whispercpp 1.3.2 → 1.3.4

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (664) hide show
  1. checksums.yaml +4 -4
  2. data/.gitignore +6 -3
  3. data/README.md +71 -14
  4. data/Rakefile +20 -7
  5. data/ext/.gitignore +4 -6
  6. data/ext/dependencies.rb +36 -24
  7. data/ext/extconf.rb +1 -1
  8. data/ext/options.rb +48 -184
  9. data/ext/ruby_whisper.c +18 -0
  10. data/ext/ruby_whisper_context.c +43 -12
  11. data/ext/ruby_whisper_model.c +1 -1
  12. data/ext/ruby_whisper_params.c +59 -27
  13. data/ext/ruby_whisper_segment.c +81 -4
  14. data/ext/ruby_whisper_transcribe.cpp +13 -7
  15. data/ext/ruby_whisper_vad_params.c +1 -1
  16. data/ext/sources/CMakeLists.txt +5 -1
  17. data/ext/sources/bindings/javascript/package.json +1 -1
  18. data/ext/sources/build-xcframework.sh +24 -0
  19. data/ext/sources/examples/CMakeLists.txt +1 -0
  20. data/ext/sources/examples/addon.node/__test__/whisper.spec.js +120 -24
  21. data/ext/sources/examples/addon.node/addon.cpp +154 -35
  22. data/ext/sources/examples/addon.node/index.js +10 -5
  23. data/ext/sources/examples/addon.node/vad-example.js +132 -0
  24. data/ext/sources/examples/bench/bench.cpp +29 -18
  25. data/ext/sources/examples/bench.wasm/index-tmpl.html +10 -9
  26. data/ext/sources/examples/cli/cli.cpp +7 -4
  27. data/ext/sources/examples/command/command.cpp +58 -32
  28. data/ext/sources/examples/command.wasm/index-tmpl.html +5 -4
  29. data/ext/sources/examples/common-ggml.cpp +2 -0
  30. data/ext/sources/examples/common-whisper.cpp +14 -7
  31. data/ext/sources/examples/lsp/lsp.cpp +21 -17
  32. data/ext/sources/examples/quantize/quantize.cpp +3 -0
  33. data/ext/sources/examples/server/CMakeLists.txt +3 -0
  34. data/ext/sources/examples/server/server.cpp +193 -35
  35. data/ext/sources/examples/server.py +6 -1
  36. data/ext/sources/examples/stream/stream.cpp +10 -2
  37. data/ext/sources/examples/stream.wasm/emscripten.cpp +6 -6
  38. data/ext/sources/examples/stream.wasm/index-tmpl.html +82 -5
  39. data/ext/sources/examples/talk-llama/CMakeLists.txt +3 -0
  40. data/ext/sources/examples/talk-llama/llama-adapter.cpp +101 -4
  41. data/ext/sources/examples/talk-llama/llama-adapter.h +6 -0
  42. data/ext/sources/examples/talk-llama/llama-arch.cpp +756 -15
  43. data/ext/sources/examples/talk-llama/llama-arch.h +85 -1
  44. data/ext/sources/examples/talk-llama/llama-batch.cpp +773 -272
  45. data/ext/sources/examples/talk-llama/llama-batch.h +126 -55
  46. data/ext/sources/examples/talk-llama/llama-chat.cpp +150 -13
  47. data/ext/sources/examples/talk-llama/llama-chat.h +8 -0
  48. data/ext/sources/examples/talk-llama/llama-context.cpp +814 -542
  49. data/ext/sources/examples/talk-llama/llama-context.h +68 -32
  50. data/ext/sources/examples/talk-llama/llama-cparams.cpp +1 -1
  51. data/ext/sources/examples/talk-llama/llama-cparams.h +4 -4
  52. data/ext/sources/examples/talk-llama/llama-graph.cpp +787 -440
  53. data/ext/sources/examples/talk-llama/llama-graph.h +333 -153
  54. data/ext/sources/examples/talk-llama/llama-hparams.cpp +128 -6
  55. data/ext/sources/examples/talk-llama/llama-hparams.h +80 -17
  56. data/ext/sources/examples/talk-llama/llama-impl.h +2 -0
  57. data/ext/sources/examples/talk-llama/llama-kv-cache-iswa.cpp +326 -0
  58. data/ext/sources/examples/talk-llama/llama-kv-cache-iswa.h +137 -0
  59. data/ext/sources/examples/talk-llama/llama-kv-cache.cpp +1248 -1967
  60. data/ext/sources/examples/talk-llama/llama-kv-cache.h +218 -345
  61. data/ext/sources/examples/talk-llama/llama-kv-cells.h +164 -52
  62. data/ext/sources/examples/talk-llama/llama-memory-hybrid.cpp +266 -0
  63. data/ext/sources/examples/talk-llama/llama-memory-hybrid.h +139 -0
  64. data/ext/sources/examples/talk-llama/llama-memory-recurrent.cpp +1154 -0
  65. data/ext/sources/examples/talk-llama/llama-memory-recurrent.h +182 -0
  66. data/ext/sources/examples/talk-llama/llama-memory.cpp +58 -0
  67. data/ext/sources/examples/talk-llama/llama-memory.h +94 -4
  68. data/ext/sources/examples/talk-llama/llama-mmap.cpp +1 -1
  69. data/ext/sources/examples/talk-llama/llama-model-loader.cpp +44 -17
  70. data/ext/sources/examples/talk-llama/llama-model-loader.h +3 -2
  71. data/ext/sources/examples/talk-llama/llama-model-saver.cpp +1 -0
  72. data/ext/sources/examples/talk-llama/llama-model.cpp +11377 -5248
  73. data/ext/sources/examples/talk-llama/llama-model.h +87 -9
  74. data/ext/sources/examples/talk-llama/llama-quant.cpp +137 -16
  75. data/ext/sources/examples/talk-llama/llama-sampling.cpp +226 -126
  76. data/ext/sources/examples/talk-llama/llama-vocab.cpp +502 -38
  77. data/ext/sources/examples/talk-llama/llama-vocab.h +46 -0
  78. data/ext/sources/examples/talk-llama/llama.cpp +76 -17
  79. data/ext/sources/examples/talk-llama/llama.h +176 -151
  80. data/ext/sources/examples/talk-llama/talk-llama.cpp +11 -6
  81. data/ext/sources/examples/talk-llama/unicode.cpp +212 -0
  82. data/ext/sources/examples/talk-llama/unicode.h +45 -0
  83. data/ext/sources/examples/vad-speech-segments/speech.cpp +6 -0
  84. data/ext/sources/examples/wchess/wchess.cmd/wchess.cmd.cpp +6 -2
  85. data/ext/sources/examples/whisper.wasm/index-tmpl.html +17 -16
  86. data/ext/sources/ggml/CMakeLists.txt +106 -33
  87. data/ext/sources/ggml/cmake/common.cmake +24 -0
  88. data/ext/sources/ggml/cmake/ggml-config.cmake.in +132 -93
  89. data/ext/sources/ggml/include/ggml-backend.h +18 -2
  90. data/ext/sources/ggml/include/ggml-cpu.h +2 -0
  91. data/ext/sources/ggml/include/ggml-metal.h +1 -6
  92. data/ext/sources/ggml/include/ggml-opt.h +25 -6
  93. data/ext/sources/ggml/include/ggml-webgpu.h +19 -0
  94. data/ext/sources/ggml/include/ggml-zdnn.h +17 -0
  95. data/ext/sources/ggml/include/ggml.h +365 -21
  96. data/ext/sources/ggml/src/CMakeLists.txt +98 -25
  97. data/ext/sources/ggml/src/ggml-alloc.c +265 -141
  98. data/ext/sources/ggml/src/ggml-backend-impl.h +4 -1
  99. data/ext/sources/ggml/src/ggml-backend-reg.cpp +35 -13
  100. data/ext/sources/ggml/src/ggml-backend.cpp +266 -60
  101. data/ext/sources/ggml/src/ggml-blas/CMakeLists.txt +4 -4
  102. data/ext/sources/ggml/src/ggml-blas/ggml-blas.cpp +5 -4
  103. data/ext/sources/ggml/src/ggml-cann/CMakeLists.txt +15 -0
  104. data/ext/sources/ggml/src/ggml-cann/acl_tensor.cpp +3 -1
  105. data/ext/sources/ggml/src/ggml-cann/aclnn_ops.cpp +903 -717
  106. data/ext/sources/ggml/src/ggml-cann/aclnn_ops.h +143 -25
  107. data/ext/sources/ggml/src/ggml-cann/common.h +149 -2
  108. data/ext/sources/ggml/src/ggml-cann/ggml-cann.cpp +521 -78
  109. data/ext/sources/ggml/src/ggml-common.h +21 -0
  110. data/ext/sources/ggml/src/ggml-cpu/CMakeLists.txt +165 -50
  111. data/ext/sources/ggml/src/ggml-cpu/amx/amx.cpp +5 -3
  112. data/ext/sources/ggml/src/ggml-cpu/amx/mmq.cpp +11 -10
  113. data/ext/sources/ggml/src/ggml-cpu/arch/arm/cpu-feats.cpp +94 -0
  114. data/ext/sources/ggml/src/ggml-cpu/arch/arm/quants.c +3650 -0
  115. data/ext/sources/ggml/src/ggml-cpu/arch/arm/repack.cpp +1891 -0
  116. data/ext/sources/ggml/src/ggml-cpu/arch/loongarch/quants.c +2160 -0
  117. data/ext/sources/ggml/src/ggml-cpu/arch/powerpc/cpu-feats.cpp +82 -0
  118. data/ext/sources/ggml/src/ggml-cpu/arch/powerpc/quants.c +2305 -0
  119. data/ext/sources/ggml/src/ggml-cpu/arch/riscv/quants.c +1897 -0
  120. data/ext/sources/ggml/src/ggml-cpu/arch/riscv/repack.cpp +342 -0
  121. data/ext/sources/ggml/src/ggml-cpu/arch/s390/quants.c +1468 -0
  122. data/ext/sources/ggml/src/ggml-cpu/arch/wasm/quants.c +1221 -0
  123. data/ext/sources/ggml/src/ggml-cpu/arch/x86/quants.c +3820 -0
  124. data/ext/sources/ggml/src/ggml-cpu/arch/x86/repack.cpp +6307 -0
  125. data/ext/sources/ggml/src/ggml-cpu/arch-fallback.h +214 -0
  126. data/ext/sources/ggml/src/ggml-cpu/common.h +18 -3
  127. data/ext/sources/ggml/src/ggml-cpu/ggml-cpu-impl.h +23 -7
  128. data/ext/sources/ggml/src/ggml-cpu/ggml-cpu.c +179 -110
  129. data/ext/sources/ggml/src/ggml-cpu/ggml-cpu.cpp +44 -33
  130. data/ext/sources/ggml/src/ggml-cpu/{ggml-cpu-hbm.cpp → hbm.cpp} +1 -1
  131. data/ext/sources/ggml/src/ggml-cpu/kleidiai/kernels.cpp +152 -18
  132. data/ext/sources/ggml/src/ggml-cpu/kleidiai/kernels.h +7 -1
  133. data/ext/sources/ggml/src/ggml-cpu/kleidiai/kleidiai.cpp +228 -98
  134. data/ext/sources/ggml/src/ggml-cpu/llamafile/sgemm.cpp +532 -1124
  135. data/ext/sources/ggml/src/ggml-cpu/llamafile/sgemm.h +5 -0
  136. data/ext/sources/ggml/src/ggml-cpu/ops.cpp +3374 -2081
  137. data/ext/sources/ggml/src/ggml-cpu/ops.h +13 -8
  138. data/ext/sources/ggml/src/ggml-cpu/quants.c +1193 -0
  139. data/ext/sources/ggml/src/ggml-cpu/{ggml-cpu-quants.h → quants.h} +34 -0
  140. data/ext/sources/ggml/src/ggml-cpu/repack.cpp +1982 -0
  141. data/ext/sources/ggml/src/ggml-cpu/repack.h +120 -0
  142. data/ext/sources/ggml/src/ggml-cpu/simd-mappings.h +367 -46
  143. data/ext/sources/ggml/src/ggml-cpu/spacemit/ime.cpp +1024 -0
  144. data/ext/sources/ggml/src/ggml-cpu/spacemit/ime.h +13 -0
  145. data/ext/sources/ggml/src/ggml-cpu/spacemit/ime1_kernels.cpp +3196 -0
  146. data/ext/sources/ggml/src/ggml-cpu/spacemit/ime_kernels.h +26 -0
  147. data/ext/sources/ggml/src/ggml-cpu/{ggml-cpu-traits.cpp → traits.cpp} +3 -3
  148. data/ext/sources/ggml/src/ggml-cpu/{ggml-cpu-traits.h → traits.h} +1 -1
