whispercpp 1.3.5 → 1.3.6

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 (610) hide show
  1. checksums.yaml +4 -4
  2. data/LICENSE +1 -1
  3. data/README.md +99 -2
  4. data/ext/extconf.rb +1 -0
  5. data/ext/ruby_whisper.c +20 -4
  6. data/ext/ruby_whisper.h +30 -2
  7. data/ext/ruby_whisper_context.c +216 -124
  8. data/ext/ruby_whisper_context_params.c +163 -0
  9. data/ext/ruby_whisper_model.c +0 -1
  10. data/ext/ruby_whisper_params.c +0 -1
  11. data/ext/ruby_whisper_segment.c +0 -1
  12. data/ext/ruby_whisper_token.c +29 -9
  13. data/ext/ruby_whisper_transcribe.cpp +4 -1
  14. data/ext/ruby_whisper_vad_context.c +48 -1
  15. data/ext/ruby_whisper_vad_context_detect.cpp +6 -5
  16. data/ext/ruby_whisper_vad_params.c +0 -1
  17. data/ext/ruby_whisper_vad_segment.c +0 -1
  18. data/ext/ruby_whisper_vad_segments.c +0 -1
  19. data/ext/sources/CMakeLists.txt +1 -1
  20. data/ext/sources/bindings/javascript/package.json +1 -1
  21. data/ext/sources/cmake/whisper-config.cmake.in +5 -40
  22. data/ext/sources/examples/bench/bench.cpp +23 -18
  23. data/ext/sources/examples/cli/cli.cpp +8 -0
  24. data/ext/sources/examples/common-ggml.cpp +2 -0
  25. data/ext/sources/examples/miniaudio.h +4507 -2131
  26. data/ext/sources/examples/server/server.cpp +18 -4
  27. data/ext/sources/examples/talk-llama/CMakeLists.txt +3 -2
  28. data/ext/sources/examples/talk-llama/llama-adapter.cpp +7 -13
  29. data/ext/sources/examples/talk-llama/llama-adapter.h +4 -3
  30. data/ext/sources/examples/talk-llama/llama-arch.cpp +335 -17
  31. data/ext/sources/examples/talk-llama/llama-arch.h +42 -0
  32. data/ext/sources/examples/talk-llama/llama-batch.cpp +3 -1
  33. data/ext/sources/examples/talk-llama/llama-chat.cpp +21 -1
  34. data/ext/sources/examples/talk-llama/llama-chat.h +1 -0
  35. data/ext/sources/examples/talk-llama/llama-context.cpp +508 -520
  36. data/ext/sources/examples/talk-llama/llama-context.h +27 -28
  37. data/ext/sources/examples/talk-llama/llama-cparams.h +5 -0
  38. data/ext/sources/examples/talk-llama/llama-ext.h +12 -0
  39. data/ext/sources/examples/talk-llama/llama-grammar.cpp +8 -8
  40. data/ext/sources/examples/talk-llama/llama-graph.cpp +583 -130
  41. data/ext/sources/examples/talk-llama/llama-graph.h +131 -10
  42. data/ext/sources/examples/talk-llama/llama-hparams.cpp +57 -40
  43. data/ext/sources/examples/talk-llama/llama-hparams.h +79 -10
  44. data/ext/sources/examples/talk-llama/llama-impl.cpp +4 -4
  45. data/ext/sources/examples/talk-llama/llama-impl.h +13 -1
  46. data/ext/sources/examples/talk-llama/llama-kv-cache-iswa.cpp +3 -1
  47. data/ext/sources/examples/talk-llama/llama-kv-cache.cpp +274 -89
  48. data/ext/sources/examples/talk-llama/llama-kv-cache.h +2 -3
  49. data/ext/sources/examples/talk-llama/llama-memory-hybrid-iswa.cpp +275 -0
  50. data/ext/sources/examples/talk-llama/llama-memory-hybrid-iswa.h +140 -0
  51. data/ext/sources/examples/talk-llama/llama-memory-recurrent.cpp +11 -13
  52. data/ext/sources/examples/talk-llama/llama-mmap.cpp +28 -11
  53. data/ext/sources/examples/talk-llama/llama-model-loader.cpp +527 -119
  54. data/ext/sources/examples/talk-llama/llama-model-loader.h +35 -5
  55. data/ext/sources/examples/talk-llama/llama-model-saver.cpp +60 -46
  56. data/ext/sources/examples/talk-llama/llama-model-saver.h +5 -2
  57. data/ext/sources/examples/talk-llama/llama-model.cpp +1365 -647
  58. data/ext/sources/examples/talk-llama/llama-model.h +72 -19
  59. data/ext/sources/examples/talk-llama/llama-quant.cpp +578 -346
  60. data/ext/sources/examples/talk-llama/{llama-sampling.cpp → llama-sampler.cpp} +190 -76
  61. data/ext/sources/examples/talk-llama/{llama-sampling.h → llama-sampler.h} +0 -2
  62. data/ext/sources/examples/talk-llama/llama-vocab.cpp +118 -48
  63. data/ext/sources/examples/talk-llama/llama-vocab.h +5 -0
  64. data/ext/sources/examples/talk-llama/llama.cpp +76 -22
  65. data/ext/sources/examples/talk-llama/llama.h +63 -30
  66. data/ext/sources/examples/talk-llama/models/afmoe.cpp +2 -3
  67. data/ext/sources/examples/talk-llama/models/apertus.cpp +3 -3
  68. data/ext/sources/examples/talk-llama/models/arcee.cpp +3 -3
  69. data/ext/sources/examples/talk-llama/models/arctic.cpp +4 -5
  70. data/ext/sources/examples/talk-llama/models/baichuan.cpp +4 -3
  71. data/ext/sources/examples/talk-llama/models/bailingmoe.cpp +1 -2
  72. data/ext/sources/examples/talk-llama/models/bailingmoe2.cpp +3 -5
  73. data/ext/sources/examples/talk-llama/models/bert.cpp +13 -7
  74. data/ext/sources/examples/talk-llama/models/bitnet.cpp +9 -24
  75. data/ext/sources/examples/talk-llama/models/bloom.cpp +2 -2
  76. data/ext/sources/examples/talk-llama/models/chameleon.cpp +3 -3
  77. data/ext/sources/examples/talk-llama/models/chatglm.cpp +2 -2
  78. data/ext/sources/examples/talk-llama/models/codeshell.cpp +3 -3
  79. data/ext/sources/examples/talk-llama/models/cogvlm.cpp +3 -3
  80. data/ext/sources/examples/talk-llama/models/cohere2-iswa.cpp +2 -2
  81. data/ext/sources/examples/talk-llama/models/command-r.cpp +2 -2
  82. data/ext/sources/examples/talk-llama/models/dbrx.cpp +4 -5
  83. data/ext/sources/examples/talk-llama/models/deci.cpp +3 -3
  84. data/ext/sources/examples/talk-llama/models/deepseek.cpp +4 -6
  85. data/ext/sources/examples/talk-llama/models/deepseek2.cpp +24 -21
  86. data/ext/sources/examples/talk-llama/models/delta-net-base.cpp +445 -0
  87. data/ext/sources/examples/talk-llama/models/dots1.cpp +4 -6
  88. data/ext/sources/examples/talk-llama/models/dream.cpp +3 -3
  89. data/ext/sources/examples/talk-llama/models/ernie4-5-moe.cpp +4 -6
  90. data/ext/sources/examples/talk-llama/models/ernie4-5.cpp +3 -3
  91. data/ext/sources/examples/talk-llama/models/eurobert.cpp +97 -0
  92. data/ext/sources/examples/talk-llama/models/exaone-moe.cpp +145 -0
  93. data/ext/sources/examples/talk-llama/models/exaone.cpp +3 -3
  94. data/ext/sources/examples/talk-llama/models/exaone4.cpp +3 -3
  95. data/ext/sources/examples/talk-llama/models/falcon-h1.cpp +2 -4
  96. data/ext/sources/examples/talk-llama/models/falcon.cpp +3 -3
  97. data/ext/sources/examples/talk-llama/models/gemma-embedding.cpp +1 -1
  98. data/ext/sources/examples/talk-llama/models/gemma.cpp +1 -1
  99. data/ext/sources/examples/talk-llama/models/gemma2-iswa.cpp +1 -1
  100. data/ext/sources/examples/talk-llama/models/gemma3.cpp +1 -1
  101. data/ext/sources/examples/talk-llama/models/gemma3n-iswa.cpp +7 -7
  102. data/ext/sources/examples/talk-llama/models/glm4-moe.cpp +3 -3
  103. data/ext/sources/examples/talk-llama/models/glm4.cpp +14 -7
  104. data/ext/sources/examples/talk-llama/models/gpt2.cpp +2 -2
  105. data/ext/sources/examples/talk-llama/models/gptneox.cpp +2 -2
  106. data/ext/sources/examples/talk-llama/models/granite-hybrid.cpp +4 -5
  107. data/ext/sources/examples/talk-llama/models/granite.cpp +4 -5
  108. data/ext/sources/examples/talk-llama/models/grok.cpp +4 -4
  109. data/ext/sources/examples/talk-llama/models/grovemoe.cpp +5 -7
  110. data/ext/sources/examples/talk-llama/models/hunyuan-dense.cpp +3 -3
  111. data/ext/sources/examples/talk-llama/models/hunyuan-moe.cpp +4 -5
  112. data/ext/sources/examples/talk-llama/models/internlm2.cpp +3 -3
  113. data/ext/sources/examples/talk-llama/models/jais.cpp +2 -2
  114. data/ext/sources/examples/talk-llama/models/jais2.cpp +123 -0
  115. data/ext/sources/examples/talk-llama/models/jamba.cpp +3 -3
  116. data/ext/sources/examples/talk-llama/models/kimi-linear.cpp +381 -0
  117. data/ext/sources/examples/talk-llama/models/lfm2.cpp +145 -124
  118. data/ext/sources/examples/talk-llama/models/llada-moe.cpp +4 -4
  119. data/ext/sources/examples/talk-llama/models/llada.cpp +3 -3
  120. data/ext/sources/examples/talk-llama/models/llama-iswa.cpp +4 -4
  121. data/ext/sources/examples/talk-llama/models/llama.cpp +18 -11
  122. data/ext/sources/examples/talk-llama/models/maincoder.cpp +3 -3
  123. data/ext/sources/examples/talk-llama/models/{graph-context-mamba.cpp → mamba-base.cpp} +9 -3
  124. data/ext/sources/examples/talk-llama/models/mamba.cpp +1 -2
  125. data/ext/sources/examples/talk-llama/models/mimo2-iswa.cpp +11 -5
  126. data/ext/sources/examples/talk-llama/models/minicpm3.cpp +14 -13
  127. data/ext/sources/examples/talk-llama/models/minimax-m2.cpp +4 -5
  128. data/ext/sources/examples/talk-llama/models/mistral3.cpp +4 -4
  129. data/ext/sources/examples/talk-llama/models/models.h +181 -46
  130. data/ext/sources/examples/talk-llama/models/modern-bert.cpp +2 -9
  131. data/ext/sources/examples/talk-llama/models/mpt.cpp +2 -2
  132. data/ext/sources/examples/talk-llama/models/nemotron-h.cpp +26 -14
  133. data/ext/sources/examples/talk-llama/models/nemotron.cpp +3 -3
  134. data/ext/sources/examples/talk-llama/models/neo-bert.cpp +2 -2
  135. data/ext/sources/examples/talk-llama/models/olmo.cpp +3 -3
  136. data/ext/sources/examples/talk-llama/models/olmo2.cpp +3 -3
  137. data/ext/sources/examples/talk-llama/models/olmoe.cpp +4 -4
  138. data/ext/sources/examples/talk-llama/models/openai-moe-iswa.cpp +1 -1
  139. data/ext/sources/examples/talk-llama/models/openelm.cpp +3 -3
  140. data/ext/sources/examples/talk-llama/models/orion.cpp +3 -3
  141. data/ext/sources/examples/talk-llama/models/paddleocr.cpp +122 -0
  142. data/ext/sources/examples/talk-llama/models/pangu-embedded.cpp +3 -3
  143. data/ext/sources/examples/talk-llama/models/phi2.cpp +2 -2
  144. data/ext/sources/examples/talk-llama/models/phi3.cpp +3 -3
  145. data/ext/sources/examples/talk-llama/models/plamo.cpp +3 -3
  146. data/ext/sources/examples/talk-llama/models/plamo2.cpp +9 -5
  147. data/ext/sources/examples/talk-llama/models/plamo3.cpp +2 -2
  148. data/ext/sources/examples/talk-llama/models/plm.cpp +15 -14
  149. data/ext/sources/examples/talk-llama/models/qwen.cpp +2 -2
  150. data/ext/sources/examples/talk-llama/models/qwen2.cpp +3 -3
  151. data/ext/sources/examples/talk-llama/models/qwen2moe.cpp +4 -4
  152. data/ext/sources/examples/talk-llama/models/qwen2vl.cpp +3 -3
  153. data/ext/sources/examples/talk-llama/models/qwen3.cpp +12 -9
  154. data/ext/sources/examples/talk-llama/models/qwen35.cpp +381 -0
  155. data/ext/sources/examples/talk-llama/models/qwen35moe.cpp +422 -0
