripten.cpp +87 -0
  44. data/ext/sources/examples/bench.wasm/index-tmpl.html +284 -0
  45. data/ext/sources/examples/cli/CMakeLists.txt +8 -0
  46. data/ext/sources/examples/cli/cli.cpp +1295 -0
  47. data/ext/sources/examples/coi-serviceworker.js +146 -0
  48. data/ext/sources/examples/command/CMakeLists.txt +10 -0
  49. data/ext/sources/examples/command/command.cpp +800 -0
  50. data/ext/sources/examples/command/commands.txt +9 -0
  51. data/ext/sources/examples/command.wasm/CMakeLists.txt +50 -0
  52. data/ext/sources/examples/command.wasm/emscripten.cpp +327 -0
  53. data/ext/sources/examples/command.wasm/index-tmpl.html +414 -0
  54. data/ext/sources/examples/common-ggml.cpp +238 -0
  55. data/ext/sources/examples/common-ggml.h +18 -0
  56. data/ext/sources/examples/common-sdl.cpp +227 -0
  57. data/ext/sources/examples/common-sdl.h +49 -0
  58. data/ext/sources/examples/common-whisper.cpp +175 -0
  59. data/ext/sources/examples/common-whisper.h +24 -0
  60. data/ext/sources/examples/common.cpp +675 -0
  61. data/ext/sources/examples/common.h +322 -0
  62. data/ext/sources/examples/deprecation-warning/CMakeLists.txt +6 -0
  63. data/ext/sources/examples/deprecation-warning/deprecation-warning.cpp +38 -0
  64. data/ext/sources/examples/ffmpeg-transcode.cpp +368 -0
  65. data/ext/sources/examples/generate-karaoke.sh +57 -0
  66. data/ext/sources/examples/grammar-parser.cpp +423 -0
  67. data/ext/sources/examples/grammar-parser.h +29 -0
  68. data/ext/sources/examples/helpers.js +191 -0
  69. data/ext/sources/examples/json.hpp +24596 -0
  70. data/ext/sources/examples/livestream.sh +112 -0
  71. data/ext/sources/examples/lsp/CMakeLists.txt +9 -0
  72. data/ext/sources/examples/lsp/lsp.cpp +469 -0
  73. data/ext/sources/examples/lsp/whisper.vim +362 -0
  74. data/ext/sources/examples/miniaudio.h +93468 -0
  75. data/ext/sources/examples/python/test_whisper_processor.py +7 -0
  76. data/ext/sources/examples/python/whisper_processor.py +54 -0
  77. data/ext/sources/examples/quantize/CMakeLists.txt +6 -0
  78. data/ext/sources/examples/quantize/quantize.cpp +226 -0
  79. data/ext/sources/examples/server/CMakeLists.txt +15 -0
  80. data/ext/sources/examples/server/bench.js +29 -0
  81. data/ext/sources/examples/server/httplib.h +10497 -0
  82. data/ext/sources/examples/server/server.cpp +1238 -0
  83. data/ext/sources/examples/server.py +115 -0
  84. data/ext/sources/examples/stb_vorbis.c +5584 -0
  85. data/ext/sources/examples/stream/CMakeLists.txt +10 -0
  86. data/ext/sources/examples/stream/stream.cpp +435 -0
  87. data/ext/sources/examples/stream.wasm/CMakeLists.txt +49 -0
  88. data/ext/sources/examples/stream.wasm/emscripten.cpp +216 -0
  89. data/ext/sources/examples/stream.wasm/index-tmpl.html +414 -0
  90. data/ext/sources/examples/sycl/CMakeLists.txt +9 -0
  91. data/ext/sources/examples/sycl/build.sh +22 -0
  92. data/ext/sources/examples/sycl/ls-sycl-device.cpp +11 -0
  93. data/ext/sources/examples/sycl/run-whisper.sh +17 -0
  94. data/ext/sources/examples/talk-llama/CMakeLists.txt +43 -0
  95. data/ext/sources/examples/talk-llama/eleven-labs.py +80 -0
  96. data/ext/sources/examples/talk-llama/llama-adapter.cpp +388 -0
  97. data/ext/sources/examples/talk-llama/llama-adapter.h +76 -0
  98. data/ext/sources/examples/talk-llama/llama-arch.cpp +1914 -0
  99. data/ext/sources/examples/talk-llama/llama-arch.h +464 -0
  100. data/ext/sources/examples/talk-llama/llama-batch.cpp +843 -0
  101. data/ext/sources/examples/talk-llama/llama-batch.h +147 -0
  102. data/ext/sources/examples/talk-llama/llama-chat.cpp +685 -0
  103. data/ext/sources/examples/talk-llama/llama-chat.h +59 -0
  104. data/ext/sources/examples/talk-llama/llama-context.cpp +2845 -0
  105. data/ext/sources/examples/talk-llama/llama-context.h +297 -0
  106. data/ext/sources/examples/talk-llama/llama-cparams.cpp +5 -0
  107. data/ext/sources/examples/talk-llama/llama-cparams.h +41 -0
  108. data/ext/sources/examples/talk-llama/llama-grammar.cpp +1229 -0
  109. data/ext/sources/examples/talk-llama/llama-grammar.h +173 -0
  110. data/ext/sources/examples/talk-llama/llama-graph.cpp +1693 -0
  111. data/ext/sources/examples/talk-llama/llama-graph.h +710 -0
  112. data/ext/sources/examples/talk-llama/llama-hparams.cpp +103 -0
  113. data/ext/sources/examples/talk-llama/llama-hparams.h +207 -0
  114. data/ext/sources/examples/talk-llama/llama-impl.cpp +167 -0
  115. data/ext/sources/examples/talk-llama/llama-impl.h +61 -0
  116. data/ext/sources/examples/talk-llama/llama-io.cpp +15 -0
  117. data/ext/sources/examples/talk-llama/llama-io.h +35 -0
  118. data/ext/sources/examples/talk-llama/llama-kv-cache-unified-iswa.cpp +279 -0
  119. data/ext/sources/examples/talk-llama/llama-kv-cache-unified-iswa.h +128 -0
  120. data/ext/sources/examples/talk-llama/llama-kv-cache-unified.cpp +1841 -0
  121. data/ext/sources/examples/talk-llama/llama-kv-cache-unified.h +303 -0
  122. data/ext/sources/examples/talk-llama/llama-kv-cache.h +44 -0
  123. data/ext/sources/examples/talk-llama/llama-kv-cells.h +439 -0
  124. data/ext/sources/examples/talk-llama/llama-memory-hybrid.cpp +246 -0
  125. data/ext/sources/examples/talk-llama/llama-memory-hybrid.h +138 -0
  126. data/ext/sources/examples/talk-llama/llama-memory-recurrent.cpp +1125 -0
  127. data/ext/sources/examples/talk-llama/llama-memory-recurrent.h +183 -0
  128. data/ext/sources/examples/talk-llama/llama-memory.cpp +59 -0
  129. data/ext/sources/examples/talk-llama/llama-memory.h +116 -0
  130. data/ext/sources/examples/talk-llama/llama-mmap.cpp +600 -0
  131. data/ext/sources/examples/talk-llama/llama-mmap.h +68 -0
  132. data/ext/sources/examples/talk-llama/llama-model-loader.cpp +1163 -0
  133. data/ext/sources/examples/talk-llama/llama-model-loader.h +169 -0
  134. data/ext/sources/examples/talk-llama/llama-model-saver.cpp +282 -0
  135. data/ext/sources/examples/talk-llama/llama-model-saver.h +37 -0
  136. data/ext/sources/examples/talk-llama/llama-model.cpp +15114 -0
  137. data/ext/sources/examples/talk-llama/llama-model.h +452 -0
  138. data/ext/sources/examples/talk-llama/llama-quant.cpp +1049 -0
  139. data/ext/sources/examples/talk-llama/llama-quant.h +1 -0
  140. data/ext/sources/examples/talk-llama/llama-sampling.cpp +2575 -0
  141. data/ext/sources/examples/talk-llama/llama-sampling.h +32 -0
  142. data/ext/sources/examples/talk-llama/llama-vocab.cpp +3377 -0
  143. data/ext/sources/examples/talk-llama/llama-vocab.h +132 -0
  144. data/ext/sources/examples/talk-llama/llama.cpp +358 -0
  145. data/ext/sources/examples/talk-llama/llama.h +1484 -0
  146. data/ext/sources/examples/talk-llama/prompts/talk-alpaca.txt +23 -0
  147. data/ext/sources/examples/talk-llama/speak +40 -0
  148. data/ext/sources/examples/talk-llama/speak.bat +1 -0
  149. data/ext/sources/examples/talk-llama/speak.ps1 +14 -0
  150. data/ext/sources/examples/talk-llama/talk-llama.cpp +810 -0
  151. data/ext/sources/examples/talk-llama/unicode-data.cpp +7034 -0
  152. data/ext/sources/examples/talk-llama/unicode-data.h +20 -0
  153. data/ext/sources/examples/talk-llama/unicode.cpp +854 -0
  154. data/ext/sources/examples/talk-llama/unicode.h +66 -0
  155. data/ext/sources/examples/vad-speech-segments/CMakeLists.txt +8 -0
  156. data/ext/sources/examples/vad-speech-segments/speech.cpp +149 -0
  157. data/ext/sources/examples/wchess/CMakeLists.txt +10 -0
  158. data/ext/sources/examples/wchess/libwchess/CMakeLists.txt +19 -0
  159. data/ext/sources/examples/wchess/libwchess/Chessboard.cpp +803 -0
  160. data/ext/sources/examples/wchess/libwchess/Chessboard.h +33 -0
  161. data/ext/sources/examples/wchess/libwchess/WChess.cpp +193 -0
  162. data/ext/sources/examples/wchess/libwchess/WChess.h +63 -0
  163. data/ext/sources/examples/wchess/libwchess/test-chessboard.cpp +117 -0
  164. data/ext/sources/examples/wchess/wchess.cmd/CMakeLists.txt +8 -0
  165. data/ext/sources/examples/wchess/wchess.cmd/wchess.cmd.cpp +251 -0
  166. data/ext/sources/examples/whisper.wasm/CMakeLists.txt +50 -0
  167. data/ext/sources/examples/whisper.wasm/emscripten.cpp +118 -0
  168. data/ext/sources/examples/whisper.wasm/index-tmpl.html +658 -0
  169. data/ext/sources/ggml/CMakeLists.txt +435 -0
  170. data/ext/sources/ggml/cmake/BuildTypes.cmake +54 -0
  171. data/ext/sources/ggml/cmake/GitVars.cmake +22 -0
  172. data/ext/sources/ggml/cmake/common.cmake +50 -0
  173. data/ext/sources/ggml/cmake/ggml-config.cmake.in +152 -0
  174. data/ext/{ggml → sources/ggml}/include/ggml-alloc.h +1 -1
  175. data/ext/{ggml → sources/ggml}/include/ggml-backend.h +10 -8
  176. data/ext/{ggml → sources/ggml}/include/ggml-cpp.h +2 -1
  177. data/ext/{ggml → sources/ggml}/include/ggml-cpu.h +11 -1
  178. data/ext/{ggml → sources/ggml}/include/ggml-metal.h +1 -1
  179. data/ext/{ggml → sources/ggml}/include/ggml-opt.h +49 -28
  180. data/ext/{ggml → sources/ggml}/include/ggml-rpc.h +6 -1
  181. data/ext/{ggml → sources/ggml}/include/ggml-vulkan.h +0 -2
  182. data/ext/{ggml → sources/ggml}/include/ggml.h +325 -269
  183. data/ext/sources/ggml/include/gguf.h +202 -0
  184. data/ext/sources/ggml/src/CMakeLists.txt +404 -0
  185. data/ext/{ggml → sources/ggml}/src/ggml-alloc.c +34 -29
  186. data/ext/sources/ggml/src/ggml-amx/CMakeLists.txt +107 -0
  187. data/ext/{ggml → sources/ggml}/src/ggml-backend-impl.h +1 -2
  188. data/ext/{ggml → sources/ggml}/src/ggml-backend-reg.cpp +92 -53
  189. data/ext/{ggml → sources/ggml}/src/ggml-backend.cpp +69 -34
  190. data/ext/sources/ggml/src/ggml-blas/CMakeLists.txt +87 -0
  191. data/ext/sources/ggml/src/ggml-cann/CMakeLists.txt +75 -0
  192. data/ext/sources/ggml/src/ggml-cann/Doxyfile +2579 -0
  193. data/ext/{ggml → sources/ggml}/src/ggml-cann/acl_tensor.cpp +10 -4
  194. data/ext/{ggml → sources/ggml}/src/ggml-cann/acl_tensor.h +5 -5
  195. data/ext/{ggml → sources/ggml}/src/ggml-cann/aclnn_ops.cpp +1272 -1506
  196. data/ext/sources/ggml/src/ggml-cann/aclnn_ops.h +1125 -0
  197. data/ext/{ggml → sources/ggml}/src/ggml-cann/common.h +140 -1
  198. data/ext/{ggml → sources/ggml}/src/ggml-cann/ggml-cann.cpp +588 -146
  199. data/ext/sources/ggml/src/ggml-cann/kernels/CMakeLists.txt +30 -0
  200. data/ext/{ggml → sources/ggml}/src/ggml-cann/kernels/dup.cpp +3 -5
  201. data/ext/{ggml → sources/ggml}/src/ggml-common.h +16 -8
  202. data/ext/sources/ggml/src/ggml-cpu/CMakeLists.txt +597 -0
  203. data/ext/{ggml → sources/ggml}/src/ggml-cpu/amx/amx.cpp +3 -2
  204. data/ext/{ggml → sources/ggml}/src/ggml-cpu/amx/mmq.cpp +11 -10
  205. data/ext/sources/ggml/src/ggml-cpu/arch/arm/cpu-feats.cpp +94 -0
  206. data/ext/sources/ggml/src/ggml-cpu/arch/arm/quants.c +4114 -0
  207. data/ext/sources/ggml/src/ggml-cpu/arch/arm/repack.cpp +2163 -0
  208. data/ext/sources/ggml/src/ggml-cpu/arch/loongarch/quants.c +2639 -0
  209. data/ext/sources/ggml/src/ggml-cpu/arch/powerpc/cpu-feats.cpp +82 -0
  210. data/ext/sources/ggml/src/ggml-cpu/arch/powerpc/quants.c +2732 -0
  211. data/ext/sources/ggml/src/ggml-cpu/arch/riscv/quants.c +2069 -0
  212. data/ext/sources/ggml/src/ggml-cpu/arch/riscv/repack.cpp +397 -0
  213. data/ext/sources/ggml/src/ggml-cpu/arch/s390/quants.c +1300 -0
  214. data/ext/sources/ggml/src/ggml-cpu/arch/wasm/quants.c +1481 -0
  215. data/ext/{ggml/src/ggml-cpu/cpu-feats-x86.cpp → sources/ggml/src/ggml-cpu/arch/x86/cpu-feats.cpp} +5 -1
  216. data/ext/sources/ggml/src/ggml-cpu/arch/x86/quants.c +4311 -0
  217. data/ext/sources/ggml/src/ggml-cpu/arch/x86/repack.cpp +3285 -0
  218. data/ext/sources/ggml/src/ggml-cpu/arch-fallback.h +184 -0
  219. data/ext/sources/ggml/src/ggml-cpu/binary-ops.cpp +158 -0
  220. data/ext/sources/ggml/src/ggml-cpu/binary-ops.h +16 -0
  221. data/ext/sources/ggml/src/ggml-cpu/cmake/FindSIMD.cmake +100 -0
  222. data/ext/sources/ggml/src/ggml-cpu/common.h +73 -0
  223. data/ext/{ggml → sources/ggml}/src/ggml-cpu/ggml-cpu-impl.h +172 -41
  224. data/ext/sources/ggml/src/ggml-cpu/ggml-cpu.c +3551 -0
  225. data/ext/{ggml → sources/ggml}/src/ggml-cpu/ggml-cpu.cpp +78 -25
  226. data/ext/{ggml/src/ggml-cpu/ggml-cpu-hbm.cpp → sources/ggml/src/ggml-cpu/hbm.cpp} +1 -1
  227. data/ext/sources/ggml/src/ggml-cpu/kleidiai/kernels.cpp +337 -0
  228. data/ext/sources/ggml/src/ggml-cpu/kleidiai/kernels.h +95 -0
  229. data/ext/sources/ggml/src/ggml-cpu/kleidiai/kleidiai.cpp +482 -0
  230. data/ext/sources/ggml/src/ggml-cpu/kleidiai/kleidiai.h +17 -0
  231. data/ext/sources/ggml/src/ggml-cpu/llamafile/sgemm.cpp +3594 -0