  149. data/ext/sources/ggml/src/ggml-cpu/vec.cpp +272 -35
  150. data/ext/sources/ggml/src/ggml-cpu/vec.h +794 -142
  151. data/ext/sources/ggml/src/ggml-cuda/CMakeLists.txt +20 -16
  152. data/ext/sources/ggml/src/ggml-cuda/add-id.cu +58 -0
  153. data/ext/sources/ggml/src/ggml-cuda/add-id.cuh +3 -0
  154. data/ext/sources/ggml/src/ggml-cuda/binbcast.cu +330 -191
  155. data/ext/sources/ggml/src/ggml-cuda/binbcast.cuh +2 -0
  156. data/ext/sources/ggml/src/ggml-cuda/common.cuh +291 -81
  157. data/ext/sources/ggml/src/ggml-cuda/conv-transpose-1d.cu +1 -4
  158. data/ext/sources/ggml/src/ggml-cuda/conv2d-dw.cu +161 -0
  159. data/ext/sources/ggml/src/ggml-cuda/conv2d-dw.cuh +5 -0
  160. data/ext/sources/ggml/src/ggml-cuda/conv2d-transpose.cu +91 -0
  161. data/ext/sources/ggml/src/ggml-cuda/conv2d-transpose.cuh +4 -0
  162. data/ext/sources/ggml/src/ggml-cuda/conv2d.cu +166 -0
  163. data/ext/sources/ggml/src/ggml-cuda/conv2d.cuh +5 -0
  164. data/ext/sources/ggml/src/ggml-cuda/convert.cu +117 -22
  165. data/ext/sources/ggml/src/ggml-cuda/convert.cuh +20 -0
  166. data/ext/sources/ggml/src/ggml-cuda/cpy-utils.cuh +217 -0
  167. data/ext/sources/ggml/src/ggml-cuda/cpy.cu +64 -307
  168. data/ext/sources/ggml/src/ggml-cuda/cross-entropy-loss.cu +2 -14
  169. data/ext/sources/ggml/src/ggml-cuda/dequantize.cuh +14 -40
  170. data/ext/sources/ggml/src/ggml-cuda/fattn-common.cuh +499 -368
  171. data/ext/sources/ggml/src/ggml-cuda/fattn-mma-f16.cuh +142 -93
  172. data/ext/sources/ggml/src/ggml-cuda/fattn-tile.cu +755 -0
  173. data/ext/sources/ggml/src/ggml-cuda/fattn-tile.cuh +3 -0
  174. data/ext/sources/ggml/src/ggml-cuda/fattn-vec.cuh +593 -0
  175. data/ext/sources/ggml/src/ggml-cuda/fattn-wmma-f16.cu +90 -50
  176. data/ext/sources/ggml/src/ggml-cuda/fattn.cu +185 -198
  177. data/ext/sources/ggml/src/ggml-cuda/fattn.cuh +2 -0
  178. data/ext/sources/ggml/src/ggml-cuda/getrows.cu +50 -39
  179. data/ext/sources/ggml/src/ggml-cuda/ggml-cuda.cu +636 -222
  180. data/ext/sources/ggml/src/ggml-cuda/im2col.cu +196 -35
  181. data/ext/sources/ggml/src/ggml-cuda/im2col.cuh +1 -0
  182. data/ext/sources/ggml/src/ggml-cuda/mean.cu +73 -0
  183. data/ext/sources/ggml/src/ggml-cuda/mean.cuh +3 -0
  184. data/ext/sources/ggml/src/ggml-cuda/mma.cuh +198 -45
  185. data/ext/sources/ggml/src/ggml-cuda/mmf.cu +123 -0
  186. data/ext/sources/ggml/src/ggml-cuda/mmf.cuh +496 -0
  187. data/ext/sources/ggml/src/ggml-cuda/mmq.cu +206 -57
  188. data/ext/sources/ggml/src/ggml-cuda/mmq.cuh +1262 -721
  189. data/ext/sources/ggml/src/ggml-cuda/mmvf.cu +506 -0
  190. data/ext/sources/ggml/src/ggml-cuda/{mmv.cuh → mmvf.cuh} +4 -5
  191. data/ext/sources/ggml/src/ggml-cuda/mmvq.cu +64 -73
  192. data/ext/sources/ggml/src/ggml-cuda/norm.cu +284 -12
  193. data/ext/sources/ggml/src/ggml-cuda/norm.cuh +7 -0
  194. data/ext/sources/ggml/src/ggml-cuda/opt-step-sgd.cu +49 -0
  195. data/ext/sources/ggml/src/ggml-cuda/opt-step-sgd.cuh +5 -0
  196. data/ext/sources/ggml/src/ggml-cuda/pad.cu +46 -23
  197. data/ext/sources/ggml/src/ggml-cuda/pad_reflect_1d.cu +91 -0
  198. data/ext/sources/ggml/src/ggml-cuda/pad_reflect_1d.cuh +5 -0
  199. data/ext/sources/ggml/src/ggml-cuda/quantize.cu +12 -10
  200. data/ext/sources/ggml/src/ggml-cuda/reduce_rows.cuh +53 -0
  201. data/ext/sources/ggml/src/ggml-cuda/roll.cu +67 -0
  202. data/ext/sources/ggml/src/ggml-cuda/roll.cuh +5 -0
  203. data/ext/sources/ggml/src/ggml-cuda/rope.cu +21 -27
  204. data/ext/sources/ggml/src/ggml-cuda/scale.cu +14 -11
  205. data/ext/sources/ggml/src/ggml-cuda/set-rows.cu +276 -0
  206. data/ext/sources/ggml/src/ggml-cuda/set-rows.cuh +7 -0
  207. data/ext/sources/ggml/src/ggml-cuda/softcap.cu +34 -0
  208. data/ext/sources/ggml/src/ggml-cuda/softcap.cuh +5 -0
  209. data/ext/sources/ggml/src/ggml-cuda/softmax.cu +126 -59
  210. data/ext/sources/ggml/src/ggml-cuda/ssm-conv.cu +10 -2
  211. data/ext/sources/ggml/src/ggml-cuda/ssm-scan.cu +322 -98
  212. data/ext/sources/ggml/src/ggml-cuda/sum.cu +6 -10
  213. data/ext/sources/ggml/src/ggml-cuda/sumrows.cu +23 -19
  214. data/ext/sources/ggml/src/ggml-cuda/sumrows.cuh +0 -1
  215. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-f16.cu +7 -0
  216. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-q4_0.cu +7 -0
  217. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-q4_1.cu +7 -0
  218. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-q5_0.cu +7 -0
  219. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-q5_1.cu +7 -0
  220. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-f16-q8_0.cu +7 -0
  221. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-f16.cu +7 -0
  222. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-q4_0.cu +7 -0
  223. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-q4_1.cu +7 -0
  224. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-q5_0.cu +7 -0
  225. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-q5_1.cu +7 -0
  226. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_0-q8_0.cu +7 -0
  227. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-f16.cu +7 -0
  228. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-q4_0.cu +7 -0
  229. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-q4_1.cu +7 -0
  230. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-q5_0.cu +7 -0
  231. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-q5_1.cu +7 -0
  232. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q4_1-q8_0.cu +7 -0
  233. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-f16.cu +7 -0
  234. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-q4_0.cu +7 -0
  235. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-q4_1.cu +7 -0
  236. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-q5_0.cu +7 -0
  237. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-q5_1.cu +7 -0
  238. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_0-q8_0.cu +7 -0
  239. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-f16.cu +7 -0
  240. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-q4_0.cu +7 -0
  241. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-q4_1.cu +7 -0
  242. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-q5_0.cu +7 -0
  243. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-q5_1.cu +7 -0
  244. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q5_1-q8_0.cu +7 -0
  245. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-f16.cu +7 -0
  246. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-q4_0.cu +7 -0
  247. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-q4_1.cu +7 -0
  248. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-q5_0.cu +7 -0
  249. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-q5_1.cu +7 -0
  250. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-instance-q8_0-q8_0.cu +7 -0
  251. data/ext/sources/ggml/src/ggml-cuda/template-instances/generate_cu_files.py +21 -18
  252. data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_1.cu +5 -0
  253. data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_10.cu +5 -0
  254. data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_11.cu +5 -0
  255. data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_12.cu +5 -0
  256. data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_13.cu +5 -0
  257. data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_14.cu +5 -0
  258. data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_15.cu +5 -0
  259. data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_16.cu +5 -0
  260. data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_2.cu +5 -0
  261. data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_3.cu +5 -0
  262. data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_4.cu +5 -0
  263. data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_5.cu +5 -0
  264. data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_6.cu +5 -0
  265. data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_7.cu +5 -0
  266. data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_8.cu +5 -0
  267. data/ext/sources/ggml/src/ggml-cuda/template-instances/mmf-instance-ncols_9.cu +5 -0
  268. data/ext/sources/ggml/src/ggml-cuda/template-instances/mmq-instance-mxfp4.cu +5 -0
  269. data/ext/sources/ggml/src/ggml-cuda/topk-moe.cu +259 -0
  270. data/ext/sources/ggml/src/ggml-cuda/topk-moe.cuh +14 -0
  271. data/ext/sources/ggml/src/ggml-cuda/tsembd.cu +3 -3
  272. data/ext/sources/ggml/src/ggml-cuda/unary.cu +179 -0
  273. data/ext/sources/ggml/src/ggml-cuda/unary.cuh +15 -0
  274. data/ext/sources/ggml/src/ggml-cuda/upscale.cu +92 -6
  275. data/ext/sources/ggml/src/ggml-cuda/vecdotq.cuh +110 -22
  276. data/ext/sources/ggml/src/ggml-cuda/vendors/cuda.h +4 -0
  277. data/ext/sources/ggml/src/ggml-cuda/vendors/hip.h +58 -36
  278. data/ext/sources/ggml/src/ggml-cuda/vendors/musa.h +4 -3
  279. data/ext/sources/ggml/src/ggml-hip/CMakeLists.txt +14 -2
  280. data/ext/sources/ggml/src/ggml-impl.h +229 -175
  281. data/ext/sources/ggml/src/ggml-metal/CMakeLists.txt +21 -17
  282. data/ext/sources/ggml/src/ggml-metal/ggml-metal-common.cpp +446 -0
  283. data/ext/sources/ggml/src/ggml-metal/ggml-metal-common.h +52 -0
  284. data/ext/sources/ggml/src/ggml-metal/ggml-metal-context.h +33 -0