  156. data/ext/sources/examples/talk-llama/models/qwen3moe.cpp +15 -8
  157. data/ext/sources/examples/talk-llama/models/qwen3next.cpp +84 -432
  158. data/ext/sources/examples/talk-llama/models/qwen3vl-moe.cpp +9 -18
  159. data/ext/sources/examples/talk-llama/models/qwen3vl.cpp +8 -17
  160. data/ext/sources/examples/talk-llama/models/refact.cpp +2 -2
  161. data/ext/sources/examples/talk-llama/models/rnd1.cpp +4 -4
  162. data/ext/sources/examples/talk-llama/models/rwkv6-base.cpp +2 -0
  163. data/ext/sources/examples/talk-llama/models/rwkv7-base.cpp +2 -0
  164. data/ext/sources/examples/talk-llama/models/seed-oss.cpp +3 -3
  165. data/ext/sources/examples/talk-llama/models/smallthinker.cpp +4 -4
  166. data/ext/sources/examples/talk-llama/models/smollm3.cpp +3 -3
  167. data/ext/sources/examples/talk-llama/models/stablelm.cpp +2 -2
  168. data/ext/sources/examples/talk-llama/models/starcoder.cpp +2 -2
  169. data/ext/sources/examples/talk-llama/models/starcoder2.cpp +3 -3
  170. data/ext/sources/examples/talk-llama/models/step35-iswa.cpp +165 -0
  171. data/ext/sources/examples/talk-llama/models/t5-dec.cpp +2 -2
  172. data/ext/sources/examples/talk-llama/models/t5-enc.cpp +2 -2
  173. data/ext/sources/examples/talk-llama/models/xverse.cpp +3 -3
  174. data/ext/sources/examples/talk-llama/unicode.cpp +21 -65
  175. data/ext/sources/ggml/CMakeLists.txt +9 -3
  176. data/ext/sources/ggml/include/ggml-backend.h +1 -1
  177. data/ext/sources/ggml/include/ggml-cann.h +1 -1
  178. data/ext/sources/ggml/include/ggml-cpu.h +5 -0
  179. data/ext/sources/ggml/include/ggml-openvino.h +37 -0
  180. data/ext/sources/ggml/include/ggml-opt.h +1 -1
  181. data/ext/sources/ggml/include/ggml-rpc.h +6 -1
  182. data/ext/sources/ggml/include/ggml-virtgpu.h +14 -0
  183. data/ext/sources/ggml/include/ggml.h +56 -9
  184. data/ext/sources/ggml/src/CMakeLists.txt +3 -0
  185. data/ext/sources/ggml/src/ggml-alloc.c +4 -9
  186. data/ext/sources/ggml/src/ggml-backend-dl.cpp +48 -0
  187. data/ext/sources/ggml/src/ggml-backend-dl.h +45 -0
  188. data/ext/sources/ggml/src/ggml-backend-reg.cpp +28 -86
  189. data/ext/sources/ggml/src/ggml-backend.cpp +5 -2
  190. data/ext/sources/ggml/src/ggml-blas/CMakeLists.txt +1 -1
  191. data/ext/sources/ggml/src/ggml-blas/ggml-blas.cpp +6 -2
  192. data/ext/sources/ggml/src/ggml-cann/acl_tensor.cpp +1 -1
  193. data/ext/sources/ggml/src/ggml-cann/acl_tensor.h +1 -1
  194. data/ext/sources/ggml/src/ggml-cann/aclnn_ops.cpp +348 -189
  195. data/ext/sources/ggml/src/ggml-cann/aclnn_ops.h +40 -85
  196. data/ext/sources/ggml/src/ggml-cann/common.h +3 -4
  197. data/ext/sources/ggml/src/ggml-cann/ggml-cann.cpp +44 -62
  198. data/ext/sources/ggml/src/ggml-common.h +11 -0
  199. data/ext/sources/ggml/src/ggml-cpu/CMakeLists.txt +16 -11
  200. data/ext/sources/ggml/src/ggml-cpu/amx/amx.cpp +42 -19
  201. data/ext/sources/ggml/src/ggml-cpu/amx/common.h +34 -10
  202. data/ext/sources/ggml/src/ggml-cpu/amx/mmq.cpp +85 -85
  203. data/ext/sources/ggml/src/ggml-cpu/arch/arm/quants.c +85 -1
  204. data/ext/sources/ggml/src/ggml-cpu/arch/arm/repack.cpp +2744 -548
  205. data/ext/sources/ggml/src/ggml-cpu/arch/riscv/quants.c +1653 -0
  206. data/ext/sources/ggml/src/ggml-cpu/arch/riscv/repack.cpp +1391 -0
  207. data/ext/sources/ggml/src/ggml-cpu/arch/s390/quants.c +8 -10
  208. data/ext/sources/ggml/src/ggml-cpu/arch/x86/quants.c +9 -9
  209. data/ext/sources/ggml/src/ggml-cpu/arch/x86/repack.cpp +118 -18
  210. data/ext/sources/ggml/src/ggml-cpu/arch-fallback.h +107 -26
  211. data/ext/sources/ggml/src/ggml-cpu/binary-ops.cpp +2 -6
  212. data/ext/sources/ggml/src/ggml-cpu/common.h +8 -0
  213. data/ext/sources/ggml/src/ggml-cpu/ggml-cpu-impl.h +3 -0
  214. data/ext/sources/ggml/src/ggml-cpu/ggml-cpu.c +59 -12
  215. data/ext/sources/ggml/src/ggml-cpu/ggml-cpu.cpp +15 -0
  216. data/ext/sources/ggml/src/ggml-cpu/kleidiai/kernels.cpp +21 -20
  217. data/ext/sources/ggml/src/ggml-cpu/kleidiai/kleidiai.cpp +965 -252
  218. data/ext/sources/ggml/src/ggml-cpu/llamafile/sgemm.cpp +584 -197
  219. data/ext/sources/ggml/src/ggml-cpu/ops.cpp +903 -188
  220. data/ext/sources/ggml/src/ggml-cpu/ops.h +1 -0
  221. data/ext/sources/ggml/src/ggml-cpu/quants.c +40 -0
  222. data/ext/sources/ggml/src/ggml-cpu/quants.h +3 -0
  223. data/ext/sources/ggml/src/ggml-cpu/repack.cpp +2890 -679
  224. data/ext/sources/ggml/src/ggml-cpu/repack.h +119 -8
  225. data/ext/sources/ggml/src/ggml-cpu/simd-gemm.h +136 -0
  226. data/ext/sources/ggml/src/ggml-cpu/simd-mappings.h +111 -3
  227. data/ext/sources/ggml/src/ggml-cpu/unary-ops.cpp +1 -1
  228. data/ext/sources/ggml/src/ggml-cpu/vec.cpp +17 -0
  229. data/ext/sources/ggml/src/ggml-cuda/CMakeLists.txt +1 -1
  230. data/ext/sources/ggml/src/ggml-cuda/argsort.cu +19 -10
  231. data/ext/sources/ggml/src/ggml-cuda/binbcast.cu +32 -30
  232. data/ext/sources/ggml/src/ggml-cuda/common.cuh +134 -18
  233. data/ext/sources/ggml/src/ggml-cuda/convert.cu +41 -27
  234. data/ext/sources/ggml/src/ggml-cuda/cpy.cu +6 -3
  235. data/ext/sources/ggml/src/ggml-cuda/fattn-common.cuh +78 -64
  236. data/ext/sources/ggml/src/ggml-cuda/fattn-mma-f16.cuh +384 -143
  237. data/ext/sources/ggml/src/ggml-cuda/fattn-tile.cuh +36 -22
  238. data/ext/sources/ggml/src/ggml-cuda/fattn-vec.cuh +3 -3
  239. data/ext/sources/ggml/src/ggml-cuda/fattn-wmma-f16.cu +26 -5
  240. data/ext/sources/ggml/src/ggml-cuda/fattn-wmma-f16.cuh +1 -1
  241. data/ext/sources/ggml/src/ggml-cuda/fattn.cu +127 -12
  242. data/ext/sources/ggml/src/ggml-cuda/gated_delta_net.cu +263 -0
  243. data/ext/sources/ggml/src/ggml-cuda/gated_delta_net.cuh +4 -0
  244. data/ext/sources/ggml/src/ggml-cuda/ggml-cuda.cu +595 -200
  245. data/ext/sources/ggml/src/ggml-cuda/mean.cu +9 -8
  246. data/ext/sources/ggml/src/ggml-cuda/mma.cuh +173 -6
  247. data/ext/sources/ggml/src/ggml-cuda/mmf.cu +30 -10
  248. data/ext/sources/ggml/src/ggml-cuda/mmf.cuh +158 -85
  249. data/ext/sources/ggml/src/ggml-cuda/mmq.cuh +34 -22
  250. data/ext/sources/ggml/src/ggml-cuda/mmvf.cu +127 -67
  251. data/ext/sources/ggml/src/ggml-cuda/mmvf.cuh +2 -0
  252. data/ext/sources/ggml/src/ggml-cuda/mmvq.cu +157 -65
  253. data/ext/sources/ggml/src/ggml-cuda/mmvq.cuh +1 -0
  254. data/ext/sources/ggml/src/ggml-cuda/norm.cu +18 -76
  255. data/ext/sources/ggml/src/ggml-cuda/pad.cu +13 -10
  256. data/ext/sources/ggml/src/ggml-cuda/quantize.cu +1 -1
  257. data/ext/sources/ggml/src/ggml-cuda/reduce_rows.cuh +2 -16
  258. data/ext/sources/ggml/src/ggml-cuda/rope.cu +233 -133
  259. data/ext/sources/ggml/src/ggml-cuda/softmax.cu +8 -83
  260. data/ext/sources/ggml/src/ggml-cuda/solve_tri.cu +1 -1
  261. data/ext/sources/ggml/src/ggml-cuda/ssm-conv.cu +56 -32
  262. data/ext/sources/ggml/src/ggml-cuda/ssm-conv.cuh +1 -1
  263. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_1-ncols2_32.cu +5 -0
  264. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_16-ncols2_4.cu +1 -0
  265. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_2-ncols2_32.cu +5 -0
  266. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_2-ncols2_4.cu +1 -0
  267. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_4-ncols2_4.cu +1 -0
  268. data/ext/sources/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_8-ncols2_4.cu +1 -0
  269. data/ext/sources/ggml/src/ggml-cuda/template-instances/generate_cu_files.py +3 -3
  270. data/ext/sources/ggml/src/ggml-cuda/top-k.cu +0 -1
  271. data/ext/sources/ggml/src/ggml-cuda/topk-moe.cu +199 -135
  272. data/ext/sources/ggml/src/ggml-cuda/topk-moe.cuh +20 -14
  273. data/ext/sources/ggml/src/ggml-cuda/unary.cu +55 -0
  274. data/ext/sources/ggml/src/ggml-cuda/unary.cuh +2 -0
  275. data/ext/sources/ggml/src/ggml-cuda/vecdotq.cuh +31 -17
  276. data/ext/sources/ggml/src/ggml-cuda/vendors/hip.h +10 -0
  277. data/ext/sources/ggml/src/ggml-hexagon/CMakeLists.txt +82 -45
  278. data/ext/sources/ggml/src/ggml-hexagon/ggml-hexagon.cpp +334 -160
  279. data/ext/sources/ggml/src/ggml-hexagon/htp/CMakeLists.txt +7 -5
  280. data/ext/sources/ggml/src/ggml-hexagon/htp/act-ops.c +328 -197
  281. data/ext/sources/ggml/src/ggml-hexagon/htp/argsort-ops.c +281 -0
  282. data/ext/sources/ggml/src/ggml-hexagon/htp/binary-ops.c +765 -234
  283. data/ext/sources/ggml/src/ggml-hexagon/htp/cpy-ops.c +252 -0
  284. data/ext/sources/ggml/src/ggml-hexagon/htp/flash-attn-ops.c +412 -265
  285. data/ext/sources/ggml/src/ggml-hexagon/htp/get-rows-ops.c +23 -23
  286. data/ext/sources/ggml/src/ggml-hexagon/htp/{htp-dma.c → hex-dma.c} +1 -1
  287. data/ext/sources/ggml/src/ggml-hexagon/htp/{htp-dma.h → hex-dma.h} +28 -3
  288. data/ext/sources/ggml/src/ggml-hexagon/htp/hex-dump.h +77 -0
  289. data/ext/sources/ggml/src/ggml-hexagon/htp/hex-fastdiv.h +37 -0
  290. data/ext/sources/ggml/src/ggml-hexagon/htp/hex-utils.h +51 -0
  291. data/ext/sources/ggml/src/ggml-hexagon/htp/htp-ctx.h +1 -1
  292. data/ext/sources/ggml/src/ggml-hexagon/htp/htp-msg.h +27 -37
  293. data/ext/sources/ggml/src/ggml-hexagon/htp/htp-ops.h +6 -35
  294. data/ext/sources/ggml/src/ggml-hexagon/htp/hvx-arith.h +443 -0
  295. data/ext/sources/ggml/src/ggml-hexagon/htp/hvx-base.h +240 -0
  296. data/ext/sources/ggml/src/ggml-hexagon/htp/hvx-copy.h +245 -0
  297. data/ext/sources/ggml/src/ggml-hexagon/htp/hvx-div.h +251 -0
  298. data/ext/sources/ggml/src/ggml-hexagon/htp/hvx-dump.h +129 -0
  299. data/ext/sources/ggml/src/ggml-hexagon/htp/hvx-exp.h +215 -0
  300. data/ext/sources/ggml/src/ggml-hexagon/htp/hvx-floor.h +100 -0
  301. data/ext/sources/ggml/src/ggml-hexagon/htp/hvx-inverse.h +210 -0
  302. data/ext/sources/ggml/src/ggml-hexagon/htp/hvx-reduce.h +296 -0
  303. data/ext/sources/ggml/src/ggml-hexagon/htp/hvx-scale.h +133 -0
  304. data/ext/sources/ggml/src/ggml-hexagon/htp/hvx-sigmoid.h +141 -0
  305. data/ext/sources/ggml/src/ggml-hexagon/htp/hvx-sqrt.h +126 -0
  306. data/ext/sources/ggml/src/ggml-hexagon/htp/hvx-types.h +36 -0
  307. data/ext/sources/ggml/src/ggml-hexagon/htp/hvx-utils.h +20 -1347
  308. data/ext/sources/ggml/src/ggml-hexagon/htp/main.c +211 -13
  309. data/ext/sources/ggml/src/ggml-hexagon/htp/matmul-ops.c +1119 -952
  310. data/ext/sources/ggml/src/ggml-hexagon/htp/rope-ops.c +254 -244
  311. data/ext/sources/ggml/src/ggml-hexagon/htp/set-rows-ops.c +36 -36
