node-llama-cpp 3.0.0-beta.9 → 3.0.0

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 (684) hide show
  1. package/README.md +42 -27
  2. package/bins/_linux-arm64.moved.txt +1 -0
  3. package/bins/_linux-armv7l.moved.txt +1 -0
  4. package/bins/_linux-x64-cuda.moved.txt +1 -0
  5. package/bins/_linux-x64-vulkan.moved.txt +1 -0
  6. package/bins/_linux-x64.moved.txt +1 -0
  7. package/bins/_mac-arm64-metal.moved.txt +1 -0
  8. package/bins/_mac-x64.moved.txt +1 -0
  9. package/bins/_win-arm64.moved.txt +1 -0
  10. package/bins/_win-x64-cuda.moved.txt +1 -0
  11. package/bins/_win-x64-vulkan.moved.txt +1 -0
  12. package/bins/_win-x64.moved.txt +1 -0
  13. package/dist/ChatWrapper.d.ts +19 -39
  14. package/dist/ChatWrapper.js +129 -72
  15. package/dist/ChatWrapper.js.map +1 -1
  16. package/dist/apiDocsIndex.d.ts +1 -0
  17. package/dist/apiDocsIndex.js +7 -0
  18. package/dist/apiDocsIndex.js.map +1 -0
  19. package/dist/bindings/AddonTypes.d.ts +88 -20
  20. package/dist/bindings/Llama.d.ts +43 -3
  21. package/dist/bindings/Llama.js +193 -23
  22. package/dist/bindings/Llama.js.map +1 -1
  23. package/dist/bindings/consts.d.ts +2 -0
  24. package/dist/bindings/consts.js +13 -0
  25. package/dist/bindings/consts.js.map +1 -0
  26. package/dist/bindings/getLlama.d.ts +123 -18
  27. package/dist/bindings/getLlama.js +264 -75
  28. package/dist/bindings/getLlama.js.map +1 -1
  29. package/dist/bindings/types.d.ts +29 -5
  30. package/dist/bindings/types.js +51 -2
  31. package/dist/bindings/types.js.map +1 -1
  32. package/dist/bindings/utils/MemoryOrchestrator.d.ts +21 -0
  33. package/dist/bindings/utils/MemoryOrchestrator.js +49 -0
  34. package/dist/bindings/utils/MemoryOrchestrator.js.map +1 -0
  35. package/dist/bindings/utils/asyncEvery.d.ts +5 -0
  36. package/dist/bindings/utils/asyncEvery.js +15 -0
  37. package/dist/bindings/utils/asyncEvery.js.map +1 -0
  38. package/dist/bindings/utils/asyncSome.d.ts +5 -0
  39. package/dist/bindings/utils/asyncSome.js +27 -0
  40. package/dist/bindings/utils/asyncSome.js.map +1 -0
  41. package/dist/bindings/utils/clearAllLocalBuilds.js.map +1 -1
  42. package/dist/bindings/utils/cloneLlamaCppRepo.d.ts +1 -1
  43. package/dist/bindings/utils/cloneLlamaCppRepo.js +39 -28
  44. package/dist/bindings/utils/cloneLlamaCppRepo.js.map +1 -1
  45. package/dist/bindings/utils/compileLLamaCpp.d.ts +11 -3
  46. package/dist/bindings/utils/compileLLamaCpp.js +250 -81
  47. package/dist/bindings/utils/compileLLamaCpp.js.map +1 -1
  48. package/dist/bindings/utils/detectAvailableComputeLayers.d.ts +14 -0
  49. package/dist/bindings/utils/detectAvailableComputeLayers.js +305 -0
  50. package/dist/bindings/utils/detectAvailableComputeLayers.js.map +1 -0
  51. package/dist/bindings/utils/detectGlibc.d.ts +4 -0
  52. package/dist/bindings/utils/detectGlibc.js +46 -0
  53. package/dist/bindings/utils/detectGlibc.js.map +1 -0
  54. package/dist/bindings/utils/getBestComputeLayersAvailable.d.ts +9 -0
  55. package/dist/bindings/utils/getBestComputeLayersAvailable.js +29 -0
  56. package/dist/bindings/utils/getBestComputeLayersAvailable.js.map +1 -0
  57. package/dist/bindings/utils/getBuildFolderNameForBuildOptions.js +14 -6
  58. package/dist/bindings/utils/getBuildFolderNameForBuildOptions.js.map +1 -1
  59. package/dist/bindings/utils/getExampleUsageCodeOfGetLlama.js.map +1 -1
  60. package/dist/bindings/utils/getGpuTypesToUseForOption.d.ts +12 -0
  61. package/dist/bindings/utils/getGpuTypesToUseForOption.js +39 -0
  62. package/dist/bindings/utils/getGpuTypesToUseForOption.js.map +1 -0
  63. package/dist/bindings/utils/getLinuxDistroInfo.d.ts +9 -0
  64. package/dist/bindings/utils/getLinuxDistroInfo.js +46 -0
  65. package/dist/bindings/utils/getLinuxDistroInfo.js.map +1 -0
  66. package/dist/bindings/utils/getLlamaWithoutBackend.d.ts +5 -0
  67. package/dist/bindings/utils/getLlamaWithoutBackend.js +27 -0
  68. package/dist/bindings/utils/getLlamaWithoutBackend.js.map +1 -0
  69. package/dist/bindings/utils/getPlatform.js.map +1 -1
  70. package/dist/bindings/utils/getPlatformInfo.d.ts +5 -0
  71. package/dist/bindings/utils/getPlatformInfo.js +28 -0
  72. package/dist/bindings/utils/getPlatformInfo.js.map +1 -0
  73. package/dist/bindings/utils/hasBuildingFromSourceDependenciesInstalled.d.ts +3 -0
  74. package/dist/bindings/utils/hasBuildingFromSourceDependenciesInstalled.js +27 -0
  75. package/dist/bindings/utils/hasBuildingFromSourceDependenciesInstalled.js.map +1 -0
  76. package/dist/bindings/utils/hasFileInPath.d.ts +2 -0
  77. package/dist/bindings/utils/hasFileInPath.js +34 -0
  78. package/dist/bindings/utils/hasFileInPath.js.map +1 -0
  79. package/dist/bindings/utils/lastBuildInfo.js.map +1 -1
  80. package/dist/bindings/utils/logBinaryUsageExampleToConsole.d.ts +1 -1
  81. package/dist/bindings/utils/logBinaryUsageExampleToConsole.js +3 -9
  82. package/dist/bindings/utils/logBinaryUsageExampleToConsole.js.map +1 -1
  83. package/dist/bindings/utils/logDistroInstallInstruction.d.ts +14 -0
  84. package/dist/bindings/utils/logDistroInstallInstruction.js +48 -0
  85. package/dist/bindings/utils/logDistroInstallInstruction.js.map +1 -0
  86. package/dist/bindings/utils/resolveCustomCmakeOptions.js +26 -26
  87. package/dist/bindings/utils/resolveCustomCmakeOptions.js.map +1 -1
  88. package/dist/bindings/utils/testBindingBinary.d.ts +1 -0
  89. package/dist/bindings/utils/testBindingBinary.js +100 -0
  90. package/dist/bindings/utils/testBindingBinary.js.map +1 -0
  91. package/dist/bindings/utils/testCmakeBinary.d.ts +5 -0
  92. package/dist/bindings/utils/testCmakeBinary.js +32 -0
  93. package/dist/bindings/utils/testCmakeBinary.js.map +1 -0
  94. package/dist/chatWrappers/AlpacaChatWrapper.d.ts +2 -1
  95. package/dist/chatWrappers/AlpacaChatWrapper.js +10 -2
  96. package/dist/chatWrappers/AlpacaChatWrapper.js.map +1 -1
  97. package/dist/chatWrappers/ChatMLChatWrapper.d.ts +2 -14
  98. package/dist/chatWrappers/ChatMLChatWrapper.js +23 -21
  99. package/dist/chatWrappers/ChatMLChatWrapper.js.map +1 -1
  100. package/dist/chatWrappers/FalconChatWrapper.d.ts +4 -10
  101. package/dist/chatWrappers/FalconChatWrapper.js +39 -21
  102. package/dist/chatWrappers/FalconChatWrapper.js.map +1 -1
  103. package/dist/chatWrappers/FunctionaryChatWrapper.d.ts +8 -32
  104. package/dist/chatWrappers/FunctionaryChatWrapper.js +514 -118
  105. package/dist/chatWrappers/FunctionaryChatWrapper.js.map +1 -1
  106. package/dist/chatWrappers/GemmaChatWrapper.d.ts +7 -0
  107. package/dist/chatWrappers/GemmaChatWrapper.js +96 -0
  108. package/dist/chatWrappers/GemmaChatWrapper.js.map +1 -0
  109. package/dist/chatWrappers/GeneralChatWrapper.d.ts +4 -10
  110. package/dist/chatWrappers/GeneralChatWrapper.js +46 -22
  111. package/dist/chatWrappers/GeneralChatWrapper.js.map +1 -1
  112. package/dist/chatWrappers/Llama2ChatWrapper.d.ts +12 -0
  113. package/dist/chatWrappers/{LlamaChatWrapper.js → Llama2ChatWrapper.js} +37 -20
  114. package/dist/chatWrappers/Llama2ChatWrapper.js.map +1 -0
  115. package/dist/chatWrappers/Llama3ChatWrapper.d.ts +16 -0
  116. package/dist/chatWrappers/Llama3ChatWrapper.js +173 -0
  117. package/dist/chatWrappers/Llama3ChatWrapper.js.map +1 -0
  118. package/dist/chatWrappers/Llama3_1ChatWrapper.d.ts +35 -0
  119. package/dist/chatWrappers/Llama3_1ChatWrapper.js +277 -0
  120. package/dist/chatWrappers/Llama3_1ChatWrapper.js.map +1 -0
  121. package/dist/chatWrappers/MistralChatWrapper.d.ts +15 -0
  122. package/dist/chatWrappers/MistralChatWrapper.js +169 -0
  123. package/dist/chatWrappers/MistralChatWrapper.js.map +1 -0
  124. package/dist/chatWrappers/generic/JinjaTemplateChatWrapper.d.ts +100 -0
  125. package/dist/chatWrappers/generic/JinjaTemplateChatWrapper.js +409 -0
  126. package/dist/chatWrappers/generic/JinjaTemplateChatWrapper.js.map +1 -0
  127. package/dist/chatWrappers/generic/TemplateChatWrapper.d.ts +60 -0
  128. package/dist/chatWrappers/generic/TemplateChatWrapper.js +204 -0
  129. package/dist/chatWrappers/generic/TemplateChatWrapper.js.map +1 -0
  130. package/dist/chatWrappers/generic/utils/chatHistoryFunctionCallMessageTemplate.d.ts +23 -0
  131. package/dist/chatWrappers/generic/utils/chatHistoryFunctionCallMessageTemplate.js +45 -0
  132. package/dist/chatWrappers/generic/utils/chatHistoryFunctionCallMessageTemplate.js.map +1 -0
  133. package/dist/chatWrappers/utils/ChatModelFunctionsDocumentationGenerator.d.ts +57 -0
  134. package/dist/chatWrappers/utils/ChatModelFunctionsDocumentationGenerator.js +119 -0
  135. package/dist/chatWrappers/utils/ChatModelFunctionsDocumentationGenerator.js.map +1 -0
  136. package/dist/chatWrappers/utils/chunkChatItems.d.ts +10 -0
  137. package/dist/chatWrappers/utils/chunkChatItems.js +44 -0
  138. package/dist/chatWrappers/utils/chunkChatItems.js.map +1 -0