  285. data/ext/sources/ggml/src/ggml-metal/ggml-metal-context.m +600 -0
  286. data/ext/sources/ggml/src/ggml-metal/ggml-metal-device.cpp +1376 -0
  287. data/ext/sources/ggml/src/ggml-metal/ggml-metal-device.h +226 -0
  288. data/ext/sources/ggml/src/ggml-metal/ggml-metal-device.m +1308 -0
  289. data/ext/sources/ggml/src/ggml-metal/ggml-metal-impl.h +163 -63
  290. data/ext/sources/ggml/src/ggml-metal/ggml-metal-ops.cpp +3158 -0
  291. data/ext/sources/ggml/src/ggml-metal/ggml-metal-ops.h +82 -0
  292. data/ext/sources/ggml/src/ggml-metal/ggml-metal.cpp +718 -0
  293. data/ext/sources/ggml/src/ggml-metal/ggml-metal.metal +3208 -1575
  294. data/ext/sources/ggml/src/ggml-musa/CMakeLists.txt +18 -8
  295. data/ext/sources/ggml/src/ggml-musa/mudnn.cuh +2 -2
  296. data/ext/sources/ggml/src/ggml-opencl/CMakeLists.txt +32 -0
  297. data/ext/sources/ggml/src/ggml-opencl/ggml-opencl.cpp +4430 -792
  298. data/ext/sources/ggml/src/ggml-opencl/kernels/add.cl +107 -0
  299. data/ext/sources/ggml/src/ggml-opencl/kernels/add_id.cl +42 -0
  300. data/ext/sources/ggml/src/ggml-opencl/kernels/argsort.cl +86 -0
  301. data/ext/sources/ggml/src/ggml-opencl/kernels/concat.cl +109 -0
  302. data/ext/sources/ggml/src/ggml-opencl/kernels/conv2d.cl +185 -0
  303. data/ext/sources/ggml/src/ggml-opencl/kernels/conv2d_f16_f32.cl +176 -0
  304. data/ext/sources/ggml/src/ggml-opencl/kernels/cvt.cl +84 -0
  305. data/ext/sources/ggml/src/ggml-opencl/kernels/div.cl +138 -0
  306. data/ext/sources/ggml/src/ggml-opencl/kernels/flash_attn_f16.cl +370 -0
  307. data/ext/sources/ggml/src/ggml-opencl/kernels/flash_attn_f32.cl +370 -0
  308. data/ext/sources/ggml/src/ggml-opencl/kernels/flash_attn_f32_f16.cl +373 -0
  309. data/ext/sources/ggml/src/ggml-opencl/kernels/gelu.cl +27 -0
  310. data/ext/sources/ggml/src/ggml-opencl/kernels/glu.cl +378 -0
  311. data/ext/sources/ggml/src/ggml-opencl/kernels/group_norm.cl +121 -0
  312. data/ext/sources/ggml/src/ggml-opencl/kernels/im2col_f16.cl +1 -1
  313. data/ext/sources/ggml/src/ggml-opencl/kernels/im2col_f32.cl +1 -1
  314. data/ext/sources/ggml/src/ggml-opencl/kernels/mul.cl +73 -0
  315. data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mat_f16_f32.cl +130 -0
  316. data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mm_f16_f32_l4_lm.cl +132 -0
  317. data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mm_f32_f32_l4_lm.cl +133 -0
  318. data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_id_mxfp4_f32.cl +189 -0
  319. data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_id_mxfp4_f32_flat.cl +176 -0
  320. data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_id_q4_0_f32_8x_flat.cl +283 -0
  321. data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_id_q8_0_f32.cl +140 -0
  322. data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_id_q8_0_f32_flat.cl +222 -0
  323. data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_mxfp4_f32.cl +144 -0
  324. data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_mxfp4_f32_flat.cl +167 -0
  325. data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_q8_0_f32.cl +125 -0
  326. data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_q8_0_f32_flat.cl +202 -0
  327. data/ext/sources/ggml/src/ggml-opencl/kernels/norm.cl +80 -0
  328. data/ext/sources/ggml/src/ggml-opencl/kernels/pad.cl +30 -0
  329. data/ext/sources/ggml/src/ggml-opencl/kernels/repeat.cl +39 -0
  330. data/ext/sources/ggml/src/ggml-opencl/kernels/rms_norm.cl +79 -0
  331. data/ext/sources/ggml/src/ggml-opencl/kernels/scale.cl +3 -2
  332. data/ext/sources/ggml/src/ggml-opencl/kernels/set_rows.cl +189 -0
  333. data/ext/sources/ggml/src/ggml-opencl/kernels/sigmoid.cl +29 -0
  334. data/ext/sources/ggml/src/ggml-opencl/kernels/softmax_4_f16.cl +34 -13
  335. data/ext/sources/ggml/src/ggml-opencl/kernels/softmax_4_f32.cl +34 -13
  336. data/ext/sources/ggml/src/ggml-opencl/kernels/softmax_f16.cl +34 -13
  337. data/ext/sources/ggml/src/ggml-opencl/kernels/softmax_f32.cl +34 -13
  338. data/ext/sources/ggml/src/ggml-opencl/kernels/sub.cl +138 -0
  339. data/ext/sources/ggml/src/ggml-opencl/kernels/sum_rows.cl +39 -0
  340. data/ext/sources/ggml/src/ggml-opencl/kernels/tanh.cl +63 -0
  341. data/ext/sources/ggml/src/ggml-opencl/kernels/transpose.cl +20 -0
  342. data/ext/sources/ggml/src/ggml-opencl/kernels/tsembd.cl +48 -0
  343. data/ext/sources/ggml/src/ggml-opencl/kernels/upscale.cl +120 -0
  344. data/ext/sources/ggml/src/ggml-opt.cpp +97 -41
  345. data/ext/sources/ggml/src/ggml-quants.c +117 -24
  346. data/ext/sources/ggml/src/ggml-quants.h +6 -0
  347. data/ext/sources/ggml/src/ggml-rpc/ggml-rpc.cpp +85 -62
  348. data/ext/sources/ggml/src/ggml-sycl/CMakeLists.txt +3 -3
  349. data/ext/sources/ggml/src/ggml-sycl/backend.hpp +2 -0
  350. data/ext/sources/ggml/src/ggml-sycl/binbcast.cpp +9 -0
  351. data/ext/sources/ggml/src/ggml-sycl/binbcast.hpp +6 -0
  352. data/ext/sources/ggml/src/ggml-sycl/common.hpp +20 -48
  353. data/ext/sources/ggml/src/ggml-sycl/concat.cpp +13 -17
  354. data/ext/sources/ggml/src/ggml-sycl/convert.cpp +21 -2
  355. data/ext/sources/ggml/src/ggml-sycl/cpy.cpp +116 -211
  356. data/ext/sources/ggml/src/ggml-sycl/cpy.hpp +213 -1
  357. data/ext/sources/ggml/src/ggml-sycl/dequantize.hpp +32 -0
  358. data/ext/sources/ggml/src/ggml-sycl/element_wise.cpp +700 -1041
  359. data/ext/sources/ggml/src/ggml-sycl/element_wise.hpp +20 -9
  360. data/ext/sources/ggml/src/ggml-sycl/gemm.hpp +17 -26
  361. data/ext/sources/ggml/src/ggml-sycl/getrows.cpp +2 -96
  362. data/ext/sources/ggml/src/ggml-sycl/ggml-sycl.cpp +393 -250
  363. data/ext/sources/ggml/src/ggml-sycl/im2col.cpp +1 -1
  364. data/ext/sources/ggml/src/ggml-sycl/mmvq.cpp +32 -8
  365. data/ext/sources/ggml/src/ggml-sycl/quantize.hpp +133 -0
  366. data/ext/sources/ggml/src/ggml-sycl/quants.hpp +38 -11
  367. data/ext/sources/ggml/src/ggml-sycl/rope.cpp +125 -21
  368. data/ext/sources/ggml/src/ggml-sycl/set_rows.cpp +234 -0
  369. data/ext/sources/ggml/src/ggml-sycl/set_rows.hpp +8 -0
  370. data/ext/sources/ggml/src/ggml-sycl/sycl_hw.cpp +3 -1
  371. data/ext/sources/ggml/src/ggml-sycl/sycl_hw.hpp +3 -0
  372. data/ext/sources/ggml/src/ggml-sycl/tsembd.cpp +4 -3
  373. data/ext/sources/ggml/src/ggml-sycl/vecdotq.hpp +105 -17
  374. data/ext/sources/ggml/src/ggml-vulkan/CMakeLists.txt +36 -32
  375. data/ext/sources/ggml/src/ggml-vulkan/ggml-vulkan.cpp +4198 -1145
  376. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/CMakeLists.txt +4 -12
  377. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/add.comp +41 -1
  378. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/add_id.comp +42 -0
  379. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/argmax.comp +13 -4
  380. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/argsort.comp +39 -29
  381. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/conv2d_mm.comp +349 -0
  382. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/conv_transpose_1d.comp +98 -0
  383. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/copy_from_quant.comp +2 -2
  384. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/copy_to_quant.comp +66 -12
  385. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_funcs.comp +154 -0
  386. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_funcs_cm2.comp +21 -0
  387. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq2_s.comp +1 -1
  388. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq2_xxs.comp +2 -1
  389. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq3_s.comp +6 -5
  390. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_iq3_xxs.comp +4 -2
  391. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_mxfp4.comp +32 -0
  392. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q2_k.comp +1 -1
  393. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q3_k.comp +1 -1
  394. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q4_k.comp +1 -1
  395. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q5_k.comp +1 -1
  396. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/dequant_q6_k.comp +1 -1
  397. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/exp.comp +21 -0
  398. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/flash_attn.comp +69 -24
  399. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_base.comp +60 -20
  400. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_cm1.comp +98 -42
  401. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_cm2.comp +64 -27
  402. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_split_k_reduce.comp +74 -13
  403. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/geglu.comp +13 -0
  404. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/geglu_erf.comp +27 -0
  405. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/geglu_quick.comp +11 -0
  406. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/gelu_erf.comp +39 -0
  407. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/generic_binary_head.comp +4 -17
  408. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/get_rows.comp +19 -10
  409. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/get_rows_quant.comp +25 -15
  410. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/glu_head.comp +19 -0