  312. data/ext/sources/ggml/src/ggml-hexagon/htp/softmax-ops.c +155 -138
  313. data/ext/sources/ggml/src/ggml-hexagon/htp/ssm-conv.c +339 -0
  314. data/ext/sources/ggml/src/ggml-hexagon/htp/sum-rows-ops.c +128 -0
  315. data/ext/sources/ggml/src/ggml-hexagon/htp/unary-ops.c +209 -114
  316. data/ext/sources/ggml/src/ggml-hexagon/htp/worker-pool.c +1 -5
  317. data/ext/sources/ggml/src/ggml-hexagon/htp-drv.cpp +418 -0
  318. data/ext/sources/ggml/src/ggml-hexagon/htp-drv.h +121 -0
  319. data/ext/sources/ggml/src/ggml-hexagon/libdl.h +79 -0
  320. data/ext/sources/ggml/src/ggml-hexagon/libggml-htp.inf +38 -0
  321. data/ext/sources/ggml/src/ggml-hip/CMakeLists.txt +6 -0
  322. data/ext/sources/ggml/src/ggml-impl.h +62 -0
  323. data/ext/sources/ggml/src/ggml-metal/CMakeLists.txt +10 -10
  324. data/ext/sources/ggml/src/ggml-metal/ggml-metal-common.cpp +13 -2
  325. data/ext/sources/ggml/src/ggml-metal/ggml-metal-context.h +8 -0
  326. data/ext/sources/ggml/src/ggml-metal/ggml-metal-context.m +147 -17
  327. data/ext/sources/ggml/src/ggml-metal/ggml-metal-device.cpp +274 -73
  328. data/ext/sources/ggml/src/ggml-metal/ggml-metal-device.h +22 -4
  329. data/ext/sources/ggml/src/ggml-metal/ggml-metal-device.m +102 -36
  330. data/ext/sources/ggml/src/ggml-metal/ggml-metal-impl.h +174 -23
  331. data/ext/sources/ggml/src/ggml-metal/ggml-metal-ops.cpp +580 -280
  332. data/ext/sources/ggml/src/ggml-metal/ggml-metal-ops.h +5 -4
  333. data/ext/sources/ggml/src/ggml-metal/ggml-metal.cpp +320 -107
  334. data/ext/sources/ggml/src/ggml-metal/ggml-metal.metal +1068 -825
  335. data/ext/sources/ggml/src/ggml-opencl/CMakeLists.txt +19 -1
  336. data/ext/sources/ggml/src/ggml-opencl/ggml-opencl.cpp +3108 -636
  337. data/ext/sources/ggml/src/ggml-opencl/kernels/concat.cl +41 -99
  338. data/ext/sources/ggml/src/ggml-opencl/kernels/cpy.cl +45 -0
  339. data/ext/sources/ggml/src/ggml-opencl/kernels/cumsum.cl +139 -0
  340. data/ext/sources/ggml/src/ggml-opencl/kernels/cvt.cl +204 -0
  341. data/ext/sources/ggml/src/ggml-opencl/kernels/diag.cl +27 -0
  342. data/ext/sources/ggml/src/ggml-opencl/kernels/exp.cl +125 -0
  343. data/ext/sources/ggml/src/ggml-opencl/kernels/expm1.cl +87 -56
  344. data/ext/sources/ggml/src/ggml-opencl/kernels/gemm_noshuffle_q4_1_f32.cl +132 -0
  345. data/ext/sources/ggml/src/ggml-opencl/kernels/gemv_noshuffle_general_q8_0_f32.cl +195 -0
  346. data/ext/sources/ggml/src/ggml-opencl/kernels/gemv_noshuffle_q4_1_f32.cl +283 -0
  347. data/ext/sources/ggml/src/ggml-opencl/kernels/l2_norm.cl +71 -0
  348. data/ext/sources/ggml/src/ggml-opencl/kernels/mean.cl +114 -13
  349. data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mm_q4_0_f32_l4_lm.cl +163 -0
  350. data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mm_q4_1_f32_l4_lm.cl +165 -0
  351. data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mm_q6_k_f32_l4_lm.cl +158 -0
  352. data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mm_q8_0_f32_8x4.cl +129 -0
  353. data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_q4_1_f32.cl +219 -0
  354. data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_q4_1_f32_flat.cl +229 -0
  355. data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_q4_k_f32.cl +180 -0
  356. data/ext/sources/ggml/src/ggml-opencl/kernels/{mul_mv_q6_k.cl → mul_mv_q6_k_f32.cl} +4 -0
  357. data/ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_q6_k_f32_flat.cl +194 -0
  358. data/ext/sources/ggml/src/ggml-opencl/kernels/neg.cl +125 -0
  359. data/ext/sources/ggml/src/ggml-opencl/kernels/repeat.cl +31 -32
  360. data/ext/sources/ggml/src/ggml-opencl/kernels/scale.cl +14 -4
  361. data/ext/sources/ggml/src/ggml-opencl/kernels/softplus.cl +88 -60
  362. data/ext/sources/ggml/src/ggml-opencl/kernels/solve_tri.cl +51 -0
  363. data/ext/sources/ggml/src/ggml-opencl/kernels/sum_rows.cl +114 -13
  364. data/ext/sources/ggml/src/ggml-opencl/kernels/tanh.cl +94 -48
  365. data/ext/sources/ggml/src/ggml-opencl/kernels/transpose.cl +26 -0
  366. data/ext/sources/ggml/src/ggml-opencl/kernels/tri.cl +32 -0
  367. data/ext/sources/ggml/src/ggml-openvino/.clang-format +154 -0
  368. data/ext/sources/ggml/src/ggml-openvino/CMakeLists.txt +22 -0
  369. data/ext/sources/ggml/src/ggml-openvino/ggml-decoder.cpp +975 -0
  370. data/ext/sources/ggml/src/ggml-openvino/ggml-decoder.h +294 -0
  371. data/ext/sources/ggml/src/ggml-openvino/ggml-openvino-extra.cpp +373 -0
  372. data/ext/sources/ggml/src/ggml-openvino/ggml-openvino-extra.h +182 -0
  373. data/ext/sources/ggml/src/ggml-openvino/ggml-openvino.cpp +1110 -0
  374. data/ext/sources/ggml/src/ggml-openvino/ggml-quants.cpp +884 -0
  375. data/ext/sources/ggml/src/ggml-openvino/ggml-quants.h +153 -0
  376. data/ext/sources/ggml/src/ggml-openvino/openvino/decoder.h +74 -0
  377. data/ext/sources/ggml/src/ggml-openvino/openvino/frontend.cpp +27 -0
  378. data/ext/sources/ggml/src/ggml-openvino/openvino/frontend.h +23 -0
  379. data/ext/sources/ggml/src/ggml-openvino/openvino/input_model.cpp +17 -0
  380. data/ext/sources/ggml/src/ggml-openvino/openvino/input_model.h +29 -0
  381. data/ext/sources/ggml/src/ggml-openvino/openvino/node_context.h +112 -0
  382. data/ext/sources/ggml/src/ggml-openvino/openvino/op/cont.cpp +48 -0
  383. data/ext/sources/ggml/src/ggml-openvino/openvino/op/cpy.cpp +21 -0
  384. data/ext/sources/ggml/src/ggml-openvino/openvino/op/flash_attn_ext.cpp +90 -0
  385. data/ext/sources/ggml/src/ggml-openvino/openvino/op/get_rows.cpp +69 -0
  386. data/ext/sources/ggml/src/ggml-openvino/openvino/op/glu_geglu.cpp +61 -0
  387. data/ext/sources/ggml/src/ggml-openvino/openvino/op/glu_swiglu.cpp +62 -0
  388. data/ext/sources/ggml/src/ggml-openvino/openvino/op/mulmat.cpp +90 -0
  389. data/ext/sources/ggml/src/ggml-openvino/openvino/op/permute.cpp +102 -0
  390. data/ext/sources/ggml/src/ggml-openvino/openvino/op/reshape.cpp +83 -0
  391. data/ext/sources/ggml/src/ggml-openvino/openvino/op/rms_norm.cpp +46 -0
  392. data/ext/sources/ggml/src/ggml-openvino/openvino/op/rope.cpp +123 -0
  393. data/ext/sources/ggml/src/ggml-openvino/openvino/op/scale.cpp +41 -0
  394. data/ext/sources/ggml/src/ggml-openvino/openvino/op/set_rows.cpp +76 -0
  395. data/ext/sources/ggml/src/ggml-openvino/openvino/op/softmax.cpp +89 -0
  396. data/ext/sources/ggml/src/ggml-openvino/openvino/op/transpose.cpp +23 -0
  397. data/ext/sources/ggml/src/ggml-openvino/openvino/op/unary_silu.cpp +27 -0
  398. data/ext/sources/ggml/src/ggml-openvino/openvino/op/view.cpp +53 -0
  399. data/ext/sources/ggml/src/ggml-openvino/openvino/op_table.cpp +46 -0
  400. data/ext/sources/ggml/src/ggml-openvino/openvino/op_table.h +39 -0
  401. data/ext/sources/ggml/src/ggml-openvino/openvino/pass/eliminate_zp.cpp +123 -0
  402. data/ext/sources/ggml/src/ggml-openvino/openvino/pass/eliminate_zp.h +17 -0
  403. data/ext/sources/ggml/src/ggml-openvino/openvino/pass/fuse_to_sdpa.cpp +60 -0
  404. data/ext/sources/ggml/src/ggml-openvino/openvino/pass/fuse_to_sdpa.h +17 -0
  405. data/ext/sources/ggml/src/ggml-openvino/openvino/pass/mark_decompression_convert_constant_folding.h +29 -0
  406. data/ext/sources/ggml/src/ggml-openvino/openvino/pass/squeeze_matmul.cpp +58 -0
  407. data/ext/sources/ggml/src/ggml-openvino/openvino/pass/squeeze_matmul.h +17 -0
  408. data/ext/sources/ggml/src/ggml-openvino/openvino/translate_session.cpp +293 -0
  409. data/ext/sources/ggml/src/ggml-openvino/openvino/translate_session.h +28 -0
  410. data/ext/sources/ggml/src/ggml-openvino/openvino/utils.cpp +226 -0
  411. data/ext/sources/ggml/src/ggml-openvino/openvino/utils.h +85 -0
  412. data/ext/sources/ggml/src/ggml-openvino/utils.cpp +823 -0
  413. data/ext/sources/ggml/src/ggml-openvino/utils.h +123 -0
  414. data/ext/sources/ggml/src/ggml-quants.c +96 -5
  415. data/ext/sources/ggml/src/ggml-quants.h +3 -0
  416. data/ext/sources/ggml/src/ggml-sycl/CMakeLists.txt +15 -88
  417. data/ext/sources/ggml/src/ggml-sycl/add-id.cpp +5 -1
  418. data/ext/sources/ggml/src/ggml-sycl/backend.hpp +1 -0
  419. data/ext/sources/ggml/src/ggml-sycl/binbcast.cpp +21 -20
  420. data/ext/sources/ggml/src/ggml-sycl/common.hpp +315 -10
  421. data/ext/sources/ggml/src/ggml-sycl/convert.cpp +69 -1
  422. data/ext/sources/ggml/src/ggml-sycl/convert.hpp +22 -1
  423. data/ext/sources/ggml/src/ggml-sycl/count-equal.cpp +1 -1
  424. data/ext/sources/ggml/src/ggml-sycl/dpct/helper.hpp +791 -47
  425. data/ext/sources/ggml/src/ggml-sycl/element_wise.cpp +78 -68
  426. data/ext/sources/ggml/src/ggml-sycl/element_wise.hpp +2 -0
  427. data/ext/sources/ggml/src/ggml-sycl/fattn-common.hpp +1179 -0
  428. data/ext/sources/ggml/src/ggml-sycl/fattn-tile.cpp +55 -0
  429. data/ext/sources/ggml/src/ggml-sycl/fattn-tile.hpp +1338 -0
  430. data/ext/sources/ggml/src/ggml-sycl/fattn-vec.hpp +667 -0
  431. data/ext/sources/ggml/src/ggml-sycl/fattn.cpp +225 -0
  432. data/ext/sources/ggml/src/ggml-sycl/fattn.hpp +22 -0
  433. data/ext/sources/ggml/src/ggml-sycl/gated_delta_net.cpp +309 -0
  434. data/ext/sources/ggml/src/ggml-sycl/gated_delta_net.hpp +8 -0
  435. data/ext/sources/ggml/src/ggml-sycl/ggml-sycl.cpp +316 -51
  436. data/ext/sources/ggml/src/ggml-sycl/norm.cpp +65 -66
  437. data/ext/sources/ggml/src/ggml-sycl/outprod.cpp +3 -3
  438. data/ext/sources/ggml/src/ggml-sycl/presets.hpp +3 -0
  439. data/ext/sources/ggml/src/ggml-sycl/quants.hpp +1 -1
  440. data/ext/sources/ggml/src/ggml-sycl/rope.cpp +450 -287
  441. data/ext/sources/ggml/src/ggml-sycl/rope.hpp +6 -0
  442. data/ext/sources/ggml/src/ggml-sycl/softmax.cpp +6 -6
  443. data/ext/sources/ggml/src/ggml-sycl/template-instances/fattn-tile-instance-dkq112-dv112.cpp +5 -0
  444. data/ext/sources/ggml/src/ggml-sycl/template-instances/fattn-tile-instance-dkq128-dv128.cpp +5 -0
  445. data/ext/sources/ggml/src/ggml-sycl/template-instances/fattn-tile-instance-dkq256-dv256.cpp +5 -0
  446. data/ext/sources/ggml/src/ggml-sycl/template-instances/fattn-tile-instance-dkq40-dv40.cpp +5 -0
  447. data/ext/sources/ggml/src/ggml-sycl/template-instances/fattn-tile-instance-dkq576-dv512.cpp +5 -0
  448. data/ext/sources/ggml/src/ggml-sycl/template-instances/fattn-tile-instance-dkq64-dv64.cpp +5 -0
  449. data/ext/sources/ggml/src/ggml-sycl/template-instances/fattn-tile-instance-dkq72-dv72.cpp +5 -0