  139. package/dist/chatWrappers/utils/isJinjaTemplateEquivalentToSpecializedChatWrapper.d.ts +4 -0
  140. package/dist/chatWrappers/utils/isJinjaTemplateEquivalentToSpecializedChatWrapper.js +221 -0
  141. package/dist/chatWrappers/utils/isJinjaTemplateEquivalentToSpecializedChatWrapper.js.map +1 -0
  142. package/dist/chatWrappers/utils/jsonDumps.d.ts +7 -0
  143. package/dist/chatWrappers/utils/jsonDumps.js +18 -0
  144. package/dist/chatWrappers/utils/jsonDumps.js.map +1 -0
  145. package/dist/chatWrappers/utils/resolveChatWrapper.d.ts +95 -0
  146. package/dist/chatWrappers/utils/resolveChatWrapper.js +335 -0
  147. package/dist/chatWrappers/utils/resolveChatWrapper.js.map +1 -0
  148. package/dist/cli/cli.js +19 -11
  149. package/dist/cli/cli.js.map +1 -1
  150. package/dist/cli/commands/ChatCommand.d.ts +16 -7
  151. package/dist/cli/commands/ChatCommand.js +321 -190
  152. package/dist/cli/commands/ChatCommand.js.map +1 -1
  153. package/dist/cli/commands/CompleteCommand.d.ts +31 -0
  154. package/dist/cli/commands/CompleteCommand.js +402 -0
  155. package/dist/cli/commands/CompleteCommand.js.map +1 -0
  156. package/dist/cli/commands/DebugCommand.d.ts +7 -0
  157. package/dist/cli/commands/DebugCommand.js +54 -0
  158. package/dist/cli/commands/DebugCommand.js.map +1 -0
  159. package/dist/cli/commands/InfillCommand.d.ts +33 -0
  160. package/dist/cli/commands/InfillCommand.js +438 -0
  161. package/dist/cli/commands/InfillCommand.js.map +1 -0
  162. package/dist/cli/commands/InitCommand.d.ts +11 -0
  163. package/dist/cli/commands/InitCommand.js +195 -0
  164. package/dist/cli/commands/InitCommand.js.map +1 -0
  165. package/dist/cli/commands/OnPostInstallCommand.js +6 -2
  166. package/dist/cli/commands/OnPostInstallCommand.js.map +1 -1
  167. package/dist/cli/commands/PullCommand.d.ts +13 -0
  168. package/dist/cli/commands/PullCommand.js +158 -0
  169. package/dist/cli/commands/PullCommand.js.map +1 -0
  170. package/dist/cli/commands/inspect/InspectCommand.d.ts +4 -0
  171. package/dist/cli/commands/inspect/InspectCommand.js +21 -0
  172. package/dist/cli/commands/inspect/InspectCommand.js.map +1 -0
  173. package/dist/cli/commands/inspect/commands/InspectEstimateCommand.d.ts +12 -0
  174. package/dist/cli/commands/inspect/commands/InspectEstimateCommand.js +225 -0
  175. package/dist/cli/commands/inspect/commands/InspectEstimateCommand.js.map +1 -0
  176. package/dist/cli/commands/inspect/commands/InspectGgufCommand.d.ts +12 -0
  177. package/dist/cli/commands/inspect/commands/InspectGgufCommand.js +149 -0
  178. package/dist/cli/commands/inspect/commands/InspectGgufCommand.js.map +1 -0
  179. package/dist/cli/commands/inspect/commands/InspectGpuCommand.d.ts +4 -0
  180. package/dist/cli/commands/inspect/commands/InspectGpuCommand.js +202 -0
  181. package/dist/cli/commands/inspect/commands/InspectGpuCommand.js.map +1 -0
  182. package/dist/cli/commands/inspect/commands/InspectMeasureCommand.d.ts +18 -0
  183. package/dist/cli/commands/inspect/commands/InspectMeasureCommand.js +629 -0
  184. package/dist/cli/commands/inspect/commands/InspectMeasureCommand.js.map +1 -0
  185. package/dist/cli/commands/source/SourceCommand.d.ts +4 -0
  186. package/dist/cli/commands/source/SourceCommand.js +19 -0
  187. package/dist/cli/commands/source/SourceCommand.js.map +1 -0
  188. package/dist/cli/commands/source/commands/BuildCommand.d.ts +16 -0
  189. package/dist/cli/commands/source/commands/BuildCommand.js +148 -0
  190. package/dist/cli/commands/source/commands/BuildCommand.js.map +1 -0
  191. package/dist/cli/commands/{ClearCommand.d.ts → source/commands/ClearCommand.d.ts} +1 -1
  192. package/dist/cli/commands/{ClearCommand.js → source/commands/ClearCommand.js} +11 -10
  193. package/dist/cli/commands/source/commands/ClearCommand.js.map +1 -0
  194. package/dist/cli/commands/{DownloadCommand.d.ts → source/commands/DownloadCommand.d.ts} +5 -4
  195. package/dist/cli/commands/source/commands/DownloadCommand.js +217 -0
  196. package/dist/cli/commands/source/commands/DownloadCommand.js.map +1 -0
  197. package/dist/cli/projectTemplates.d.ts +7 -0
  198. package/dist/cli/projectTemplates.js +10 -0
  199. package/dist/cli/projectTemplates.js.map +1 -0
  200. package/dist/cli/recommendedModels.d.ts +2 -0
  201. package/dist/cli/recommendedModels.js +585 -0
  202. package/dist/cli/recommendedModels.js.map +1 -0
  203. package/dist/cli/startCreateCli.d.ts +2 -0
  204. package/dist/cli/startCreateCli.js +26 -0
  205. package/dist/cli/startCreateCli.js.map +1 -0
  206. package/dist/cli/utils/ConsoleInteraction.d.ts +22 -0
  207. package/dist/cli/utils/ConsoleInteraction.js +122 -0
  208. package/dist/cli/utils/ConsoleInteraction.js.map +1 -0
  209. package/dist/cli/utils/ConsoleTable.d.ts +23 -0
  210. package/dist/cli/utils/ConsoleTable.js +86 -0
  211. package/dist/cli/utils/ConsoleTable.js.map +1 -0
  212. package/dist/cli/utils/basicChooseFromListConsoleInteraction.d.ts +13 -0
  213. package/dist/cli/utils/basicChooseFromListConsoleInteraction.js +111 -0
  214. package/dist/cli/utils/basicChooseFromListConsoleInteraction.js.map +1 -0
  215. package/dist/cli/utils/consolePromptQuestion.d.ts +6 -0
  216. package/dist/cli/utils/consolePromptQuestion.js +82 -0
  217. package/dist/cli/utils/consolePromptQuestion.js.map +1 -0
  218. package/dist/cli/utils/getReadablePath.d.ts +1 -0
  219. package/dist/cli/utils/getReadablePath.js +14 -0
  220. package/dist/cli/utils/getReadablePath.js.map +1 -0
  221. package/dist/cli/utils/interactivelyAskForModel.d.ts +8 -0
  222. package/dist/cli/utils/interactivelyAskForModel.js +450 -0
  223. package/dist/cli/utils/interactivelyAskForModel.js.map +1 -0
  224. package/dist/cli/utils/logUsedGpuTypeOption.d.ts +2 -0
  225. package/dist/cli/utils/logUsedGpuTypeOption.js +9 -0
  226. package/dist/cli/utils/logUsedGpuTypeOption.js.map +1 -0
  227. package/dist/cli/utils/printCommonInfoLines.d.ts +9 -0
  228. package/dist/cli/utils/printCommonInfoLines.js +82 -0
  229. package/dist/cli/utils/printCommonInfoLines.js.map +1 -0
  230. package/dist/cli/utils/printInfoLine.d.ts +12 -0
  231. package/dist/cli/utils/printInfoLine.js +54 -0
  232. package/dist/cli/utils/printInfoLine.js.map +1 -0
  233. package/dist/cli/utils/projectTemplates.d.ts +19 -0
  234. package/dist/cli/utils/projectTemplates.js +47 -0
  235. package/dist/cli/utils/projectTemplates.js.map +1 -0
  236. package/dist/cli/utils/renderModelCompatibilityPercentageWithColors.d.ts +6 -0
  237. package/dist/cli/utils/renderModelCompatibilityPercentageWithColors.js +14 -0
  238. package/dist/cli/utils/renderModelCompatibilityPercentageWithColors.js.map +1 -0
  239. package/dist/cli/utils/resolveCommandGgufPath.d.ts +5 -0
  240. package/dist/cli/utils/resolveCommandGgufPath.js +72 -0
  241. package/dist/cli/utils/resolveCommandGgufPath.js.map +1 -0
  242. package/dist/cli/utils/resolveHeaderFlag.d.ts +1 -0
  243. package/dist/cli/utils/resolveHeaderFlag.js +21 -0
  244. package/dist/cli/utils/resolveHeaderFlag.js.map +1 -0
  245. package/dist/cli/utils/resolveModelRecommendationFileOptions.d.ts +19 -0
  246. package/dist/cli/utils/resolveModelRecommendationFileOptions.js +7 -0
  247. package/dist/cli/utils/resolveModelRecommendationFileOptions.js.map +1 -0
  248. package/dist/cli/utils/splitAnsiToLines.d.ts +1 -0
  249. package/dist/cli/utils/splitAnsiToLines.js +32 -0
  250. package/dist/cli/utils/splitAnsiToLines.js.map +1 -0
  251. package/dist/cli/utils/withCliCommandDescriptionDocsUrl.d.ts +2 -0
  252. package/dist/cli/utils/withCliCommandDescriptionDocsUrl.js +23 -0
  253. package/dist/cli/utils/withCliCommandDescriptionDocsUrl.js.map +1 -0
  254. package/dist/commands.d.ts +4 -3
  255. package/dist/commands.js +6 -3
  256. package/dist/commands.js.map +1 -1
  257. package/dist/config.d.ts +35 -4
  258. package/dist/config.js +58 -17
  259. package/dist/config.js.map +1 -1
  260. package/dist/consts.d.ts +4 -0
  261. package/dist/consts.js +11 -0
  262. package/dist/consts.js.map +1 -0
  263. package/dist/evaluator/LlamaChat/LlamaChat.d.ts +151 -41
  264. package/dist/evaluator/LlamaChat/LlamaChat.js +1289 -437
  265. package/dist/evaluator/LlamaChat/LlamaChat.js.map +1 -1
  266. package/dist/evaluator/LlamaChat/utils/FunctionCallNameGrammar.d.ts +11 -0
  267. package/dist/evaluator/LlamaChat/utils/FunctionCallNameGrammar.js +55 -0
  268. package/dist/evaluator/LlamaChat/utils/FunctionCallNameGrammar.js.map +1 -0
  269. package/dist/evaluator/LlamaChat/utils/FunctionCallParamsGrammar.d.ts +16 -0
  270. package/dist/evaluator/LlamaChat/utils/FunctionCallParamsGrammar.js +45 -0
  271. package/dist/evaluator/LlamaChat/utils/FunctionCallParamsGrammar.js.map +1 -0
  272. package/dist/evaluator/LlamaChat/utils/LlamaFunctionCallValidationError.d.ts +8 -0