  411. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/glu_main.comp +29 -0
  412. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/hardsigmoid.comp +22 -0
  413. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/hardswish.comp +22 -0
  414. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/im2col.comp +18 -14
  415. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/im2col_3d.comp +126 -0
  416. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_base.comp +65 -1
  417. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_nc.comp +11 -7
  418. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vecq.comp +140 -0
  419. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mm.comp +144 -531
  420. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mm_cm2.comp +206 -38
  421. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mm_funcs.comp +556 -0
  422. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mmq.comp +12 -5
  423. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mmq_funcs.comp +15 -9
  424. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/multi_add.comp +111 -0
  425. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/opt_step_sgd.comp +22 -0
  426. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/pad.comp +24 -3
  427. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/quantize_q8_1.comp +53 -3
  428. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/reglu.comp +9 -0
  429. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rms_norm.comp +64 -11
  430. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rms_norm_partials.comp +65 -0
  431. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/roll.comp +46 -0
  432. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rope_head.comp +1 -4
  433. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rope_multi.comp +7 -9
  434. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rope_neox.comp +7 -9
  435. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rope_norm.comp +7 -9
  436. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rte.comp +5 -0
  437. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/scale.comp +1 -1
  438. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/soft_max.comp +29 -7
  439. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/soft_max_back.comp +4 -0
  440. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/sqrt.comp +17 -0
  441. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/sum_rows.comp +38 -5
  442. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/swiglu.comp +9 -0
  443. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/swiglu_oai.comp +14 -0
  444. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/timestep_embedding.comp +4 -3
  445. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/types.comp +101 -9
  446. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/upscale.comp +69 -5
  447. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/utils.comp +25 -0
  448. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/vulkan-shaders-gen.cpp +338 -71
  449. data/ext/sources/ggml/src/ggml-webgpu/CMakeLists.txt +54 -0
  450. data/ext/sources/ggml/src/ggml-webgpu/ggml-webgpu.cpp +1558 -0
  451. data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/add.tmpl.wgsl +44 -0
  452. data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/add_in_place.tmpl.wgsl +41 -0
  453. data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/binary_head.tmpl +45 -0
  454. data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/common_decls.tmpl +930 -0
  455. data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/cpy.wgsl +60 -0
  456. data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/embed_wgsl.py +124 -0
  457. data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/get_rows.tmpl.wgsl +874 -0
  458. data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/memset.wgsl +40 -0
  459. data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/mul.tmpl.wgsl +44 -0
  460. data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/mul_in_place.tmpl.wgsl +41 -0
  461. data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/mul_mat.tmpl.wgsl +907 -0
  462. data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/rms_norm.wgsl +57 -0
  463. data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/rms_norm_in_place.wgsl +48 -0
  464. data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/set_rows.wgsl +81 -0
  465. data/ext/sources/ggml/src/ggml-zdnn/CMakeLists.txt +36 -0
  466. data/ext/sources/ggml/src/ggml-zdnn/common.hpp +59 -0
  467. data/ext/sources/ggml/src/ggml-zdnn/ggml-zdnn.cpp +628 -0
  468. data/ext/sources/ggml/src/ggml-zdnn/mmf.cpp +80 -0
  469. data/ext/sources/ggml/src/ggml-zdnn/mmf.hpp +12 -0
  470. data/ext/sources/ggml/src/ggml-zdnn/utils.cpp +79 -0
  471. data/ext/sources/ggml/src/ggml-zdnn/utils.hpp +19 -0
  472. data/ext/sources/ggml/src/ggml.c +802 -142
  473. data/ext/sources/ggml/src/ggml.cpp +26 -0
  474. data/ext/sources/ggml/src/gguf.cpp +32 -4
  475. data/ext/sources/include/whisper.h +2 -0
  476. data/ext/sources/src/CMakeLists.txt +2 -0
  477. data/ext/sources/src/coreml/whisper-compat.h +10 -0
  478. data/ext/sources/src/coreml/whisper-compat.m +35 -0
  479. data/ext/sources/src/coreml/whisper-decoder-impl.m +1 -0
  480. data/ext/sources/src/coreml/whisper-encoder-impl.m +1 -0
  481. data/ext/sources/src/whisper.cpp +241 -215
  482. data/ext/sources/tests/CMakeLists.txt +8 -1
  483. data/ext/sources/tests/test-vad-full.cpp +3 -3
  484. data/ext/sources/tests/test-vad.cpp +2 -2
  485. data/extsources.rb +15 -9
  486. data/lib/whisper/context.rb +15 -0
  487. data/lib/whisper/model/uri.rb +57 -2
  488. data/lib/whisper/segment.rb +58 -0
  489. data/sig/whisper.rbs +75 -38
  490. data/{tests → test}/helper.rb +1 -12
  491. data/{tests → test}/test_model.rb +9 -0
  492. data/test/test_package.rb +51 -0
  493. data/{tests → test}/test_params.rb +8 -0
  494. data/test/test_segment.rb +146 -0
  495. data/{tests → test}/test_whisper.rb +70 -0
  496. data/whispercpp.gemspec +2 -3
  497. metadata +246 -191
  498. data/ext/sources/.dockerignore +0 -3
  499. data/ext/sources/.github/workflows/bindings-ruby.yml +0 -21
  500. data/ext/sources/ci/run.sh +0 -336
  501. data/ext/sources/close-issue.yml +0 -28
  502. data/ext/sources/ggml/include/ggml-kompute.h +0 -50
  503. data/ext/sources/ggml/src/ggml-amx/CMakeLists.txt +0 -107
  504. data/ext/sources/ggml/src/ggml-amx/common.h +0 -94
  505. data/ext/sources/ggml/src/ggml-amx/ggml-amx.cpp +0 -446
  506. data/ext/sources/ggml/src/ggml-amx/mmq.cpp +0 -2510
  507. data/ext/sources/ggml/src/ggml-amx/mmq.h +0 -17
  508. data/ext/sources/ggml/src/ggml-cann/kernels/CMakeLists.txt +0 -30
  509. data/ext/sources/ggml/src/ggml-cann/kernels/ascendc_kernels.h +0 -19
  510. data/ext/sources/ggml/src/ggml-cann/kernels/dup.cpp +0 -234
  511. data/ext/sources/ggml/src/ggml-cann/kernels/get_row_f16.cpp +0 -197
  512. data/ext/sources/ggml/src/ggml-cann/kernels/get_row_f32.cpp +0 -190
  513. data/ext/sources/ggml/src/ggml-cann/kernels/get_row_q4_0.cpp +0 -204
  514. data/ext/sources/ggml/src/ggml-cann/kernels/get_row_q8_0.cpp +0 -191
  515. data/ext/sources/ggml/src/ggml-cann/kernels/quantize_f16_q8_0.cpp +0 -218
  516. data/ext/sources/ggml/src/ggml-cann/kernels/quantize_f32_q8_0.cpp +0 -216
  517. data/ext/sources/ggml/src/ggml-cann/kernels/quantize_float_to_q4_0.cpp +0 -295
  518. data/ext/sources/ggml/src/ggml-cpu/ggml-cpu-aarch64.cpp +0 -6431
  519. data/ext/sources/ggml/src/ggml-cpu/ggml-cpu-aarch64.h +0 -8
  520. data/ext/sources/ggml/src/ggml-cpu/ggml-cpu-quants.c +0 -13747
  521. data/ext/sources/ggml/src/ggml-cuda/fattn-tile-f16.cu +0 -357
  522. data/ext/sources/ggml/src/ggml-cuda/fattn-tile-f16.cuh +0 -3
  523. data/ext/sources/ggml/src/ggml-cuda/fattn-tile-f32.cu +0 -365
  524. data/ext/sources/ggml/src/ggml-cuda/fattn-tile-f32.cuh +0 -3
  525. data/ext/sources/ggml/src/ggml-cuda/fattn-vec-f16.cuh +0 -482
  526. data/ext/sources/ggml/src/ggml-cuda/fattn-vec-f32.cuh +0 -472
  527. data/ext/sources/ggml/src/ggml-cuda/mmv.cu +0 -336
  528. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-f16.cu +0 -5
  529. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q4_0.cu +0 -5
  530. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q4_1.cu +0 -5
  531. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q5_0.cu +0 -5
  532. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q5_1.cu +0 -5
  533. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q8_0.cu +0 -5
  534. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-f16.cu +0 -5
  535. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q4_0.cu +0 -5
  536. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q4_1.cu +0 -5
  537. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q5_0.cu +0 -5
  538. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q5_1.cu +0 -5
  539. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q8_0.cu +0 -5
  540. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-f16.cu +0 -5
  541. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q4_0.cu +0 -5
  542. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q4_1.cu +0 -5
  543. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q5_0.cu +0 -5