  450. data/ext/sources/ggml/src/ggml-sycl/template-instances/fattn-tile-instance-dkq80-dv80.cpp +5 -0
  451. data/ext/sources/ggml/src/ggml-sycl/template-instances/fattn-tile-instance-dkq96-dv96.cpp +5 -0
  452. data/ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-f16-f16.cpp +7 -0
  453. data/ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-f16-q4_0.cpp +7 -0
  454. data/ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-f16-q4_1.cpp +7 -0
  455. data/ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-f16-q5_0.cpp +7 -0
  456. data/ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-f16-q5_1.cpp +7 -0
  457. data/ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-f16-q8_0.cpp +7 -0
  458. data/ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_0-f16.cpp +7 -0
  459. data/ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_0-q4_0.cpp +7 -0
  460. data/ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_0-q4_1.cpp +7 -0
  461. data/ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_0-q5_0.cpp +7 -0
  462. data/ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_0-q5_1.cpp +7 -0
  463. data/ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_0-q8_0.cpp +7 -0
  464. data/ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_1-f16.cpp +7 -0
  465. data/ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_1-q4_0.cpp +7 -0
  466. data/ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_1-q4_1.cpp +7 -0
  467. data/ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_1-q5_0.cpp +7 -0
  468. data/ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_1-q5_1.cpp +7 -0
  469. data/ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_1-q8_0.cpp +7 -0
  470. data/ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_0-f16.cpp +7 -0
  471. data/ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_0-q4_0.cpp +7 -0
  472. data/ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_0-q4_1.cpp +7 -0
  473. data/ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_0-q5_0.cpp +7 -0
  474. data/ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_0-q5_1.cpp +7 -0
  475. data/ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_0-q8_0.cpp +7 -0
  476. data/ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_1-f16.cpp +7 -0
  477. data/ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_1-q4_0.cpp +7 -0
  478. data/ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_1-q4_1.cpp +7 -0
  479. data/ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_1-q5_0.cpp +7 -0
  480. data/ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_1-q5_1.cpp +7 -0
  481. data/ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_1-q8_0.cpp +7 -0
  482. data/ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q8_0-f16.cpp +7 -0
  483. data/ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q8_0-q4_0.cpp +7 -0
  484. data/ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q8_0-q4_1.cpp +7 -0
  485. data/ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q8_0-q5_0.cpp +7 -0
  486. data/ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q8_0-q5_1.cpp +7 -0
  487. data/ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q8_0-q8_0.cpp +7 -0
  488. data/ext/sources/ggml/src/ggml-sycl/vecdotq.hpp +13 -0
  489. data/ext/sources/ggml/src/ggml-sycl/wkv.cpp +1 -1
  490. data/ext/sources/ggml/src/ggml-virtgpu/CMakeLists.txt +70 -0
  491. data/ext/sources/ggml/src/ggml-virtgpu/apir_cs_ggml-rpc-front.cpp +87 -0
  492. data/ext/sources/ggml/src/ggml-virtgpu/backend/CMakeLists.txt +21 -0
  493. data/ext/sources/ggml/src/ggml-virtgpu/backend/apir_cs_ggml-rpc-back.cpp +115 -0
  494. data/ext/sources/ggml/src/ggml-virtgpu/backend/backend-convert.h +13 -0
  495. data/ext/sources/ggml/src/ggml-virtgpu/backend/backend-dispatched-backend.cpp +102 -0
  496. data/ext/sources/ggml/src/ggml-virtgpu/backend/backend-dispatched-buffer-type.cpp +105 -0
  497. data/ext/sources/ggml/src/ggml-virtgpu/backend/backend-dispatched-buffer.cpp +179 -0
  498. data/ext/sources/ggml/src/ggml-virtgpu/backend/backend-dispatched-device.cpp +148 -0
  499. data/ext/sources/ggml/src/ggml-virtgpu/backend/backend-dispatched.cpp +51 -0
  500. data/ext/sources/ggml/src/ggml-virtgpu/backend/backend-dispatched.gen.h +73 -0
  501. data/ext/sources/ggml/src/ggml-virtgpu/backend/backend-dispatched.h +27 -0
  502. data/ext/sources/ggml/src/ggml-virtgpu/backend/backend-virgl-apir.h +32 -0
  503. data/ext/sources/ggml/src/ggml-virtgpu/backend/backend.cpp +144 -0
  504. data/ext/sources/ggml/src/ggml-virtgpu/backend/shared/api_remoting.h +95 -0
  505. data/ext/sources/ggml/src/ggml-virtgpu/backend/shared/apir_backend.gen.h +94 -0
  506. data/ext/sources/ggml/src/ggml-virtgpu/backend/shared/apir_backend.h +50 -0
  507. data/ext/sources/ggml/src/ggml-virtgpu/backend/shared/apir_cs.h +378 -0
  508. data/ext/sources/ggml/src/ggml-virtgpu/backend/shared/apir_cs_ggml.h +232 -0
  509. data/ext/sources/ggml/src/ggml-virtgpu/backend/shared/apir_cs_rpc.h +58 -0
  510. data/ext/sources/ggml/src/ggml-virtgpu/ggml-backend-buffer-type.cpp +81 -0
  511. data/ext/sources/ggml/src/ggml-virtgpu/ggml-backend-buffer.cpp +119 -0
  512. data/ext/sources/ggml/src/ggml-virtgpu/ggml-backend-device.cpp +158 -0
  513. data/ext/sources/ggml/src/ggml-virtgpu/ggml-backend-reg.cpp +213 -0
  514. data/ext/sources/ggml/src/ggml-virtgpu/ggml-backend.cpp +69 -0
  515. data/ext/sources/ggml/src/ggml-virtgpu/ggml-remoting.h +71 -0
  516. data/ext/sources/ggml/src/ggml-virtgpu/ggmlremoting_functions.yaml +166 -0
  517. data/ext/sources/ggml/src/ggml-virtgpu/include/apir_hw.h +9 -0
  518. data/ext/sources/ggml/src/ggml-virtgpu/regenerate_remoting.py +333 -0
  519. data/ext/sources/ggml/src/ggml-virtgpu/virtgpu-apir.h +15 -0
  520. data/ext/sources/ggml/src/ggml-virtgpu/virtgpu-forward-backend.cpp +58 -0
  521. data/ext/sources/ggml/src/ggml-virtgpu/virtgpu-forward-buffer-type.cpp +110 -0
  522. data/ext/sources/ggml/src/ggml-virtgpu/virtgpu-forward-buffer.cpp +173 -0
  523. data/ext/sources/ggml/src/ggml-virtgpu/virtgpu-forward-device.cpp +192 -0
  524. data/ext/sources/ggml/src/ggml-virtgpu/virtgpu-forward-impl.h +36 -0
  525. data/ext/sources/ggml/src/ggml-virtgpu/virtgpu-forward.gen.h +53 -0
  526. data/ext/sources/ggml/src/ggml-virtgpu/virtgpu-shm.cpp +98 -0
  527. data/ext/sources/ggml/src/ggml-virtgpu/virtgpu-shm.h +23 -0
  528. data/ext/sources/ggml/src/ggml-virtgpu/virtgpu-utils.cpp +179 -0
  529. data/ext/sources/ggml/src/ggml-virtgpu/virtgpu-utils.h +86 -0
  530. data/ext/sources/ggml/src/ggml-virtgpu/virtgpu.cpp +544 -0
  531. data/ext/sources/ggml/src/ggml-virtgpu/virtgpu.h +117 -0
  532. data/ext/sources/ggml/src/ggml-vulkan/CMakeLists.txt +1 -1
  533. data/ext/sources/ggml/src/ggml-vulkan/ggml-vulkan.cpp +1250 -465
  534. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/acc.comp +16 -8
  535. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/elu.comp +27 -0
  536. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/flash_attn.comp +374 -170
  537. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_base.glsl +66 -22
  538. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_cm1.comp +389 -201
  539. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_cm2.comp +106 -58
  540. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_mask_opt.comp +162 -0
  541. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_split_k_reduce.comp +9 -8
  542. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/gated_delta_net.comp +128 -0
  543. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/l2_norm.comp +12 -9
  544. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_base.glsl +20 -17
  545. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mm.comp +11 -3
  546. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mm_cm2.comp +8 -4
  547. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mm_id_funcs.glsl +3 -1
  548. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mmq.comp +5 -3
  549. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/mul_mmq_funcs.glsl +3 -3
  550. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rms_norm.comp +2 -3
  551. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rope_funcs.glsl +36 -63
  552. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rope_multi.comp +7 -4
  553. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rope_neox.comp +7 -4
  554. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rope_norm.comp +7 -4
  555. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rope_params.glsl +10 -5
  556. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rope_vision.comp +7 -4
  557. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/sgn.comp +21 -0
  558. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/ssm_conv.comp +16 -10
  559. data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/vulkan-shaders-gen.cpp +55 -35
  560. data/ext/sources/ggml/src/ggml-webgpu/ggml-webgpu-shader-lib.hpp +1314 -109
  561. data/ext/sources/ggml/src/ggml-webgpu/ggml-webgpu.cpp +1660 -1371
  562. data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/argmax.wgsl +72 -0
  563. data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/argsort.wgsl +106 -0
  564. data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/argsort_merge.wgsl +134 -0
  565. data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/binary.wgsl +141 -0
  566. data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/common_decls.tmpl +65 -72
  567. data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/concat.wgsl +75 -0
  568. data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/cpy.tmpl.wgsl +6 -0
  569. data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/cumsum.wgsl +66 -0
  570. data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/embed_wgsl.py +40 -5
  571. data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/flash_attn.wgsl +105 -60
  572. data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/{get_rows.tmpl.wgsl → get_rows.wgsl} +53 -259
  573. data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/{mul_mat.tmpl.wgsl → mul_mat.wgsl} +68 -257
  574. data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/mul_mat_decls.tmpl +692 -23