  232. data/ext/sources/ggml/src/ggml-cpu/llamafile/sgemm.h +19 -0
  233. data/ext/sources/ggml/src/ggml-cpu/ops.cpp +9786 -0
  234. data/ext/sources/ggml/src/ggml-cpu/ops.h +118 -0
  235. data/ext/sources/ggml/src/ggml-cpu/quants.c +1158 -0
  236. data/ext/{ggml/src/ggml-cpu/ggml-cpu-quants.h → sources/ggml/src/ggml-cpu/quants.h} +26 -0
  237. data/ext/sources/ggml/src/ggml-cpu/repack.cpp +1571 -0
  238. data/ext/sources/ggml/src/ggml-cpu/repack.h +98 -0
  239. data/ext/sources/ggml/src/ggml-cpu/simd-mappings.h +1184 -0
  240. data/ext/{ggml/src/ggml-cpu/ggml-cpu-traits.cpp → sources/ggml/src/ggml-cpu/traits.cpp} +1 -1
  241. data/ext/sources/ggml/src/ggml-cpu/unary-ops.cpp +186 -0
  242. data/ext/sources/ggml/src/ggml-cpu/unary-ops.h +28 -0
  243. data/ext/sources/ggml/src/ggml-cpu/vec.cpp +345 -0
  244. data/ext/sources/ggml/src/ggml-cpu/vec.h +1027 -0
  245. data/ext/sources/ggml/src/ggml-cuda/CMakeLists.txt +184 -0
  246. data/ext/sources/ggml/src/ggml-cuda/acc.cu +61 -0
  247. data/ext/sources/ggml/src/ggml-cuda/acc.cuh +5 -0
  248. data/ext/sources/ggml/src/ggml-cuda/arange.cu +34 -0
  249. data/ext/sources/ggml/src/ggml-cuda/arange.cuh +5 -0
  250. data/ext/sources/ggml/src/ggml-cuda/argmax.cu +91 -0
  251. data/ext/sources/ggml/src/ggml-cuda/argmax.cuh +3 -0
  252. data/ext/sources/ggml/src/ggml-cuda/argsort.cu +104 -0
  253. data/ext/sources/ggml/src/ggml-cuda/argsort.cuh +3 -0
  254. data/ext/sources/ggml/src/ggml-cuda/binbcast.cu +363 -0
  255. data/ext/sources/ggml/src/ggml-cuda/binbcast.cuh +9 -0
  256. data/ext/sources/ggml/src/ggml-cuda/clamp.cu +45 -0
  257. data/ext/sources/ggml/src/ggml-cuda/clamp.cuh +5 -0
  258. data/ext/sources/ggml/src/ggml-cuda/common.cuh +851 -0
  259. data/ext/sources/ggml/src/ggml-cuda/concat.cu +221 -0
  260. data/ext/sources/ggml/src/ggml-cuda/concat.cuh +5 -0
  261. data/ext/sources/ggml/src/ggml-cuda/conv-transpose-1d.cu +89 -0
  262. data/ext/sources/ggml/src/ggml-cuda/conv-transpose-1d.cuh +5 -0
  263. data/ext/sources/ggml/src/ggml-cuda/conv2d-dw.cu +161 -0
  264. data/ext/sources/ggml/src/ggml-cuda/conv2d-dw.cuh +5 -0
  265. data/ext/sources/ggml/src/ggml-cuda/conv2d-transpose.cu +91 -0
  266. data/ext/sources/ggml/src/ggml-cuda/conv2d-transpose.cuh +4 -0
  267. data/ext/sources/ggml/src/ggml-cuda/convert.cu +752 -0
  268. data/ext/sources/ggml/src/ggml-cuda/convert.cuh +31 -0
  269. data/ext/sources/ggml/src/ggml-cuda/count-equal.cu +64 -0
  270. data/ext/sources/ggml/src/ggml-cuda/count-equal.cuh +5 -0
  271. data/ext/sources/ggml/src/ggml-cuda/cp-async.cuh +57 -0
  272. data/ext/sources/ggml/src/ggml-cuda/cpy.cu +705 -0
  273. data/ext/sources/ggml/src/ggml-cuda/cpy.cuh +11 -0
  274. data/ext/sources/ggml/src/ggml-cuda/cross-entropy-loss.cu +189 -0
  275. data/ext/sources/ggml/src/ggml-cuda/cross-entropy-loss.cuh +7 -0
  276. data/ext/sources/ggml/src/ggml-cuda/dequantize.cuh +103 -0
  277. data/ext/sources/ggml/src/ggml-cuda/diagmask.cu +40 -0
  278. data/ext/sources/ggml/src/ggml-cuda/diagmask.cuh +5 -0
  279. data/ext/sources/ggml/src/ggml-cuda/fattn-common.cuh +881 -0
  280. data/ext/sources/ggml/src/ggml-cuda/fattn-mma-f16.cuh +1474 -0
  281. data/ext/sources/ggml/src/ggml-cuda/fattn-tile-f16.cu +357 -0
  282. data/ext/sources/ggml/src/ggml-cuda/fattn-tile-f16.cuh +3 -0
  283. data/ext/sources/ggml/src/ggml-cuda/fattn-tile-f32.cu +365 -0
  284. data/ext/sources/ggml/src/ggml-cuda/fattn-tile-f32.cuh +3 -0
  285. data/ext/sources/ggml/src/ggml-cuda/fattn-vec-f16.cuh +482 -0
  286. data/ext/sources/ggml/src/ggml-cuda/fattn-vec-f32.cuh +472 -0
  287. data/ext/sources/ggml/src/ggml-cuda/fattn-wmma-f16.cu +638 -0
  288. data/ext/sources/ggml/src/ggml-cuda/fattn-wmma-f16.cuh +3 -0
  289. data/ext/sources/ggml/src/ggml-cuda/fattn.cu +346 -0
  290. data/ext/sources/ggml/src/ggml-cuda/fattn.cuh +3 -0
  291. data/ext/sources/ggml/src/ggml-cuda/getrows.cu +275 -0
  292. data/ext/sources/ggml/src/ggml-cuda/getrows.cuh +15 -0
  293. data/ext/sources/ggml/src/ggml-cuda/ggml-cuda.cu +3647 -0
  294. data/ext/sources/ggml/src/ggml-cuda/gla.cu +93 -0
  295. data/ext/sources/ggml/src/ggml-cuda/gla.cuh +3 -0
  296. data/ext/sources/ggml/src/ggml-cuda/im2col.cu +103 -0
  297. data/ext/sources/ggml/src/ggml-cuda/im2col.cuh +5 -0
  298. data/ext/sources/ggml/src/ggml-cuda/mean.cu +19 -0
  299. data/ext/sources/ggml/src/ggml-cuda/mean.cuh +3 -0
  300. data/ext/sources/ggml/src/ggml-cuda/mma.cuh +396 -0
  301. data/ext/sources/ggml/src/ggml-cuda/mmq.cu +324 -0
  302. data/ext/sources/ggml/src/ggml-cuda/mmq.cuh +3217 -0
  303. data/ext/sources/ggml/src/ggml-cuda/mmv.cu +506 -0
  304. data/ext/sources/ggml/src/ggml-cuda/mmv.cuh +11 -0
  305. data/ext/sources/ggml/src/ggml-cuda/mmvq.cu +595 -0
  306. data/ext/sources/ggml/src/ggml-cuda/mmvq.cuh +12 -0
  307. data/ext/sources/ggml/src/ggml-cuda/norm.cu +458 -0
  308. data/ext/sources/ggml/src/ggml-cuda/norm.cuh +11 -0
  309. data/ext/sources/ggml/src/ggml-cuda/opt-step-adamw.cu +78 -0
  310. data/ext/sources/ggml/src/ggml-cuda/opt-step-adamw.cuh +5 -0
  311. data/ext/sources/ggml/src/ggml-cuda/out-prod.cu +68 -0
  312. data/ext/sources/ggml/src/ggml-cuda/out-prod.cuh +3 -0
  313. data/ext/sources/ggml/src/ggml-cuda/pad.cu +49 -0
  314. data/ext/sources/ggml/src/ggml-cuda/pad.cuh +5 -0
  315. data/ext/sources/ggml/src/ggml-cuda/pool2d.cu +94 -0
  316. data/ext/sources/ggml/src/ggml-cuda/pool2d.cuh +5 -0
  317. data/ext/sources/ggml/src/ggml-cuda/quantize.cu +190 -0
  318. data/ext/sources/ggml/src/ggml-cuda/quantize.cuh +27 -0
  319. data/ext/sources/ggml/src/ggml-cuda/rope.cu +456 -0
  320. data/ext/sources/ggml/src/ggml-cuda/rope.cuh +7 -0
  321. data/ext/sources/ggml/src/ggml-cuda/scale.cu +31 -0
  322. data/ext/sources/ggml/src/ggml-cuda/scale.cuh +5 -0
  323. data/ext/sources/ggml/src/ggml-cuda/softmax.cu +283 -0
  324. data/ext/sources/ggml/src/ggml-cuda/softmax.cuh +7 -0
  325. data/ext/sources/ggml/src/ggml-cuda/ssm-conv.cu +148 -0
  326. data/ext/sources/ggml/src/ggml-cuda/ssm-conv.cuh +3 -0
  327. data/ext/sources/ggml/src/ggml-cuda/ssm-scan.cu +155 -0
  328. data/ext/sources/ggml/src/ggml-cuda/ssm-scan.cuh +3 -0
  329. data/ext/sources/ggml/src/ggml-cuda/sum.cu +45 -0
  330. data/ext/sources/ggml/src/ggml-cuda/sum.cuh +5 -0
  331. data/ext/sources/ggml/src/ggml-cuda/sumrows.cu +26 -0
  332. data/ext/sources/ggml/src/ggml-cuda/sumrows.cuh +4 -0
  333. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_1-ncols2_16.cu +5 -0
  334. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_1-ncols2_8.cu +10 -0
  335. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_16-ncols2_1.cu +10 -0
  336. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_16-ncols2_2.cu +10 -0
  337. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_16-ncols2_4.cu +10 -0
  338. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_2-ncols2_16.cu +5 -0
  339. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_2-ncols2_4.cu +10 -0
  340. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_2-ncols2_8.cu +10 -0
  341. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_32-ncols2_1.cu +10 -0
  342. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_32-ncols2_2.cu +10 -0
  343. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_4-ncols2_16.cu +5 -0
  344. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_4-ncols2_2.cu +10 -0
  345. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_4-ncols2_4.cu +10 -0
  346. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_4-ncols2_8.cu +10 -0
  347. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_64-ncols2_1.cu +10 -0
  348. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_8-ncols2_1.cu +10 -0
  349. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_8-ncols2_2.cu +10 -0
  350. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_8-ncols2_4.cu +10 -0
  351. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_8-ncols2_8.cu +10 -0
  352. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-f16.cu +5 -0
  353. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q4_0.cu +5 -0
  354. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q4_1.cu +5 -0
  355. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q5_0.cu +5 -0
  356. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q5_1.cu +5 -0
  357. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q8_0.cu +5 -0
  358. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-f16.cu +5 -0
  359. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q4_0.cu +5 -0
  360. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q4_1.cu +5 -0
  361. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q5_0.cu +5 -0
  362. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q5_1.cu +5 -0
  363. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q8_0.cu +5 -0
  364. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-f16.cu +5 -0
  365. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q4_0.cu +5 -0
  366. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q4_1.cu +5 -0
  367. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q5_0.cu +5 -0
  368. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q5_1.cu +5 -0
  369. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q8_0.cu +5 -0
  370. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-f16.cu +5 -0
  371. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q4_0.cu +5 -0
  372. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q4_1.cu +5 -0
  373. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q5_0.cu +5 -0
  374. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q5_1.cu +5 -0
  375. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q8_0.cu +5 -0
  376. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-f16.cu +5 -0
  377. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q4_0.cu +5 -0
  378. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q4_1.cu +5 -0
  379. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q5_0.cu +5 -0
  380. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q5_1.cu +5 -0
  381. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q8_0.cu +5 -0
  382. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-f16.cu +5 -0
  383. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q4_0.cu +5 -0
  384. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q4_1.cu +5 -0
  385. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q5_0.cu +5 -0
  386. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q5_1.cu +5 -0
  387. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q8_0.cu +5 -0
  388. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs256-f16-f16.cu +5 -0
  389. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-f16.cu +5 -0
  390. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q4_0.cu +5 -0
  391. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q4_1.cu +5 -0
  392. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q5_0.cu +5 -0
  393. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q5_1.cu +5 -0
  394. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q8_0.cu +5 -0
  395. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-f16.cu +5 -0
  396. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q4_0.cu +5 -0
  397. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q4_1.cu +5 -0
  398. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q5_0.cu +5 -0
  399. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q5_1.cu +5 -0
  400. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q8_0.cu +5 -0
  401. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-f16.cu +5 -0
  402. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q4_0.cu +5 -0
  403. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q4_1.cu +5 -0
  404. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q5_0.cu +5 -0
  405. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q5_1.cu +5 -0
  406. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q8_0.cu +5 -0
  407. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-f16.cu +5 -0
  408. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q4_0.cu +5 -0
  409. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q4_1.cu +5 -0
  410. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q5_0.cu +5 -0
  411. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q5_1.cu +5 -0
  412. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q8_0.cu +5 -0
  413. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-f16.cu +5 -0