  544. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q5_1.cu +0 -5
  545. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q8_0.cu +0 -5
  546. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-f16.cu +0 -5
  547. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q4_0.cu +0 -5
  548. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q4_1.cu +0 -5
  549. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q5_0.cu +0 -5
  550. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q5_1.cu +0 -5
  551. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q8_0.cu +0 -5
  552. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-f16.cu +0 -5
  553. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q4_0.cu +0 -5
  554. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q4_1.cu +0 -5
  555. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q5_0.cu +0 -5
  556. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q5_1.cu +0 -5
  557. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q8_0.cu +0 -5
  558. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-f16.cu +0 -5
  559. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q4_0.cu +0 -5
  560. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q4_1.cu +0 -5
  561. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q5_0.cu +0 -5
  562. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q5_1.cu +0 -5
  563. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q8_0.cu +0 -5
  564. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs256-f16-f16.cu +0 -5
  565. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-f16.cu +0 -5
  566. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q4_0.cu +0 -5
  567. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q4_1.cu +0 -5
  568. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q5_0.cu +0 -5
  569. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q5_1.cu +0 -5
  570. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q8_0.cu +0 -5
  571. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-f16.cu +0 -5
  572. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q4_0.cu +0 -5
  573. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q4_1.cu +0 -5
  574. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q5_0.cu +0 -5
  575. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q5_1.cu +0 -5
  576. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q8_0.cu +0 -5
  577. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-f16.cu +0 -5
  578. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q4_0.cu +0 -5
  579. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q4_1.cu +0 -5
  580. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q5_0.cu +0 -5
  581. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q5_1.cu +0 -5
  582. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q8_0.cu +0 -5
  583. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-f16.cu +0 -5
  584. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q4_0.cu +0 -5
  585. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q4_1.cu +0 -5
  586. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q5_0.cu +0 -5
  587. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q5_1.cu +0 -5
  588. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q8_0.cu +0 -5
  589. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-f16.cu +0 -5
  590. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q4_0.cu +0 -5
  591. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q4_1.cu +0 -5
  592. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q5_0.cu +0 -5
  593. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q5_1.cu +0 -5
  594. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q8_0.cu +0 -5
  595. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-f16.cu +0 -5
  596. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q4_0.cu +0 -5
  597. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q4_1.cu +0 -5
  598. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q5_0.cu +0 -5
  599. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q5_1.cu +0 -5
  600. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q8_0.cu +0 -5
  601. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-f16.cu +0 -5
  602. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q4_0.cu +0 -5
  603. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q4_1.cu +0 -5
  604. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q5_0.cu +0 -5
  605. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q5_1.cu +0 -5
  606. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q8_0.cu +0 -5
  607. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs256-f16-f16.cu +0 -5
  608. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-f16.cu +0 -5
  609. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q4_0.cu +0 -5
  610. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q4_1.cu +0 -5
  611. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q5_0.cu +0 -5
  612. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q5_1.cu +0 -5
  613. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q8_0.cu +0 -5
  614. data/ext/sources/ggml/src/ggml-kompute/CMakeLists.txt +0 -166
  615. data/ext/sources/ggml/src/ggml-kompute/ggml-kompute.cpp +0 -2251
  616. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/common.comp +0 -112
  617. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_add.comp +0 -58
  618. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_addrow.comp +0 -25
  619. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_cpy_f16_f16.comp +0 -52
  620. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_cpy_f16_f32.comp +0 -52
  621. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_cpy_f32_f16.comp +0 -52
  622. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_cpy_f32_f32.comp +0 -52
  623. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_diagmask.comp +0 -30
  624. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_gelu.comp +0 -22
  625. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_getrows.comp +0 -17
  626. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_getrows_f16.comp +0 -31
  627. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_getrows_f32.comp +0 -31
  628. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_getrows_q4_0.comp +0 -38
  629. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_getrows_q4_1.comp +0 -39
  630. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_getrows_q6_k.comp +0 -44
  631. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_mul.comp +0 -52
  632. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_mul_mat_f16.comp +0 -69
  633. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_mul_mat_mat_f32.comp +0 -51
  634. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_mul_mat_q4_0.comp +0 -33
  635. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_mul_mat_q4_1.comp +0 -35
  636. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_mul_mat_q4_k.comp +0 -140
  637. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_mul_mat_q6_k.comp +0 -106
  638. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_mul_mat_q8_0.comp +0 -73
  639. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_mul_mv_q_n.comp +0 -52
  640. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_mul_mv_q_n_pre.comp +0 -28
  641. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_norm.comp +0 -84
  642. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_relu.comp +0 -21
  643. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_rmsnorm.comp +0 -53
  644. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_rope_neox_f16.comp +0 -52
  645. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_rope_neox_f32.comp +0 -52
  646. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_rope_norm_f16.comp +0 -52
  647. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_rope_norm_f32.comp +0 -52
  648. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_scale.comp +0 -19
  649. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_scale_8.comp +0 -23
  650. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_silu.comp +0 -22
  651. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/op_softmax.comp +0 -72
  652. data/ext/sources/ggml/src/ggml-kompute/kompute-shaders/rope_common.comp +0 -71
  653. data/ext/sources/ggml/src/ggml-metal/ggml-metal.m +0 -5998
  654. data/tests/test_package.rb +0 -46
  655. data/tests/test_segment.rb +0 -74
  656. /data/ext/sources/ggml/src/ggml-cpu/{cpu-feats-x86.cpp → arch/x86/cpu-feats.cpp} +0 -0
  657. /data/ext/sources/ggml/src/ggml-cpu/{ggml-cpu-hbm.h → hbm.h} +0 -0
  658. /data/{tests → test}/jfk_reader/.gitignore +0 -0
  659. /data/{tests → test}/jfk_reader/extconf.rb +0 -0
  660. /data/{tests → test}/jfk_reader/jfk_reader.c +0 -0
  661. /data/{tests → test}/test_callback.rb +0 -0
  662. /data/{tests → test}/test_error.rb +0 -0
  663. /data/{tests → test}/test_vad.rb +0 -0
  664. /data/{tests → test}/test_vad_params.rb +0 -0
@@ -11,6 +11,9 @@ extern ID id_new;
11
11
  extern ID id_to_path;
12
12
  extern ID id_URI;
13
13
  extern ID id_pre_converted_models;
14
+ extern ID id_coreml_compiled_models;
15
+ extern ID id_cache;
16
+ extern ID id_n_processors;
14
17
 