  273. package/dist/evaluator/LlamaChat/utils/LlamaFunctionCallValidationError.js +12 -0
  274. package/dist/evaluator/LlamaChat/utils/LlamaFunctionCallValidationError.js.map +1 -0
  275. package/dist/evaluator/LlamaChat/utils/contextShiftStrategies/eraseFirstResponseAndKeepFirstSystemChatContextShiftStrategy.js +27 -17
  276. package/dist/evaluator/LlamaChat/utils/contextShiftStrategies/eraseFirstResponseAndKeepFirstSystemChatContextShiftStrategy.js.map +1 -1
  277. package/dist/evaluator/LlamaChatSession/LlamaChatSession.d.ts +187 -13
  278. package/dist/evaluator/LlamaChatSession/LlamaChatSession.js +280 -53
  279. package/dist/evaluator/LlamaChatSession/LlamaChatSession.js.map +1 -1
  280. package/dist/evaluator/LlamaChatSession/utils/LlamaChatSessionPromptCompletionEngine.d.ts +40 -0
  281. package/dist/evaluator/LlamaChatSession/utils/LlamaChatSessionPromptCompletionEngine.js +186 -0
  282. package/dist/evaluator/LlamaChatSession/utils/LlamaChatSessionPromptCompletionEngine.js.map +1 -0
  283. package/dist/evaluator/LlamaChatSession/utils/defineChatSessionFunction.d.ts +10 -2
  284. package/dist/evaluator/LlamaChatSession/utils/defineChatSessionFunction.js +8 -0
  285. package/dist/evaluator/LlamaChatSession/utils/defineChatSessionFunction.js.map +1 -1
  286. package/dist/evaluator/LlamaCompletion.d.ts +168 -0
  287. package/dist/evaluator/LlamaCompletion.js +470 -0
  288. package/dist/evaluator/LlamaCompletion.js.map +1 -0
  289. package/dist/evaluator/LlamaContext/LlamaContext.d.ts +62 -21
  290. package/dist/evaluator/LlamaContext/LlamaContext.js +501 -120
  291. package/dist/evaluator/LlamaContext/LlamaContext.js.map +1 -1
  292. package/dist/evaluator/LlamaContext/LlamaSampler.d.ts +1 -0
  293. package/dist/evaluator/LlamaContext/LlamaSampler.js +31 -0
  294. package/dist/evaluator/LlamaContext/LlamaSampler.js.map +1 -0
  295. package/dist/evaluator/LlamaContext/types.d.ts +177 -16
  296. package/dist/evaluator/LlamaContext/utils/batchItemsPrioritizationStrategies/firstInFirstOutStrategy.js.map +1 -0
  297. package/dist/evaluator/LlamaContext/utils/batchItemsPrioritizationStrategies/maximumParallelismStrategy.js.map +1 -0
  298. package/dist/evaluator/LlamaContext/utils/resolveBatchItemsPrioritizationStrategy.d.ts +2 -0
  299. package/dist/evaluator/LlamaContext/utils/{resolveBatchItemsPrioritizingStrategy.js → resolveBatchItemsPrioritizationStrategy.js} +5 -5
  300. package/dist/evaluator/LlamaContext/utils/resolveBatchItemsPrioritizationStrategy.js.map +1 -0
  301. package/dist/evaluator/LlamaEmbedding.d.ts +21 -0
  302. package/dist/evaluator/LlamaEmbedding.js +53 -0
  303. package/dist/evaluator/LlamaEmbedding.js.map +1 -0
  304. package/dist/evaluator/LlamaEmbeddingContext.d.ts +29 -19
  305. package/dist/evaluator/LlamaEmbeddingContext.js +36 -43
  306. package/dist/evaluator/LlamaEmbeddingContext.js.map +1 -1
  307. package/dist/evaluator/LlamaGrammar.d.ts +16 -13
  308. package/dist/evaluator/LlamaGrammar.js +17 -10
  309. package/dist/evaluator/LlamaGrammar.js.map +1 -1
  310. package/dist/evaluator/LlamaGrammarEvaluationState.d.ts +7 -3
  311. package/dist/evaluator/LlamaGrammarEvaluationState.js +8 -4
  312. package/dist/evaluator/LlamaGrammarEvaluationState.js.map +1 -1
  313. package/dist/evaluator/LlamaJsonSchemaGrammar.d.ts +3 -0
  314. package/dist/evaluator/LlamaJsonSchemaGrammar.js +5 -3
  315. package/dist/evaluator/LlamaJsonSchemaGrammar.js.map +1 -1
  316. package/dist/evaluator/LlamaModel/LlamaModel.d.ts +255 -0
  317. package/dist/evaluator/LlamaModel/LlamaModel.js +780 -0
  318. package/dist/evaluator/LlamaModel/LlamaModel.js.map +1 -0
  319. package/dist/evaluator/LlamaModel/utils/TokenAttributes.d.ts +29 -0
  320. package/dist/evaluator/LlamaModel/utils/TokenAttributes.js +65 -0
  321. package/dist/evaluator/LlamaModel/utils/TokenAttributes.js.map +1 -0
  322. package/dist/evaluator/TokenBias.d.ts +34 -0
  323. package/dist/evaluator/TokenBias.js +65 -0
  324. package/dist/evaluator/TokenBias.js.map +1 -0
  325. package/dist/evaluator/TokenMeter.d.ts +45 -0
  326. package/dist/evaluator/TokenMeter.js +74 -0
  327. package/dist/evaluator/TokenMeter.js.map +1 -0
  328. package/dist/gguf/consts.d.ts +4 -0
  329. package/dist/gguf/consts.js +12 -0
  330. package/dist/gguf/consts.js.map +1 -0
  331. package/dist/gguf/errors/InvalidGgufMagicError.d.ts +3 -0
  332. package/dist/gguf/errors/InvalidGgufMagicError.js +6 -0
  333. package/dist/gguf/errors/InvalidGgufMagicError.js.map +1 -0
  334. package/dist/gguf/errors/UnsupportedGgufValueTypeError.d.ts +4 -0
  335. package/dist/gguf/errors/UnsupportedGgufValueTypeError.js +9 -0
  336. package/dist/gguf/errors/UnsupportedGgufValueTypeError.js.map +1 -0
  337. package/dist/gguf/fileReaders/GgufFileReader.d.ts +36 -0
  338. package/dist/gguf/fileReaders/GgufFileReader.js +109 -0
  339. package/dist/gguf/fileReaders/GgufFileReader.js.map +1 -0
  340. package/dist/gguf/fileReaders/GgufFsFileReader.d.ts +16 -0
  341. package/dist/gguf/fileReaders/GgufFsFileReader.js +62 -0
  342. package/dist/gguf/fileReaders/GgufFsFileReader.js.map +1 -0
  343. package/dist/gguf/fileReaders/GgufNetworkFetchFileReader.d.ts +25 -0
  344. package/dist/gguf/fileReaders/GgufNetworkFetchFileReader.js +92 -0
  345. package/dist/gguf/fileReaders/GgufNetworkFetchFileReader.js.map +1 -0
  346. package/dist/gguf/insights/GgufInsights.d.ts +50 -0
  347. package/dist/gguf/insights/GgufInsights.js +401 -0
  348. package/dist/gguf/insights/GgufInsights.js.map +1 -0
  349. package/dist/gguf/insights/GgufInsightsConfigurationResolver.d.ts +146 -0
  350. package/dist/gguf/insights/GgufInsightsConfigurationResolver.js +226 -0
  351. package/dist/gguf/insights/GgufInsightsConfigurationResolver.js.map +1 -0
  352. package/dist/gguf/insights/utils/resolveContextContextSizeOption.d.ts +19 -0
  353. package/dist/gguf/insights/utils/resolveContextContextSizeOption.js +78 -0
  354. package/dist/gguf/insights/utils/resolveContextContextSizeOption.js.map +1 -0
  355. package/dist/gguf/insights/utils/resolveModelGpuLayersOption.d.ts +15 -0
  356. package/dist/gguf/insights/utils/resolveModelGpuLayersOption.js +183 -0
  357. package/dist/gguf/insights/utils/resolveModelGpuLayersOption.js.map +1 -0
  358. package/dist/gguf/insights/utils/scoreLevels.d.ts +5 -0
  359. package/dist/gguf/insights/utils/scoreLevels.js +16 -0
  360. package/dist/gguf/insights/utils/scoreLevels.js.map +1 -0
  361. package/dist/gguf/parser/GgufV2Parser.d.ts +20 -0
  362. package/dist/gguf/parser/GgufV2Parser.js +156 -0
  363. package/dist/gguf/parser/GgufV2Parser.js.map +1 -0
  364. package/dist/gguf/parser/GgufV3Parser.d.ts +3 -0
  365. package/dist/gguf/parser/GgufV3Parser.js +4 -0
  366. package/dist/gguf/parser/GgufV3Parser.js.map +1 -0
  367. package/dist/gguf/parser/parseGguf.d.ts +8 -0
  368. package/dist/gguf/parser/parseGguf.js +61 -0
  369. package/dist/gguf/parser/parseGguf.js.map +1 -0
  370. package/dist/gguf/readGgufFileInfo.d.ts +45 -0
  371. package/dist/gguf/readGgufFileInfo.js +71 -0
  372. package/dist/gguf/readGgufFileInfo.js.map +1 -0
  373. package/dist/gguf/types/GgufFileInfoTypes.d.ts +84 -0
  374. package/dist/gguf/types/GgufFileInfoTypes.js +18 -0
  375. package/dist/gguf/types/GgufFileInfoTypes.js.map +1 -0
  376. package/dist/gguf/types/GgufMetadataTypes.d.ts +372 -0
  377. package/dist/gguf/types/GgufMetadataTypes.js +114 -0
  378. package/dist/gguf/types/GgufMetadataTypes.js.map +1 -0
  379. package/dist/gguf/types/GgufTensorInfoTypes.d.ts +37 -0
  380. package/dist/gguf/types/GgufTensorInfoTypes.js +33 -0
  381. package/dist/gguf/types/GgufTensorInfoTypes.js.map +1 -0
  382. package/dist/gguf/utils/GgufReadOffset.d.ts +6 -0
  383. package/dist/gguf/utils/GgufReadOffset.js +18 -0
  384. package/dist/gguf/utils/GgufReadOffset.js.map +1 -0
  385. package/dist/gguf/utils/convertMetadataKeyValueRecordToNestedObject.d.ts +6 -0
  386. package/dist/gguf/utils/convertMetadataKeyValueRecordToNestedObject.js +76 -0
  387. package/dist/gguf/utils/convertMetadataKeyValueRecordToNestedObject.js.map +1 -0
  388. package/dist/gguf/utils/getGgufFileTypeName.d.ts +4 -0
  389. package/dist/gguf/utils/getGgufFileTypeName.js +13 -0
  390. package/dist/gguf/utils/getGgufFileTypeName.js.map +1 -0
  391. package/dist/gguf/utils/getGgufMetadataArchitectureData.d.ts +3 -0
  392. package/dist/gguf/utils/getGgufMetadataArchitectureData.js +4 -0
  393. package/dist/gguf/utils/getGgufMetadataArchitectureData.js.map +1 -0
  394. package/dist/gguf/utils/normalizeGgufDownloadUrl.d.ts +1 -0
  395. package/dist/gguf/utils/normalizeGgufDownloadUrl.js +16 -0
  396. package/dist/gguf/utils/normalizeGgufDownloadUrl.js.map +1 -0
  397. package/dist/gguf/utils/resolveBinarySplitGgufPartUrls.d.ts +2 -0
  398. package/dist/gguf/utils/resolveBinarySplitGgufPartUrls.js +39 -0