  414. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q4_0.cu +5 -0
  415. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q4_1.cu +5 -0
  416. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q5_0.cu +5 -0
  417. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q5_1.cu +5 -0
  418. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q8_0.cu +5 -0
  419. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-f16.cu +5 -0
  420. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q4_0.cu +5 -0
  421. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q4_1.cu +5 -0
  422. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q5_0.cu +5 -0
  423. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q5_1.cu +5 -0
  424. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q8_0.cu +5 -0
  425. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-f16.cu +5 -0
  426. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q4_0.cu +5 -0
  427. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q4_1.cu +5 -0
  428. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q5_0.cu +5 -0
  429. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q5_1.cu +5 -0
  430. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q8_0.cu +5 -0
  431. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs256-f16-f16.cu +5 -0
  432. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-f16.cu +5 -0
  433. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q4_0.cu +5 -0
  434. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q4_1.cu +5 -0
  435. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q5_0.cu +5 -0
  436. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q5_1.cu +5 -0
  437. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q8_0.cu +5 -0
  438. data/ext/sources/ggml/src/ggml-cuda/template-instances/generate_cu_files.py +78 -0
  439. data/ext/sources/ggml/src/ggml-cuda/template-instances/mmq-instance-iq1_s.cu +5 -0
  440. data/ext/sources/ggml/src/ggml-cuda/template-instances/mmq-instance-iq2_s.cu +5 -0
  441. data/ext/sources/ggml/src/ggml-cuda/template-instances/mmq-instance-iq2_xs.cu +5 -0
  442. data/ext/sources/ggml/src/ggml-cuda/template-instances/mmq-instance-iq2_xxs.cu +5 -0
  443. data/ext/sources/ggml/src/ggml-cuda/template-instances/mmq-instance-iq3_s.cu +5 -0
  444. data/ext/sources/ggml/src/ggml-cuda/template-instances/mmq-instance-iq3_xxs.cu +5 -0
  445. data/ext/sources/ggml/src/ggml-cuda/template-instances/mmq-instance-iq4_nl.cu +5 -0
  446. data/ext/sources/ggml/src/ggml-cuda/template-instances/mmq-instance-iq4_xs.cu +5 -0
  447. data/ext/sources/ggml/src/ggml-cuda/template-instances/mmq-instance-q2_k.cu +5 -0
  448. data/ext/sources/ggml/src/ggml-cuda/template-instances/mmq-instance-q3_k.cu +5 -0
  449. data/ext/sources/ggml/src/ggml-cuda/template-instances/mmq-instance-q4_0.cu +5 -0
  450. data/ext/sources/ggml/src/ggml-cuda/template-instances/mmq-instance-q4_1.cu +5 -0
  451. data/ext/sources/ggml/src/ggml-cuda/template-instances/mmq-instance-q4_k.cu +5 -0
  452. data/ext/sources/ggml/src/ggml-cuda/template-instances/mmq-instance-q5_0.cu +5 -0
  453. data/ext/sources/ggml/src/ggml-cuda/template-instances/mmq-instance-q5_1.cu +5 -0
  454. data/ext/sources/ggml/src/ggml-cuda/template-instances/mmq-instance-q5_k.cu +5 -0
  455. data/ext/sources/ggml/src/ggml-cuda/template-instances/mmq-instance-q6_k.cu +5 -0
  456. data/ext/sources/ggml/src/ggml-cuda/template-instances/mmq-instance-q8_0.cu +5 -0
  457. data/ext/sources/ggml/src/ggml-cuda/tsembd.cu +47 -0
  458. data/ext/sources/ggml/src/ggml-cuda/tsembd.cuh +5 -0
  459. data/ext/sources/ggml/src/ggml-cuda/unary.cu +378 -0
  460. data/ext/sources/ggml/src/ggml-cuda/unary.cuh +66 -0
  461. data/ext/sources/ggml/src/ggml-cuda/upscale.cu +51 -0
  462. data/ext/sources/ggml/src/ggml-cuda/upscale.cuh +5 -0
  463. data/ext/sources/ggml/src/ggml-cuda/vecdotq.cuh +1135 -0
  464. data/ext/{ggml → sources/ggml}/src/ggml-cuda/vendors/cuda.h +1 -0
  465. data/ext/{ggml → sources/ggml}/src/ggml-cuda/vendors/hip.h +57 -0
  466. data/ext/{ggml → sources/ggml}/src/ggml-cuda/vendors/musa.h +7 -1
  467. data/ext/sources/ggml/src/ggml-cuda/wkv.cu +199 -0
  468. data/ext/sources/ggml/src/ggml-cuda/wkv.cuh +7 -0
  469. data/ext/sources/ggml/src/ggml-hip/CMakeLists.txt +135 -0
  470. data/ext/{ggml → sources/ggml}/src/ggml-impl.h +147 -158
  471. data/ext/sources/ggml/src/ggml-kompute/CMakeLists.txt +166 -0
  472. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/common.comp +112 -0
  473. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_add.comp +58 -0
  474. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_addrow.comp +25 -0
  475. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_cpy_f16_f16.comp +52 -0
  476. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_cpy_f16_f32.comp +52 -0
  477. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_cpy_f32_f16.comp +52 -0
  478. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_cpy_f32_f32.comp +52 -0
  479. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_diagmask.comp +30 -0
  480. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_gelu.comp +22 -0
  481. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_getrows.comp +17 -0
  482. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_getrows_f16.comp +31 -0
  483. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_getrows_f32.comp +31 -0
  484. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_getrows_q4_0.comp +38 -0
  485. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_getrows_q4_1.comp +39 -0
  486. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_getrows_q6_k.comp +44 -0
  487. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_mul.comp +52 -0
  488. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_mul_mat_f16.comp +69 -0
  489. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_mul_mat_mat_f32.comp +51 -0
  490. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_mul_mat_q4_0.comp +33 -0
  491. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_mul_mat_q4_1.comp +35 -0
  492. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_mul_mat_q4_k.comp +140 -0
  493. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_mul_mat_q6_k.comp +106 -0
  494. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_mul_mat_q8_0.comp +73 -0
  495. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_mul_mv_q_n.comp +52 -0
  496. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_mul_mv_q_n_pre.comp +28 -0
  497. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_norm.comp +84 -0
  498. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_relu.comp +21 -0
  499. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_rmsnorm.comp +53 -0
  500. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_rope_neox_f16.comp +52 -0
  501. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_rope_neox_f32.comp +52 -0
  502. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_rope_norm_f16.comp +52 -0
  503. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_rope_norm_f32.comp +52 -0
  504. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_scale.comp +19 -0
  505. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_scale_8.comp +23 -0
  506. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_silu.comp +22 -0
  507. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_softmax.comp +72 -0
  508. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/rope_common.comp +71 -0
  509. data/ext/sources/ggml/src/ggml-metal/CMakeLists.txt +121 -0
  510. data/ext/sources/ggml/src/ggml-metal/ggml-metal-impl.h +649 -0
  511. data/ext/{ggml → sources/ggml}/src/ggml-metal/ggml-metal.m +2504 -1108
  512. data/ext/{ggml → sources/ggml}/src/ggml-metal/ggml-metal.metal +2102 -1463
  513. data/ext/sources/ggml/src/ggml-musa/CMakeLists.txt +113 -0
  514. data/ext/sources/ggml/src/ggml-musa/mudnn.cu +112 -0
  515. data/ext/sources/ggml/src/ggml-musa/mudnn.cuh +12 -0
  516. data/ext/sources/ggml/src/ggml-opencl/CMakeLists.txt +110 -0
  517. data/ext/sources/ggml/src/ggml-opencl/ggml-opencl.cpp +6494 -0
  518. data/ext/sources/ggml/src/ggml-opencl/kernels/add.cl +83 -0
  519. data/ext/sources/ggml/src/ggml-opencl/kernels/argsort.cl +86 -0
  520. data/ext/sources/ggml/src/ggml-opencl/kernels/clamp.cl +20 -0
  521. data/ext/sources/ggml/src/ggml-opencl/kernels/concat.cl +109 -0
  522. data/ext/sources/ggml/src/ggml-opencl/kernels/cpy.cl +184 -0
  523. data/ext/sources/ggml/src/ggml-opencl/kernels/cvt.cl +118 -0
  524. data/ext/sources/ggml/src/ggml-opencl/kernels/diag_mask_inf.cl +58 -0
  525. data/ext/sources/ggml/src/ggml-opencl/kernels/div.cl +72 -0
  526. data/ext/sources/ggml/src/ggml-opencl/kernels/embed_kernel.py +26 -0
  527. data/ext/sources/ggml/src/ggml-opencl/kernels/gelu.cl +62 -0
  528. data/ext/sources/ggml/src/ggml-opencl/kernels/gemv_noshuffle.cl +268 -0
  529. data/ext/sources/ggml/src/ggml-opencl/kernels/gemv_noshuffle_general.cl +274 -0
  530. data/ext/sources/ggml/src/ggml-opencl/kernels/get_rows.cl +163 -0
  531. data/ext/sources/ggml/src/ggml-opencl/kernels/glu.cl +201 -0
  532. data/ext/sources/ggml/src/ggml-opencl/kernels/group_norm.cl +72 -0
  533. data/ext/sources/ggml/src/ggml-opencl/kernels/im2col_f16.cl +57 -0
  534. data/ext/sources/ggml/src/ggml-opencl/kernels/im2col_f32.cl +57 -0
  535. data/ext/sources/ggml/src/ggml-opencl/kernels/mul.cl +79 -0
  536. data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mat_Ab_Bi_8x4.cl +139 -0
  537. data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_f16_f16.cl +118 -0
  538. data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_f16_f32.cl +118 -0
  539. data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_f16_f32_1row.cl +94 -0
  540. data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_f16_f32_l4.cl +84 -0
  541. data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_f32_f32.cl +118 -0
  542. data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_id_q4_0_f32_8x_flat.cl +283 -0
  543. data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_q4_0_f32.cl +192 -0
  544. data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_q4_0_f32_1d_16x_flat.cl +307 -0
  545. data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_q4_0_f32_1d_8x_flat.cl +265 -0
  546. data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_q4_0_f32_8x_flat.cl +272 -0
  547. data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_q4_0_f32_v.cl +254 -0
  548. data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_q6_k.cl +190 -0
  549. data/ext/sources/ggml/src/ggml-opencl/kernels/norm.cl +81 -0
  550. data/ext/sources/ggml/src/ggml-opencl/kernels/pad.cl +30 -0
  551. data/ext/sources/ggml/src/ggml-opencl/kernels/relu.cl +16 -0
  552. data/ext/sources/ggml/src/ggml-opencl/kernels/repeat.cl +39 -0
  553. data/ext/sources/ggml/src/ggml-opencl/kernels/rms_norm.cl +96 -0
  554. data/ext/sources/ggml/src/ggml-opencl/kernels/rope.cl +721 -0
  555. data/ext/sources/ggml/src/ggml-opencl/kernels/scale.cl +16 -0
  556. data/ext/sources/ggml/src/ggml-opencl/kernels/sigmoid.cl +29 -0
  557. data/ext/sources/ggml/src/ggml-opencl/kernels/silu.cl +30 -0
  558. data/ext/sources/ggml/src/ggml-opencl/kernels/softmax_4_f16.cl +87 -0
  559. data/ext/sources/ggml/src/ggml-opencl/kernels/softmax_4_f32.cl +87 -0
  560. data/ext/sources/ggml/src/ggml-opencl/kernels/softmax_f16.cl +86 -0
  561. data/ext/sources/ggml/src/ggml-opencl/kernels/softmax_f32.cl +86 -0
  562. data/ext/sources/ggml/src/ggml-opencl/kernels/sub.cl +72 -0
  563. data/ext/sources/ggml/src/ggml-opencl/kernels/sum_rows.cl +39 -0
  564. data/ext/sources/ggml/src/ggml-opencl/kernels/tanh.cl +63 -0
  565. data/ext/sources/ggml/src/ggml-opencl/kernels/transpose.cl +84 -0
  566. data/ext/sources/ggml/src/ggml-opencl/kernels/tsembd.cl +48 -0
  567. data/ext/sources/ggml/src/ggml-opencl/kernels/upscale.cl +121 -0
  568. data/ext/{ggml → sources/ggml}/src/ggml-opt.cpp +373 -190
  569. data/ext/{ggml → sources/ggml}/src/ggml-quants.c +120 -128
  570. data/ext/sources/ggml/src/ggml-rpc/CMakeLists.txt +9 -0
  571. data/ext/{ggml → sources/ggml}/src/ggml-rpc/ggml-rpc.cpp +494 -84
  572. data/ext/sources/ggml/src/ggml-sycl/CMakeLists.txt +189 -0
  573. data/ext/sources/ggml/src/ggml-sycl/backend.hpp +37 -0
  574. data/ext/sources/ggml/src/ggml-sycl/binbcast.cpp +344 -0
  575. data/ext/sources/ggml/src/ggml-sycl/binbcast.hpp +39 -0