  575. data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/{mul_mat_reg_tile.tmpl.wgsl → mul_mat_reg_tile.wgsl} +28 -128
  576. data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/{mul_mat_subgroup_matrix.tmpl.wgsl → mul_mat_subgroup_matrix.wgsl} +31 -137
  577. data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/mul_mat_vec.wgsl +480 -0
  578. data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/pad.wgsl +86 -0
  579. data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/repeat.wgsl +67 -0
  580. data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/{scale.tmpl.wgsl → scale.wgsl} +9 -36
  581. data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/set_rows.wgsl +40 -12
  582. data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/sum_rows.wgsl +55 -0
  583. data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/unary.wgsl +193 -0
  584. data/ext/sources/ggml/src/ggml-zdnn/ggml-zdnn.cpp +6 -1
  585. data/ext/sources/ggml/src/ggml-zendnn/CMakeLists.txt +31 -32
  586. data/ext/sources/ggml/src/ggml-zendnn/ggml-zendnn.cpp +9 -6
  587. data/ext/sources/ggml/src/ggml.c +167 -33
  588. data/ext/sources/ggml/src/gguf.cpp +229 -44
  589. data/ext/sources/src/whisper.cpp +6 -28
  590. data/sig/whisper.rbs +43 -2
  591. data/test/test_context_params.rb +82 -0
  592. data/test/test_token.rb +11 -0
  593. data/test/test_vad_context.rb +58 -8
  594. data/test/test_whisper.rb +20 -0
  595. data/whispercpp.gemspec +1 -1
  596. metadata +240 -28
  597. data/ext/sources/ggml/cmake/BuildTypes.cmake +0 -54
  598. data/ext/sources/ggml/src/ggml-cpu/llamafile/sgemm-ppc.h +0 -333
  599. data/ext/sources/ggml/src/ggml-hexagon/htp/hvx-exp.c +0 -94
  600. data/ext/sources/ggml/src/ggml-hexagon/htp/hvx-inverse.c +0 -72
  601. data/ext/sources/ggml/src/ggml-hexagon/htp/hvx-sigmoid.c +0 -49
  602. data/ext/sources/ggml/src/ggml-hexagon/htp/hvx-utils.c +0 -1020
  603. data/ext/sources/ggml/src/ggml-hexagon/htp/ops-utils.h +0 -149
  604. data/ext/sources/ggml/src/ggml-hexagon/htp-utils.c +0 -454
  605. data/ext/sources/ggml/src/ggml-hexagon/htp-utils.h +0 -221
  606. data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/bin_op.tmpl.wgsl +0 -188
  607. data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/binary_head.tmpl +0 -45
  608. data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/mul_mat_vec.tmpl.wgsl +0 -267
  609. data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/set_rows.tmpl.wgsl +0 -112
  610. data/ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/unary_op.wgsl +0 -483
metadata CHANGED
@@ -1,14 +1,15 @@
1
1
  --- !ruby/object:Gem::Specification
2
2
  name: whispercpp
3
3
  version: !ruby/object:Gem::Version
4
- version: 1.3.5
4
+ version: 1.3.6
5
5
  platform: ruby
6
6
  authors:
7
7
  - Georgi Gerganov
8
8
  - Todd A. Fisher
9
+ autorequire:
9
10
  bindir: bin
10
11
  cert_chain: []
11
- date: 1980-01-02 00:00:00.000000000 Z
12
+ date: 2026-03-19 00:00:00.000000000 Z
12
13
  dependencies: []
13
14
  description: High-performance inference of OpenAI's Whisper automatic speech recognition
14
15
  (ASR) model via Ruby
@@ -31,6 +32,7 @@ files:
31
32
  - ext/ruby_whisper.c
32
33
  - ext/ruby_whisper.h
33
34
  - ext/ruby_whisper_context.c
35
+ - ext/ruby_whisper_context_params.c
34
36
  - ext/ruby_whisper_error.c
35
37
  - ext/ruby_whisper_model.c
36
38
  - ext/ruby_whisper_params.c
@@ -135,6 +137,7 @@ files:
135
137
  - ext/sources/examples/talk-llama/llama-context.h
136
138
  - ext/sources/examples/talk-llama/llama-cparams.cpp
137
139
  - ext/sources/examples/talk-llama/llama-cparams.h
140
+ - ext/sources/examples/talk-llama/llama-ext.h
138
141
  - ext/sources/examples/talk-llama/llama-grammar.cpp
139
142
  - ext/sources/examples/talk-llama/llama-grammar.h
140
143
  - ext/sources/examples/talk-llama/llama-graph.cpp
@@ -150,6 +153,8 @@ files:
150
153
  - ext/sources/examples/talk-llama/llama-kv-cache.cpp
151
154
  - ext/sources/examples/talk-llama/llama-kv-cache.h
152
155
  - ext/sources/examples/talk-llama/llama-kv-cells.h
156
+ - ext/sources/examples/talk-llama/llama-memory-hybrid-iswa.cpp
157
+ - ext/sources/examples/talk-llama/llama-memory-hybrid-iswa.h
153
158
  - ext/sources/examples/talk-llama/llama-memory-hybrid.cpp
154
159
  - ext/sources/examples/talk-llama/llama-memory-hybrid.h
155
160
  - ext/sources/examples/talk-llama/llama-memory-recurrent.cpp
@@ -166,8 +171,8 @@ files:
166
171
  - ext/sources/examples/talk-llama/llama-model.h
167
172
  - ext/sources/examples/talk-llama/llama-quant.cpp
168
173
  - ext/sources/examples/talk-llama/llama-quant.h
169
- - ext/sources/examples/talk-llama/llama-sampling.cpp
170
- - ext/sources/examples/talk-llama/llama-sampling.h
174
+ - ext/sources/examples/talk-llama/llama-sampler.cpp
175
+ - ext/sources/examples/talk-llama/llama-sampler.h
171
176
  - ext/sources/examples/talk-llama/llama-vocab.cpp
172
177
  - ext/sources/examples/talk-llama/llama-vocab.h
173
178
  - ext/sources/examples/talk-llama/llama.cpp
@@ -193,10 +198,13 @@ files:
193
198
  - ext/sources/examples/talk-llama/models/deci.cpp
194
199
  - ext/sources/examples/talk-llama/models/deepseek.cpp
195
200
  - ext/sources/examples/talk-llama/models/deepseek2.cpp
201
+ - ext/sources/examples/talk-llama/models/delta-net-base.cpp
196
202
  - ext/sources/examples/talk-llama/models/dots1.cpp
197
203
  - ext/sources/examples/talk-llama/models/dream.cpp
198
204
  - ext/sources/examples/talk-llama/models/ernie4-5-moe.cpp
199
205
  - ext/sources/examples/talk-llama/models/ernie4-5.cpp
206
+ - ext/sources/examples/talk-llama/models/eurobert.cpp
207
+ - ext/sources/examples/talk-llama/models/exaone-moe.cpp
200
208
  - ext/sources/examples/talk-llama/models/exaone.cpp
201
209
  - ext/sources/examples/talk-llama/models/exaone4.cpp
202
210
  - ext/sources/examples/talk-llama/models/falcon-h1.cpp
@@ -212,20 +220,22 @@ files:
212
220
  - ext/sources/examples/talk-llama/models/gptneox.cpp
213
221
  - ext/sources/examples/talk-llama/models/granite-hybrid.cpp
214
222
  - ext/sources/examples/talk-llama/models/granite.cpp
215
- - ext/sources/examples/talk-llama/models/graph-context-mamba.cpp
216
223
  - ext/sources/examples/talk-llama/models/grok.cpp
217
224
  - ext/sources/examples/talk-llama/models/grovemoe.cpp
218
225
  - ext/sources/examples/talk-llama/models/hunyuan-dense.cpp
219
226
  - ext/sources/examples/talk-llama/models/hunyuan-moe.cpp
220
227
  - ext/sources/examples/talk-llama/models/internlm2.cpp
221
228
  - ext/sources/examples/talk-llama/models/jais.cpp
229
+ - ext/sources/examples/talk-llama/models/jais2.cpp
222
230
  - ext/sources/examples/talk-llama/models/jamba.cpp
231
+ - ext/sources/examples/talk-llama/models/kimi-linear.cpp
223
232
  - ext/sources/examples/talk-llama/models/lfm2.cpp
224
233
  - ext/sources/examples/talk-llama/models/llada-moe.cpp
225
234
  - ext/sources/examples/talk-llama/models/llada.cpp
226
235
  - ext/sources/examples/talk-llama/models/llama-iswa.cpp
227
236
  - ext/sources/examples/talk-llama/models/llama.cpp
228
237
  - ext/sources/examples/talk-llama/models/maincoder.cpp
238
+ - ext/sources/examples/talk-llama/models/mamba-base.cpp
229
239
  - ext/sources/examples/talk-llama/models/mamba.cpp
230
240
  - ext/sources/examples/talk-llama/models/mimo2-iswa.cpp
231
241
  - ext/sources/examples/talk-llama/models/minicpm3.cpp
@@ -243,6 +253,7 @@ files:
243
253
  - ext/sources/examples/talk-llama/models/openai-moe-iswa.cpp
244
254
  - ext/sources/examples/talk-llama/models/openelm.cpp
245
255
  - ext/sources/examples/talk-llama/models/orion.cpp
256
+ - ext/sources/examples/talk-llama/models/paddleocr.cpp
246
257
  - ext/sources/examples/talk-llama/models/pangu-embedded.cpp
247
258
  - ext/sources/examples/talk-llama/models/phi2.cpp
248
259
  - ext/sources/examples/talk-llama/models/phi3.cpp
@@ -255,6 +266,8 @@ files:
255
266
  - ext/sources/examples/talk-llama/models/qwen2moe.cpp
256
267
  - ext/sources/examples/talk-llama/models/qwen2vl.cpp
257
268
  - ext/sources/examples/talk-llama/models/qwen3.cpp
269
+ - ext/sources/examples/talk-llama/models/qwen35.cpp
270
+ - ext/sources/examples/talk-llama/models/qwen35moe.cpp
258
271
  - ext/sources/examples/talk-llama/models/qwen3moe.cpp
259
272
  - ext/sources/examples/talk-llama/models/qwen3next.cpp
260
273
  - ext/sources/examples/talk-llama/models/qwen3vl-moe.cpp
@@ -272,6 +285,7 @@ files:
272
285
  - ext/sources/examples/talk-llama/models/stablelm.cpp
273
286
  - ext/sources/examples/talk-llama/models/starcoder.cpp
274
287
  - ext/sources/examples/talk-llama/models/starcoder2.cpp
288
+ - ext/sources/examples/talk-llama/models/step35-iswa.cpp
275
289
  - ext/sources/examples/talk-llama/models/t5-dec.cpp
276
290
  - ext/sources/examples/talk-llama/models/t5-enc.cpp
277
291
  - ext/sources/examples/talk-llama/models/wavtokenizer-dec.cpp
@@ -300,7 +314,6 @@ files:
300
314
  - ext/sources/examples/whisper.wasm/emscripten.cpp
301
315
  - ext/sources/examples/whisper.wasm/index-tmpl.html
302
316
  - ext/sources/ggml/CMakeLists.txt
303
- - ext/sources/ggml/cmake/BuildTypes.cmake
304
317
  - ext/sources/ggml/cmake/GitVars.cmake
305
318
  - ext/sources/ggml/cmake/common.cmake
306
319
  - ext/sources/ggml/cmake/ggml-config.cmake.in
@@ -314,9 +327,11 @@ files:
314
327
  - ext/sources/ggml/include/ggml-hexagon.h
315
328
  - ext/sources/ggml/include/ggml-metal.h
316
329
  - ext/sources/ggml/include/ggml-opencl.h
330
+ - ext/sources/ggml/include/ggml-openvino.h
317
331
  - ext/sources/ggml/include/ggml-opt.h
318
332
  - ext/sources/ggml/include/ggml-rpc.h
319
333
  - ext/sources/ggml/include/ggml-sycl.h
334
+ - ext/sources/ggml/include/ggml-virtgpu.h
320
335
  - ext/sources/ggml/include/ggml-vulkan.h
321
336
  - ext/sources/ggml/include/ggml-webgpu.h
322
337
  - ext/sources/ggml/include/ggml-zdnn.h
@@ -325,6 +340,8 @@ files:
325
340
  - ext/sources/ggml/include/gguf.h
326
341
  - ext/sources/ggml/src/CMakeLists.txt
327
342
  - ext/sources/ggml/src/ggml-alloc.c
343
+ - ext/sources/ggml/src/ggml-backend-dl.cpp
344
+ - ext/sources/ggml/src/ggml-backend-dl.h
328
345
  - ext/sources/ggml/src/ggml-backend-impl.h
329
346
  - ext/sources/ggml/src/ggml-backend-reg.cpp
330
347
  - ext/sources/ggml/src/ggml-backend.cpp
@@ -373,7 +390,6 @@ files:
373
390
  - ext/sources/ggml/src/ggml-cpu/kleidiai/kernels.h