15
18
  extern VALUE cContext;
16
19
  extern VALUE eError;
@@ -18,10 +21,12 @@ extern VALUE cModel;
18
21
 
19
22
  extern const rb_data_type_t ruby_whisper_params_type;
20
23
  extern VALUE ruby_whisper_transcribe(int argc, VALUE *argv, VALUE self);
21
- extern VALUE rb_whisper_model_initialize(VALUE context);
22
- extern VALUE rb_whisper_segment_initialize(VALUE context, int index);
24
+ extern VALUE rb_whisper_model_s_new(VALUE context);
25
+ extern VALUE rb_whisper_segment_s_new(VALUE context, int index);
23
26
  extern void prepare_transcription(ruby_whisper_params *rwp, VALUE *context);
24
27
 
28
+ ID transcribe_option_names[1];
29
+
25
30
  static void
26
31
  ruby_whisper_free(ruby_whisper *rw)
27
32
  {
@@ -53,6 +58,9 @@ ruby_whisper_memsize(const void *p)
53
58
  if (!rw) {
54
59
  return 0;
55
60
  }
61
+ if (rw->context) {
62
+ size += sizeof(rw->context);
63
+ }
56
64
  return size;
57
65
  }
58
66
 
@@ -79,6 +87,13 @@ ruby_whisper_normalize_model_path(VALUE model_path)
79
87
  VALUE pre_converted_model = rb_hash_aref(pre_converted_models, model_path);
80
88
  if (!NIL_P(pre_converted_model)) {
81
89
  model_path = pre_converted_model;
90
+ #ifdef RUBY_WHISPER_USE_COREML
91
+ VALUE coreml_converted_models = rb_funcall(cModel, id_coreml_compiled_models, 0);
92
+ VALUE coreml_converted_model = rb_hash_aref(coreml_converted_models, pre_converted_model);
93
+ if (!NIL_P(coreml_converted_model)) {
94
+ rb_funcall(coreml_converted_model, id_cache, 0);
95
+ }
96
+ #endif
82
97
  }
83
98
  else if (TYPE(model_path) == T_STRING) {
84
99
  const char * model_path_str = StringValueCStr(model_path);
@@ -293,13 +308,20 @@ VALUE ruby_whisper_full(int argc, VALUE *argv, VALUE self)
293
308
  // Should check when samples.respond_to?(:length)?
294
309
  } else {
295
310
  if (TYPE(samples) == T_ARRAY) {
296
- n_samples = RARRAY_LEN(samples);
311
+ if (RARRAY_LEN(samples) > INT_MAX) {
312
+ rb_raise(rb_eArgError, "samples are too long");
313
+ }
314
+ n_samples = (int)RARRAY_LEN(samples);
297
315
  } else if (memory_view_available_p) {
298
316
  if (!rb_memory_view_get(samples, &view, RUBY_MEMORY_VIEW_SIMPLE)) {
299
317
  view.obj = Qnil;
300
318
  rb_raise(rb_eArgError, "unable to get a memory view");
301
319
  }
302
- n_samples = view.byte_size / view.item_size;
320
+ ssize_t n_samples_size = view.byte_size / view.item_size;
321
+ if (n_samples_size > INT_MAX) {
322
+ rb_raise(rb_eArgError, "samples are too long");
323
+ }
324
+ n_samples = (int)n_samples_size;
303
325
  } else if (rb_respond_to(samples, id_length)) {
304
326
  n_samples = NUM2INT(rb_funcall(samples, id_length, 0));
305
327
  } else {
@@ -387,10 +409,17 @@ ruby_whisper_full_parallel(int argc, VALUE *argv,VALUE self)
387
409
  view.obj = Qnil;
388
410
  rb_raise(rb_eArgError, "unable to get a memory view");
389
411
  }
390
- n_samples = view.byte_size / view.item_size;
412
+ ssize_t n_samples_size = view.byte_size / view.item_size;
413
+ if (n_samples_size > INT_MAX) {
414
+ rb_raise(rb_eArgError, "samples are too long");
415
+ }
416
+ n_samples = (int)n_samples_size;
391
417
  } else {
392
418
  if (TYPE(samples) == T_ARRAY) {
393
- n_samples = RARRAY_LEN(samples);
419
+ if (RARRAY_LEN(samples) > INT_MAX) {
420
+ rb_raise(rb_eArgError, "samples are too long");
421
+ }
422
+ n_samples = (int)RARRAY_LEN(samples);
394
423
  } else if (rb_respond_to(samples, id_length)) {
395
424
  n_samples = NUM2INT(rb_funcall(samples, id_length, 0));
396
425
  } else {
@@ -476,7 +505,7 @@ ruby_whisper_full_get_segment_t0(VALUE self, VALUE i_segment)
476
505
  TypedData_Get_Struct(self, ruby_whisper, &ruby_whisper_type, rw);
477
506
  const int c_i_segment = ruby_whisper_full_check_segment_index(rw, i_segment);
478
507
  const int64_t t0 = whisper_full_get_segment_t0(rw->context, c_i_segment);
479
- return INT2NUM(t0);
508
+ return LONG2NUM(t0);
480
509
  }
481
510
 
482
511
  /*
@@ -494,7 +523,7 @@ ruby_whisper_full_get_segment_t1(VALUE self, VALUE i_segment)
494
523
  TypedData_Get_Struct(self, ruby_whisper, &ruby_whisper_type, rw);
495
524
  const int c_i_segment = ruby_whisper_full_check_segment_index(rw, i_segment);
496
525
  const int64_t t1 = whisper_full_get_segment_t1(rw->context, c_i_segment);
497
- return INT2NUM(t1);
526
+ return LONG2NUM(t1);
498
527
  }
499
528
 
500
529
  /*
@@ -552,7 +581,7 @@ ruby_whisper_full_get_segment_no_speech_prob(VALUE self, VALUE i_segment)
552
581
  static VALUE
553
582
  ruby_whisper_full_get_segment(VALUE self, VALUE i_segment)
554
583
  {
555
- return rb_whisper_segment_initialize(self, NUM2INT(i_segment));
584
+ return rb_whisper_segment_s_new(self, NUM2INT(i_segment));
556
585
  }
557
586
 
558
587
  /*
@@ -586,7 +615,7 @@ ruby_whisper_each_segment(VALUE self)
586
615
 
587
616
  const int n_segments = whisper_full_n_segments(rw->context);
588
617
  for (int i = 0; i < n_segments; ++i) {
589
- rb_yield(rb_whisper_segment_initialize(self, i));
618
+ rb_yield(rb_whisper_segment_s_new(self, i));
590
619
  }
591
620
 
592
621
  return self;
@@ -599,7 +628,7 @@ ruby_whisper_each_segment(VALUE self)
599
628
  static VALUE
600
629
  ruby_whisper_get_model(VALUE self)
601
630
  {
602
- return rb_whisper_model_initialize(self);
631
+ return rb_whisper_model_s_new(self);
603
632
  }
604
633
 
605
634
  void
@@ -607,6 +636,8 @@ init_ruby_whisper_context(VALUE *mWhisper)
607
636
  {
608
637
  cContext = rb_define_class_under(*mWhisper, "Context", rb_cObject);
609
638
 
639
+ transcribe_option_names[0] = id_n_processors;
640
+
610
641
  rb_define_alloc_func(cContext, ruby_whisper_allocate);
611
642
  rb_define_method(cContext, "initialize", ruby_whisper_initialize, -1);
612
643
 
@@ -633,7 +664,7 @@ init_ruby_whisper_context(VALUE *mWhisper)
633
664
  rb_define_method(cContext, "full", ruby_whisper_full, -1);
634
665
  rb_define_method(cContext, "full_parallel", ruby_whisper_full_parallel, -1);
635
666
 
636
- // High leve
667
+ // High level
637
668
  rb_define_method(cContext, "full_get_segment", ruby_whisper_full_get_segment, 1);
638
669
  rb_define_method(cContext, "each_segment", ruby_whisper_each_segment, 0);
639
670
 
@@ -35,7 +35,7 @@ static VALUE ruby_whisper_model_allocate(VALUE klass) {
35
35
  return TypedData_Make_Struct(klass, ruby_whisper_model, &rb_whisper_model_type, rwm);
36
36
  }
37
37
 
38
- VALUE rb_whisper_model_initialize(VALUE context) {
38
+ VALUE rb_whisper_model_s_new(VALUE context) {
39
39
  ruby_whisper_model *rwm;
40
40
  const VALUE model = ruby_whisper_model_allocate(cModel);
41
41
  TypedData_Get_Struct(model, ruby_whisper_model, &rb_whisper_model_type, rwm);
@@ -26,7 +26,7 @@
26
26
  rb_define_method(cParams, #param_name, ruby_whisper_params_get_ ## param_name, 0); \
27
27
  rb_define_method(cParams, #param_name "=", ruby_whisper_params_set_ ## param_name, 1);
28
28
 
29
- #define RUBY_WHISPER_PARAMS_PARAM_NAMES_COUNT 35
29
+ #define RUBY_WHISPER_PARAMS_PARAM_NAMES_COUNT 36
30
30
 
31
31
  extern VALUE cParams;
32
32
  extern VALUE cVADParams;
@@ -34,7 +34,7 @@ extern VALUE cVADParams;
34
34
  extern ID id_call;
35
35
 
36
36
  extern VALUE ruby_whisper_normalize_model_path(VALUE model_path);
37
- extern VALUE rb_whisper_segment_initialize(VALUE context, int index);
37
+ extern VALUE rb_whisper_segment_s_new(VALUE context, int index);
38
38
  extern const rb_data_type_t ruby_whisper_vad_params_type;
39
39
 