  576. data/ext/{ggml → sources/ggml}/src/ggml-sycl/common.cpp +20 -32
  577. data/ext/sources/ggml/src/ggml-sycl/common.hpp +561 -0
  578. data/ext/{ggml → sources/ggml}/src/ggml-sycl/concat.cpp +56 -70
  579. data/ext/sources/ggml/src/ggml-sycl/concat.hpp +20 -0
  580. data/ext/{ggml → sources/ggml}/src/ggml-sycl/conv.cpp +8 -12
  581. data/ext/sources/ggml/src/ggml-sycl/conv.hpp +20 -0
  582. data/ext/sources/ggml/src/ggml-sycl/convert.cpp +575 -0
  583. data/ext/sources/ggml/src/ggml-sycl/convert.hpp +34 -0
  584. data/ext/sources/ggml/src/ggml-sycl/cpy.cpp +839 -0
  585. data/ext/sources/ggml/src/ggml-sycl/cpy.hpp +11 -0
  586. data/ext/sources/ggml/src/ggml-sycl/dequantize.hpp +823 -0
  587. data/ext/{ggml → sources/ggml}/src/ggml-sycl/dmmv.cpp +188 -67
  588. data/ext/sources/ggml/src/ggml-sycl/dmmv.hpp +27 -0
  589. data/ext/sources/ggml/src/ggml-sycl/dpct/helper.hpp +2987 -0
  590. data/ext/sources/ggml/src/ggml-sycl/element_wise.cpp +1120 -0
  591. data/ext/sources/ggml/src/ggml-sycl/element_wise.hpp +84 -0
  592. data/ext/sources/ggml/src/ggml-sycl/gemm.hpp +102 -0
  593. data/ext/sources/ggml/src/ggml-sycl/getrows.cpp +212 -0
  594. data/ext/sources/ggml/src/ggml-sycl/getrows.hpp +20 -0
  595. data/ext/{ggml → sources/ggml}/src/ggml-sycl/ggml-sycl.cpp +1197 -1295
  596. data/ext/sources/ggml/src/ggml-sycl/gla.cpp +106 -0
  597. data/ext/sources/ggml/src/ggml-sycl/gla.hpp +8 -0
  598. data/ext/sources/ggml/src/ggml-sycl/im2col.cpp +136 -0
  599. data/ext/sources/ggml/src/ggml-sycl/im2col.hpp +21 -0
  600. data/ext/{ggml → sources/ggml}/src/ggml-sycl/mmq.cpp +60 -81
  601. data/ext/sources/ggml/src/ggml-sycl/mmq.hpp +33 -0
  602. data/ext/sources/ggml/src/ggml-sycl/mmvq.cpp +1065 -0
  603. data/ext/sources/ggml/src/ggml-sycl/mmvq.hpp +27 -0
  604. data/ext/sources/ggml/src/ggml-sycl/norm.cpp +482 -0
  605. data/ext/sources/ggml/src/ggml-sycl/norm.hpp +26 -0
  606. data/ext/{ggml → sources/ggml}/src/ggml-sycl/outprod.cpp +8 -17
  607. data/ext/sources/ggml/src/ggml-sycl/outprod.hpp +10 -0
  608. data/ext/sources/ggml/src/ggml-sycl/presets.hpp +74 -0
  609. data/ext/sources/ggml/src/ggml-sycl/quants.hpp +111 -0
  610. data/ext/sources/ggml/src/ggml-sycl/rope.cpp +472 -0
  611. data/ext/sources/ggml/src/ggml-sycl/rope.hpp +20 -0
  612. data/ext/{ggml → sources/ggml}/src/ggml-sycl/softmax.cpp +38 -28
  613. data/ext/sources/ggml/src/ggml-sycl/softmax.hpp +20 -0
  614. data/ext/sources/ggml/src/ggml-sycl/sycl_hw.cpp +15 -0
  615. data/ext/sources/ggml/src/ggml-sycl/sycl_hw.hpp +26 -0
  616. data/ext/{ggml → sources/ggml}/src/ggml-sycl/tsembd.cpp +6 -11
  617. data/ext/sources/ggml/src/ggml-sycl/tsembd.hpp +20 -0
  618. data/ext/sources/ggml/src/ggml-sycl/vecdotq.hpp +1307 -0
  619. data/ext/sources/ggml/src/ggml-sycl/wkv.cpp +289 -0
  620. data/ext/sources/ggml/src/ggml-sycl/wkv.hpp +10 -0
  621. data/ext/sources/ggml/src/ggml-vulkan/CMakeLists.txt +200 -0
  622. data/ext/sources/ggml/src/ggml-vulkan/cmake/host-toolchain.cmake.in +15 -0
  623. data/ext/{ggml → sources/ggml}/src/ggml-vulkan/ggml-vulkan.cpp +3822 -1335
  624. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/CMakeLists.txt +31 -0
  625. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/acc.comp +29 -0
  626. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/add.comp +29 -0
  627. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/argmax.comp +51 -0
  628. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/argsort.comp +69 -0
  629. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/clamp.comp +17 -0
  630. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/concat.comp +41 -0
  631. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/contig_copy.comp +49 -0
  632. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/conv2d_dw.comp +105 -0
  633. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/conv_transpose_1d.comp +98 -0
  634. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/copy.comp +23 -0
  635. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/copy_from_quant.comp +51 -0
  636. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/copy_to_quant.comp +242 -0
  637. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/cos.comp +17 -0
  638. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/count_equal.comp +31 -0
  639. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_f32.comp +20 -0
  640. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_funcs.comp +462 -0
  641. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_funcs_cm2.comp +699 -0
  642. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_head.comp +13 -0
  643. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq1_m.comp +42 -0
  644. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq1_s.comp +35 -0
  645. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq2_s.comp +44 -0
  646. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq2_xs.comp +43 -0
  647. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq2_xxs.comp +48 -0
  648. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq3_s.comp +39 -0
  649. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq3_xxs.comp +49 -0
  650. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq4_nl.comp +32 -0
  651. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq4_xs.comp +34 -0
  652. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q2_k.comp +34 -0
  653. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q3_k.comp +42 -0
  654. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q4_0.comp +30 -0
  655. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q4_1.comp +32 -0
  656. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q4_k.comp +68 -0
  657. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q5_0.comp +34 -0
  658. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q5_1.comp +35 -0
  659. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q5_k.comp +70 -0
  660. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q6_k.comp +33 -0
  661. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q8_0.comp +31 -0
  662. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/diag_mask_inf.comp +34 -0
  663. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/div.comp +27 -0
  664. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/flash_attn.comp +337 -0
  665. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_base.comp +162 -0
  666. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_cm1.comp +360 -0
  667. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_cm2.comp +267 -0
  668. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_split_k_reduce.comp +59 -0
  669. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/geglu.comp +13 -0
  670. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/gelu.comp +25 -0
  671. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/gelu_quick.comp +23 -0
  672. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/generic_binary_head.comp +64 -0
  673. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/generic_head.comp +9 -0
  674. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/generic_unary_head.comp +76 -0
  675. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/get_rows.comp +33 -0
  676. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/get_rows_quant.comp +41 -0
  677. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/glu_head.comp +15 -0
  678. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/glu_main.comp +29 -0
  679. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/group_norm.comp +66 -0
  680. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/im2col.comp +100 -0
  681. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/l2_norm.comp +41 -0
  682. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/leaky_relu.comp +22 -0
  683. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul.comp +27 -0
  684. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_split_k_reduce.comp +48 -0
  685. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec.comp +169 -0
  686. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_base.comp +118 -0
  687. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iq1_m.comp +82 -0
  688. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iq1_s.comp +79 -0
  689. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iq2_s.comp +90 -0
  690. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iq2_xs.comp +87 -0
  691. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iq2_xxs.comp +87 -0
  692. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iq3_s.comp +90 -0
  693. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iq3_xxs.comp +88 -0
  694. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_nc.comp +118 -0
  695. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_p021.comp +154 -0
  696. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_q2_k.comp +130 -0
  697. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_q3_k.comp +132 -0
  698. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_q4_k.comp +136 -0
  699. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_q5_k.comp +167 -0
  700. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_q6_k.comp +130 -0
  701. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mm.comp +868 -0
  702. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mm_cm2.comp +441 -0
  703. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mmq.comp +442 -0
  704. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mmq_funcs.comp +99 -0
  705. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/norm.comp +44 -0
  706. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/opt_step_adamw.comp +42 -0
  707. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/pad.comp +28 -0
  708. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/pool2d.comp +74 -0
  709. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/quantize_q8_1.comp +77 -0
  710. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/reglu.comp +9 -0
  711. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/relu.comp +21 -0
  712. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/repeat.comp +26 -0
  713. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/repeat_back.comp +37 -0
  714. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rms_norm.comp +61 -0
  715. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rms_norm_back.comp +55 -0
  716. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rope_head.comp +58 -0
  717. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rope_multi.comp +60 -0
  718. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rope_neox.comp +43 -0
  719. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rope_norm.comp +43 -0
  720. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rope_vision.comp +47 -0
  721. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/scale.comp +24 -0
  722. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/sigmoid.comp +20 -0
  723. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/silu.comp +22 -0
  724. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/silu_back.comp +26 -0
  725. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/sin.comp +17 -0
  726. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/soft_max.comp +173 -0
  727. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/soft_max_back.comp +50 -0
  728. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/square.comp +17 -0
  729. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/sub.comp +29 -0
  730. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/sum_rows.comp +37 -0
  731. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/swiglu.comp +9 -0
  732. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/tanh.comp +20 -0
  733. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/test_bfloat16_support.comp +7 -0
  734. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/test_coopmat2_support.comp +7 -0
  735. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/test_coopmat_support.comp +7 -0
  736. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/test_integer_dot_support.comp +7 -0
  737. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/timestep_embedding.comp +41 -0
  738. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/types.comp +1373 -0
  739. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/upscale.comp +36 -0
  740. data/ext/{ggml → sources/ggml}/src/ggml-vulkan/vulkan-shaders/vulkan-shaders-gen.cpp +203 -36