374
391
  - ext/sources/ggml/src/ggml-cpu/kleidiai/kleidiai.cpp
375
392
  - ext/sources/ggml/src/ggml-cpu/kleidiai/kleidiai.h
376
- - ext/sources/ggml/src/ggml-cpu/llamafile/sgemm-ppc.h
377
393
  - ext/sources/ggml/src/ggml-cpu/llamafile/sgemm.cpp
378
394
  - ext/sources/ggml/src/ggml-cpu/llamafile/sgemm.h
379
395
  - ext/sources/ggml/src/ggml-cpu/ops.cpp
@@ -382,6 +398,7 @@ files:
382
398
  - ext/sources/ggml/src/ggml-cpu/quants.h
383
399
  - ext/sources/ggml/src/ggml-cpu/repack.cpp
384
400
  - ext/sources/ggml/src/ggml-cpu/repack.h
401
+ - ext/sources/ggml/src/ggml-cpu/simd-gemm.h
385
402
  - ext/sources/ggml/src/ggml-cpu/simd-mappings.h
386
403
  - ext/sources/ggml/src/ggml-cpu/spacemit/ime.cpp
387
404
  - ext/sources/ggml/src/ggml-cpu/spacemit/ime.h
@@ -447,6 +464,8 @@ files:
447
464
  - ext/sources/ggml/src/ggml-cuda/fattn.cuh
448
465
  - ext/sources/ggml/src/ggml-cuda/fill.cu
449
466
  - ext/sources/ggml/src/ggml-cuda/fill.cuh
467
+ - ext/sources/ggml/src/ggml-cuda/gated_delta_net.cu
468
+ - ext/sources/ggml/src/ggml-cuda/gated_delta_net.cuh
450
469
  - ext/sources/ggml/src/ggml-cuda/getrows.cu
451
470
  - ext/sources/ggml/src/ggml-cuda/getrows.cuh
452
471
  - ext/sources/ggml/src/ggml-cuda/ggml-cuda.cu
@@ -509,11 +528,13 @@ files:
509
528
  - ext/sources/ggml/src/ggml-cuda/sumrows.cu
510
529
  - ext/sources/ggml/src/ggml-cuda/sumrows.cuh
511
530
  - ext/sources/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_1-ncols2_16.cu
531
+ - ext/sources/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_1-ncols2_32.cu
512
532
  - ext/sources/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_1-ncols2_8.cu
513
533
  - ext/sources/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_16-ncols2_1.cu
514
534
  - ext/sources/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_16-ncols2_2.cu
515
535
  - ext/sources/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_16-ncols2_4.cu
516
536
  - ext/sources/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_2-ncols2_16.cu
537
+ - ext/sources/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_2-ncols2_32.cu
517
538
  - ext/sources/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_2-ncols2_4.cu
518
539
  - ext/sources/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_2-ncols2_8.cu
519
540
  - ext/sources/ggml/src/ggml-cuda/template-instances/fattn-mma-f16-instance-ncols1_32-ncols2_1.cu
@@ -628,34 +649,51 @@ files:
628
649
  - ext/sources/ggml/src/ggml-cuda/wkv.cuh
629
650
  - ext/sources/ggml/src/ggml-hexagon/CMakeLists.txt
630
651
  - ext/sources/ggml/src/ggml-hexagon/ggml-hexagon.cpp
631
- - ext/sources/ggml/src/ggml-hexagon/htp-utils.c
632
- - ext/sources/ggml/src/ggml-hexagon/htp-utils.h
652
+ - ext/sources/ggml/src/ggml-hexagon/htp-drv.cpp
653
+ - ext/sources/ggml/src/ggml-hexagon/htp-drv.h
633
654
  - ext/sources/ggml/src/ggml-hexagon/htp/CMakeLists.txt
634
655
  - ext/sources/ggml/src/ggml-hexagon/htp/act-ops.c
656
+ - ext/sources/ggml/src/ggml-hexagon/htp/argsort-ops.c
635
657
  - ext/sources/ggml/src/ggml-hexagon/htp/binary-ops.c
636
658
  - ext/sources/ggml/src/ggml-hexagon/htp/cmake-toolchain.cmake
659
+ - ext/sources/ggml/src/ggml-hexagon/htp/cpy-ops.c
637
660
  - ext/sources/ggml/src/ggml-hexagon/htp/flash-attn-ops.c
638
661
  - ext/sources/ggml/src/ggml-hexagon/htp/get-rows-ops.c
662
+ - ext/sources/ggml/src/ggml-hexagon/htp/hex-dma.c
663
+ - ext/sources/ggml/src/ggml-hexagon/htp/hex-dma.h
664
+ - ext/sources/ggml/src/ggml-hexagon/htp/hex-dump.h
665
+ - ext/sources/ggml/src/ggml-hexagon/htp/hex-fastdiv.h
666
+ - ext/sources/ggml/src/ggml-hexagon/htp/hex-utils.h
639
667
  - ext/sources/ggml/src/ggml-hexagon/htp/htp-ctx.h
640
- - ext/sources/ggml/src/ggml-hexagon/htp/htp-dma.c
641
- - ext/sources/ggml/src/ggml-hexagon/htp/htp-dma.h
642
668
  - ext/sources/ggml/src/ggml-hexagon/htp/htp-msg.h
643
669
  - ext/sources/ggml/src/ggml-hexagon/htp/htp-ops.h
644
670
  - ext/sources/ggml/src/ggml-hexagon/htp/htp_iface.idl
645
- - ext/sources/ggml/src/ggml-hexagon/htp/hvx-exp.c
646
- - ext/sources/ggml/src/ggml-hexagon/htp/hvx-inverse.c
647
- - ext/sources/ggml/src/ggml-hexagon/htp/hvx-sigmoid.c
648
- - ext/sources/ggml/src/ggml-hexagon/htp/hvx-utils.c
671
+ - ext/sources/ggml/src/ggml-hexagon/htp/hvx-arith.h
672
+ - ext/sources/ggml/src/ggml-hexagon/htp/hvx-base.h
673
+ - ext/sources/ggml/src/ggml-hexagon/htp/hvx-copy.h
674
+ - ext/sources/ggml/src/ggml-hexagon/htp/hvx-div.h
675
+ - ext/sources/ggml/src/ggml-hexagon/htp/hvx-dump.h
676
+ - ext/sources/ggml/src/ggml-hexagon/htp/hvx-exp.h
677
+ - ext/sources/ggml/src/ggml-hexagon/htp/hvx-floor.h
678
+ - ext/sources/ggml/src/ggml-hexagon/htp/hvx-inverse.h
679
+ - ext/sources/ggml/src/ggml-hexagon/htp/hvx-reduce.h
680
+ - ext/sources/ggml/src/ggml-hexagon/htp/hvx-scale.h
681
+ - ext/sources/ggml/src/ggml-hexagon/htp/hvx-sigmoid.h
682
+ - ext/sources/ggml/src/ggml-hexagon/htp/hvx-sqrt.h
683
+ - ext/sources/ggml/src/ggml-hexagon/htp/hvx-types.h
649
684
  - ext/sources/ggml/src/ggml-hexagon/htp/hvx-utils.h
650
685
  - ext/sources/ggml/src/ggml-hexagon/htp/main.c
651
686
  - ext/sources/ggml/src/ggml-hexagon/htp/matmul-ops.c
652
- - ext/sources/ggml/src/ggml-hexagon/htp/ops-utils.h
653
687
  - ext/sources/ggml/src/ggml-hexagon/htp/rope-ops.c
654
688
  - ext/sources/ggml/src/ggml-hexagon/htp/set-rows-ops.c
655
689
  - ext/sources/ggml/src/ggml-hexagon/htp/softmax-ops.c
690
+ - ext/sources/ggml/src/ggml-hexagon/htp/ssm-conv.c
691
+ - ext/sources/ggml/src/ggml-hexagon/htp/sum-rows-ops.c
656
692
  - ext/sources/ggml/src/ggml-hexagon/htp/unary-ops.c
657
693
  - ext/sources/ggml/src/ggml-hexagon/htp/worker-pool.c
658
694
  - ext/sources/ggml/src/ggml-hexagon/htp/worker-pool.h
695
+ - ext/sources/ggml/src/ggml-hexagon/libdl.h
696
+ - ext/sources/ggml/src/ggml-hexagon/libggml-htp.inf
659
697
  - ext/sources/ggml/src/ggml-hexagon/op-desc.h
660
698
  - ext/sources/ggml/src/ggml-hip/CMakeLists.txt
661
699
  - ext/sources/ggml/src/ggml-impl.h
@@ -685,10 +723,13 @@ files:
685
723
  - ext/sources/ggml/src/ggml-opencl/kernels/conv2d.cl
686
724
  - ext/sources/ggml/src/ggml-opencl/kernels/conv2d_f16_f32.cl
687
725
  - ext/sources/ggml/src/ggml-opencl/kernels/cpy.cl
726
+ - ext/sources/ggml/src/ggml-opencl/kernels/cumsum.cl
688
727
  - ext/sources/ggml/src/ggml-opencl/kernels/cvt.cl
728
+ - ext/sources/ggml/src/ggml-opencl/kernels/diag.cl
689
729
  - ext/sources/ggml/src/ggml-opencl/kernels/diag_mask_inf.cl
690
730
  - ext/sources/ggml/src/ggml-opencl/kernels/div.cl
691
731
  - ext/sources/ggml/src/ggml-opencl/kernels/embed_kernel.py
732
+ - ext/sources/ggml/src/ggml-opencl/kernels/exp.cl
692
733
  - ext/sources/ggml/src/ggml-opencl/kernels/expm1.cl
693
734
  - ext/sources/ggml/src/ggml-opencl/kernels/fill.cl
694
735
  - ext/sources/ggml/src/ggml-opencl/kernels/flash_attn_f16.cl
@@ -696,14 +737,18 @@ files:
696
737
  - ext/sources/ggml/src/ggml-opencl/kernels/flash_attn_f32_f16.cl
697
738
  - ext/sources/ggml/src/ggml-opencl/kernels/gelu.cl
698
739
  - ext/sources/ggml/src/ggml-opencl/kernels/gemm_moe_mxfp4_f32.cl
740
+ - ext/sources/ggml/src/ggml-opencl/kernels/gemm_noshuffle_q4_1_f32.cl
699
741
  - ext/sources/ggml/src/ggml-opencl/kernels/gemv_moe_mxfp4_f32.cl
700
742
  - ext/sources/ggml/src/ggml-opencl/kernels/gemv_noshuffle.cl
701
743
  - ext/sources/ggml/src/ggml-opencl/kernels/gemv_noshuffle_general.cl
744
+ - ext/sources/ggml/src/ggml-opencl/kernels/gemv_noshuffle_general_q8_0_f32.cl
745
+ - ext/sources/ggml/src/ggml-opencl/kernels/gemv_noshuffle_q4_1_f32.cl
702
746
  - ext/sources/ggml/src/ggml-opencl/kernels/get_rows.cl
703
747
  - ext/sources/ggml/src/ggml-opencl/kernels/glu.cl
704
748
  - ext/sources/ggml/src/ggml-opencl/kernels/group_norm.cl
705
749
  - ext/sources/ggml/src/ggml-opencl/kernels/im2col_f16.cl
706
750
  - ext/sources/ggml/src/ggml-opencl/kernels/im2col_f32.cl
751
+ - ext/sources/ggml/src/ggml-opencl/kernels/l2_norm.cl
707
752
  - ext/sources/ggml/src/ggml-opencl/kernels/mean.cl
708
753
  - ext/sources/ggml/src/ggml-opencl/kernels/mul.cl
709
754
  - ext/sources/ggml/src/ggml-opencl/kernels/mul_mat_Ab_Bi_8x4.cl
@@ -711,6 +756,10 @@ files:
711
756
  - ext/sources/ggml/src/ggml-opencl/kernels/mul_mm_f16_f32_kq_kqv.cl
712
757
  - ext/sources/ggml/src/ggml-opencl/kernels/mul_mm_f16_f32_l4_lm.cl
713
758
  - ext/sources/ggml/src/ggml-opencl/kernels/mul_mm_f32_f32_l4_lm.cl
759
+ - ext/sources/ggml/src/ggml-opencl/kernels/mul_mm_q4_0_f32_l4_lm.cl
760
+ - ext/sources/ggml/src/ggml-opencl/kernels/mul_mm_q4_1_f32_l4_lm.cl
761
+ - ext/sources/ggml/src/ggml-opencl/kernels/mul_mm_q6_k_f32_l4_lm.cl
762
+ - ext/sources/ggml/src/ggml-opencl/kernels/mul_mm_q8_0_f32_8x4.cl
714
763
  - ext/sources/ggml/src/ggml-opencl/kernels/mul_mm_q8_0_f32_l4_lm.cl
715
764
  - ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_f16_f16.cl
716
765
  - ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_f16_f32.cl
@@ -729,9 +778,14 @@ files:
729
778
  - ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_q4_0_f32_1d_8x_flat.cl
730
779
  - ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_q4_0_f32_8x_flat.cl
731
780
  - ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_q4_0_f32_v.cl
732
- - ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_q6_k.cl
781
+ - ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_q4_1_f32.cl
782
+ - ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_q4_1_f32_flat.cl
783
+ - ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_q4_k_f32.cl
784
+ - ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_q6_k_f32.cl
785
+ - ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_q6_k_f32_flat.cl
733
786
  - ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_q8_0_f32.cl
734
787
  - ext/sources/ggml/src/ggml-opencl/kernels/mul_mv_q8_0_f32_flat.cl
788
+ - ext/sources/ggml/src/ggml-opencl/kernels/neg.cl
735
789
  - ext/sources/ggml/src/ggml-opencl/kernels/norm.cl
736
790
  - ext/sources/ggml/src/ggml-opencl/kernels/pad.cl
737
791
  - ext/sources/ggml/src/ggml-opencl/kernels/relu.cl
@@ -747,6 +801,7 @@ files:
747
801
  - ext/sources/ggml/src/ggml-opencl/kernels/softmax_f16.cl
748
802
  - ext/sources/ggml/src/ggml-opencl/kernels/softmax_f32.cl
749
803
  - ext/sources/ggml/src/ggml-opencl/kernels/softplus.cl
804
+ - ext/sources/ggml/src/ggml-opencl/kernels/solve_tri.cl