40
40
  static ID param_names[RUBY_WHISPER_PARAMS_PARAM_NAMES_COUNT];
@@ -49,6 +49,7 @@ static ID id_print_timestamps;
49
49
  static ID id_suppress_blank;
50
50
  static ID id_suppress_nst;
51
51
  static ID id_token_timestamps;
52
+ static ID id_max_len;
52
53
  static ID id_split_on_word;
53
54
  static ID id_initial_prompt;
54
55
  static ID id_diarize;
@@ -77,6 +78,8 @@ static ID id_vad_params;
77
78
  static void
78
79
  rb_whisper_callbcack_container_mark(ruby_whisper_callback_container *rwc)
79
80
  {
81
+ if (rwc == NULL) return;
82
+
80
83
  rb_gc_mark(rwc->user_data);
81
84
  rb_gc_mark(rwc->callback);
82
85
  rb_gc_mark(rwc->callbacks);
@@ -108,7 +111,7 @@ static void new_segment_callback(struct whisper_context *ctx, struct whisper_sta
108
111
  const int n_segments = whisper_full_n_segments_from_state(state);
109
112
  for (int i = n_new; i > 0; i--) {
110
113
  int i_segment = n_segments - i;
111
- VALUE segment = rb_whisper_segment_initialize(*container->context, i_segment);
114
+ VALUE segment = rb_whisper_segment_s_new(*container->context, i_segment);
112
115
  for (int j = 0; j < callbacks_len; j++) {
113
116
  VALUE cb = rb_ary_entry(container->callbacks, j);
114
117
  rb_funcall(cb, id_call, 1, segment);
@@ -512,6 +515,33 @@ ruby_whisper_params_set_token_timestamps(VALUE self, VALUE value)
512
515
  {
513
516
  BOOL_PARAMS_SETTER(self, token_timestamps, value)
514
517
  }
518
+
519
+ /*
520
+ * max segment length in characters.
521
+ *
522
+ * call-seq:
523
+ * max_len -> Integer
524
+ */
525
+ static VALUE
526
+ ruby_whisper_params_get_max_len(VALUE self)
527
+ {
528
+ ruby_whisper_params *rwp;
529
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
530
+ return INT2NUM(rwp->params.max_len);
531
+ }
532
+ /*
533
+ * call-seq:
534
+ * max_len = length -> length
535
+ */
536
+ static VALUE
537
+ ruby_whisper_params_set_max_len(VALUE self, VALUE value)
538
+ {
539
+ ruby_whisper_params *rwp;
540
+ TypedData_Get_Struct(self, ruby_whisper_params, &ruby_whisper_params_type, rwp);
541
+ rwp->params.max_len = NUM2INT(value);
542
+ return value;
543
+ }
544
+
515
545
  /*
516
546
  * If true, split on word rather than on token (when used with max_len).
517
547
  *
@@ -1135,6 +1165,7 @@ ruby_whisper_params_initialize(int argc, VALUE *argv, VALUE self)
1135
1165
  SET_PARAM_IF_SAME(suppress_blank)
1136
1166
  SET_PARAM_IF_SAME(suppress_nst)
1137
1167
  SET_PARAM_IF_SAME(token_timestamps)
1168
+ SET_PARAM_IF_SAME(max_len)
1138
1169
  SET_PARAM_IF_SAME(split_on_word)
1139
1170
  SET_PARAM_IF_SAME(initial_prompt)
1140
1171
  SET_PARAM_IF_SAME(offset)
@@ -1269,30 +1300,31 @@ init_ruby_whisper_params(VALUE *mWhisper)
1269
1300
  DEFINE_PARAM(suppress_blank, 8)
1270
1301
  DEFINE_PARAM(suppress_nst, 9)
1271
1302
  DEFINE_PARAM(token_timestamps, 10)
1272
- DEFINE_PARAM(split_on_word, 11)
1273
- DEFINE_PARAM(initial_prompt, 12)
1274
- DEFINE_PARAM(diarize, 13)
1275
- DEFINE_PARAM(offset, 14)
1276
- DEFINE_PARAM(duration, 15)
1277
- DEFINE_PARAM(max_text_tokens, 16)
1278
- DEFINE_PARAM(temperature, 17)
1279
- DEFINE_PARAM(max_initial_ts, 18)
1280
- DEFINE_PARAM(length_penalty, 19)
1281
- DEFINE_PARAM(temperature_inc, 20)
1282
- DEFINE_PARAM(entropy_thold, 21)
1283
- DEFINE_PARAM(logprob_thold, 22)
1284
- DEFINE_PARAM(no_speech_thold, 23)
1285
- DEFINE_PARAM(new_segment_callback, 24)
1286
- DEFINE_PARAM(new_segment_callback_user_data, 25)
1287
- DEFINE_PARAM(progress_callback, 26)
1288
- DEFINE_PARAM(progress_callback_user_data, 27)
1289
- DEFINE_PARAM(encoder_begin_callback, 28)
1290
- DEFINE_PARAM(encoder_begin_callback_user_data, 29)
1291
- DEFINE_PARAM(abort_callback, 30)
1292
- DEFINE_PARAM(abort_callback_user_data, 31)
1293
- DEFINE_PARAM(vad, 32)
1294
- DEFINE_PARAM(vad_model_path, 33)
1295
- DEFINE_PARAM(vad_params, 34)
1303
+ DEFINE_PARAM(max_len, 11)
1304
+ DEFINE_PARAM(split_on_word, 12)
1305
+ DEFINE_PARAM(initial_prompt, 13)
1306
+ DEFINE_PARAM(diarize, 14)
1307
+ DEFINE_PARAM(offset, 15)
1308
+ DEFINE_PARAM(duration, 16)
1309
+ DEFINE_PARAM(max_text_tokens, 17)
1310
+ DEFINE_PARAM(temperature, 18)
1311
+ DEFINE_PARAM(max_initial_ts, 19)
1312
+ DEFINE_PARAM(length_penalty, 20)
1313
+ DEFINE_PARAM(temperature_inc, 21)
1314
+ DEFINE_PARAM(entropy_thold, 22)
1315
+ DEFINE_PARAM(logprob_thold, 23)
1316
+ DEFINE_PARAM(no_speech_thold, 24)
1317
+ DEFINE_PARAM(new_segment_callback, 25)
1318
+ DEFINE_PARAM(new_segment_callback_user_data, 26)
1319
+ DEFINE_PARAM(progress_callback, 27)
1320
+ DEFINE_PARAM(progress_callback_user_data, 28)
1321
+ DEFINE_PARAM(encoder_begin_callback, 29)
1322
+ DEFINE_PARAM(encoder_begin_callback_user_data, 30)
1323
+ DEFINE_PARAM(abort_callback, 31)
1324
+ DEFINE_PARAM(abort_callback_user_data, 32)
1325
+ DEFINE_PARAM(vad, 33)
1326
+ DEFINE_PARAM(vad_model_path, 34)
1327
+ DEFINE_PARAM(vad_params, 35)
1296
1328
 
1297
1329
  rb_define_method(cParams, "on_new_segment", ruby_whisper_params_on_new_segment, 0);
1298
1330
  rb_define_method(cParams, "on_progress", ruby_whisper_params_on_progress, 0);
@@ -1,6 +1,15 @@
1
1
  #include <ruby.h>
2
2
  #include "ruby_whisper.h"
3
3
 
4
+ #define N_KEY_NAMES 5
5
+
6
+ static VALUE sym_start_time;
7
+ static VALUE sym_end_time;
8
+ static VALUE sym_text;
9
+ static VALUE sym_no_speech_prob;
10
+ static VALUE sym_speaker_turn_next;
11
+ static VALUE key_names;
12
+
4
13
  extern const rb_data_type_t ruby_whisper_type;
5
14
 
6
15
  extern VALUE cSegment;
@@ -38,7 +47,7 @@ ruby_whisper_segment_allocate(VALUE klass)
38
47
  }
39
48
 
40
49
  VALUE
41
- rb_whisper_segment_initialize(VALUE context, int index)
50
+ rb_whisper_segment_s_new(VALUE context, int index)
42
51
  {
43
52
  ruby_whisper_segment *rws;
44
53
  const VALUE segment = ruby_whisper_segment_allocate(cSegment);
@@ -63,7 +72,7 @@ ruby_whisper_segment_get_start_time(VALUE self)
63
72
  TypedData_Get_Struct(rws->context, ruby_whisper, &ruby_whisper_type, rw);
64
73
  const int64_t t0 = whisper_full_get_segment_t0(rw->context, rws->index);
65
74
  // able to multiply 10 without overflow because to_timestamp() in whisper.cpp does it
66
- return INT2NUM(t0 * 10);
75
+ return LONG2NUM(t0 * 10);
67
76
  }
68
77
 
69
78
  /*
@@ -81,7 +90,7 @@ ruby_whisper_segment_get_end_time(VALUE self)
81
90
  TypedData_Get_Struct(rws->context, ruby_whisper, &ruby_whisper_type, rw);
82
91
  const int64_t t1 = whisper_full_get_segment_t1(rw->context, rws->index);
83
92
  // able to multiply 10 without overflow because to_timestamp() in whisper.cpp does it
84
- return INT2NUM(t1 * 10);
93
+ return LONG2NUM(t1 * 10);
85
94
  }
86
95
 
87
96
  /*
@@ -129,15 +138,83 @@ ruby_whisper_segment_get_no_speech_prob(VALUE self)
129
138
  return DBL2NUM(whisper_full_get_segment_no_speech_prob(rw->context, rws->index));
130
139
  }
131
140
 
141
+ /*
142
+ * call-seq:
143
+ * deconstruct_keys(keys) -> hash
144
+ *
145
+ * Possible keys: :start_time, :end_time, :text, :no_speech_prob, :speaker_turn_next
146
+ *
147
+ * whisper.each_segment do |segment|
148
+ * segment => {start_time:, end_time:, text:, no_speech_prob:, speaker_turn_next:}
149
+ *
150
+ * puts "[#{start_time} --> #{end_time}] #{text} (no speech prob: #{no_speech_prob}#{speaker_turn_next ? ', speaker turns next' : ''})"
151
+ * end
152
+ */
153
+ static VALUE
154
+ ruby_whisper_segment_deconstruct_keys(VALUE self, VALUE keys)
155
+ {
156
+ ruby_whisper_segment *rws;
157
+ TypedData_Get_Struct(self, ruby_whisper_segment, &ruby_whisper_segment_type, rws);
158
+ ruby_whisper *rw;
159
+ TypedData_Get_Struct(rws->context, ruby_whisper, &ruby_whisper_type, rw);
160
+
161
+ VALUE hash = rb_hash_new();
162
+ long n_keys;
163
+ if (NIL_P(keys)) {
164
+ keys = key_names;
165
+ n_keys = N_KEY_NAMES;
166
+ } else {
167
+ n_keys = RARRAY_LEN(keys);
168
+ if (n_keys > N_KEY_NAMES) {
169
+ return hash;
170
+ }
171
+ }
172
+ for (int i = 0; i < n_keys; i++) {
173
+ VALUE key = rb_ary_entry(keys, i);
174
+ if (key == sym_start_time) {
175
+ rb_hash_aset(hash, key, ruby_whisper_segment_get_start_time(self));
176
+ }
177
+ if (key == sym_end_time) {
178
+ rb_hash_aset(hash, key, ruby_whisper_segment_get_end_time(self));
179
+ }
180
+ if (key == sym_text) {
181
+ rb_hash_aset(hash, key, ruby_whisper_segment_get_text(self));
182
+ }
183
+ if (key == sym_no_speech_prob) {
184
+ rb_hash_aset(hash, key, ruby_whisper_segment_get_no_speech_prob(self));
185
+ }
186
+ if (key == sym_speaker_turn_next) {
187
+ rb_hash_aset(hash, key, ruby_whisper_segment_get_speaker_turn_next(self));
188
+ }
189
+ }
190
+
191
+ return hash;
192
+ }
193
+
132
194
  void
133
195
  init_ruby_whisper_segment(VALUE *mWhisper, VALUE *cContext)
134
196
  {
135
197
  cSegment = rb_define_class_under(*mWhisper, "Segment", rb_cObject);
136
198
 