  741. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/wkv6.comp +87 -0
  742. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/wkv7.comp +91 -0
  743. data/ext/{ggml → sources/ggml}/src/ggml.c +918 -1782
  744. data/ext/sources/ggml/src/ggml.cpp +26 -0
  745. data/ext/sources/ggml/src/gguf.cpp +1351 -0
  746. data/ext/{include → sources/include}/whisper.h +70 -2
  747. data/ext/sources/src/CMakeLists.txt +145 -0
  748. data/ext/sources/src/coreml/whisper-compat.h +10 -0
  749. data/ext/sources/src/coreml/whisper-compat.m +35 -0
  750. data/ext/{src → sources/src}/coreml/whisper-decoder-impl.h +27 -15
  751. data/ext/{src → sources/src}/coreml/whisper-decoder-impl.m +36 -10
  752. data/ext/{src → sources/src}/coreml/whisper-encoder-impl.h +21 -9
  753. data/ext/{src → sources/src}/coreml/whisper-encoder-impl.m +29 -3
  754. data/ext/sources/src/coreml/whisper-encoder.mm +73 -0
  755. data/ext/sources/src/whisper-arch.h +197 -0
  756. data/ext/{src → sources/src}/whisper.cpp +1966 -386
  757. data/ext/sources/tests/CMakeLists.txt +105 -0
  758. data/ext/sources/tests/earnings21/eval.mk +58 -0
  759. data/ext/sources/tests/earnings21/eval.py +68 -0
  760. data/ext/sources/tests/earnings21/normalizers/__init__.py +2 -0
  761. data/ext/sources/tests/earnings21/normalizers/basic.py +80 -0
  762. data/ext/sources/tests/earnings21/normalizers/english.json +1741 -0
  763. data/ext/sources/tests/earnings21/normalizers/english.py +550 -0
  764. data/ext/sources/tests/earnings21/requirements.txt +6 -0
  765. data/ext/sources/tests/en-0-ref.txt +1 -0
  766. data/ext/sources/tests/en-1-ref.txt +1 -0
  767. data/ext/sources/tests/en-2-ref.txt +1 -0
  768. data/ext/sources/tests/es-0-ref.txt +1 -0
  769. data/ext/sources/tests/librispeech/eval.mk +39 -0
  770. data/ext/sources/tests/librispeech/eval.py +47 -0
  771. data/ext/sources/tests/librispeech/normalizers/__init__.py +2 -0
  772. data/ext/sources/tests/librispeech/normalizers/basic.py +80 -0
  773. data/ext/sources/tests/librispeech/normalizers/english.json +1741 -0
  774. data/ext/sources/tests/librispeech/normalizers/english.py +550 -0
  775. data/ext/sources/tests/librispeech/requirements.txt +6 -0
  776. data/ext/sources/tests/run-tests.sh +130 -0
  777. data/ext/sources/tests/test-c.c +3 -0
  778. data/ext/sources/tests/test-vad-full.cpp +54 -0
  779. data/ext/sources/tests/test-vad.cpp +83 -0
  780. data/ext/sources/tests/test-whisper.js +58 -0
  781. data/extsources.rb +39 -5
  782. data/lib/whisper/context.rb +15 -0
  783. data/lib/whisper/model/uri.rb +202 -126
  784. data/lib/whisper/segment.rb +58 -0
  785. data/sig/whisper.rbs +510 -0
  786. data/test/helper.rb +24 -0
  787. data/{tests → test}/test_callback.rb +45 -3
  788. data/{tests → test}/test_error.rb +2 -2
  789. data/{tests → test}/test_model.rb +47 -0
  790. data/test/test_package.rb +51 -0
  791. data/test/test_params.rb +297 -0
  792. data/test/test_segment.rb +146 -0
  793. data/test/test_vad.rb +19 -0
  794. data/test/test_vad_params.rb +103 -0
  795. data/{tests → test}/test_whisper.rb +106 -36
  796. data/whispercpp.gemspec +5 -5
  797. metadata +837 -134
  798. data/ext/cpu.mk +0 -9
  799. data/ext/examples/dr_wav.h +0 -8815
  800. data/ext/ggml/src/ggml-cann/aclnn_ops.h +0 -592
  801. data/ext/ggml/src/ggml-cpu/ggml-cpu-aarch64.cpp +0 -4262
  802. data/ext/ggml/src/ggml-cpu/ggml-cpu-aarch64.h +0 -8
  803. data/ext/ggml/src/ggml-cpu/ggml-cpu-quants.c +0 -10835
  804. data/ext/ggml/src/ggml-cpu/ggml-cpu.c +0 -14123
  805. data/ext/ggml/src/ggml-cpu/llamafile/sgemm.cpp +0 -1884
  806. data/ext/ggml/src/ggml-cpu/llamafile/sgemm.h +0 -14
  807. data/ext/ggml/src/ggml-metal/ggml-metal-impl.h +0 -288
  808. data/ext/ggml/src/ggml-sycl/convert.cpp +0 -547
  809. data/ext/ggml/src/ggml-sycl/element_wise.cpp +0 -1030
  810. data/ext/ggml/src/ggml-sycl/im2col.cpp +0 -126
  811. data/ext/ggml/src/ggml-sycl/mmvq.cpp +0 -1015
  812. data/ext/ggml/src/ggml-sycl/norm.cpp +0 -378
  813. data/ext/ggml/src/ggml-sycl/rope.cpp +0 -276
  814. data/ext/ggml/src/ggml-sycl/wkv6.cpp +0 -141
  815. data/ext/metal-embed.mk +0 -17
  816. data/ext/metal.mk +0 -6
  817. data/ext/ruby_whisper.cpp +0 -1909
  818. data/ext/scripts/get-flags.mk +0 -38
  819. data/lib/whisper.rb +0 -2
  820. data/tests/helper.rb +0 -7
  821. data/tests/test_package.rb +0 -31
  822. data/tests/test_params.rb +0 -160
  823. data/tests/test_segment.rb +0 -83
  824. /data/ext/{ggml → sources/ggml}/include/ggml-blas.h +0 -0
  825. /data/ext/{ggml → sources/ggml}/include/ggml-cann.h +0 -0
  826. /data/ext/{ggml → sources/ggml}/include/ggml-cuda.h +0 -0
  827. /data/ext/{ggml → sources/ggml}/include/ggml-kompute.h +0 -0
  828. /data/ext/{ggml → sources/ggml}/include/ggml-opencl.h +0 -0
  829. /data/ext/{ggml → sources/ggml}/include/ggml-sycl.h +0 -0
  830. /data/ext/{ggml → sources/ggml}/src/ggml-amx/common.h +0 -0
  831. /data/ext/{ggml → sources/ggml}/src/ggml-amx/ggml-amx.cpp +0 -0
  832. /data/ext/{ggml → sources/ggml}/src/ggml-amx/mmq.cpp +0 -0
  833. /data/ext/{ggml → sources/ggml}/src/ggml-amx/mmq.h +0 -0
  834. /data/ext/{ggml → sources/ggml}/src/ggml-blas/ggml-blas.cpp +0 -0
  835. /data/ext/{ggml → sources/ggml}/src/ggml-cann/kernels/ascendc_kernels.h +0 -0
  836. /data/ext/{ggml → sources/ggml}/src/ggml-cann/kernels/get_row_f16.cpp +0 -0
  837. /data/ext/{ggml → sources/ggml}/src/ggml-cann/kernels/get_row_f32.cpp +0 -0
  838. /data/ext/{ggml → sources/ggml}/src/ggml-cann/kernels/get_row_q4_0.cpp +0 -0
  839. /data/ext/{ggml → sources/ggml}/src/ggml-cann/kernels/get_row_q8_0.cpp +0 -0
  840. /data/ext/{ggml → sources/ggml}/src/ggml-cann/kernels/quantize_f16_q8_0.cpp +0 -0
  841. /data/ext/{ggml → sources/ggml}/src/ggml-cann/kernels/quantize_f32_q8_0.cpp +0 -0
  842. /data/ext/{ggml → sources/ggml}/src/ggml-cann/kernels/quantize_float_to_q4_0.cpp +0 -0
  843. /data/ext/{ggml → sources/ggml}/src/ggml-cpu/amx/amx.h +0 -0
  844. /data/ext/{ggml → sources/ggml}/src/ggml-cpu/amx/common.h +0 -0
  845. /data/ext/{ggml → sources/ggml}/src/ggml-cpu/amx/mmq.h +0 -0
  846. /data/ext/{ggml/src/ggml-cpu/ggml-cpu-hbm.h → sources/ggml/src/ggml-cpu/hbm.h} +0 -0
  847. /data/ext/{ggml/src/ggml-cpu/ggml-cpu-traits.h → sources/ggml/src/ggml-cpu/traits.h} +0 -0
  848. /data/ext/{ggml → sources/ggml}/src/ggml-kompute/ggml-kompute.cpp +0 -0
  849. /data/ext/{ggml → sources/ggml}/src/ggml-quants.h +0 -0
  850. /data/ext/{ggml → sources/ggml}/src/ggml-threading.cpp +0 -0
  851. /data/ext/{ggml → sources/ggml}/src/ggml-threading.h +0 -0
  852. /data/ext/{src → sources/src}/coreml/whisper-encoder.h +0 -0
  853. /data/ext/{src → sources/src}/openvino/whisper-openvino-encoder.cpp +0 -0
  854. /data/ext/{src → sources/src}/openvino/whisper-openvino-encoder.h +0 -0
  855. /data/{tests → test}/jfk_reader/.gitignore +0 -0
  856. /data/{tests → test}/jfk_reader/extconf.rb +0 -0
  857. /data/{tests → test}/jfk_reader/jfk_reader.c +0 -0
@@ -0,0 +1,1303 @@
1
+ #include <ruby.h>
2
+ #include "ruby_whisper.h"
3
+
4
+ #define BOOL_PARAMS_SETTER(self, prop, value) \
5
+ ruby_whisper_params *rwp; \
6
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp); \
7
+ if (value == Qfalse || value == Qnil) { \
8
+ rwp->params.prop = false; \
9
+ } else { \
10
+ rwp->params.prop = true; \
11
+ } \
12
+ return value; \
13
+
14
+ #define BOOL_PARAMS_GETTER(self, prop) \
15
+ ruby_whisper_params *rwp; \
16
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp); \
17
+ if (rwp->params.prop) { \
18
+ return Qtrue; \
19
+ } else { \
20
+ return Qfalse; \
21
+ }
22
+
23
+ #define DEFINE_PARAM(param_name, nth) \
24
+ id_ ## param_name = rb_intern(#param_name); \
25
+ param_names[nth] = id_ ## param_name; \
26
+ rb_define_method(cParams, #param_name, ruby_whisper_params_get_ ## param_name, 0); \
27
+ rb_define_method(cParams, #param_name "=", ruby_whisper_params_set_ ## param_name, 1);
28
+
29
+ #define RUBY_WHISPER_PARAMS_PARAM_NAMES_COUNT 35
30
+
31
+ extern VALUE cParams;
32
+ extern VALUE cVADParams;
33
+
34
+ extern ID id_call;
35
+
36
+ extern VALUE ruby_whisper_normalize_model_path(VALUE model_path);
37
+ extern VALUE rb_whisper_segment_s_new(VALUE context, int index);
38
+ extern const rb_data_type_t ruby_whisper_vad_params_type;
39
+
40
+ static ID param_names[RUBY_WHISPER_PARAMS_PARAM_NAMES_COUNT];
41
+ static ID id_language;
42
+ static ID id_translate;
43
+ static ID id_no_context;
44
+ static ID id_single_segment;
45
+ static ID id_print_special;
46
+ static ID id_print_progress;
47
+ static ID id_print_realtime;
48
+ static ID id_print_timestamps;
49
+ static ID id_suppress_blank;
50
+ static ID id_suppress_nst;
51
+ static ID id_token_timestamps;
52
+ static ID id_split_on_word;
53
+ static ID id_initial_prompt;
54
+ static ID id_diarize;
55
+ static ID id_offset;
56
+ static ID id_duration;
57
+ static ID id_max_text_tokens;
58
+ static ID id_temperature;
59
+ static ID id_max_initial_ts;
60
+ static ID id_length_penalty;
61
+ static ID id_temperature_inc;
62
+ static ID id_entropy_thold;
63
+ static ID id_logprob_thold;
64
+ static ID id_no_speech_thold;
65
+ static ID id_new_segment_callback;
66
+ static ID id_new_segment_callback_user_data;
67
+ static ID id_progress_callback;
68
+ static ID id_progress_callback_user_data;
69
+ static ID id_encoder_begin_callback;
70
+ static ID id_encoder_begin_callback_user_data;
71
+ static ID id_abort_callback;
72
+ static ID id_abort_callback_user_data;
73
+ static ID id_vad;
74
+ static ID id_vad_model_path;
75
+ static ID id_vad_params;
76
+
77
+ static void
78
+ rb_whisper_callbcack_container_mark(ruby_whisper_callback_container *rwc)
79
+ {
80
+ if (rwc == NULL) return;
81
+
82
+ rb_gc_mark(rwc->user_data);
83
+ rb_gc_mark(rwc->callback);
84
+ rb_gc_mark(rwc->callbacks);
85
+ }
86
+
87
+ static ruby_whisper_callback_container*
88
+ rb_whisper_callback_container_allocate() {
89
+ ruby_whisper_callback_container *container;
90
+ container = ALLOC(ruby_whisper_callback_container);
91
+ container->context = NULL;
92
+ container->user_data = Qnil;
93
+ container->callback = Qnil;
94
+ container->callbacks = rb_ary_new();
95
+ return container;
96
+ }
97
+
98
+ static void new_segment_callback(struct whisper_context *ctx, struct whisper_state *state, int n_new, void *user_data) {
99
+ const ruby_whisper_callback_container *container = (ruby_whisper_callback_container *)user_data;
100
+
101
+ // Currently, doesn't support state because
102
+ // those require to resolve GC-related problems.
103
+ if (!NIL_P(container->callback)) {
104
+ rb_funcall(container->callback, id_call, 4, *container->context, Qnil, INT2NUM(n_new), container->user_data);
105
+ }
106
+ const long callbacks_len = RARRAY_LEN(container->callbacks);
107
+ if (0 == callbacks_len) {
108
+ return;
109
+ }
110
+ const int n_segments = whisper_full_n_segments_from_state(state);
111
+ for (int i = n_new; i > 0; i--) {
112
+ int i_segment = n_segments - i;
113
+ VALUE segment = rb_whisper_segment_s_new(*container->context, i_segment);
114
+ for (int j = 0; j < callbacks_len; j++) {
115
+ VALUE cb = rb_ary_entry(container->callbacks, j);
116
+ rb_funcall(cb, id_call, 1, segment);
117
+ }
118
+ }
119
+ }
120
+
121
+ static void progress_callback(struct whisper_context *ctx, struct whisper_state *state, int progress_cur, void *user_data) {
122
+ const ruby_whisper_callback_container *container = (ruby_whisper_callback_container *)user_data;
123
+ const VALUE progress = INT2NUM(progress_cur);
124
+ // Currently, doesn't support state because
125
+ // those require to resolve GC-related problems.
126
+ if (!NIL_P(container->callback)) {
127
+ rb_funcall(container->callback, id_call, 4, *container->context, Qnil, progress, container->user_data);
128
+ }
129
+ const long callbacks_len = RARRAY_LEN(container->callbacks);
130
+ if (0 == callbacks_len) {
131
+ return;
132
+ }
133
+ for (int j = 0; j < callbacks_len; j++) {
134
+ VALUE cb = rb_ary_entry(container->callbacks, j);
135
+ rb_funcall(cb, id_call, 1, progress);
136
+ }
137
+ }
138
+
139
+ static bool encoder_begin_callback(struct whisper_context *ctx, struct whisper_state *state, void *user_data) {
140
+ const ruby_whisper_callback_container *container = (ruby_whisper_callback_container *)user_data;
141
+ bool is_aborted = false;
142
+ VALUE result;
143
+
144
+ // Currently, doesn't support state because
145
+ // those require to resolve GC-related problems.