  399. package/dist/gguf/utils/resolveBinarySplitGgufPartUrls.js.map +1 -0
  400. package/dist/gguf/utils/resolveSplitGgufParts.d.ts +7 -0
  401. package/dist/gguf/utils/resolveSplitGgufParts.js +55 -0
  402. package/dist/gguf/utils/resolveSplitGgufParts.js.map +1 -0
  403. package/dist/index.d.ts +39 -14
  404. package/dist/index.js +29 -8
  405. package/dist/index.js.map +1 -1
  406. package/dist/state.d.ts +2 -0
  407. package/dist/state.js +7 -0
  408. package/dist/state.js.map +1 -1
  409. package/dist/tsconfig.tsbuildinfo +1 -0
  410. package/dist/types.d.ts +131 -5
  411. package/dist/types.js.map +1 -1
  412. package/dist/utils/DisposeGuard.d.ts +13 -0
  413. package/dist/utils/DisposeGuard.js +120 -0
  414. package/dist/utils/DisposeGuard.js.map +1 -0
  415. package/dist/utils/InsufficientMemoryError.d.ts +3 -0
  416. package/dist/utils/InsufficientMemoryError.js +6 -0
  417. package/dist/utils/InsufficientMemoryError.js.map +1 -0
  418. package/dist/utils/LlamaText.d.ts +73 -26
  419. package/dist/utils/LlamaText.js +475 -157
  420. package/dist/utils/LlamaText.js.map +1 -1
  421. package/dist/utils/LruCache.d.ts +12 -0
  422. package/dist/utils/LruCache.js +44 -0
  423. package/dist/utils/LruCache.js.map +1 -0
  424. package/dist/utils/OverridesObject.d.ts +7 -0
  425. package/dist/utils/OverridesObject.js +2 -0
  426. package/dist/utils/OverridesObject.js.map +1 -0
  427. package/dist/utils/ReplHistory.js +5 -1
  428. package/dist/utils/ReplHistory.js.map +1 -1
  429. package/dist/utils/StopGenerationDetector.d.ts +27 -8
  430. package/dist/utils/StopGenerationDetector.js +108 -22
  431. package/dist/utils/StopGenerationDetector.js.map +1 -1
  432. package/dist/utils/ThreadsSplitter.d.ts +32 -0
  433. package/dist/utils/ThreadsSplitter.js +177 -0
  434. package/dist/utils/ThreadsSplitter.js.map +1 -0
  435. package/dist/utils/TokenStreamRegulator.d.ts +10 -4
  436. package/dist/utils/TokenStreamRegulator.js +102 -10
  437. package/dist/utils/TokenStreamRegulator.js.map +1 -1
  438. package/dist/utils/UnsupportedError.d.ts +2 -0
  439. package/dist/utils/UnsupportedError.js +7 -0
  440. package/dist/utils/UnsupportedError.js.map +1 -0
  441. package/dist/utils/appendUserMessageToChatHistory.d.ts +4 -0
  442. package/dist/utils/appendUserMessageToChatHistory.js +4 -0
  443. package/dist/utils/appendUserMessageToChatHistory.js.map +1 -1
  444. package/dist/utils/clearTempFolder.js.map +1 -1
  445. package/dist/utils/cmake.js +23 -10
  446. package/dist/utils/cmake.js.map +1 -1
  447. package/dist/utils/compareTokens.d.ts +1 -1
  448. package/dist/utils/compareTokens.js.map +1 -1
  449. package/dist/utils/createModelDownloader.d.ts +199 -0
  450. package/dist/utils/createModelDownloader.js +405 -0
  451. package/dist/utils/createModelDownloader.js.map +1 -0
  452. package/dist/utils/findBestOption.d.ts +4 -0
  453. package/dist/utils/findBestOption.js +15 -0
  454. package/dist/utils/findBestOption.js.map +1 -0
  455. package/dist/utils/findCharacterRemovalCountToFitChatHistoryInContext.d.ts +1 -0
  456. package/dist/utils/findCharacterRemovalCountToFitChatHistoryInContext.js +23 -12
  457. package/dist/utils/findCharacterRemovalCountToFitChatHistoryInContext.js.map +1 -1
  458. package/dist/utils/gbnfJson/GbnfGrammarGenerator.js.map +1 -1
  459. package/dist/utils/gbnfJson/getGbnfGrammarForGbnfJsonSchema.d.ts +5 -0
  460. package/dist/utils/gbnfJson/getGbnfGrammarForGbnfJsonSchema.js +11 -0
  461. package/dist/utils/gbnfJson/getGbnfGrammarForGbnfJsonSchema.js.map +1 -0
  462. package/dist/utils/gbnfJson/terminals/GbnfArray.d.ts +3 -1
  463. package/dist/utils/gbnfJson/terminals/GbnfArray.js +10 -5
  464. package/dist/utils/gbnfJson/terminals/GbnfArray.js.map +1 -1
  465. package/dist/utils/gbnfJson/terminals/GbnfBoolean.d.ts +1 -1
  466. package/dist/utils/gbnfJson/terminals/GbnfBoolean.js.map +1 -1
  467. package/dist/utils/gbnfJson/terminals/GbnfBooleanValue.js.map +1 -1
  468. package/dist/utils/gbnfJson/terminals/GbnfGrammar.js.map +1 -1
  469. package/dist/utils/gbnfJson/terminals/GbnfNull.d.ts +1 -1
  470. package/dist/utils/gbnfJson/terminals/GbnfNull.js.map +1 -1
  471. package/dist/utils/gbnfJson/terminals/GbnfNumber.d.ts +1 -1
  472. package/dist/utils/gbnfJson/terminals/GbnfNumber.js.map +1 -1
  473. package/dist/utils/gbnfJson/terminals/GbnfNumberValue.js.map +1 -1
  474. package/dist/utils/gbnfJson/terminals/GbnfObjectMap.d.ts +3 -1
  475. package/dist/utils/gbnfJson/terminals/GbnfObjectMap.js +9 -4
  476. package/dist/utils/gbnfJson/terminals/GbnfObjectMap.js.map +1 -1
  477. package/dist/utils/gbnfJson/terminals/GbnfOr.js.map +1 -1
  478. package/dist/utils/gbnfJson/terminals/GbnfRepetition.d.ts +9 -0
  479. package/dist/utils/gbnfJson/terminals/GbnfRepetition.js +37 -0
  480. package/dist/utils/gbnfJson/terminals/GbnfRepetition.js.map +1 -0
  481. package/dist/utils/gbnfJson/terminals/GbnfString.d.ts +1 -1
  482. package/dist/utils/gbnfJson/terminals/GbnfString.js +23 -5
  483. package/dist/utils/gbnfJson/terminals/GbnfString.js.map +1 -1
  484. package/dist/utils/gbnfJson/terminals/GbnfStringValue.js.map +1 -1
  485. package/dist/utils/gbnfJson/terminals/GbnfVerbatimText.js.map +1 -1
  486. package/dist/utils/gbnfJson/terminals/GbnfWhitespace.d.ts +7 -4
  487. package/dist/utils/gbnfJson/terminals/GbnfWhitespace.js +37 -9
  488. package/dist/utils/gbnfJson/terminals/GbnfWhitespace.js.map +1 -1
  489. package/dist/utils/gbnfJson/terminals/gbnfConsts.d.ts +5 -4
  490. package/dist/utils/gbnfJson/terminals/gbnfConsts.js +14 -3
  491. package/dist/utils/gbnfJson/terminals/gbnfConsts.js.map +1 -1
  492. package/dist/utils/gbnfJson/types.d.ts +3 -0
  493. package/dist/utils/gbnfJson/types.js.map +1 -1
  494. package/dist/utils/gbnfJson/utils/GbnfJsonScopeState.d.ts +10 -0
  495. package/dist/utils/gbnfJson/utils/GbnfJsonScopeState.js +15 -0
  496. package/dist/utils/gbnfJson/utils/GbnfJsonScopeState.js.map +1 -0
  497. package/dist/utils/gbnfJson/utils/getGbnfJsonTerminalForGbnfJsonSchema.d.ts +2 -1
  498. package/dist/utils/gbnfJson/utils/getGbnfJsonTerminalForGbnfJsonSchema.js +6 -5
  499. package/dist/utils/gbnfJson/utils/getGbnfJsonTerminalForGbnfJsonSchema.js.map +1 -1
  500. package/dist/utils/gbnfJson/utils/validateObjectAgainstGbnfSchema.js +3 -3
  501. package/dist/utils/gbnfJson/utils/validateObjectAgainstGbnfSchema.js.map +1 -1
  502. package/dist/utils/getBuildDefaults.d.ts +1 -2
  503. package/dist/utils/getBuildDefaults.js +2 -3
  504. package/dist/utils/getBuildDefaults.js.map +1 -1
  505. package/dist/utils/getConsoleLogPrefix.d.ts +1 -1
  506. package/dist/utils/getConsoleLogPrefix.js +5 -4
  507. package/dist/utils/getConsoleLogPrefix.js.map +1 -1
  508. package/dist/utils/getGrammarsFolder.js +1 -1
  509. package/dist/utils/getGrammarsFolder.js.map +1 -1
  510. package/dist/utils/getQueuedTokensBeforeStopTrigger.d.ts +6 -0
  511. package/dist/utils/getQueuedTokensBeforeStopTrigger.js +22 -0
  512. package/dist/utils/getQueuedTokensBeforeStopTrigger.js.map +1 -0
  513. package/dist/utils/getReadableContextSize.d.ts +1 -0
  514. package/dist/utils/getReadableContextSize.js +7 -0
  515. package/dist/utils/getReadableContextSize.js.map +1 -0
  516. package/dist/utils/getTypeScriptTypeStringForGbnfJsonSchema.js +15 -11
  517. package/dist/utils/getTypeScriptTypeStringForGbnfJsonSchema.js.map +1 -1
  518. package/dist/utils/gitReleaseBundles.js +68 -1
  519. package/dist/utils/gitReleaseBundles.js.map +1 -1
  520. package/dist/utils/isToken.d.ts +2 -0
  521. package/dist/utils/isToken.js +4 -0
  522. package/dist/utils/isToken.js.map +1 -0
  523. package/dist/utils/isUrl.d.ts +1 -0
  524. package/dist/utils/isUrl.js +15 -0
  525. package/dist/utils/isUrl.js.map +1 -0
  526. package/dist/utils/mergeUnionTypes.d.ts +10 -0
  527. package/dist/utils/mergeUnionTypes.js +2 -0
  528. package/dist/utils/mergeUnionTypes.js.map +1 -0
  529. package/dist/utils/modelFileAccesTokens.d.ts +4 -0
  530. package/dist/utils/modelFileAccesTokens.js +40 -0
  531. package/dist/utils/modelFileAccesTokens.js.map +1 -0
  532. package/dist/utils/parseModelFileName.d.ts +1 -0
  533. package/dist/utils/parseModelFileName.js +6 -1
  534. package/dist/utils/parseModelFileName.js.map +1 -1
  535. package/dist/utils/parseTextTemplate.d.ts +66 -0
  536. package/dist/utils/parseTextTemplate.js +116 -0
  537. package/dist/utils/parseTextTemplate.js.map +1 -0
  538. package/dist/utils/prettyPrintObject.d.ts +10 -1
  539. package/dist/utils/prettyPrintObject.js +61 -15
  540. package/dist/utils/prettyPrintObject.js.map +1 -1
  541. package/dist/utils/pushAll.d.ts +6 -0
  542. package/dist/utils/pushAll.js +11 -0
  543. package/dist/utils/pushAll.js.map +1 -0
  544. package/dist/utils/removeNullFields.d.ts +2 -2
  545. package/dist/utils/removeNullFields.js.map +1 -1