199
+ sym_start_time = ID2SYM(rb_intern("start_time"));
200
+ sym_end_time = ID2SYM(rb_intern("end_time"));
201
+ sym_text = ID2SYM(rb_intern("text"));
202
+ sym_no_speech_prob = ID2SYM(rb_intern("no_speech_prob"));
203
+ sym_speaker_turn_next = ID2SYM(rb_intern("speaker_turn_next"));
204
+ key_names = rb_ary_new3(
205
+ N_KEY_NAMES,
206
+ sym_start_time,
207
+ sym_end_time,
208
+ sym_text,
209
+ sym_no_speech_prob,
210
+ sym_speaker_turn_next
211
+ );
212
+
137
213
  rb_define_alloc_func(cSegment, ruby_whisper_segment_allocate);
138
214
  rb_define_method(cSegment, "start_time", ruby_whisper_segment_get_start_time, 0);
139
215
  rb_define_method(cSegment, "end_time", ruby_whisper_segment_get_end_time, 0);
140
- rb_define_method(cSegment, "speaker_next_turn?", ruby_whisper_segment_get_speaker_turn_next, 0);
216
+ rb_define_method(cSegment, "speaker_turn_next?", ruby_whisper_segment_get_speaker_turn_next, 0);
141
217
  rb_define_method(cSegment, "text", ruby_whisper_segment_get_text, 0);
142
218
  rb_define_method(cSegment, "no_speech_prob", ruby_whisper_segment_get_no_speech_prob, 0);
219
+ rb_define_method(cSegment, "deconstruct_keys", ruby_whisper_segment_deconstruct_keys, 1);
143
220
  }
@@ -13,6 +13,7 @@ extern const rb_data_type_t ruby_whisper_params_type;
13
13
 
14
14
  extern ID id_to_s;
15
15
  extern ID id_call;
16
+ extern ID transcribe_option_names[1];
16
17
 
17
18
  extern void
18
19
  prepare_transcription(ruby_whisper_params * rwp, VALUE * self);
@@ -34,9 +35,14 @@ VALUE
34
35
  ruby_whisper_transcribe(int argc, VALUE *argv, VALUE self) {
35
36
  ruby_whisper *rw;
36
37
  ruby_whisper_params *rwp;
37
- VALUE wave_file_path, blk, params;
38
+ VALUE wave_file_path, blk, params, kws;
39
+ VALUE opts[1];
40
+
41
+ rb_scan_args_kw(RB_SCAN_ARGS_LAST_HASH_KEYWORDS, argc, argv, "2:&", &wave_file_path, &params, &kws, &blk);
42
+ rb_get_kwargs(kws, transcribe_option_names, 0, 1, opts);
43
+
44
+ int n_processors = opts[0] == Qundef ? 1 : NUM2INT(opts[0]);
38
45
 
39
- rb_scan_args(argc, argv, "02&", &wave_file_path, &params, &blk);
40
46
  TypedData_Get_Struct(self, ruby_whisper, &ruby_whisper_type, rw);
41
47
  TypedData_Get_Struct(params, ruby_whisper_params, &ruby_whisper_params_type, rwp);
42
48
 
@@ -66,20 +72,20 @@ ruby_whisper_transcribe(int argc, VALUE *argv, VALUE self) {
66
72
 
67
73
  prepare_transcription(rwp, &self);
68
74
 
69
- if (whisper_full_parallel(rw->context, rwp->params, pcmf32.data(), pcmf32.size(), 1) != 0) {
75
+ if (whisper_full_parallel(rw->context, rwp->params, pcmf32.data(), pcmf32.size(), n_processors) != 0) {
70
76
  fprintf(stderr, "failed to process audio\n");
71
77
  return self;
72
78
  }
79
+ if (NIL_P(blk)) {
80
+ return self;
81
+ }
73
82
  const int n_segments = whisper_full_n_segments(rw->context);
74
83
  VALUE output = rb_str_new2("");
75
84
  for (int i = 0; i < n_segments; ++i) {
76
85
  const char * text = whisper_full_get_segment_text(rw->context, i);
77
86
  output = rb_str_concat(output, rb_str_new2(text));
78
87
  }
79
- VALUE idCall = id_call;
80
- if (blk != Qnil) {
81
- rb_funcall(blk, idCall, 1, output);
82
- }
88
+ rb_funcall(blk, id_call, 1, output);
83
89
  return self;
84
90
  }
85
91
  #ifdef __cplusplus
@@ -249,7 +249,7 @@ ruby_whisper_vad_params_initialize(int argc, VALUE *argv, VALUE self)
249
249
  rb_get_kwargs(kw_hash, param_names, 0, NUM_PARAMS, values);
250
250
 
251
251
  for (i = 0; i < NUM_PARAMS; i++) {
252
- id= param_names[i];
252
+ id = param_names[i];
253
253
  value = values[i];
254
254
  if (value == Qundef) {
255
255
  continue;
@@ -1,6 +1,6 @@
1
1
  cmake_minimum_required(VERSION 3.5) # for add_link_options and implicit target directories.
2
2
  project("whisper.cpp" C CXX)
3
- project("whisper.cpp" VERSION 1.7.5)
3
+ project("whisper.cpp" VERSION 1.8.0)
4
4
  include(CheckIncludeFileCXX)
5
5
 
6
6
  set(SOVERSION 1)
@@ -178,6 +178,10 @@ get_directory_property(WHISPER_TRANSIENT_DEFINES COMPILE_DEFINITIONS)
178
178
  set_target_properties(whisper PROPERTIES PUBLIC_HEADER ${CMAKE_CURRENT_SOURCE_DIR}/include/whisper.h)
179
179
  install(TARGETS whisper LIBRARY PUBLIC_HEADER)
180
180
 
181
+ target_compile_definitions(whisper PRIVATE
182
+ WHISPER_VERSION="${PROJECT_VERSION}"
183
+ )
184
+
181
185
  configure_package_config_file(
182
186
  ${CMAKE_CURRENT_SOURCE_DIR}/cmake/whisper-config.cmake.in
183
187
  ${CMAKE_CURRENT_BINARY_DIR}/whisper-config.cmake
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "whisper.cpp",
3
- "version": "1.7.5",
3
+ "version": "1.8.0",
4
4
  "description": "Whisper speech recognition",
5
5
  "main": "whisper.js",
6
6
  "scripts": {
@@ -15,6 +15,7 @@ GGML_METAL_EMBED_LIBRARY=ON
15
15
  GGML_BLAS_DEFAULT=ON
16
16
  GGML_METAL_USE_BF16=ON
17
17
  GGML_OPENMP=OFF
18
+ BUILD_STATIC_XCFRAMEWORK=${BUILD_STATIC_XCFRAMEWORK:-OFF}
18
19
 
19
20
  COMMON_C_FLAGS="-Wno-macro-redefined -Wno-shorten-64-to-32 -Wno-unused-command-line-argument -g"
20
21
  COMMON_CXX_FLAGS="-Wno-macro-redefined -Wno-shorten-64-to-32 -Wno-unused-command-line-argument -g"
@@ -327,6 +328,15 @@ combine_static_libraries() {
327
328
  arch_flags+=" -arch $arch"
328
329
  done
329
330
 
331
+
332
+ if [[ "${BUILD_STATIC_XCFRAMEWORK}" == "ON" ]]; then
333
+ echo "Packaging static framework for ${platform}."
334
+ mkdir -p "$(dirname "${base_dir}/${output_lib}")"
335
+ cp "${temp_dir}/combined.a" "${base_dir}/${output_lib}"
336
+ rm -rf "${temp_dir}"
337
+ return
338
+ fi
339
+
330
340
  # Create dynamic library
331
341
  echo "Creating dynamic library for ${platform}."
332
342
  xcrun -sdk $sdk clang++ -dynamiclib \
@@ -529,6 +539,20 @@ combine_static_libraries "build-tvos-device" "Release-appletvos" "tvos" "false"
529
539
 
530
540
  # Create XCFramework with correct debug symbols paths
531
541
  echo "Creating XCFramework..."
542
+
543
+ if [[ "${BUILD_STATIC_XCFRAMEWORK}" == "ON" ]]; then
544
+ xcodebuild -create-xcframework \
545
+ -framework $(pwd)/build-ios-sim/framework/whisper.framework \
546
+ -framework $(pwd)/build-ios-device/framework/whisper.framework \
547
+ -framework $(pwd)/build-macos/framework/whisper.framework \
548
+ -framework $(pwd)/build-visionos/framework/whisper.framework \
549
+ -framework $(pwd)/build-visionos-sim/framework/whisper.framework \
550
+ -framework $(pwd)/build-tvos-device/framework/whisper.framework \
551
+ -framework $(pwd)/build-tvos-sim/framework/whisper.framework \
552
+ -output $(pwd)/build-apple/whisper.xcframework
553
+ exit 0
554
+ fi
555
+
532
556
  xcodebuild -create-xcframework \
533
557
  -framework $(pwd)/build-ios-sim/framework/whisper.framework \
534
558
  -debug-symbols $(pwd)/build-ios-sim/dSYMs/whisper.dSYM \
@@ -98,6 +98,7 @@ if (EMSCRIPTEN)
98
98
  add_subdirectory(stream.wasm)
99
99
  add_subdirectory(command.wasm)
100
100
  add_subdirectory(bench.wasm)
101
+ add_subdirectory(wchess)
101
102
  elseif(CMAKE_JS_VERSION)
102
103
  add_subdirectory(addon.node)
103
104
  else()