750
805
  - ext/sources/ggml/src/ggml-opencl/kernels/sqr.cl
751
806
  - ext/sources/ggml/src/ggml-opencl/kernels/sqrt.cl
752
807
  - ext/sources/ggml/src/ggml-opencl/kernels/ssm_conv.cl
@@ -754,8 +809,56 @@ files:
754
809
  - ext/sources/ggml/src/ggml-opencl/kernels/sum_rows.cl
755
810
  - ext/sources/ggml/src/ggml-opencl/kernels/tanh.cl
756
811
  - ext/sources/ggml/src/ggml-opencl/kernels/transpose.cl
812
+ - ext/sources/ggml/src/ggml-opencl/kernels/tri.cl
757
813
  - ext/sources/ggml/src/ggml-opencl/kernels/tsembd.cl
758
814
  - ext/sources/ggml/src/ggml-opencl/kernels/upscale.cl
815
+ - ext/sources/ggml/src/ggml-openvino/.clang-format
816
+ - ext/sources/ggml/src/ggml-openvino/CMakeLists.txt
817
+ - ext/sources/ggml/src/ggml-openvino/ggml-decoder.cpp
818
+ - ext/sources/ggml/src/ggml-openvino/ggml-decoder.h
819
+ - ext/sources/ggml/src/ggml-openvino/ggml-openvino-extra.cpp
820
+ - ext/sources/ggml/src/ggml-openvino/ggml-openvino-extra.h
821
+ - ext/sources/ggml/src/ggml-openvino/ggml-openvino.cpp
822
+ - ext/sources/ggml/src/ggml-openvino/ggml-quants.cpp
823
+ - ext/sources/ggml/src/ggml-openvino/ggml-quants.h
824
+ - ext/sources/ggml/src/ggml-openvino/openvino/decoder.h
825
+ - ext/sources/ggml/src/ggml-openvino/openvino/frontend.cpp
826
+ - ext/sources/ggml/src/ggml-openvino/openvino/frontend.h
827
+ - ext/sources/ggml/src/ggml-openvino/openvino/input_model.cpp
828
+ - ext/sources/ggml/src/ggml-openvino/openvino/input_model.h
829
+ - ext/sources/ggml/src/ggml-openvino/openvino/node_context.h
830
+ - ext/sources/ggml/src/ggml-openvino/openvino/op/cont.cpp
831
+ - ext/sources/ggml/src/ggml-openvino/openvino/op/cpy.cpp
832
+ - ext/sources/ggml/src/ggml-openvino/openvino/op/flash_attn_ext.cpp
833
+ - ext/sources/ggml/src/ggml-openvino/openvino/op/get_rows.cpp
834
+ - ext/sources/ggml/src/ggml-openvino/openvino/op/glu_geglu.cpp
835
+ - ext/sources/ggml/src/ggml-openvino/openvino/op/glu_swiglu.cpp
836
+ - ext/sources/ggml/src/ggml-openvino/openvino/op/mulmat.cpp
837
+ - ext/sources/ggml/src/ggml-openvino/openvino/op/permute.cpp
838
+ - ext/sources/ggml/src/ggml-openvino/openvino/op/reshape.cpp
839
+ - ext/sources/ggml/src/ggml-openvino/openvino/op/rms_norm.cpp
840
+ - ext/sources/ggml/src/ggml-openvino/openvino/op/rope.cpp
841
+ - ext/sources/ggml/src/ggml-openvino/openvino/op/scale.cpp
842
+ - ext/sources/ggml/src/ggml-openvino/openvino/op/set_rows.cpp
843
+ - ext/sources/ggml/src/ggml-openvino/openvino/op/softmax.cpp
844
+ - ext/sources/ggml/src/ggml-openvino/openvino/op/transpose.cpp
845
+ - ext/sources/ggml/src/ggml-openvino/openvino/op/unary_silu.cpp
846
+ - ext/sources/ggml/src/ggml-openvino/openvino/op/view.cpp
847
+ - ext/sources/ggml/src/ggml-openvino/openvino/op_table.cpp
848
+ - ext/sources/ggml/src/ggml-openvino/openvino/op_table.h
849
+ - ext/sources/ggml/src/ggml-openvino/openvino/pass/eliminate_zp.cpp
850
+ - ext/sources/ggml/src/ggml-openvino/openvino/pass/eliminate_zp.h
851
+ - ext/sources/ggml/src/ggml-openvino/openvino/pass/fuse_to_sdpa.cpp
852
+ - ext/sources/ggml/src/ggml-openvino/openvino/pass/fuse_to_sdpa.h
853
+ - ext/sources/ggml/src/ggml-openvino/openvino/pass/mark_decompression_convert_constant_folding.h
854
+ - ext/sources/ggml/src/ggml-openvino/openvino/pass/squeeze_matmul.cpp
855
+ - ext/sources/ggml/src/ggml-openvino/openvino/pass/squeeze_matmul.h
856
+ - ext/sources/ggml/src/ggml-openvino/openvino/translate_session.cpp
857
+ - ext/sources/ggml/src/ggml-openvino/openvino/translate_session.h
858
+ - ext/sources/ggml/src/ggml-openvino/openvino/utils.cpp
859
+ - ext/sources/ggml/src/ggml-openvino/openvino/utils.h
860
+ - ext/sources/ggml/src/ggml-openvino/utils.cpp
861
+ - ext/sources/ggml/src/ggml-openvino/utils.h
759
862
  - ext/sources/ggml/src/ggml-opt.cpp
760
863
  - ext/sources/ggml/src/ggml-quants.c
761
864
  - ext/sources/ggml/src/ggml-quants.h
@@ -785,6 +888,14 @@ files:
785
888
  - ext/sources/ggml/src/ggml-sycl/dpct/helper.hpp
786
889
  - ext/sources/ggml/src/ggml-sycl/element_wise.cpp
787
890
  - ext/sources/ggml/src/ggml-sycl/element_wise.hpp
891
+ - ext/sources/ggml/src/ggml-sycl/fattn-common.hpp
892
+ - ext/sources/ggml/src/ggml-sycl/fattn-tile.cpp
893
+ - ext/sources/ggml/src/ggml-sycl/fattn-tile.hpp
894
+ - ext/sources/ggml/src/ggml-sycl/fattn-vec.hpp
895
+ - ext/sources/ggml/src/ggml-sycl/fattn.cpp
896
+ - ext/sources/ggml/src/ggml-sycl/fattn.hpp
897
+ - ext/sources/ggml/src/ggml-sycl/gated_delta_net.cpp
898
+ - ext/sources/ggml/src/ggml-sycl/gated_delta_net.hpp
788
899
  - ext/sources/ggml/src/ggml-sycl/gemm.hpp
789
900
  - ext/sources/ggml/src/ggml-sycl/getrows.cpp
790
901
  - ext/sources/ggml/src/ggml-sycl/getrows.hpp
@@ -824,6 +935,51 @@ files:
824
935
  - ext/sources/ggml/src/ggml-sycl/ssm_conv.hpp
825
936
  - ext/sources/ggml/src/ggml-sycl/sycl_hw.cpp
826
937
  - ext/sources/ggml/src/ggml-sycl/sycl_hw.hpp
938
+ - ext/sources/ggml/src/ggml-sycl/template-instances/fattn-tile-instance-dkq112-dv112.cpp
939
+ - ext/sources/ggml/src/ggml-sycl/template-instances/fattn-tile-instance-dkq128-dv128.cpp
940
+ - ext/sources/ggml/src/ggml-sycl/template-instances/fattn-tile-instance-dkq256-dv256.cpp
941
+ - ext/sources/ggml/src/ggml-sycl/template-instances/fattn-tile-instance-dkq40-dv40.cpp
942
+ - ext/sources/ggml/src/ggml-sycl/template-instances/fattn-tile-instance-dkq576-dv512.cpp
943
+ - ext/sources/ggml/src/ggml-sycl/template-instances/fattn-tile-instance-dkq64-dv64.cpp
944
+ - ext/sources/ggml/src/ggml-sycl/template-instances/fattn-tile-instance-dkq72-dv72.cpp
945
+ - ext/sources/ggml/src/ggml-sycl/template-instances/fattn-tile-instance-dkq80-dv80.cpp
946
+ - ext/sources/ggml/src/ggml-sycl/template-instances/fattn-tile-instance-dkq96-dv96.cpp
947
+ - ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-f16-f16.cpp
948
+ - ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-f16-q4_0.cpp
949
+ - ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-f16-q4_1.cpp
950
+ - ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-f16-q5_0.cpp
951
+ - ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-f16-q5_1.cpp
952
+ - ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-f16-q8_0.cpp
953
+ - ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_0-f16.cpp
954
+ - ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_0-q4_0.cpp
955
+ - ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_0-q4_1.cpp
956
+ - ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_0-q5_0.cpp
957
+ - ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_0-q5_1.cpp
958
+ - ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_0-q8_0.cpp
959
+ - ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_1-f16.cpp
960
+ - ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_1-q4_0.cpp
961
+ - ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_1-q4_1.cpp
962
+ - ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_1-q5_0.cpp
963
+ - ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_1-q5_1.cpp
964
+ - ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q4_1-q8_0.cpp
965
+ - ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_0-f16.cpp
966
+ - ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_0-q4_0.cpp
967
+ - ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_0-q4_1.cpp
968
+ - ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_0-q5_0.cpp
969
+ - ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_0-q5_1.cpp
970
+ - ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_0-q8_0.cpp
971
+ - ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_1-f16.cpp
972
+ - ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_1-q4_0.cpp
973
+ - ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_1-q4_1.cpp
974
+ - ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_1-q5_0.cpp
975
+ - ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_1-q5_1.cpp
976
+ - ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q5_1-q8_0.cpp
977
+ - ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q8_0-f16.cpp
978
+ - ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q8_0-q4_0.cpp
979
+ - ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q8_0-q4_1.cpp
980
+ - ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q8_0-q5_0.cpp
981
+ - ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q8_0-q5_1.cpp
982
+ - ext/sources/ggml/src/ggml-sycl/template-instances/fattn-vec-instance-q8_0-q8_0.cpp
827
983
  - ext/sources/ggml/src/ggml-sycl/tsembd.cpp
828
984
  - ext/sources/ggml/src/ggml-sycl/tsembd.hpp
829
985
  - ext/sources/ggml/src/ggml-sycl/vecdotq.hpp
@@ -831,6 +987,48 @@ files:
831
987
  - ext/sources/ggml/src/ggml-sycl/wkv.hpp
832
988
  - ext/sources/ggml/src/ggml-threading.cpp
833
989
  - ext/sources/ggml/src/ggml-threading.h
990
+ - ext/sources/ggml/src/ggml-virtgpu/CMakeLists.txt
991
+ - ext/sources/ggml/src/ggml-virtgpu/apir_cs_ggml-rpc-front.cpp
992
+ - ext/sources/ggml/src/ggml-virtgpu/backend/CMakeLists.txt
993
+ - ext/sources/ggml/src/ggml-virtgpu/backend/apir_cs_ggml-rpc-back.cpp
994
+ - ext/sources/ggml/src/ggml-virtgpu/backend/backend-convert.h
995
+ - ext/sources/ggml/src/ggml-virtgpu/backend/backend-dispatched-backend.cpp
996
+ - ext/sources/ggml/src/ggml-virtgpu/backend/backend-dispatched-buffer-type.cpp
997
+ - ext/sources/ggml/src/ggml-virtgpu/backend/backend-dispatched-buffer.cpp
998
+ - ext/sources/ggml/src/ggml-virtgpu/backend/backend-dispatched-device.cpp
999
+ - ext/sources/ggml/src/ggml-virtgpu/backend/backend-dispatched.cpp
1000
+ - ext/sources/ggml/src/ggml-virtgpu/backend/backend-dispatched.gen.h
1001
+ - ext/sources/ggml/src/ggml-virtgpu/backend/backend-dispatched.h
1002
+ - ext/sources/ggml/src/ggml-virtgpu/backend/backend-virgl-apir.h
1003
+ - ext/sources/ggml/src/ggml-virtgpu/backend/backend.cpp
1004
+ - ext/sources/ggml/src/ggml-virtgpu/backend/shared/api_remoting.h
1005
+ - ext/sources/ggml/src/ggml-virtgpu/backend/shared/apir_backend.gen.h
1006
+ - ext/sources/ggml/src/ggml-virtgpu/backend/shared/apir_backend.h
1007
+ - ext/sources/ggml/src/ggml-virtgpu/backend/shared/apir_cs.h
1008
+ - ext/sources/ggml/src/ggml-virtgpu/backend/shared/apir_cs_ggml.h
1009
+ - ext/sources/ggml/src/ggml-virtgpu/backend/shared/apir_cs_rpc.h
1010
+ - ext/sources/ggml/src/ggml-virtgpu/ggml-backend-buffer-type.cpp
1011