146
+ if (!NIL_P(container->callback)) {
147
+ result = rb_funcall(container->callback, id_call, 3, *container->context, Qnil, container->user_data);
148
+ if (result == Qfalse) {
149
+ is_aborted = true;
150
+ }
151
+ }
152
+ const long callbacks_len = RARRAY_LEN(container->callbacks);
153
+ if (0 == callbacks_len) {
154
+ return !is_aborted;
155
+ }
156
+ for (int j = 0; j < callbacks_len; j++) {
157
+ VALUE cb = rb_ary_entry(container->callbacks, j);
158
+ result = rb_funcall(cb, id_call, 0);
159
+ if (result == Qfalse) {
160
+ is_aborted = true;
161
+ }
162
+ }
163
+ return !is_aborted;
164
+ }
165
+
166
+ static bool abort_callback(void * user_data) {
167
+ const ruby_whisper_callback_container *container = (ruby_whisper_callback_container *)user_data;
168
+ if (!NIL_P(container->callback)) {
169
+ VALUE result = rb_funcall(container->callback, id_call, 1, container->user_data);
170
+ if (!NIL_P(result) && Qfalse != result) {
171
+ return true;
172
+ }
173
+ }
174
+ const long callbacks_len = RARRAY_LEN(container->callbacks);
175
+ if (0 == callbacks_len) {
176
+ return false;
177
+ }
178
+ for (int j = 0; j < callbacks_len; j++) {
179
+ VALUE cb = rb_ary_entry(container->callbacks, j);
180
+ VALUE result = rb_funcall(cb, id_call, 1, container->user_data);
181
+ if (!NIL_P(result) && Qfalse != result) {
182
+ return true;
183
+ }
184
+ }
185
+ return false;
186
+ }
187
+
188
+ static void register_callbacks(ruby_whisper_params * rwp, VALUE * context) {
189
+ if (!NIL_P(rwp->new_segment_callback_container->callback) || 0 != RARRAY_LEN(rwp->new_segment_callback_container->callbacks)) {
190
+ rwp->new_segment_callback_container->context = context;
191
+ rwp->params.new_segment_callback = new_segment_callback;
192
+ rwp->params.new_segment_callback_user_data = rwp->new_segment_callback_container;
193
+ }
194
+
195
+ if (!NIL_P(rwp->progress_callback_container->callback) || 0 != RARRAY_LEN(rwp->progress_callback_container->callbacks)) {
196
+ rwp->progress_callback_container->context = context;
197
+ rwp->params.progress_callback = progress_callback;
198
+ rwp->params.progress_callback_user_data = rwp->progress_callback_container;
199
+ }
200
+
201
+ if (!NIL_P(rwp->encoder_begin_callback_container->callback) || 0 != RARRAY_LEN(rwp->encoder_begin_callback_container->callbacks)) {
202
+ rwp->encoder_begin_callback_container->context = context;
203
+ rwp->params.encoder_begin_callback = encoder_begin_callback;
204
+ rwp->params.encoder_begin_callback_user_data = rwp->encoder_begin_callback_container;
205
+ }
206
+
207
+ if (!NIL_P(rwp->abort_callback_container->callback) || 0 != RARRAY_LEN(rwp->abort_callback_container->callbacks)) {
208
+ rwp->abort_callback_container->context = context;
209
+ rwp->params.abort_callback = abort_callback;
210
+ rwp->params.abort_callback_user_data = rwp->abort_callback_container;
211
+ }
212
+ }
213
+
214
+ static void set_vad_params(ruby_whisper_params *rwp)
215
+ {
216
+ ruby_whisper_vad_params * rwvp;
217
+ TypedData_Get_Struct(rwp->vad_params, ruby_whisper_vad_params, &ruby_whisper_vad_params_type, rwvp);
218
+ rwp->params.vad_params = rwvp->params;
219
+ }
220
+
221
+ void
222
+ prepare_transcription(ruby_whisper_params *rwp, VALUE *context)
223
+ {
224
+ register_callbacks(rwp, context);
225
+ set_vad_params(rwp);
226
+ }
227
+
228
+ void
229
+ rb_whisper_params_mark(void *p)
230
+ {
231
+ ruby_whisper_params *rwp = (ruby_whisper_params *)p;
232
+ rb_whisper_callbcack_container_mark(rwp->new_segment_callback_container);
233
+ rb_whisper_callbcack_container_mark(rwp->progress_callback_container);
234
+ rb_whisper_callbcack_container_mark(rwp->encoder_begin_callback_container);
235
+ rb_whisper_callbcack_container_mark(rwp->abort_callback_container);
236
+ rb_gc_mark(rwp->vad_params);
237
+ }
238
+
239
+ void
240
+ ruby_whisper_params_free(ruby_whisper_params *rwp)
241
+ {
242
+ }
243
+
244
+ void
245
+ rb_whisper_params_free(void *p)
246
+ {
247
+ ruby_whisper_params *rwp = (ruby_whisper_params *)p;
248
+ // How to free user_data and callback only when not referred to by others?
249
+ ruby_whisper_params_free(rwp);
250
+ free(rwp);
251
+ }
252
+
253
+ static size_t
254
+ ruby_whisper_params_memsize(const void *p)
255
+ {
256
+ const ruby_whisper_params *rwp = (const ruby_whisper_params *)p;
257
+
258
+ return sizeof(ruby_whisper_params) + sizeof(rwp->params) + sizeof(rwp->vad_params);
259
+ }
260
+
261
+ const rb_data_type_t ruby_whisper_params_type = {
262
+ "ruby_whisper_params",
263
+ {
264
+ rb_whisper_params_mark,
265
+ rb_whisper_params_free,
266
+ ruby_whisper_params_memsize,
267
+ },
268
+ 0, 0,
269
+ 0
270
+ };
271
+
272
+ static VALUE
273
+ ruby_whisper_params_allocate(VALUE klass)
274
+ {
275
+ ruby_whisper_params *rwp;
276
+ VALUE obj = TypedData_Make_Struct(klass, ruby_whisper_params, &ruby_whisper_params_type, rwp);
277
+ rwp->params = whisper_full_default_params(WHISPER_SAMPLING_GREEDY);
278
+ rwp->diarize = false;
279
+ rwp->vad_params = TypedData_Wrap_Struct(cVADParams, &ruby_whisper_vad_params_type, (void *)&rwp->params.vad_params);
280
+ rwp->new_segment_callback_container = rb_whisper_callback_container_allocate();
281
+ rwp->progress_callback_container = rb_whisper_callback_container_allocate();
282
+ rwp->encoder_begin_callback_container = rb_whisper_callback_container_allocate();
283
+ rwp->abort_callback_container = rb_whisper_callback_container_allocate();
284
+ return obj;
285
+ }
286
+
287
+ /*
288
+ * params.language = "auto" | "en", etc...
289
+ *
290
+ * call-seq:
291
+ * language = lang_name -> lang_name
292
+ */
293
+ static VALUE
294
+ ruby_whisper_params_set_language(VALUE self, VALUE value)
295
+ {
296
+ ruby_whisper_params *rwp;
297
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
298
+ if (value == Qfalse || value == Qnil) {
299
+ rwp->params.language = "auto";
300
+ } else {
301
+ rwp->params.language = StringValueCStr(value);
302
+ }
303
+ return value;
304
+ }
305
+ /*
306
+ * call-seq:
307
+ * language -> String
308
+ */
309
+ static VALUE
310
+ ruby_whisper_params_get_language(VALUE self)
311
+ {
312
+ ruby_whisper_params *rwp;
313
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
314
+ if (rwp->params.language) {
315
+ return rb_str_new2(rwp->params.language);
316
+ } else {
317
+ return rb_str_new2("auto");
318
+ }
319
+ }
320
+ /*
321
+ * call-seq:
322
+ * translate = do_translate -> do_translate
323
+ */
324
+ static VALUE
325
+ ruby_whisper_params_set_translate(VALUE self, VALUE value)
326
+ {
327
+ BOOL_PARAMS_SETTER(self, translate, value)
328
+ }
329
+ /*
330
+ * call-seq:
331
+ * translate -> bool
332
+ */
333
+ static VALUE
334
+ ruby_whisper_params_get_translate(VALUE self)
335
+ {
336
+ BOOL_PARAMS_GETTER(self, translate)
337
+ }
338
+ /*
339
+ * call-seq:
340
+ * no_context = dont_use_context -> dont_use_context
341
+ */
342
+ static VALUE
343
+ ruby_whisper_params_set_no_context(VALUE self, VALUE value)
344
+ {
345
+ BOOL_PARAMS_SETTER(self, no_context, value)
346
+ }
347
+ /*
348
+ * If true, does not use past transcription (if any) as initial prompt for the decoder.
349
+ *
350
+ * call-seq:
351
+ * no_context -> bool
352
+ */
353
+ static VALUE
354
+ ruby_whisper_params_get_no_context(VALUE self)
355
+ {
356
+ BOOL_PARAMS_GETTER(self, no_context)
357
+ }
358
+ /*
359
+ * call-seq:
360
+ * single_segment = force_single -> force_single
361
+ */
362
+ static VALUE
363
+ ruby_whisper_params_set_single_segment(VALUE self, VALUE value)
364
+ {
365
+ BOOL_PARAMS_SETTER(self, single_segment, value)
366
+ }
367
+ /*
368
+ * If true, forces single segment output (useful for streaming).
369
+ *
370
+ * call-seq:
371
+ * single_segment -> bool
372
+ */
373
+ static VALUE
374
+ ruby_whisper_params_get_single_segment(VALUE self)
375
+ {
376
+ BOOL_PARAMS_GETTER(self, single_segment)
377
+ }
378
+ /*
379
+ * call-seq:
380
+ * print_special = force_print -> force_print
381
+ */
382
+ static VALUE
383
+ ruby_whisper_params_set_print_special(VALUE self, VALUE value)
384
+ {
385
+ BOOL_PARAMS_SETTER(self, print_special, value)
386
+ }
387
+ /*
388
+ * If true, prints special tokens (e.g. <SOT>, <EOT>, <BEG>, etc.).
389
+ *
390
+ * call-seq:
391
+ * print_special -> bool
392
+ */
393
+ static VALUE
394
+ ruby_whisper_params_get_print_special(VALUE self)
395
+ {
396
+ BOOL_PARAMS_GETTER(self, print_special)
397
+ }
398
+ /*
399
+ * call-seq:
400
+ * print_progress = force_print -> force_print
401
+ */
402
+ static VALUE
403
+ ruby_whisper_params_set_print_progress(VALUE self, VALUE value)
404
+ {
405
+ BOOL_PARAMS_SETTER(self, print_progress, value)
406
+ }
407
+ /*
408
+ * If true, prints progress information.
409
+ *
410
+ * call-seq:
411
+ * print_progress -> bool
412
+ */
413
+ static VALUE
414
+ ruby_whisper_params_get_print_progress(VALUE self)
415
+ {
416
+ BOOL_PARAMS_GETTER(self, print_progress)
417
+ }
418
+ /*
419
+ * call-seq:
420
+ * print_realtime = force_print -> force_print
421
+ */
422
+ static VALUE
423
+ ruby_whisper_params_set_print_realtime(VALUE self, VALUE value)
424
+ {
425
+ BOOL_PARAMS_SETTER(self, print_realtime, value)
426
+ }
427
+ /*
428
+ * If true, prints results from within whisper.cpp. (avoid it, use callback instead)
429
+ * call-seq:
430
+ * print_realtime -> bool
431
+ */
432
+ static VALUE
433
+ ruby_whisper_params_get_print_realtime(VALUE self)
434
+ {
435
+ BOOL_PARAMS_GETTER(self, print_realtime)
436
+ }
437
+ /*
438
+ * call-seq:
439
+ * print_timestamps = force_print -> force_print
440
+ */
441
+ static VALUE
442
+ ruby_whisper_params_set_print_timestamps(VALUE self, VALUE value)
443
+ {
444
+ BOOL_PARAMS_SETTER(self, print_timestamps, value)
445
+ }
446
+ /*
447
+ * If true, prints timestamps for each text segment when printing realtime.
448
+ *
449
+ * call-seq:
450
+ * print_timestamps -> bool
451
+ */
452
+ static VALUE
453
+ ruby_whisper_params_get_print_timestamps(VALUE self)
454
+ {
455
+ BOOL_PARAMS_GETTER(self, print_timestamps)
456
+ }
457
+ /*
458
+ * call-seq:
459
+ * suppress_blank = force_suppress -> force_suppress
460
+ */
461
+ static VALUE
462
+ ruby_whisper_params_set_suppress_blank(VALUE self, VALUE value)
463
+ {
464
+ BOOL_PARAMS_SETTER(self, suppress_blank, value)
465
+ }
466
+ /*
467
+ * If true, suppresses blank outputs.
468
+ *
469
+ * call-seq:
470
+ * suppress_blank -> bool
471
+ */
472
+ static VALUE
473
+ ruby_whisper_params_get_suppress_blank(VALUE self)
474
+ {
475
+ BOOL_PARAMS_GETTER(self, suppress_blank)
476
+ }
477
+ /*
478
+ * call-seq:
479
+ * suppress_nst = force_suppress -> force_suppress
480
+ */
481
+ static VALUE
482
+ ruby_whisper_params_set_suppress_nst(VALUE self, VALUE value)
483
+ {
484
+ BOOL_PARAMS_SETTER(self, suppress_nst, value)
485
+ }
486
+ /*
487
+ * If true, suppresses non-speech-tokens.
488
+ *
489
+ * call-seq:
490
+ * suppress_nst -> bool
491
+ */
492
+ static VALUE
493
+ ruby_whisper_params_get_suppress_nst(VALUE self)
494
+ {
495
+ BOOL_PARAMS_GETTER(self, suppress_nst)
496
+ }
497
+ /*
498
+ * If true, enables token-level timestamps.
499
+ *
500
+ * call-seq:
501
+ * token_timestamps -> bool
502
+ */
503
+ static VALUE
504
+ ruby_whisper_params_get_token_timestamps(VALUE self)
505
+ {
506
+ BOOL_PARAMS_GETTER(self, token_timestamps)
507
+ }
508
+ /*
509
+ * call-seq:
510
+ * token_timestamps = force_timestamps -> force_timestamps
511
+ */
512
+ static VALUE
513
+ ruby_whisper_params_set_token_timestamps(VALUE self, VALUE value)
514
+ {
515
+ BOOL_PARAMS_SETTER(self, token_timestamps, value)
516
+ }
517
+ /*
518
+ * If true, split on word rather than on token (when used with max_len).
519
+ *
520
+ * call-seq:
521
+ * translate -> bool
522
+ */
523
+ static VALUE
524
+ ruby_whisper_params_get_split_on_word(VALUE self)
525
+ {
526
+ BOOL_PARAMS_GETTER(self, split_on_word)
527
+ }
528
+ /*
529
+ * call-seq:
530
+ * split_on_word = force_split -> force_split
531
+ */
532
+ static VALUE
533
+ ruby_whisper_params_set_split_on_word(VALUE self, VALUE value)
534
+ {
535
+ BOOL_PARAMS_SETTER(self, split_on_word, value)
536
+ }
537
+ /*
538
+ * Tokens to provide to the whisper decoder as initial prompt
539
+ * these are prepended to any existing text context from a previous call
540
+ * use whisper_tokenize() to convert text to tokens.
541
+ * Maximum of whisper_n_text_ctx()/2 tokens are used (typically 224).