  546. package/dist/utils/resolveGithubRelease.d.ts +2 -2
  547. package/dist/utils/resolveGithubRelease.js.map +1 -1
  548. package/dist/utils/resolveLastTokens.d.ts +2 -0
  549. package/dist/utils/resolveLastTokens.js +12 -0
  550. package/dist/utils/resolveLastTokens.js.map +1 -0
  551. package/dist/utils/runtime.d.ts +4 -0
  552. package/dist/utils/runtime.js +8 -0
  553. package/dist/utils/runtime.js.map +1 -0
  554. package/dist/utils/safeEventCallback.d.ts +6 -0
  555. package/dist/utils/safeEventCallback.js +29 -0
  556. package/dist/utils/safeEventCallback.js.map +1 -0
  557. package/dist/utils/spawnCommand.d.ts +11 -2
  558. package/dist/utils/spawnCommand.js +55 -7
  559. package/dist/utils/spawnCommand.js.map +1 -1
  560. package/dist/utils/tokenizeInput.d.ts +1 -1
  561. package/dist/utils/tokenizeInput.js +6 -3
  562. package/dist/utils/tokenizeInput.js.map +1 -1
  563. package/dist/utils/transformPromisable.d.ts +40 -0
  564. package/dist/utils/transformPromisable.js +64 -0
  565. package/dist/utils/transformPromisable.js.map +1 -0
  566. package/dist/utils/truncateTextAndRoundToWords.d.ts +2 -0
  567. package/dist/utils/truncateTextAndRoundToWords.js +32 -0
  568. package/dist/utils/truncateTextAndRoundToWords.js.map +1 -1
  569. package/dist/utils/utilTypes.d.ts +3 -0
  570. package/dist/utils/utilTypes.js +2 -0
  571. package/dist/utils/utilTypes.js.map +1 -0
  572. package/dist/utils/waitForLockfileRelease.js.map +1 -1
  573. package/dist/utils/withLockfile.js.map +1 -1
  574. package/dist/utils/withOra.d.ts +2 -0
  575. package/dist/utils/withOra.js +16 -6
  576. package/dist/utils/withOra.js.map +1 -1
  577. package/dist/utils/withProgressLog.d.ts +22 -0
  578. package/dist/utils/withProgressLog.js +211 -0
  579. package/dist/utils/withProgressLog.js.map +1 -0
  580. package/dist/utils/withStatusLogs.js +1 -1
  581. package/dist/utils/withStatusLogs.js.map +1 -1
  582. package/dist/utils/wrapAbortSignal.d.ts +1 -0
  583. package/dist/utils/wrapAbortSignal.js +9 -0
  584. package/dist/utils/wrapAbortSignal.js.map +1 -0
  585. package/llama/CMakeLists.txt +134 -5
  586. package/llama/addon/AddonContext.cpp +629 -0
  587. package/llama/addon/AddonContext.h +52 -0
  588. package/llama/addon/AddonGrammar.cpp +39 -0
  589. package/llama/addon/AddonGrammar.h +19 -0
  590. package/llama/addon/AddonGrammarEvaluationState.cpp +25 -0
  591. package/llama/addon/AddonGrammarEvaluationState.h +17 -0
  592. package/llama/addon/AddonModel.cpp +672 -0
  593. package/llama/addon/AddonModel.h +61 -0
  594. package/llama/addon/AddonModelData.cpp +25 -0
  595. package/llama/addon/AddonModelData.h +15 -0
  596. package/llama/addon/AddonModelLora.cpp +105 -0
  597. package/llama/addon/AddonModelLora.h +28 -0
  598. package/llama/addon/AddonSampler.cpp +513 -0
  599. package/llama/addon/AddonSampler.h +65 -0
  600. package/llama/addon/RingBuffer.h +109 -0
  601. package/llama/addon/addon.cpp +223 -0
  602. package/llama/addon/addonGlobals.cpp +22 -0
  603. package/llama/addon/addonGlobals.h +12 -0
  604. package/llama/addon/globals/addonLog.cpp +136 -0
  605. package/llama/addon/globals/addonLog.h +21 -0
  606. package/llama/addon/globals/addonProgress.cpp +15 -0
  607. package/llama/addon/globals/addonProgress.h +15 -0
  608. package/llama/addon/globals/getGpuInfo.cpp +108 -0
  609. package/llama/addon/globals/getGpuInfo.h +6 -0
  610. package/llama/binariesGithubRelease.json +1 -1
  611. package/llama/gitRelease.bundle +0 -0
  612. package/llama/gpuInfo/cuda-gpu-info.cu +120 -0
  613. package/llama/gpuInfo/cuda-gpu-info.h +10 -0
  614. package/llama/gpuInfo/metal-gpu-info.h +8 -0
  615. package/llama/gpuInfo/metal-gpu-info.mm +30 -0
  616. package/llama/gpuInfo/vulkan-gpu-info.cpp +83 -0
  617. package/llama/gpuInfo/vulkan-gpu-info.h +9 -0
  618. package/llama/grammars/README.md +297 -6
  619. package/llama/grammars/json.gbnf +4 -4
  620. package/llama/grammars/json_arr.gbnf +4 -4
  621. package/llama/llama.cpp.info.json +1 -1
  622. package/llama/toolchains/win32.host-x64.target-arm64.cmake +41 -0
  623. package/package.json +109 -59
  624. package/templates/packed/electron-typescript-react.json +1 -0
  625. package/templates/packed/node-typescript.json +1 -0
  626. package/dist/AbortError.d.ts +0 -2
  627. package/dist/AbortError.js +0 -7
  628. package/dist/AbortError.js.map +0 -1
  629. package/dist/chatWrappers/LlamaChatWrapper.d.ts +0 -13
  630. package/dist/chatWrappers/LlamaChatWrapper.js.map +0 -1
  631. package/dist/chatWrappers/resolveChatWrapperBasedOnModel.d.ts +0 -13
  632. package/dist/chatWrappers/resolveChatWrapperBasedOnModel.js +0 -57
  633. package/dist/chatWrappers/resolveChatWrapperBasedOnModel.js.map +0 -1
  634. package/dist/cli/commands/BuildCommand.d.ts +0 -11
  635. package/dist/cli/commands/BuildCommand.js +0 -106
  636. package/dist/cli/commands/BuildCommand.js.map +0 -1
  637. package/dist/cli/commands/ClearCommand.js.map +0 -1
  638. package/dist/cli/commands/DownloadCommand.js +0 -169
  639. package/dist/cli/commands/DownloadCommand.js.map +0 -1
  640. package/dist/evaluator/LlamaChat/utils/FunctionCallGrammar.d.ts +0 -22
  641. package/dist/evaluator/LlamaChat/utils/FunctionCallGrammar.js +0 -121
  642. package/dist/evaluator/LlamaChat/utils/FunctionCallGrammar.js.map +0 -1
  643. package/dist/evaluator/LlamaContext/utils/batchItemsPrioritizingStrategies/firstInFirstOutStrategy.js.map +0 -1
  644. package/dist/evaluator/LlamaContext/utils/batchItemsPrioritizingStrategies/maximumParallelismStrategy.js.map +0 -1
  645. package/dist/evaluator/LlamaContext/utils/resolveBatchItemsPrioritizingStrategy.d.ts +0 -2
  646. package/dist/evaluator/LlamaContext/utils/resolveBatchItemsPrioritizingStrategy.js.map +0 -1
  647. package/dist/evaluator/LlamaModel.d.ts +0 -120
  648. package/dist/evaluator/LlamaModel.js +0 -320
  649. package/dist/evaluator/LlamaModel.js.map +0 -1
  650. package/dist/utils/getGbnfGrammarForGbnfJsonSchema.d.ts +0 -2
  651. package/dist/utils/getGbnfGrammarForGbnfJsonSchema.js +0 -9
  652. package/dist/utils/getGbnfGrammarForGbnfJsonSchema.js.map +0 -1
  653. package/dist/utils/parseModelTypeDescription.d.ts +0 -6
  654. package/dist/utils/parseModelTypeDescription.js +0 -9
  655. package/dist/utils/parseModelTypeDescription.js.map +0 -1
  656. package/dist/utils/resolveChatWrapper.d.ts +0 -4
  657. package/dist/utils/resolveChatWrapper.js +0 -16
  658. package/dist/utils/resolveChatWrapper.js.map +0 -1
  659. package/llama/addon.cpp +0 -950
  660. package/llamaBins/linux-arm64/.buildMetadata.json +0 -1
  661. package/llamaBins/linux-arm64/llama-addon.node +0 -0
  662. package/llamaBins/linux-armv7l/.buildMetadata.json +0 -1
  663. package/llamaBins/linux-armv7l/llama-addon.node +0 -0
  664. package/llamaBins/linux-x64/.buildMetadata.json +0 -1
  665. package/llamaBins/linux-x64/llama-addon.node +0 -0
  666. package/llamaBins/linux-x64-cuda/.buildMetadata.json +0 -1
  667. package/llamaBins/linux-x64-cuda/llama-addon.node +0 -0
  668. package/llamaBins/mac-arm64-metal/.buildMetadata.json +0 -1
  669. package/llamaBins/mac-arm64-metal/ggml-metal.metal +0 -6119
  670. package/llamaBins/mac-arm64-metal/llama-addon.node +0 -0
  671. package/llamaBins/mac-x64/.buildMetadata.json +0 -1
  672. package/llamaBins/mac-x64/llama-addon.node +0 -0
  673. package/llamaBins/win-x64/.buildMetadata.json +0 -1
  674. package/llamaBins/win-x64/llama-addon.exp +0 -0
  675. package/llamaBins/win-x64/llama-addon.lib +0 -0
  676. package/llamaBins/win-x64/llama-addon.node +0 -0
  677. package/llamaBins/win-x64-cuda/.buildMetadata.json +0 -1
  678. package/llamaBins/win-x64-cuda/llama-addon.exp +0 -0
  679. package/llamaBins/win-x64-cuda/llama-addon.lib +0 -0
  680. package/llamaBins/win-x64-cuda/llama-addon.node +0 -0
  681. /package/dist/evaluator/LlamaContext/utils/{batchItemsPrioritizingStrategies → batchItemsPrioritizationStrategies}/firstInFirstOutStrategy.d.ts +0 -0
  682. /package/dist/evaluator/LlamaContext/utils/{batchItemsPrioritizingStrategies → batchItemsPrioritizationStrategies}/firstInFirstOutStrategy.js +0 -0
  683. /package/dist/evaluator/LlamaContext/utils/{batchItemsPrioritizingStrategies → batchItemsPrioritizationStrategies}/maximumParallelismStrategy.d.ts +0 -0
  684. /package/dist/evaluator/LlamaContext/utils/{batchItemsPrioritizingStrategies → batchItemsPrioritizationStrategies}/maximumParallelismStrategy.js +0 -0
@@ -0,0 +1 @@
1