+ - ext/sources/ggml/src/ggml-virtgpu/ggml-backend-buffer.cpp
1012
+ - ext/sources/ggml/src/ggml-virtgpu/ggml-backend-device.cpp
1013
+ - ext/sources/ggml/src/ggml-virtgpu/ggml-backend-reg.cpp
1014
+ - ext/sources/ggml/src/ggml-virtgpu/ggml-backend.cpp
1015
+ - ext/sources/ggml/src/ggml-virtgpu/ggml-remoting.h
1016
+ - ext/sources/ggml/src/ggml-virtgpu/ggmlremoting_functions.yaml
1017
+ - ext/sources/ggml/src/ggml-virtgpu/include/apir_hw.h
1018
+ - ext/sources/ggml/src/ggml-virtgpu/regenerate_remoting.py
1019
+ - ext/sources/ggml/src/ggml-virtgpu/virtgpu-apir.h
1020
+ - ext/sources/ggml/src/ggml-virtgpu/virtgpu-forward-backend.cpp
1021
+ - ext/sources/ggml/src/ggml-virtgpu/virtgpu-forward-buffer-type.cpp
1022
+ - ext/sources/ggml/src/ggml-virtgpu/virtgpu-forward-buffer.cpp
1023
+ - ext/sources/ggml/src/ggml-virtgpu/virtgpu-forward-device.cpp
1024
+ - ext/sources/ggml/src/ggml-virtgpu/virtgpu-forward-impl.h
1025
+ - ext/sources/ggml/src/ggml-virtgpu/virtgpu-forward.gen.h
1026
+ - ext/sources/ggml/src/ggml-virtgpu/virtgpu-shm.cpp
1027
+ - ext/sources/ggml/src/ggml-virtgpu/virtgpu-shm.h
1028
+ - ext/sources/ggml/src/ggml-virtgpu/virtgpu-utils.cpp
1029
+ - ext/sources/ggml/src/ggml-virtgpu/virtgpu-utils.h
1030
+ - ext/sources/ggml/src/ggml-virtgpu/virtgpu.cpp
1031
+ - ext/sources/ggml/src/ggml-virtgpu/virtgpu.h
834
1032
  - ext/sources/ggml/src/ggml-vulkan/CMakeLists.txt
835
1033
  - ext/sources/ggml/src/ggml-vulkan/cmake/host-toolchain.cmake.in
836
1034
  - ext/sources/ggml/src/ggml-vulkan/ggml-vulkan.cpp
@@ -888,6 +1086,7 @@ files:
888
1086
  - ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/diag.comp
889
1087
  - ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/diag_mask_inf.comp
890
1088
  - ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/div.comp
1089
+ - ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/elu.comp
891
1090
  - ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/exp.comp
892
1091
  - ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/feature-tests/bfloat16.comp
893
1092
  - ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/feature-tests/coopmat.comp
@@ -898,8 +1097,10 @@ files:
898
1097
  - ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_base.glsl
899
1098
  - ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_cm1.comp
900
1099
  - ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_cm2.comp
1100
+ - ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_mask_opt.comp
901
1101
  - ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_split_k_reduce.comp
902
1102
  - ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/floor.comp
1103
+ - ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/gated_delta_net.comp
903
1104
  - ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/geglu.comp
904
1105
  - ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/geglu_erf.comp
905
1106
  - ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/geglu_quick.comp
@@ -975,6 +1176,7 @@ files:
975
1176
  - ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/round.comp
976
1177
  - ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rte.glsl
977
1178
  - ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/scale.comp
1179
+ - ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/sgn.comp
978
1180
  - ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/sigmoid.comp
979
1181
  - ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/silu.comp
980
1182
  - ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/silu_back.comp
@@ -1015,27 +1217,33 @@ files:
1015
1217
  - ext/sources/ggml/src/ggml-webgpu/ggml-webgpu-shader-lib.hpp
1016
1218
  - ext/sources/ggml/src/ggml-webgpu/ggml-webgpu.cpp
1017
1219
  - ext/sources/ggml/src/ggml-webgpu/pre_wgsl.hpp
1018
- - ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/bin_op.tmpl.wgsl
1019
- - ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/binary_head.tmpl
1220
+ - ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/argmax.wgsl
1221
+ - ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/argsort.wgsl
1222
+ - ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/argsort_merge.wgsl
1223
+ - ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/binary.wgsl
1020
1224
  - ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/common_decls.tmpl
1225
+ - ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/concat.wgsl
1021
1226
  - ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/cpy.tmpl.wgsl
1227
+ - ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/cumsum.wgsl
1022
1228
  - ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/embed_wgsl.py
1023
1229
  - ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/flash_attn.wgsl
1024
- - ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/get_rows.tmpl.wgsl
1230
+ - ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/get_rows.wgsl
1025
1231
  - ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/glu.tmpl.wgsl
1026
1232
  - ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/memset.wgsl
1027
- - ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/mul_mat.tmpl.wgsl
1233
+ - ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/mul_mat.wgsl
1028
1234
  - ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/mul_mat_decls.tmpl
1029
- - ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/mul_mat_reg_tile.tmpl.wgsl
1030
- - ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/mul_mat_subgroup_matrix.tmpl.wgsl
1031
- - ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/mul_mat_vec.tmpl.wgsl
1235
+ - ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/mul_mat_reg_tile.wgsl
1236
+ - ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/mul_mat_subgroup_matrix.wgsl
1237
+ - ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/mul_mat_vec.wgsl
1238
+ - ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/pad.wgsl
1239
+ - ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/repeat.wgsl
1032
1240
  - ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/rms_norm.wgsl
1033
1241
  - ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/rope.tmpl.wgsl
1034
- - ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/scale.tmpl.wgsl
1035
- - ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/set_rows.tmpl.wgsl
1242
+ - ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/scale.wgsl
1036
1243
  - ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/set_rows.wgsl
1037
1244
  - ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/soft_max.tmpl.wgsl
1038
- - ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/unary_op.wgsl
1245
+ - ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/sum_rows.wgsl
1246
+ - ext/sources/ggml/src/ggml-webgpu/wgsl-shaders/unary.wgsl
1039
1247
  - ext/sources/ggml/src/ggml-zdnn/CMakeLists.txt
1040
1248
  - ext/sources/ggml/src/ggml-zdnn/common.hpp
1041
1249
  - ext/sources/ggml/src/ggml-zdnn/ggml-zdnn.cpp
@@ -1096,6 +1304,7 @@ files:
1096
1304
  - test/jfk_reader/extconf.rb
1097
1305
  - test/jfk_reader/jfk_reader.c
1098
1306
  - test/test_callback.rb
1307
+ - test/test_context_params.rb
1099
1308
  - test/test_error.rb
1100
1309
  - test/test_model.rb
1101
1310
  - test/test_package.rb
@@ -1113,6 +1322,7 @@ homepage: https://github.com/ggml-org/whisper.cpp
1113
1322
  licenses:
1114
1323
  - MIT
1115
1324
  metadata: {}
1325
+ post_install_message:
1116
1326
  rdoc_options:
1117
1327
  - "--main"
1118
1328
  - README.md
@@ -1129,7 +1339,8 @@ required_rubygems_version: !ruby/object:Gem::Requirement
1129
1339
  - !ruby/object:Gem::Version
1130
1340
  version: '0'
1131
1341
  requirements: []
1132
- rubygems_version: 3.6.9
1342
+ rubygems_version: 3.5.22
1343
+ signing_key:
1133
1344
  specification_version: 4
1134
1345
  summary: Ruby whisper.cpp bindings
1135
1346
  test_files:
@@ -1138,6 +1349,7 @@ test_files:
1138
1349
  - test/jfk_reader/extconf.rb
1139
1350
  - test/jfk_reader/jfk_reader.c
1140
1351
  - test/test_callback.rb
1352
+ - test/test_context_params.rb
1141
1353
  - test/test_error.rb
1142
1354
  - test/test_model.rb
1143
1355
  - test/test_package.rb
@@ -1,54 +0,0 @@
1
- # Add new build types
2
-
3
- # ReleaseGG - Release with enabled asserts
4
-
5
- SET(CMAKE_CXX_FLAGS_RELEASEGG
6
- "-O3"
7
- CACHE STRING "Flags used by the c++ compiler during release builds with enabled asserts."
8
- FORCE )
9
- SET(CMAKE_C_FLAGS_RELEASEGG
10
- "-O3"
11
- CACHE STRING "Flags used by the compiler during release builds with enabled asserts."
12
- FORCE )
13
- SET(CMAKE_EXE_LINKER_FLAGS_RELEASEGG
14
- ""
15
- CACHE STRING "Flags used for linking binaries during release builds with enabled asserts."
16
- FORCE )
17
- SET(CMAKE_SHARED_LINKER_FLAGS_RELEASEGG
18
- ""
19
- CACHE STRING "Flags used by the shared libraries linker during release builds with enabled asserts."
20
- FORCE )
21
- MARK_AS_ADVANCED(
22
- CMAKE_CXX_FLAGS_RELEASEGG
23
- CMAKE_C_FLAGS_RELEASEGG
24
- CMAKE_EXE_LINKER_FLAGS_RELEASEGG
25
- CMAKE_SHARED_LINKER_FLAGS_RELEASEGG )
26
-
27
- # RelWithDebInfoGG - RelWithDebInfo with enabled asserts
28
-
29
- SET(CMAKE_CXX_FLAGS_RELWITHDEBINFOGG
30
- "-O2 -g"
31
- CACHE STRING "Flags used by the c++ compiler during release builds with debug symbols and enabled asserts."
32
- FORCE )
33
- SET(CMAKE_C_FLAGS_RELWITHDEBINFOGG
34
- "-O2 -g"
35
- CACHE STRING "Flags used by the compiler during release builds with debug symbols and enabled asserts."
36
- FORCE )
37
- SET(CMAKE_EXE_LINKER_FLAGS_RELWITHDEBINFOGG
38
- ""
39
- CACHE STRING "Flags used for linking binaries during release builds with debug symbols and enabled asserts."
40
- FORCE )
41
- SET(CMAKE_SHARED_LINKER_FLAGS_RELWITHDEBINFOGG
42
- ""
43
- CACHE STRING "Flags used by the shared libraries linker during release builds with debug symbols and enabled asserts."
44
- FORCE )
45
- MARK_AS_ADVANCED(
46
- CMAKE_CXX_FLAGS_RELWITHDEBINFOGG
47
- CMAKE_C_FLAGS_RELWITHDEBINFOGG
48
- CMAKE_EXE_LINKER_FLAGS_RELWITHDEBINFOGG
49
- CMAKE_SHARED_LINKER_FLAGS_RELWITHDEBINFOGG )
50
-
51
- if (NOT XCODE AND NOT MSVC AND NOT CMAKE_BUILD_TYPE)
52
- set(CMAKE_BUILD_TYPE Release CACHE STRING "Build type" FORCE)
53
- set_property(CACHE CMAKE_BUILD_TYPE PROPERTY STRINGS "Debug" "Release" "MinSizeRel" "RelWithDebInfo" "ReleaseGG" "RelWithDebInfoGG")
54
- endif()