542
+ *
543
+ * call-seq:
544
+ * initial_prompt -> String
545
+ */
546
+ static VALUE
547
+ ruby_whisper_params_get_initial_prompt(VALUE self)
548
+ {
549
+ ruby_whisper_params *rwp;
550
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
551
+ return rwp->params.initial_prompt == NULL ? Qnil : rb_str_new2(rwp->params.initial_prompt);
552
+ }
553
+ /*
554
+ * call-seq:
555
+ * initial_prompt = prompt -> prompt
556
+ */
557
+ static VALUE
558
+ ruby_whisper_params_set_initial_prompt(VALUE self, VALUE value)
559
+ {
560
+ ruby_whisper_params *rwp;
561
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
562
+ rwp->params.initial_prompt = StringValueCStr(value);
563
+ return value;
564
+ }
565
+ /*
566
+ * If true, enables diarization.
567
+ *
568
+ * call-seq:
569
+ * diarize -> bool
570
+ */
571
+ static VALUE
572
+ ruby_whisper_params_get_diarize(VALUE self)
573
+ {
574
+ ruby_whisper_params *rwp;
575
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
576
+ if (rwp->diarize) {
577
+ return Qtrue;
578
+ } else {
579
+ return Qfalse;
580
+ }
581
+ }
582
+ /*
583
+ * call-seq:
584
+ * diarize = force_diarize -> force_diarize
585
+ */
586
+ static VALUE
587
+ ruby_whisper_params_set_diarize(VALUE self, VALUE value)
588
+ {
589
+ ruby_whisper_params *rwp;
590
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
591
+ if (value == Qfalse || value == Qnil) {
592
+ rwp->diarize = false;
593
+ } else {
594
+ rwp->diarize = true;
595
+ } \
596
+ return value;
597
+ }
598
+
599
+ /*
600
+ * Start offset in ms.
601
+ *
602
+ * call-seq:
603
+ * offset -> Integer
604
+ */
605
+ static VALUE
606
+ ruby_whisper_params_get_offset(VALUE self)
607
+ {
608
+ ruby_whisper_params *rwp;
609
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
610
+ return INT2NUM(rwp->params.offset_ms);
611
+ }
612
+ /*
613
+ * call-seq:
614
+ * offset = offset_ms -> offset_ms
615
+ */
616
+ static VALUE
617
+ ruby_whisper_params_set_offset(VALUE self, VALUE value)
618
+ {
619
+ ruby_whisper_params *rwp;
620
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
621
+ rwp->params.offset_ms = NUM2INT(value);
622
+ return value;
623
+ }
624
+ /*
625
+ * Audio duration to process in ms.
626
+ *
627
+ * call-seq:
628
+ * duration -> Integer
629
+ */
630
+ static VALUE
631
+ ruby_whisper_params_get_duration(VALUE self)
632
+ {
633
+ ruby_whisper_params *rwp;
634
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
635
+ return INT2NUM(rwp->params.duration_ms);
636
+ }
637
+ /*
638
+ * call-seq:
639
+ * duration = duration_ms -> duration_ms
640
+ */
641
+ static VALUE
642
+ ruby_whisper_params_set_duration(VALUE self, VALUE value)
643
+ {
644
+ ruby_whisper_params *rwp;
645
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
646
+ rwp->params.duration_ms = NUM2INT(value);
647
+ return value;
648
+ }
649
+
650
+ /*
651
+ * Max tokens to use from past text as prompt for the decoder.
652
+ *
653
+ * call-seq:
654
+ * max_text_tokens -> Integer
655
+ */
656
+ static VALUE
657
+ ruby_whisper_params_get_max_text_tokens(VALUE self)
658
+ {
659
+ ruby_whisper_params *rwp;
660
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
661
+ return INT2NUM(rwp->params.n_max_text_ctx);
662
+ }
663
+ /*
664
+ * call-seq:
665
+ * max_text_tokens = n_tokens -> n_tokens
666
+ */
667
+ static VALUE
668
+ ruby_whisper_params_set_max_text_tokens(VALUE self, VALUE value)
669
+ {
670
+ ruby_whisper_params *rwp;
671
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
672
+ rwp->params.n_max_text_ctx = NUM2INT(value);
673
+ return value;
674
+ }
675
+ /*
676
+ * call-seq:
677
+ * temperature -> Float
678
+ */
679
+ static VALUE
680
+ ruby_whisper_params_get_temperature(VALUE self)
681
+ {
682
+ ruby_whisper_params *rwp;
683
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
684
+ return DBL2NUM(rwp->params.temperature);
685
+ }
686
+ /*
687
+ * call-seq:
688
+ * temperature = temp -> temp
689
+ */
690
+ static VALUE
691
+ ruby_whisper_params_set_temperature(VALUE self, VALUE value)
692
+ {
693
+ ruby_whisper_params *rwp;
694
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
695
+ rwp->params.temperature = RFLOAT_VALUE(value);
696
+ return value;
697
+ }
698
+ /*
699
+ * See https://github.com/openai/whisper/blob/f82bc59f5ea234d4b97fb2860842ed38519f7e65/whisper/decoding.py#L97
700
+ *
701
+ * call-seq:
702
+ * max_initial_ts -> Flaot
703
+ */
704
+ static VALUE
705
+ ruby_whisper_params_get_max_initial_ts(VALUE self)
706
+ {
707
+ ruby_whisper_params *rwp;
708
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
709
+ return DBL2NUM(rwp->params.max_initial_ts);
710
+ }
711
+ /*
712
+ * call-seq:
713
+ * max_initial_ts = timestamp -> timestamp
714
+ */
715
+ static VALUE
716
+ ruby_whisper_params_set_max_initial_ts(VALUE self, VALUE value)
717
+ {
718
+ ruby_whisper_params *rwp;
719
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
720
+ rwp->params.max_initial_ts = RFLOAT_VALUE(value);
721
+ return value;
722
+ }
723
+ /*
724
+ * call-seq:
725
+ * length_penalty -> Float
726
+ */
727
+ static VALUE
728
+ ruby_whisper_params_get_length_penalty(VALUE self)
729
+ {
730
+ ruby_whisper_params *rwp;
731
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
732
+ return DBL2NUM(rwp->params.length_penalty);
733
+ }
734
+ /*
735
+ * call-seq:
736
+ * length_penalty = penalty -> penalty
737
+ */
738
+ static VALUE
739
+ ruby_whisper_params_set_length_penalty(VALUE self, VALUE value)
740
+ {
741
+ ruby_whisper_params *rwp;
742
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
743
+ rwp->params.length_penalty = RFLOAT_VALUE(value);
744
+ return value;
745
+ }
746
+ /*
747
+ * call-seq:
748
+ * temperature_inc -> Float
749
+ */
750
+ static VALUE
751
+ ruby_whisper_params_get_temperature_inc(VALUE self)
752
+ {
753
+ ruby_whisper_params *rwp;
754
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
755
+ return DBL2NUM(rwp->params.temperature_inc);
756
+ }
757
+ /*
758
+ * call-seq:
759
+ * temperature_inc = inc -> inc
760
+ */
761
+ static VALUE
762
+ ruby_whisper_params_set_temperature_inc(VALUE self, VALUE value)
763
+ {
764
+ ruby_whisper_params *rwp;
765
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
766
+ rwp->params.temperature_inc = RFLOAT_VALUE(value);
767
+ return value;
768
+ }
769
+ /*
770
+ * Similar to OpenAI's "compression_ratio_threshold"
771
+ *
772
+ * call-seq:
773
+ * entropy_thold -> Float
774
+ */
775
+ static VALUE
776
+ ruby_whisper_params_get_entropy_thold(VALUE self)
777
+ {
778
+ ruby_whisper_params *rwp;
779
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
780
+ return DBL2NUM(rwp->params.entropy_thold);
781
+ }
782
+ /*
783
+ * call-seq:
784
+ * entropy_thold = threshold -> threshold
785
+ */
786
+ static VALUE
787
+ ruby_whisper_params_set_entropy_thold(VALUE self, VALUE value)
788
+ {
789
+ ruby_whisper_params *rwp;
790
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
791
+ rwp->params.entropy_thold = RFLOAT_VALUE(value);
792
+ return value;
793
+ }
794
+ /*
795
+ * call-seq:
796
+ * logprob_thold -> Float
797
+ */
798
+ static VALUE
799
+ ruby_whisper_params_get_logprob_thold(VALUE self)
800
+ {
801
+ ruby_whisper_params *rwp;
802
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
803
+ return DBL2NUM(rwp->params.logprob_thold);
804
+ }
805
+ /*
806
+ * call-seq:
807
+ * logprob_thold = threshold -> threshold
808
+ */
809
+ static VALUE
810
+ ruby_whisper_params_set_logprob_thold(VALUE self, VALUE value)
811
+ {
812
+ ruby_whisper_params *rwp;
813
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
814
+ rwp->params.logprob_thold = RFLOAT_VALUE(value);
815
+ return value;
816
+ }
817
+ /*
818
+ * call-seq:
819
+ * no_speech_thold -> Float
820
+ */
821
+ static VALUE
822
+ ruby_whisper_params_get_no_speech_thold(VALUE self)
823
+ {
824
+ ruby_whisper_params *rwp;
825
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
826
+ return DBL2NUM(rwp->params.no_speech_thold);
827
+ }
828
+ /*
829
+ * call-seq:
830
+ * no_speech_thold = threshold -> threshold
831
+ */
832
+ static VALUE
833
+ ruby_whisper_params_set_no_speech_thold(VALUE self, VALUE value)
834
+ {
835
+ ruby_whisper_params *rwp;
836
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
837
+ rwp->params.no_speech_thold = RFLOAT_VALUE(value);
838
+ return value;
839
+ }
840
+ static VALUE
841
+ ruby_whisper_params_get_new_segment_callback(VALUE self)
842
+ {
843
+ ruby_whisper_params *rwp;
844
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
845
+ return rwp->new_segment_callback_container->callback;
846
+ }
847
+ /*
848
+ * Sets new segment callback, called for every newly generated text segment.
849
+ *
850
+ * params.new_segment_callback = ->(context, _, n_new, user_data) {
851
+ * # ...
852
+ * }
853
+ *
854
+ * call-seq:
855
+ * new_segment_callback = callback -> callback
856
+ */
857
+ static VALUE
858
+ ruby_whisper_params_set_new_segment_callback(VALUE self, VALUE value)
859
+ {
860
+ ruby_whisper_params *rwp;
861
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
862
+ rwp->new_segment_callback_container->callback = value;
863
+ return value;
864
+ }
865
+ static VALUE
866
+ ruby_whisper_params_get_new_segment_callback_user_data(VALUE self)
867
+ {
868
+ ruby_whisper_params *rwp;
869
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
870
+ return rwp->new_segment_callback_container->user_data;
871
+ }
872
+ /*
873
+ * Sets user data passed to the last argument of new segment callback.
874
+ *
875
+ * call-seq:
876
+ * new_segment_callback_user_data = user_data -> use_data
877
+ */
878
+ static VALUE
879
+ ruby_whisper_params_set_new_segment_callback_user_data(VALUE self, VALUE value)
880
+ {
881
+ ruby_whisper_params *rwp;
882
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
883
+ rwp->new_segment_callback_container->user_data = value;
884
+ return value;
885
+ }
886
+ static VALUE
887
+ ruby_whisper_params_get_progress_callback(VALUE self)
888
+ {
889
+ ruby_whisper_params *rwp;
890
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
891
+ return rwp->progress_callback_container->callback;
892
+ }
893
+ /*
894
+ * Sets progress callback, called on each progress update.
895
+ *
896
+ * params.new_segment_callback = ->(context, _, progress, user_data) {
897
+ * # ...
898
+ * }
899
+ *
900
+ * +progress+ is an Integer between 0 and 100.
901
+ *
902
+ * call-seq:
903
+ * progress_callback = callback -> callback
904
+ */
905
+ static VALUE
906
+ ruby_whisper_params_set_progress_callback(VALUE self, VALUE value)
907
+ {
908
+ ruby_whisper_params *rwp;
909
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
910
+ rwp->progress_callback_container->callback = value;
911
+ return value;
912
+ }
913
+ static VALUE
914
+ ruby_whisper_params_get_progress_callback_user_data(VALUE self)
915
+ {
916
+ ruby_whisper_params *rwp;
917
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
918
+ return rwp->progress_callback_container->user_data;
919
+ }
920
+ /*
921
+ * Sets user data passed to the last argument of progress callback.
922
+ *
923
+ * call-seq:
924
+ * progress_callback_user_data = user_data -> use_data
925
+ */
926
+ static VALUE
927
+ ruby_whisper_params_set_progress_callback_user_data(VALUE self, VALUE value)
928
+ {
929
+ ruby_whisper_params *rwp;
930
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
931
+ rwp->progress_callback_container->user_data = value;
932
+ return value;
933
+ }
934
+
935
+ static VALUE
936
+ ruby_whisper_params_get_encoder_begin_callback(VALUE self)
937
+ {
938
+ ruby_whisper_params *rwp;
939
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
940
+ return rwp->encoder_begin_callback_container->callback;
941
+ }
942
+
943
+ /*
944
+ * Sets encoder begin callback, called when the encoder starts.
945
+ *
946
+ * params.encoder_begin_callback = ->(context, _, user_data) {
947
+ * # ...
948
+ * }
949
+ *
950
+ * call-seq:
951
+ * encoder_begin_callback = callback -> callback
952
+ */
953
+ static VALUE
954
+ ruby_whisper_params_set_encoder_begin_callback(VALUE self, VALUE value)
955
+ {
956
+ ruby_whisper_params *rwp;
957
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
958
+ rwp->encoder_begin_callback_container->callback = value;
959
+ return value;
960
+ }
961
+
962
+ static VALUE
963
+ ruby_whisper_params_get_encoder_begin_callback_user_data(VALUE self)
964
+ {
965
+ ruby_whisper_params *rwp;
966
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
967
+ return rwp->encoder_begin_callback_container->user_data;
968
+ }
969
+
970
+ /*
971
+ * Sets user data passed to the last argument of encoder begin callback.