+ {"version":3,"file":"GgufInsights.js","sourceRoot":"","sources":["../../../src/gguf/insights/GgufInsights.ts"],"names":[],"mappings":"AACA,OAAO,EAAC,sBAAsB,EAAC,MAAM,gDAAgD,CAAC;AACtF,OAAO,EAAC,0BAA0B,EAAE,0BAA0B,EAAC,MAAM,8CAA8C,CAAC;AAGpH,OAAO,EAAC,oBAAoB,EAAC,MAAM,+BAA+B,CAAC;AACnE,OAAO,EAAC,eAAe,EAAC,MAAM,oCAAoC,CAAC;AACnE,OAAO,EAAC,iCAAiC,EAAC,MAAM,wCAAwC,CAAC;AAOzF,MAAM,OAAO,YAAY;IACrB,gBAAgB,CAAiB,MAAM,CAAQ;IAC/C,gBAAgB,CAAkB,UAAU,CAAS;IACrD,gBAAgB,CAAS,YAAY,GAAkB,IAAI,CAAC;IAC5D,gBAAgB,CAAkB,aAAa,CAAe;IAC9D,gBAAgB,CAAkB,sBAAsB,CAAoC;IAE5F,YAAoB,YAA0B,EAAE,KAAY;QACxD,IAAI,CAAC,MAAM,GAAG,KAAK,CAAC;QACpB,IAAI,CAAC,aAAa,GAAG,YAAY,CAAC;QAElC,IAAI,CAAC,UAAU,GAAG,oBAAoB,CAAC,YAAY,CAAC,cAAc,IAAI,EAAE,EAAE,KAAK,CAAC,CAAC;QACjF,IAAI,CAAC,sBAAsB,GAAG,iCAAiC,CAAC,OAAO,CAAC,IAAI,CAAC,CAAC;IAClF,CAAC;IAED;;;;OAIG;IACI,WAAW,CAAC,aAAsB;QACrC,MAAM,QAAQ,GAAa,EAAE,CAAC;QAC9B,MAAM,iBAAiB,GAAG,CAAC,aAAa,IAAI,IAAI,IAAI,aAAa,KAAK,EAAE,CAAC;YACrE,CAAC,CAAC,MAAM,eAAe,CAAC,aAAa,CAAC,IAAI;YAC1C,CAAC,CAAC,EAAE,CAAC;QAET,IAAI,IAAI,CAAC,aAAa,EAAE,QAAQ,EAAE,SAAS,EAAE,IAAI,EAAE,KAAK,KAAK,MAAM;YAC/D,IAAI,CAAC,aAAa,EAAE,QAAQ,EAAE,SAAS,EAAE,IAAI,EAAE,KAAK,IAAI,IAAI,EAC9D,CAAC;YACC,kHAAkH;YAClH,QAAQ,CAAC,IAAI,CACT,kBAAkB,iBAAiB,6CAA6C;gBAChF,iFAAiF;gBACjF,mEAAmE,CACtE,CAAC;QACN,CAAC;QAED,OAAO,QAAQ,CAAC;IACpB,CAAC;IAED,IAAW,YAAY;QACnB,OAAO,IAAI,CAAC,aAAa,CAAC;IAC9B,CAAC;IAED,IAAW,qBAAqB;QAC5B,OAAO,IAAI,CAAC,sBAAsB,CAAC;IACvC,CAAC;IAED,gDAAgD;IAChD,IAAW,gBAAgB;QACvB,OAAO,IAAI,CAAC,aAAa,CAAC,oBAAoB,CAAC,cAAc,CAAC;IAClE,CAAC;IAED,4DAA4D;IAC5D,IAAW,mBAAmB;QAC1B,OAAO,IAAI,CAAC,aAAa,CAAC,oBAAoB,CAAC,gBAAgB,CAAC;IACpE,CAAC;IAED,IAAW,WAAW;QAClB,IAAI,IAAI,CAAC,YAAY,IAAI,IAAI;YACzB,OAAO,IAAI,CAAC,YAAY,CAAC;QAE7B,MAAM,YAAY,GAAG,CAAC,CAAC;QACvB,IAAI,CAAC,YAAY,GAAG,IAAI,CAAC,cAAc,EAAE,GAAG,YAAY,CAAC;QAEzD,OAAO,IAAI,CAAC,YAAY,CAAC;IAC7B,CAAC;IAED,IAAW,SAAS;QAChB,OAAO,IAAI,CAAC,UAAU,CAAC;IAC3B,CAAC;IAED,IAAW,uBAAuB;QAC9B,wDAAwD;QAExD,IAAI,IAAI,CAAC,aAAa,CAAC,QAAQ,EAAE,OAAO,EAAE,YAAY,KAAK,oBAAoB,CAAC,IAAI;YAChF,OAAO,KAAK,CAAC;aACZ,IAAI,IAAI,CAAC,aAAa,CAAC,QAAQ,EAAE,OAAO,EAAE,YAAY,KAAK,oBAAoB,CAAC,MAAM;YACvF,OAAO,KAAK,CAAC;aACZ,CAAC;YACF,MAAM,KAAK,GAAG,IAAI,CAAC,aAAa,CAAC,oBAAoB,EAAE,SAAS,EAAE,UAAU,IAAI,CAAC,CAAC;YAClF,MAAM,KAAK,GAAG,IAAI,CAAC,aAAa,CAAC,oBAAoB,EAAE,gBAAgB,IAAI,CAAC,CAAC;YAC7E,MAAM,UAAU,GAAG,IAAI,CAAC,aAAa,CAAC,oBAAoB,EAAE,SAAS,EAAE,UAAU,IAAI,CAAC,CAAC,KAAK,IAAI,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,KAAK,GAAG,KAAK,CAAC,CAAC,CAAC;YAC1H,MAAM,UAAU,GAAG,IAAI,CAAC,aAAa,CAAC,oBAAoB,EAAE,SAAS,EAAE,YAAY,IAAI,CAAC,CAAC,KAAK,IAAI,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,KAAK,GAAG,KAAK,CAAC,CAAC;YAE1H,IAAI,UAAU,KAAK,UAAU;gBACzB,OAAO,KAAK,CAAC;QACrB,CAAC;QAED,OAAO,IAAI,CAAC;IAChB,CAAC;IAEM,iCAAiC,CAAC,EAAC,SAAS,EAAsB;QACrE,MAAM,EAAC,GAAG,EAAE,GAAG,EAAC,GAAG,IAAI,CAAC,uBAAuB,CAAC,SAAS,CAAC,CAAC;QAE3D,OAAO;YACH,MAAM,EAAE,oBAAoB,CAAC,GAAG,EAAE,IAAI,CAAC,MAAM,CAAC;YAC9C,OAAO,EAAE,oBAAoB,CAAC,GAAG,EAAE,IAAI,CAAC,MAAM,CAAC;SAClD,CAAC;IACN,CAAC;IAED;;;;OAIG;IACI,mCAAmC,CAAC,EACvC,WAAW,EAAE,cAAc,EAAE,SAAS,EAAE,SAAS,EAAE,kBAAkB,GAAG,KAAK,EAAE,oBAAoB,GAAG,IAAI,EAAE,cAAc,GAAG,KAAK,EAIrI;QACG,IAAI,SAAS,IAAI,IAAI;YAAE,SAAS,GAAG,0BAA0B,EAAE,CAAC;QAChE,IAAI,SAAS,IAAI,IAAI;YAAE,SAAS,GAAG,0BAA0B,CAAC,EAAC,WAAW,EAAE,SAAS,EAAC,CAAC,CAAC;QAExF,MAAM,iBAAiB,GAAG,WAAW,GAAG,SAAS,CAAC;QAElD,MAAM,WAAW,GAAG,IAAI,CAAC,WAAW,CAAC;QACrC,MAAM,cAAc,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,IAAI,CAAC,GAAG,CAAC,cAAc,IAAI,WAAW,EAAE,WAAW,CAAC,CAAC,CAAC;QACzF,MAAM,cAAc,GAAG,WAAW,GAAG,cAAc,CAAC;QACpD,MAAM,OAAO,GAAG,IAAI,CAAC,aAAa,CAAC,oBAAoB,CAAC;QAExD,MAAM,cAAc,GAAG,OAAO,CAAC,UAAU,IAAI,IAAI,CAAC,aAAa,CAAC,QAAQ,CAAC,SAAS,EAAE,IAAI,EAAE,MAAM,EAAE,MAAM,IAAI,CAAC,CAAC;QAC9G,MAAM,UAAU,GAAG,cAAc,GAAG,SAAS,CAAC;QAC9C,MAAM,SAAS,GAAG,kBAAkB;YAChC,CAAC,CAAC,CAAC,OAAO,CAAC,gBAAgB,IAAI,CAAC,CAAC,GAAG,SAAS;YAC7C,CAAC,CAAC,CAAC,CAAC;QAER,MAAM,UAAU,GAAG,CAAC,CAAC,CAAC,iBAAiB;QACvC,MAAM,UAAU,GAAG,CAAC,CAAC,CAAC,gBAAgB;QACtC,MAAM,YAAY,GAAG,CAAC,CAAC,CAAC,mBAAmB;QAC3C,MAAM,WAAW,GAAG,CAAC,CAAC,CAAC,kBAAkB;QAEzC,gDAAgD;QAChD,MAAM,QAAQ,GAAG,UAAU,CAAC;QAC5B,MAAM,IAAI,GAAG,EAAE,GAAG,IAAI,CAAC,CAAC,sBAAsB;QAC9C,MAAM,SAAS,GAAG,UAAU,CAAC;QAC7B,MAAM,WAAW,GAAG,SAAS,GAAG,WAAW,CAAC;QAC5C,MAAM,WAAW,GAAG,UAAU,CAAC;QAC/B,MAAM,OAAO,GAAG,UAAU,GAAG,UAAU,CAAC;QACxC,MAAM,cAAc,GAAG,UAAU,CAAC;QAClC,MAAM,UAAU,GAAG,SAAS,GAAG,UAAU,CAAC;QAC1C,MAAM,UAAU,GAAG,UAAU,CAAC;QAC9B,MAAM,OAAO,GAAG,YAAY,CAAC;QAC7B,MAAM,OAAO,GAAG,YAAY,CAAC;QAC7B,MAAM,OAAO,GAAG,YAAY,CAAC;QAC7B,MAAM,GAAG,GAAG,CAAC,GAAG,WAAW,GAAG,cAAc,GAAG,IAAI,CAAC,MAAM,CAAC,OAAO,CAAC,kBAAkB,CAAC;QACtF,MAAM,OAAO,GAAG,IAAI,CAAC,MAAM,CAAC,OAAO,CAAC,YAAY,GAAG,UAAU,GAAG,IAAI,CAAC,MAAM,CAAC,OAAO,CAAC,cAAc,CAAC;QACnG,MAAM,YAAY,GAAG,IAAI,CAAC,aAAa,CAAC,QAAQ,CAAC,OAAO,EAAE,YAAY,KAAK,oBAAoB,CAAC,KAAK;YACjG,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,SAAS,CAAC;YACxB,CAAC,CAAC,iBAAiB,CAAC;QACxB,MAAM,QAAQ,GAAG,YAAY,GAAG,OAAO,CAAC;QAExC,MAAM,cAAc,GAAG,CACnB,QAAQ;YACR,IAAI;YACJ,SAAS;YACT,WAAW;YACX,WAAW;YACX,OAAO;YACP,cAAc;YACd,UAAU;YACV,UAAU;YACV,OAAO;YACP,OAAO;YACP,OAAO;YACP,GAAG;YACH,QAAQ,CACX,CAAC;QAEF,mHAAmH;QACnH,mHAAmH;QACnH,MAAM,2BAA2B,GAAG,GAAG,EAAE;YACrC,MAAM,IAAI,GAAG,IAAI,CAAC,GAAG,CAAC,IAAI,EAAE,CAAC,CAAC,CAAC;YAC/B,MAAM,UAAU,GAAG,IAAI,CAAC,aAAa,CAAC,cAAc,IAAI,EAAE,CAAC;YAE3D,IAAI,4BAA4B,GAAG,CAAC,CAAC;YAErC,IAAI,SAAS,IAAI,IAAI;gBACjB,OAAO,CAAC,CAAC;YAEb,IAAI,IAAI,CAAC,aAAa,CAAC,QAAQ,CAAC,OAAO,EAAE,YAAY,KAAK,oBAAoB,CAAC,KAAK,EAAE,CAAC;gBACnF,MAAM,WAAW,GAAG,IAAI,CAAC,aAAa,CAAC,oBAAoB,CAAC,YAAY,IAAI,CAAC,CAAC;gBAC9E,MAAM,SAAS,GAAG,IAAI,CAAC,aAAa,CAAC,oBAAoB,CAAC,SAAS,EAAE,UAAU,IAAI,CAAC,CAAC;gBACrF,MAAM,eAAe,GAAG,OAAO,CAAC,gBAAgB,IAAI,CAAC,CAAC;gBAEtD,IAAI,WAAW,GAAG,CAAC,EAAE,CAAC;oBAClB,MAAM,gBAAgB,GAAG,IAAI,CAAC,aAAa,CAAC,oBAAoB,CAAC,iBAAiB,IAAI,CAAC,CAAC;oBAExF,OAAO,WAAW,GAAG,SAAS,GAAG,CAAC,CAAC,CAAC,gBAAgB,GAAG,CAAC,CAAC,GAAG,eAAe,CAAC,GAAG,CAAC,iBAAiB,GAAG,SAAS,CAAC,CAAC,CAAC;gBACpH,CAAC;gBAED,OAAO,WAAW,GAAG,SAAS,GAAG,CAAC,eAAe,GAAG,CAAC,iBAAiB,GAAG,SAAS,CAAC,CAAC,CAAC;YACzF,CAAC;iBAAM,IAAI,IAAI,CAAC,aAAa,CAAC,QAAQ,CAAC,OAAO,EAAE,YAAY,KAAK,oBAAoB,CAAC,KAAK,EAAE,CAAC;gBAC1F,IAAI,cAAc,KAAK,IAAI,CAAC,WAAW,EAAE,CAAC;oBACtC,4BAA4B,IAAI,CAAC,IAAI,GAAG,GAAG,CAAC,GAAG,CAC3C,IAAI,CAAC,gBAAgB,IAAI,IAAI;wBACzB,CAAC,CAAC,CAAC;wBACH,CAAC,CAAC,iBAAiB,GAAG,IAAI,CAAC,gBAAgB,CAClD,CAAC;gBACN,CAAC;qBAAM,CAAC;oBACJ,4BAA4B,IAAI,CAAC,IAAI,GAAG,GAAG,CAAC,GAAG,CAC3C,CAAC,IAAI,GAAG,EAAE,CAAC,GAAG,CACV,IAAI,CAAC,gBAAgB,IAAI,IAAI;wBACzB,CAAC,CAAC,CAAC;wBACH,CAAC,CAAC,iBAAiB,GAAG,IAAI,CAAC,gBAAgB,CAClD,CACJ,CAAC;gBACN,CAAC;YACL,CAAC;iBAAM,IAAI,IAAI,CAAC,aAAa,CAAC,QAAQ,CAAC,OAAO,EAAE,YAAY,KAAK,oBAAoB,CAAC,KAAK,EAAE,CAAC;gBAC1F,uFAAuF;gBACvF,sEAAsE;gBAEtE,IAAI,cAAc,KAAK,IAAI,CAAC,WAAW,EAAE,CAAC;oBACtC,4BAA4B,IAAI,CAAC,IAAI,GAAG,EAAE,CAAC,GAAG,CAC1C,CAAC,IAAI,GAAG,GAAG,CAAC,GAAG,CACX,IAAI,CAAC,gBAAgB,IAAI,IAAI;wBACzB,CAAC,CAAC,CAAC;wBACH,CAAC,CAAC,iBAAiB,GAAG,IAAI,CAAC,gBAAgB,CAClD,CACJ,CAAC;gBACN,CAAC;qBAAM,CAAC;oBACJ,4BAA4B,IAAI,CAAC,CAAC,IAAI,GAAG,GAAG,CAAC,GAAG,CAC5C,CAAC,IAAI,GAAG,GAAG,CAAC,GAAG,CACX,IAAI,CAAC,gBAAgB,IAAI,IAAI;wBACzB,CAAC,CAAC,CAAC;wBACH,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,CAAC,CAAC,GAAG,CAAC,iBAAiB,GAAG,IAAI,CAAC,gBAAgB,CAAC,CAAC,CAAC,CACvE,CACJ,CAAC;gBACN,CAAC;YACL,CAAC;iBAAM,IAAI,IAAI,CAAC,aAAa,CAAC,QAAQ,CAAC,OAAO,EAAE,YAAY,KAAK,oBAAoB,CAAC,QAAQ,EAAE,CAAC;gBAC7F,MAAM,SAAS,GAAG,IAAI,CAAC,aAAa,CAAC,oBAAoB,CAAC,SAAS,EAAE,UAAU,IAAI,CAAC,CAAC;gBAErF,OAAO,CAAC,WAAW,GAAG,SAAS,GAAG,iBAAiB,GAAG,SAAS,CAAC,GAAG,CAAC,EAAE,GAAG,IAAI,CAAC,CAAC;gBAE/E,6CAA6C;gBAC7C,uDAAuD;gBACvD,2BAA2B;gBAC3B,4CAA4C;gBAC5C,sBAAsB;gBACtB,8DAA8D;gBAC9D,YAAY;gBACZ,SAAS;gBACT,WAAW;gBACX,uDAAuD;gBACvD,2BAA2B;gBAC3B,4CAA4C;gBAC5C,sBAAsB;gBACtB,8DAA8D;gBAC9D,YAAY;gBACZ,SAAS;gBACT,IAAI;YACR,CAAC;YAED,MAAM,aAAa,GAAG,UAAU,CAAC,MAAM,KAAK,CAAC;gBACzC,CAAC,CAAC,IAAI,CAAC,WAAW,GAAG,CACjB,CACI,CAAC,OAAO,CAAC,gBAAgB,IAAI,CAAC,CAAC;oBAC/B,CAAC,OAAO,CAAC,mBAAmB,IAAI,CAAC,CAAC,CACrC,GAAG,CAAC,CACR;gBACD,CAAC,CAAC,UAAU,CAAC,MAAM,CAAC,CAAC,GAAG,EAAE,MAAM,EAAE,EAAE;oBAChC,OAAO,GAAG,GAAG,MAAM,CAAC,UAAU,CAAC,MAAM,CAAC,CAAC,GAAW,EAAE,GAAG,EAAE,EAAE,CAAC,GAAG,GAAG,MAAM,CAAC,GAAG,CAAC,EAAE,CAAC,CAAC,CAAC;gBACtF,CAAC,EAAE,CAAC,CAAC,CAAC;YAEV,IAAI,IAAI,CAAC,aAAa,CAAC,QAAQ,CAAC,OAAO,EAAE,YAAY,KAAK,oBAAoB,CAAC,IAAI,EAAE,CAAC;gBAClF,+DAA+D;gBAC/D,OAAO,CAAC,aAAa,GAAG,GAAG,GAAG,CAAC,iBAAiB,GAAG,IAAI,CAAC,CAAC,GAAG,4BAA4B,CAAC;YAC7F,CAAC;YAED,+DAA+D;YAC/D,OAAO,CAAC,aAAa,GAAG,MAAM,GAAG,CAAC,iBAAiB,GAAG,IAAI,CAAC,CAAC,GAAG,4BAA4B,CAAC;QAChG,CAAC,CAAC;QAEF,MAAM,mBAAmB,GAAG,CAAC,cAAc,IAAI,CAAC,oBAAoB,CAAC;YACjE,CAAC,CAAC,CAAC;YACH,CAAC,CAAC,2BAA2B,EAAE,CAAC;QAEpC,MAAM,QAAQ,GAAG,cAAc,KAAK,CAAC,CAAC;QAEtC,MAAM,MAAM,GAAG,CACX,CAAC,QAAQ;YACL,CAAC,CAAC,CAAC,cAAc,GAAG,mBAAmB,CAAC;YACxC,CAAC,CAAC,CAAC,CACV;YACG,IAAI,CAAC,4BAA4B,CAAC,iBAAiB,EAAE,cAAc,CAAC,CAAC;QACzE,MAAM,OAAO,GAAG,QAAQ;YACpB,CAAC,CAAC,CACE,cAAc;gBACd,mBAAmB;gBACnB,IAAI,CAAC,4BAA4B,CAC7B,iBAAiB,EACjB,cAAc,GAAG,WAAW;oBACxB,CAAC,CAAC,CAAC,cAAc,GAAG,CAAC,CAAC;oBACtB,CAAC,CAAC,cAAc,CACvB,CACJ;YACD,CAAC,CAAC,CAAC,CAAC;QAER,OAAO;YACH,MAAM;YACN,OAAO;SACV,CAAC;IACN,CAAC;IAED;;;OAGG;IACI,uBAAuB,CAAC,SAAiB;QAI5C,MAAM,UAAU,GAAG,IAAI,CAAC,aAAa,CAAC,cAAc,IAAI,EAAE,CAAC;QAE3D,IAAI,SAAS,KAAK,CAAC,EAAE,CAAC;YAClB,OAAO;gBACH,GAAG,EAAE,UAAU;gBACf,GAAG,EAAE,EAAE;aACV,CAAC;QACN,CAAC;QAED,MAAM,UAAU,GAAG,IAAI,CAAC,cAAc,EAAE,CAAC;QACzC,MAAM,aAAa,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,UAAU,GAAG,SAAS,CAAC,CAAC;QAE1D,MAAM,UAAU,GAAqB,EAAE,CAAC;QACxC,MAAM,UAAU,GAAqB,EAAE,CAAC;QAExC,KAAK,MAAM,gBAAgB,IAAI,UAAU,EAAE,CAAC;YACxC,MAAM,EAAC,WAAW,EAAC,GAAG,eAAe,CAAC,gBAAgB,CAAC,IAAI,CAAC,CAAC;YAE7D,IAAI,SAAS,KAAK,IAAI,CAAC,WAAW,EAAE,CAAC;gBACjC,MAAM,YAAY,GAAG,IAAI,CAAC,aAAa,CAAC,QAAQ,EAAE,OAAO,EAAE,YAAY,CAAC;gBAExE,IAAI,YAAY,KAAK,oBAAoB,CAAC,KAAK,IAAI,YAAY,KAAK,oBAAoB,CAAC,KAAK,EAAE,CAAC;oBAC7F,IAAI,WAAW,IAAI,IAAI,IAAI,WAAW,IAAI,aAAa;wBACnD,UAAU,CAAC,IAAI,CAAC,gBAAgB,CAAC,CAAC;;wBAElC,UAAU,CAAC,IAAI,CAAC,gBAAgB,CAAC,CAAC;oBAEtC,SAAS;gBACb,CAAC;YACL,CAAC;YAED,IAAI,WAAW,IAAI,IAAI,IAAI,WAAW,IAAI,aAAa;gBACnD,UAAU,CAAC,IAAI,CAAC,gBAAgB,CAAC,CAAC;;gBAElC,UAAU,CAAC,IAAI,CAAC,gBAAgB,CAAC,CAAC;QAC1C,CAAC;QAED,OAAO;YACH,GAAG,EAAE,UAAU;YACf,GAAG,EAAE,UAAU;SAClB,CAAC;IACN,CAAC;IAED,gBAAgB;IACT,sCAAsC;QACzC,MAAM,YAAY,GAAG,IAAI,GAAG,EAAU,CAAC;QAEvC,KAAK,MAAM,gBAAgB,IAAI,CAAC,IAAI,CAAC,aAAa,CAAC,cAAc,IAAI,EAAE,CAAC,EAAE,CAAC;YACvE,MAAM,EAAC,WAAW,EAAC,GAAG,eAAe,CAAC,gBAAgB,CAAC,IAAI,CAAC,CAAC;YAE7D,IAAI,WAAW,IAAI,IAAI;gBACnB,YAAY,CAAC,GAAG,CAAC,WAAW,CAAC,CAAC;QACtC,CAAC;QAED,OAAO,YAAY,CAAC,IAAI,CAAC;IAC7B,CAAC;IAED,gBAAgB;IACT,cAAc;QACjB,OAAO,IAAI,CAAC,aAAa,CAAC,oBAAoB,CAAC,WAAW,IAAI,IAAI,CAAC,sCAAsC,EAAE,CAAC;IAChH,CAAC;IAED,gBAAgB;IACT,4BAA4B,CAAC,WAAmB,EAAE,MAAc;QACnE,+CAA+C;QAC/C,MAAM,KAAK,GAAG,IAAI,CAAC,aAAa,CAAC,oBAAoB,CAAC,SAAS,EAAE,UAAU,IAAI,CAAC,CAAC;QACjF,MAAM,KAAK,GAAG,IAAI,CAAC,aAAa,CAAC,oBAAoB,CAAC,gBAAgB,IAAI,CAAC,CAAC;QAC5E,MAAM,UAAU,GAAG,IAAI,CAAC,aAAa,CAAC,oBAAoB,CAAC,SAAS,EAAE,UAAU,IAAI,CAAC,CAAC,KAAK,IAAI,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,KAAK,GAAG,KAAK,CAAC,CAAC,CAAC;QACzH,MAAM,OAAO,GAAG,IAAI,CAAC,aAAa,CAAC,oBAAoB,CAAC,SAAS,EAAE,aAAa,IAAI,KAAK,CAAC;QAC1F,MAAM,cAAc,GAAG,UAAU,GAAG,OAAO,CAAC;QAE5C,MAAM,QAAQ,GAAG,IAAI,CAAC,aAAa,CAAC,oBAAoB,CAAC,GAAG,EAAE,WAAW,IAAI,CAAC,CAAC;QAC/E,MAAM,SAAS,GAAG,IAAI,CAAC,aAAa,CAAC,oBAAoB,CAAC,GAAG,EAAE,UAAU,IAAI,CAAC,CAAC;QAC/E,MAAM,YAAY,GAAG,CAAC,QAAQ,GAAG,CAAC,CAAC,CAAC,CAAC,CAAC,QAAQ,GAAG,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,GAAG,SAAS,CAAC;QAErE,MAAM,UAAU,GAAG,IAAI,CAAC,aAAa,CAAC,oBAAoB,CAAC,SAAS,EAAE,YAAY,IAAI,CAAC,CAAC,KAAK,IAAI,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,KAAK,GAAG,KAAK,CAAC,CAAC;QACzH,MAAM,cAAc,GAAG,UAAU,GAAG,OAAO,CAAC;QAE5C,MAAM,SAAS,GAAG,IAAI,CAAC,aAAa,CAAC,oBAAoB,CAAC,GAAG,EAAE,UAAU,IAAI,CAAC,CAAC;QAC/E,MAAM,YAAY,GAAG,SAAS,GAAG,SAAS,CAAC;QAE3C,MAAM,cAAc,GAAG,cAAc,GAAG,YAAY,CAAC;QACrD,MAAM,cAAc,GAAG,cAAc,GAAG,YAAY,CAAC;QAErD,MAAM,WAAW,GAAG,IAAI,CAAC,aAAa,CAAC,QAAQ,CAAC,OAAO,EAAE,YAAY,KAAK,oBAAoB,CAAC,KAAK;YAChG,sFAAsF;YACtF,0CAA0C;YAC1C,CAAC,CAAC,IAAI,CAAC,MAAM,CAAC,OAAO,CAAC,eAAe;YACrC,CAAC,CAAC,IAAI,CAAC,MAAM,CAAC,OAAO,CAAC,eAAe,CAAC;QAC1C,MAAM,aAAa,GAAG,IAAI,CAAC,aAAa,CAAC,QAAQ,CAAC,OAAO,EAAE,YAAY,KAAK,oBAAoB,CAAC,KAAK;YAClG,sFAAsF;YACtF,0CAA0C;YAC1C,CAAC,CAAC,IAAI,CAAC,MAAM,CAAC,OAAO,CAAC,eAAe;YACrC,CAAC,CAAC,IAAI,CAAC,MAAM,CAAC,OAAO,CAAC,eAAe,CAAC;QAE1C,MAAM,cAAc,GAAG,MAAM,GAAG,cAAc,GAAG,WAAW,GAAG,WAAW,CAAC;QAC3E,MAAM,gBAAgB,GAAG,MAAM,GAAG,cAAc,GAAG,WAAW,GAAG,aAAa,CAAC;QAE/E,OAAO,cAAc,GAAG,gBAAgB,CAAC;IAC7C,CAAC;IAED;;;;;;OAMG;IACI,MAAM,CAAC,KAAK,CAAC,IAAI,CAAC,YAA0B,EAAE,KAAa;QAC9D,IAAI,aAAa,GAAG,KAAK,CAAC;QAC1B,IAAI,aAAa,IAAI,IAAI;YACrB,aAAa,GAAG,MAAM,sBAAsB,EAAE,CAAC;QAEnD,OAAO,IAAI,YAAY,CAAC,YAAY,EAAE,aAAa,CAAC,CAAC;IACzD,CAAC;CACJ;AAED,SAAS,eAAe,CAAC,UAAmB;IAGxC,IAAI,UAAU,IAAI,IAAI;QAClB,OAAO,EAAC,WAAW,EAAE,SAAS,EAAC,CAAC;IAEpC,MAAM,iBAAiB,GAAG,MAAM,CAAC;IACjC,IAAI,CAAC,UAAU,CAAC,UAAU,CAAC,iBAAiB,CAAC;QACzC,OAAO,EAAC,WAAW,EAAE,SAAS,EAAC,CAAC;IAEpC,MAAM,QAAQ,GAAG,UAAU,CAAC,OAAO,CAAC,GAAG,EAAE,iBAAiB,CAAC,MAAM,CAAC,CAAC;IACnE,MAAM,iBAAiB,GAAG,UAAU,CAAC,KAAK,CACtC,iBAAiB,CAAC,MAAM,EACxB,QAAQ,GAAG,CAAC;QACR,CAAC,CAAC,UAAU,CAAC,MAAM;QACnB,CAAC,CAAC,QAAQ,CACjB,CAAC;IAEF,MAAM,WAAW,GAAG,QAAQ,CAAC,iBAAiB,CAAC,CAAC;IAChD,IAAI,MAAM,CAAC,QAAQ,CAAC,WAAW,CAAC;QAC5B,OAAO,EAAC,WAAW,EAAC,CAAC;IAEzB,OAAO,EAAC,WAAW,EAAE,SAAS,EAAC,CAAC;AACpC,CAAC;AAED,SAAS,oBAAoB,CAAC,WAA6B,EAAE,KAAY;IACrE,IAAI,IAAI,GAAG,CAAC,CAAC;IACb,KAAK,MAAM,UAAU,IAAI,WAAW;QAChC,IAAI,IAAI,mBAAmB,CAAC,UAAU,EAAE,KAAK,CAAC,CAAC;IAEnD,OAAO,IAAI,CAAC;AAChB,CAAC;AAED,SAAS,mBAAmB,CAAC,MAAsB,EAAE,KAAY;IAC7D,MAAM,QAAQ,GAAG,KAAK,CAAC,SAAS,CAAC,sBAAsB,CAAC,MAAM,CAAC,QAAQ,CAAC,CAAC;IACzE,MAAM,SAAS,GAAG,KAAK,CAAC,SAAS,CAAC,uBAAuB,CAAC,MAAM,CAAC,QAAQ,CAAC,CAAC;IAC3E,MAAM,WAAW,GAAG,KAAK,CAAC,OAAO,CAAC,WAAW,CAAC;IAE9C,IAAI,QAAQ,IAAI,IAAI,IAAI,SAAS,IAAI,IAAI;QACrC,MAAM,IAAI,KAAK,CAAC,4BAA4B,CAAC,CAAC;IAElD,MAAM,EAAC,EAAE,EAAE,EAAE,EAAC,GAAG,gBAAgB,CAAC,MAAM,EAAE,EAAC,QAAQ,EAAE,SAAS,EAAE,WAAW,EAAC,CAAC,CAAC;IAE9E,IAAI,SAAS,KAAK,CAAC,EAAE,CAAC;QAClB,IAAI,UAAU,GAAG,QAAQ,CAAC;QAC1B,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,WAAW,EAAE,CAAC,EAAE,EAAE,CAAC;YACnC,UAAU,IAAI,CAAC,EAAE,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,GAAG,EAAE,CAAC,CAAC,CAAC,CAAC;QACtC,CAAC;QAED,OAAO,UAAU,CAAC;IACtB,CAAC;SAAM,CAAC;QACJ,IAAI,UAAU,GAAG,IAAI,CAAC,KAAK,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,GAAG,EAAE,CAAC,CAAC,CAAC,CAAC,GAAG,SAAS,CAAC,CAAC;QACzD,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,WAAW,EAAE,CAAC,EAAE,EAAE,CAAC;YACnC,UAAU,IAAI,CAAC,EAAE,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,GAAG,EAAE,CAAC,CAAC,CAAC,CAAC;QACtC,CAAC;QAED,OAAO,UAAU,CAAC;IACtB,CAAC;AACL,CAAC;AAED,SAAS,gBAAgB,CAAC,MAAsB,EAAE,EAC9C,QAAQ,EAAE,SAAS,EAAE,WAAW,EAGnC;IACG,qBAAqB;IACrB,6CAA6C;IAC7C,MAAM,EAAE,GAAG;QACP,GAAG,MAAM,CAAC,UAAU;QACpB,GAAG,CAAC,KAAK,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,WAAW,GAAG,MAAM,CAAC,UAAU,CAAC,MAAM,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;KAC1E,CAAC,KAAK,CAAC,CAAC,EAAE,WAAW,CAAC,CAAC;IAExB,kBAAkB;IAClB,6CAA6C;IAC7C,MAAM,EAAE,GAAG;QACP,QAAQ;QACR,IAAI,CAAC,KAAK,CAAC,QAAQ,GAAG,CAAC,EAAE,CAAC,CAAC,CAAC,GAAG,SAAS,CAAC,CAAC;QAC1C,GAAG,KAAK,CAAC,WAAW,GAAG,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC;KACpC,CAAC;IACF,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,WAAW,EAAE,CAAC,EAAE,EAAE,CAAC;QACnC,EAAE,CAAC,CAAC,CAAC,GAAG,EAAE,CAAC,CAAC,GAAG,CAAC,CAAC,GAAG,EAAE,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC;IAClC,CAAC;IAED,OAAO;QACH,EAAE;QACF,EAAE;KACL,CAAC;AACN,CAAC"}
@@ -0,0 +1,146 @@
1
+ import { BuildGpu } from "../../bindings/types.js";
2
+ import { LlamaModelOptions } from "../../evaluator/LlamaModel/LlamaModel.js";
3
+ import { LlamaContextOptions } from "../../evaluator/LlamaContext/types.js";
4
+ import type { GgufInsights } from "./GgufInsights.js";
5
+ export declare const defaultTrainContextSizeForEstimationPurposes = 4096;
6
+ export declare class GgufInsightsConfigurationResolver {
7
+ private constructor();
8
+ get ggufInsights(): GgufInsights;
9
+ /**
10
+ * Resolve the best configuration for loading a model and creating a context using the current hardware.
11
+ *
12
+ * Specifying a `targetGpuLayers` and/or `targetContextSize` will ensure the resolved configuration matches those values,
13
+ * but note it can lower the compatibility score if the hardware doesn't support it.
14
+ *
15
+ * Overriding hardware values it possible by configuring `hardwareOverrides`.
16
+ * @param options
17
+ * @param hardwareOverrides
18
+ */
19
+ resolveAndScoreConfig({ targetGpuLayers, targetContextSize, embeddingContext, flashAttention }?: {
20
+ targetGpuLayers?: number | "max";
21
+ targetContextSize?: number;
22
+ embeddingContext?: boolean;
23
+ flashAttention?: boolean;
24
+ }, { getVramState, getRamState, llamaVramPaddingSize, llamaGpu, llamaSupportsGpuOffloading }?: {
25
+ getVramState?(): Promise<{
26
+ total: number;
27
+ free: number;
28
+ }>;
29
+ getRamState?(): Promise<{
30
+ total: number;
31
+ free: number;
32
+ }>;
33
+ llamaVramPaddingSize?: number;
34
+ llamaGpu?: BuildGpu;
35
+ llamaSupportsGpuOffloading?: boolean;
36
+ }): Promise<{
37
+ /**
38
+ * A number between `0` (inclusive) and `1` (inclusive) representing the compatibility score.
39
+ */
40
+ compatibilityScore: number;
41
+ /**
42
+ * A number starting at `0` with no upper limit representing the bonus score.
43
+ * For each multiplier of the specified `contextSize` that the resolved context size is larger by, 1 bonus point is given.
44
+ */
45
+ bonusScore: number;
46
+ /**
47
+ * The total score, which is the sum of the compatibility and bonus scores.
48
+ */
49
+ totalScore: number;
50
+ /**
51
+ * The resolved values used to calculate the scores.
52
+ */
53
+ resolvedValues: {
54
+ gpuLayers: number;
55
+ contextSize: number;
56
+ modelRamUsage: number;
57
+ contextRamUsage: number;
58
+ totalRamUsage: number;
59
+ modelVramUsage: number;
60
+ contextVramUsage: number;
61
+ totalVramUsage: number;
62
+ };
63
+ }>;
64
+ /**
65
+ * Score the compatibility of the model configuration with the current GPU and VRAM state.