972
+ *
973
+ * call-seq:
974
+ * encoder_begin_callback_user_data = user_data -> use_data
975
+ */
976
+ static VALUE
977
+ ruby_whisper_params_set_encoder_begin_callback_user_data(VALUE self, VALUE value)
978
+ {
979
+ ruby_whisper_params *rwp;
980
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
981
+ rwp->encoder_begin_callback_container->user_data = value;
982
+ return value;
983
+ }
984
+
985
+ static VALUE
986
+ ruby_whisper_params_get_abort_callback(VALUE self)
987
+ {
988
+ ruby_whisper_params *rwp;
989
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
990
+ return rwp->abort_callback_container->callback;
991
+ }
992
+ /*
993
+ * Sets abort callback, called to check if the process should be aborted.
994
+ *
995
+ * params.abort_callback = ->(user_data) {
996
+ * # ...
997
+ * }
998
+ *
999
+ * call-seq:
1000
+ * abort_callback = callback -> callback
1001
+ */
1002
+ static VALUE
1003
+ ruby_whisper_params_set_abort_callback(VALUE self, VALUE value)
1004
+ {
1005
+ ruby_whisper_params *rwp;
1006
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
1007
+ rwp->abort_callback_container->callback = value;
1008
+ return value;
1009
+ }
1010
+ static VALUE
1011
+ ruby_whisper_params_get_abort_callback_user_data(VALUE self)
1012
+ {
1013
+ ruby_whisper_params *rwp;
1014
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
1015
+ return rwp->abort_callback_container->user_data;
1016
+ }
1017
+ /*
1018
+ * Sets user data passed to the last argument of abort callback.
1019
+ *
1020
+ * call-seq:
1021
+ * abort_callback_user_data = user_data -> use_data
1022
+ */
1023
+ static VALUE
1024
+ ruby_whisper_params_set_abort_callback_user_data(VALUE self, VALUE value)
1025
+ {
1026
+ ruby_whisper_params *rwp;
1027
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
1028
+ rwp->abort_callback_container->user_data = value;
1029
+ return value;
1030
+ }
1031
+
1032
+ /*
1033
+ * call-seq:
1034
+ * vad = use_vad -> use_vad
1035
+ */
1036
+ static VALUE
1037
+ ruby_whisper_params_get_vad(VALUE self)
1038
+ {
1039
+ BOOL_PARAMS_GETTER(self, vad)
1040
+ }
1041
+
1042
+ static VALUE
1043
+ ruby_whisper_params_set_vad(VALUE self, VALUE value)
1044
+ {
1045
+ BOOL_PARAMS_SETTER(self, vad, value)
1046
+ }
1047
+
1048
+ /*
1049
+ * call-seq:
1050
+ * vad_model_path = model_path -> model_path
1051
+ */
1052
+ static VALUE
1053
+ ruby_whisper_params_set_vad_model_path(VALUE self, VALUE value)
1054
+ {
1055
+ ruby_whisper_params *rwp;
1056
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
1057
+ if (NIL_P(value)) {
1058
+ rwp->params.vad_model_path = NULL;
1059
+ return value;
1060
+ }
1061
+ VALUE path = ruby_whisper_normalize_model_path(value);
1062
+ rwp->params.vad_model_path = StringValueCStr(path);
1063
+ return value;
1064
+ }
1065
+
1066
+ static VALUE
1067
+ ruby_whisper_params_get_vad_model_path(VALUE self)
1068
+ {
1069
+ ruby_whisper_params *rwp;
1070
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
1071
+ return rwp->params.vad_model_path == NULL ? Qnil : rb_str_new2(rwp->params.vad_model_path);
1072
+ }
1073
+
1074
+ /*
1075
+ * call-seq:
1076
+ * vad_params = params -> params
1077
+ */
1078
+ static VALUE
1079
+ ruby_whisper_params_set_vad_params(VALUE self, VALUE value)
1080
+ {
1081
+ ruby_whisper_params *rwp;
1082
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
1083
+ rwp->vad_params = value;
1084
+ return value;
1085
+ }
1086
+
1087
+ static VALUE
1088
+ ruby_whisper_params_get_vad_params(VALUE self)
1089
+ {
1090
+ ruby_whisper_params *rwp;
1091
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
1092
+ return rwp->vad_params;
1093
+ }
1094
+
1095
+ #define SET_PARAM_IF_SAME(param_name) \
1096
+ if (id == id_ ## param_name) { \
1097
+ ruby_whisper_params_set_ ## param_name(self, value); \
1098
+ continue; \
1099
+ }
1100
+
1101
+ static VALUE
1102
+ ruby_whisper_params_initialize(int argc, VALUE *argv, VALUE self)
1103
+ {
1104
+ VALUE kw_hash;
1105
+ VALUE values[RUBY_WHISPER_PARAMS_PARAM_NAMES_COUNT] = {Qundef};
1106
+ VALUE value;
1107
+ ruby_whisper_params *rwp;
1108
+ ID id;
1109
+ int i;
1110
+
1111
+ rb_scan_args_kw(RB_SCAN_ARGS_KEYWORDS, argc, argv, ":", &kw_hash);
1112
+ if (NIL_P(kw_hash)) {
1113
+ return self;
1114
+ }
1115
+
1116
+ rb_get_kwargs(kw_hash, param_names, 0, RUBY_WHISPER_PARAMS_PARAM_NAMES_COUNT, values);
1117
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
1118
+
1119
+ for (i = 0; i < RUBY_WHISPER_PARAMS_PARAM_NAMES_COUNT; i++) {
1120
+ id = param_names[i];
1121
+ value = values[i];
1122
+ if (value == Qundef) {
1123
+ continue;
1124
+ }
1125
+ if (id == id_diarize) {
1126
+ rwp->diarize = value;
1127
+ continue;
1128
+ } else {
1129
+ SET_PARAM_IF_SAME(language)
1130
+ SET_PARAM_IF_SAME(translate)
1131
+ SET_PARAM_IF_SAME(no_context)
1132
+ SET_PARAM_IF_SAME(single_segment)
1133
+ SET_PARAM_IF_SAME(print_special)
1134
+ SET_PARAM_IF_SAME(print_progress)
1135
+ SET_PARAM_IF_SAME(print_realtime)
1136
+ SET_PARAM_IF_SAME(print_timestamps)
1137
+ SET_PARAM_IF_SAME(suppress_blank)
1138
+ SET_PARAM_IF_SAME(suppress_nst)
1139
+ SET_PARAM_IF_SAME(token_timestamps)
1140
+ SET_PARAM_IF_SAME(split_on_word)
1141
+ SET_PARAM_IF_SAME(initial_prompt)
1142
+ SET_PARAM_IF_SAME(offset)
1143
+ SET_PARAM_IF_SAME(duration)
1144
+ SET_PARAM_IF_SAME(max_text_tokens)
1145
+ SET_PARAM_IF_SAME(temperature)
1146
+ SET_PARAM_IF_SAME(max_initial_ts)
1147
+ SET_PARAM_IF_SAME(length_penalty)
1148
+ SET_PARAM_IF_SAME(temperature_inc)
1149
+ SET_PARAM_IF_SAME(entropy_thold)
1150
+ SET_PARAM_IF_SAME(logprob_thold)
1151
+ SET_PARAM_IF_SAME(no_speech_thold)
1152
+ SET_PARAM_IF_SAME(new_segment_callback)
1153
+ SET_PARAM_IF_SAME(new_segment_callback_user_data)
1154
+ SET_PARAM_IF_SAME(progress_callback)
1155
+ SET_PARAM_IF_SAME(progress_callback_user_data)
1156
+ SET_PARAM_IF_SAME(encoder_begin_callback)
1157
+ SET_PARAM_IF_SAME(encoder_begin_callback_user_data)
1158
+ SET_PARAM_IF_SAME(abort_callback)
1159
+ SET_PARAM_IF_SAME(abort_callback_user_data)
1160
+ SET_PARAM_IF_SAME(vad)
1161
+ SET_PARAM_IF_SAME(vad_model_path)
1162
+ SET_PARAM_IF_SAME(vad_params)
1163
+ }
1164
+ }
1165
+
1166
+ return self;
1167
+ }
1168
+
1169
+ #undef SET_PARAM_IF_SAME
1170
+
1171
+ /*
1172
+ * Hook called on new segment. Yields each Whisper::Segment.
1173
+ *
1174
+ * whisper.on_new_segment do |segment|
1175
+ * # ...
1176
+ * end
1177
+ *
1178
+ * call-seq:
1179
+ * on_new_segment {|segment| ... }
1180
+ */
1181
+ static VALUE
1182
+ ruby_whisper_params_on_new_segment(VALUE self)
1183
+ {
1184
+ ruby_whisper_params *rwp;
1185
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
1186
+ const VALUE blk = rb_block_proc();
1187
+ rb_ary_push(rwp->new_segment_callback_container->callbacks, blk);
1188
+ return Qnil;
1189
+ }
1190
+
1191
+ /*
1192
+ * Hook called on progress update. Yields each progress Integer between 0 and 100.
1193
+ *
1194
+ * whisper.on_progress do |progress|
1195
+ * # ...
1196
+ * end
1197
+ *
1198
+ * call-seq:
1199
+ * on_progress {|progress| ... }
1200
+ */
1201
+ static VALUE
1202
+ ruby_whisper_params_on_progress(VALUE self)
1203
+ {
1204
+ ruby_whisper_params *rwp;
1205
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
1206
+ const VALUE blk = rb_block_proc();
1207
+ rb_ary_push(rwp->progress_callback_container->callbacks, blk);
1208
+ return Qnil;
1209
+ }
1210
+
1211
+ /*
1212
+ * Hook called when the encoder starts.
1213
+ *
1214
+ * whisper.on_encoder_begin do
1215
+ * # ...
1216
+ * end
1217
+ *
1218
+ * call-seq:
1219
+ * on_encoder_begin { ... }
1220
+ */
1221
+ static VALUE
1222
+ ruby_whisper_params_on_encoder_begin(VALUE self)
1223
+ {
1224
+ ruby_whisper_params *rwp;
1225
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
1226
+ const VALUE blk = rb_block_proc();
1227
+ rb_ary_push(rwp->encoder_begin_callback_container->callbacks, blk);
1228
+ return Qnil;
1229
+ }
1230
+
1231
+ /*
1232
+ * Call block to determine whether abort or not. Return +true+ when you want to abort.
1233
+ *
1234
+ * params.abort_on do
1235
+ * if some_condition
1236
+ * true # abort
1237
+ * else
1238
+ * false # continue
1239
+ * end
1240
+ * end
1241
+ *
1242
+ * call-seq:
1243
+ * abort_on { ... }
1244
+ */
1245
+ static VALUE
1246
+ ruby_whisper_params_abort_on(VALUE self)
1247
+ {
1248
+ ruby_whisper_params *rwp;
1249
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
1250
+ const VALUE blk = rb_block_proc();
1251
+ rb_ary_push(rwp->abort_callback_container->callbacks, blk);
1252
+ return Qnil;
1253
+ }
1254
+
1255
+ void
1256
+ init_ruby_whisper_params(VALUE *mWhisper)
1257
+ {
1258
+ cParams = rb_define_class_under(*mWhisper, "Params", rb_cObject);
1259
+
1260
+ rb_define_alloc_func(cParams, ruby_whisper_params_allocate);
1261
+ rb_define_method(cParams, "initialize", ruby_whisper_params_initialize, -1);
1262
+
1263
+ DEFINE_PARAM(language, 0)
1264
+ DEFINE_PARAM(translate, 1)
1265
+ DEFINE_PARAM(no_context, 2)
1266
+ DEFINE_PARAM(single_segment, 3)
1267
+ DEFINE_PARAM(print_special, 4)
1268
+ DEFINE_PARAM(print_progress, 5)
1269
+ DEFINE_PARAM(print_realtime, 6)
1270
+ DEFINE_PARAM(print_timestamps, 7)
1271
+ DEFINE_PARAM(suppress_blank, 8)
1272
+ DEFINE_PARAM(suppress_nst, 9)
1273
+ DEFINE_PARAM(token_timestamps, 10)
1274
+ DEFINE_PARAM(split_on_word, 11)
1275
+ DEFINE_PARAM(initial_prompt, 12)
1276
+ DEFINE_PARAM(diarize, 13)
1277
+ DEFINE_PARAM(offset, 14)
1278
+ DEFINE_PARAM(duration, 15)
1279
+ DEFINE_PARAM(max_text_tokens, 16)
1280
+ DEFINE_PARAM(temperature, 17)
1281
+ DEFINE_PARAM(max_initial_ts, 18)
1282
+ DEFINE_PARAM(length_penalty, 19)
1283
+ DEFINE_PARAM(temperature_inc, 20)
1284
+ DEFINE_PARAM(entropy_thold, 21)
1285
+ DEFINE_PARAM(logprob_thold, 22)
1286
+ DEFINE_PARAM(no_speech_thold, 23)
1287
+ DEFINE_PARAM(new_segment_callback, 24)
1288
+ DEFINE_PARAM(new_segment_callback_user_data, 25)
1289
+ DEFINE_PARAM(progress_callback, 26)
1290
+ DEFINE_PARAM(progress_callback_user_data, 27)
1291
+ DEFINE_PARAM(encoder_begin_callback, 28)
1292
+ DEFINE_PARAM(encoder_begin_callback_user_data, 29)
1293
+ DEFINE_PARAM(abort_callback, 30)
1294
+ DEFINE_PARAM(abort_callback_user_data, 31)
1295
+ DEFINE_PARAM(vad, 32)
1296
+ DEFINE_PARAM(vad_model_path, 33)
1297
+ DEFINE_PARAM(vad_params, 34)
1298
+
1299
+ rb_define_method(cParams, "on_new_segment", ruby_whisper_params_on_new_segment, 0);
1300
+ rb_define_method(cParams, "on_progress", ruby_whisper_params_on_progress, 0);
1301
+ rb_define_method(cParams, "on_encoder_begin", ruby_whisper_params_on_encoder_begin, 0);
1302
+ rb_define_method(cParams, "abort_on", ruby_whisper_params_abort_on, 0);
1303
+ }