66
+ * Assumes a model is loaded with the default `"auto"` configurations.
67
+ * Scored based on the following criteria:
68
+ * - The number of GPU layers that can be offloaded to the GPU (only if there's a GPU. If there's no GPU then by how small the model is)
69
+ * - Whether all layers can be offloaded to the GPU (gives additional points)
70
+ * - Whether the resolved context size is at least as large as the specified `contextSize`
71
+ *
72
+ * IF the resolved context size is larger than the specified context size, for each multiplier of the specified `contextSize`
73
+ * that the resolved context size is larger by, 1 bonus point is given in the `bonusScore`.
74
+ *
75
+ * `contextSize` defaults to `4096` (if the model train context size is lower than this, the model train context size is used instead).
76
+ */
77
+ scoreModelConfigurationCompatibility({ contextSize, embeddingContext, flashAttention }?: {
78
+ contextSize?: number;
79
+ embeddingContext?: boolean;
80
+ flashAttention?: boolean;
81
+ }, { getVramState, getRamState, llamaVramPaddingSize, llamaGpu, llamaSupportsGpuOffloading }?: {
82
+ getVramState?(): Promise<{
83
+ total: number;
84
+ free: number;
85
+ }>;
86
+ getRamState?(): Promise<{
87
+ total: number;
88
+ free: number;
89
+ }>;
90
+ llamaVramPaddingSize?: number;
91
+ llamaGpu?: BuildGpu;
92
+ llamaSupportsGpuOffloading?: boolean;
93
+ }): Promise<{
94
+ /**
95
+ * A number between `0` (inclusive) and `1` (inclusive) representing the compatibility score.
96
+ */
97
+ compatibilityScore: number;
98
+ /**
99
+ * A number starting at `0` with no upper limit representing the bonus score.
100
+ * For each multiplier of the specified `contextSize` that the resolved context size is larger by, 1 bonus point is given.
101
+ */
102
+ bonusScore: number;
103
+ /**
104
+ * The total score, which is the sum of the compatibility and bonus scores.
105
+ */
106
+ totalScore: number;
107
+ /**
108
+ * The resolved values used to calculate the scores.
109
+ */
110
+ resolvedValues: {
111
+ gpuLayers: number;
112
+ contextSize: number;
113
+ modelRamUsage: number;
114
+ contextRamUsage: number;
115
+ totalRamUsage: number;
116
+ modelVramUsage: number;
117
+ contextVramUsage: number;
118
+ totalVramUsage: number;
119
+ };
120
+ }>;
121
+ resolveModelGpuLayers(gpuLayers?: LlamaModelOptions["gpuLayers"], { ignoreMemorySafetyChecks, getVramState, llamaVramPaddingSize, llamaGpu, llamaSupportsGpuOffloading, defaultContextFlashAttention }?: {
122
+ ignoreMemorySafetyChecks?: boolean;
123
+ getVramState?(): Promise<{
124
+ total: number;
125
+ free: number;
126
+ }>;
127
+ llamaVramPaddingSize?: number;
128
+ llamaGpu?: BuildGpu;
129
+ llamaSupportsGpuOffloading?: boolean;
130
+ defaultContextFlashAttention?: boolean;
131
+ }): Promise<number>;
132
+ resolveContextContextSize(contextSize: LlamaContextOptions["contextSize"], { modelGpuLayers, batchSize, modelTrainContextSize, flashAttention, getVramState, llamaGpu, ignoreMemorySafetyChecks, isEmbeddingContext, sequences }: {
133
+ modelGpuLayers: number;
134
+ modelTrainContextSize: number;
135
+ flashAttention?: boolean;
136
+ batchSize?: LlamaContextOptions["batchSize"];
137
+ sequences?: number;
138
+ getVramState?(): Promise<{
139
+ total: number;
140
+ free: number;
141
+ }>;
142
+ llamaGpu?: BuildGpu;
143
+ ignoreMemorySafetyChecks?: boolean;
144
+ isEmbeddingContext?: boolean;
145
+ }): Promise<number>;
146
+ }
@@ -0,0 +1,226 @@
1
+ import os from "os";
2
+ import { getDefaultContextSequences } from "../../evaluator/LlamaContext/LlamaContext.js";
3
+ import { resolveModelGpuLayersOption } from "./utils/resolveModelGpuLayersOption.js";
4
+ import { resolveContextContextSizeOption } from "./utils/resolveContextContextSizeOption.js";
5
+ import { scoreLevels } from "./utils/scoreLevels.js";
6
+ export const defaultTrainContextSizeForEstimationPurposes = 4096;
7
+ export class GgufInsightsConfigurationResolver {
8
+ /** @internal */ _ggufInsights;
9
+ constructor(ggufInsights) {
10
+ this._ggufInsights = ggufInsights;
11
+ }
12
+ get ggufInsights() {
13
+ return this._ggufInsights;
14
+ }
15
+ /**
16
+ * Resolve the best configuration for loading a model and creating a context using the current hardware.
17
+ *
18
+ * Specifying a `targetGpuLayers` and/or `targetContextSize` will ensure the resolved configuration matches those values,
19
+ * but note it can lower the compatibility score if the hardware doesn't support it.
20
+ *
21
+ * Overriding hardware values it possible by configuring `hardwareOverrides`.
22
+ * @param options
23
+ * @param hardwareOverrides
24
+ */
25
+ async resolveAndScoreConfig({ targetGpuLayers, targetContextSize, embeddingContext = false, flashAttention = false } = {}, { getVramState = (() => this._ggufInsights._llama._vramOrchestrator.getMemoryState()), getRamState = (async () => ({ total: os.totalmem(), free: os.freemem() })), llamaVramPaddingSize = this._ggufInsights._llama.vramPaddingSize, llamaGpu = this._ggufInsights._llama.gpu, llamaSupportsGpuOffloading = this._ggufInsights._llama.supportsGpuOffloading } = {}) {
26
+ const compatibilityScore = await this.scoreModelConfigurationCompatibility({
27
+ flashAttention,
28
+ contextSize: targetContextSize,
29
+ embeddingContext
30
+ }, {
31
+ getVramState,
32
+ getRamState,
33
+ llamaVramPaddingSize,
34
+ llamaGpu,
35
+ llamaSupportsGpuOffloading
36
+ });
37
+ if (targetContextSize != null || targetGpuLayers != null) {
38
+ const vramState = await getVramState();
39
+ const resolvedGpuLayers = await this.resolveModelGpuLayers(targetGpuLayers == null
40
+ ? {
41
+ fitContext: {
42
+ contextSize: targetContextSize,
43
+ embeddingContext
44
+ }
45
+ }
46
+ : targetGpuLayers, {
47
+ getVramState: async () => vramState,
48
+ defaultContextFlashAttention: flashAttention,
49
+ ignoreMemorySafetyChecks: targetGpuLayers != null,
50
+ llamaGpu,
51
+ llamaSupportsGpuOffloading,
52
+ llamaVramPaddingSize
53
+ });
54
+ const estimatedModelResourceUsage = this._ggufInsights.estimateModelResourceRequirements({
55
+ gpuLayers: resolvedGpuLayers
56
+ });
57
+ const resolvedContextSize = await this._ggufInsights.configurationResolver.resolveContextContextSize(targetContextSize ?? "auto", {
58
+ getVramState: async () => ({
59
+ total: vramState.total,
60
+ free: Math.max(0, vramState.free - estimatedModelResourceUsage.gpuVram)
61
+ }),
62
+ isEmbeddingContext: embeddingContext,
63
+ modelGpuLayers: resolvedGpuLayers,
64
+ modelTrainContextSize: this._ggufInsights.trainContextSize ?? defaultTrainContextSizeForEstimationPurposes,
65
+ flashAttention,
66
+ ignoreMemorySafetyChecks: targetContextSize != null,
67
+ llamaGpu
68
+ });
69
+ const estimatedContextResourceUsage = this._ggufInsights.estimateContextResourceRequirements({
70
+ contextSize: resolvedContextSize,
71
+ isEmbeddingContext: embeddingContext,
72
+ modelGpuLayers: resolvedGpuLayers,
73
+ flashAttention
74
+ });
75
+ compatibilityScore.resolvedValues = {
76
+ gpuLayers: resolvedGpuLayers,
77
+ contextSize: resolvedContextSize,
78
+ modelRamUsage: estimatedModelResourceUsage.cpuRam,
79
+ contextRamUsage: estimatedContextResourceUsage.cpuRam,
80
+ totalRamUsage: estimatedModelResourceUsage.cpuRam + estimatedContextResourceUsage.cpuRam,
81
+ modelVramUsage: estimatedModelResourceUsage.gpuVram,
82
+ contextVramUsage: estimatedContextResourceUsage.gpuVram,
83
+ totalVramUsage: estimatedModelResourceUsage.gpuVram + estimatedContextResourceUsage.gpuVram
84
+ };
85
+ if (compatibilityScore.resolvedValues.totalVramUsage > vramState.total) {
86
+ compatibilityScore.compatibilityScore = 0;
87
+ compatibilityScore.bonusScore = 0;
88
+ compatibilityScore.totalScore = 0;
89
+ }
90
+ }
91
+ return compatibilityScore;
92
+ }
93
+ /**
94
+ * Score the compatibility of the model configuration with the current GPU and VRAM state.
95
+ * Assumes a model is loaded with the default `"auto"` configurations.
96
+ * Scored based on the following criteria:
97
+ * - The number of GPU layers that can be offloaded to the GPU (only if there's a GPU. If there's no GPU then by how small the model is)
98
+ * - Whether all layers can be offloaded to the GPU (gives additional points)
99
+ * - Whether the resolved context size is at least as large as the specified `contextSize`
100
+ *
101
+ * IF the resolved context size is larger than the specified context size, for each multiplier of the specified `contextSize`
102
+ * that the resolved context size is larger by, 1 bonus point is given in the `bonusScore`.
103
+ *
104
+ * `contextSize` defaults to `4096` (if the model train context size is lower than this, the model train context size is used instead).
105
+ */
106
+ async scoreModelConfigurationCompatibility({ contextSize = Math.min(4096, this._ggufInsights.trainContextSize ?? 4096), embeddingContext = false, flashAttention = false } = {}, { getVramState = (() => this._ggufInsights._llama._vramOrchestrator.getMemoryState()), getRamState = (async () => ({ total: os.totalmem(), free: os.freemem() })), llamaVramPaddingSize = this._ggufInsights._llama.vramPaddingSize, llamaGpu = this._ggufInsights._llama.gpu, llamaSupportsGpuOffloading = this._ggufInsights._llama.supportsGpuOffloading } = {}) {
107
+ const [vramState, ramState] = await Promise.all([
108
+ getVramState(),
109
+ getRamState()
110
+ ]);
111
+ const resolvedGpuLayers = await this.resolveModelGpuLayers(embeddingContext
112
+ ? { fitContext: { embeddingContext: true } }
113
+ : "auto", {
114
+ getVramState: async () => vramState,
115
+ llamaVramPaddingSize,
116
+ llamaGpu,
117
+ llamaSupportsGpuOffloading,
118
+ defaultContextFlashAttention: flashAttention
119
+ });
120
+ const canUseGpu = llamaSupportsGpuOffloading && llamaGpu !== false;
121
+ const estimatedModelResourceUsage = this._ggufInsights.estimateModelResourceRequirements({
122
+ gpuLayers: resolvedGpuLayers
123
+ });
124
+ const resolvedContextSize = await this.resolveContextContextSize("auto", {
125
+ getVramState: async () => ({
126
+ total: vramState.total,
127
+ free: Math.max(0, vramState.free - estimatedModelResourceUsage.gpuVram)
128
+ }),
129
+ llamaGpu,
130
+ isEmbeddingContext: embeddingContext,
131
+ modelGpuLayers: resolvedGpuLayers,
132
+ modelTrainContextSize: this._ggufInsights.trainContextSize ?? defaultTrainContextSizeForEstimationPurposes,
133
+ flashAttention
134
+ });
135
+ const estimatedContextResourceUsage = this._ggufInsights.estimateContextResourceRequirements({
136
+ contextSize: resolvedContextSize,
137
+ isEmbeddingContext: embeddingContext,
138
+ modelGpuLayers: resolvedGpuLayers,
139
+ flashAttention
140
+ });
141
+ const rankPoints = {
142
+ gpuLayers: 60,
143
+ allLayersAreOffloaded: 10,
144
+ contextSize: 30,
145
+ ramUsageFitsInRam: 10,
146
+ cpuOnlySmallModelSize: 60, // also defined inside `scoreModelSizeForCpuOnlyUsage`
147
+ bonusContextSize: 10
148
+ };
149
+ const gpuLayersPoints = rankPoints.gpuLayers * Math.min(1, resolvedGpuLayers / this._ggufInsights.totalLayers);
150
+ const allLayersAreOffloadedPoints = rankPoints.allLayersAreOffloaded * (resolvedGpuLayers === this._ggufInsights.totalLayers ? 1 : 0);
151
+ const contextSizePoints = rankPoints.contextSize * Math.min(1, resolvedContextSize / contextSize);
152
+ const ramUsageFitsInRamPoints = rankPoints.ramUsageFitsInRam * (estimatedModelResourceUsage.cpuRam <= ramState.free
153
+ ? 1
154
+ : estimatedModelResourceUsage.cpuRam <= ramState.total
155
+ ? 0.5
156
+ : (0.5 - Math.min(0.5, 0.5 * ((estimatedModelResourceUsage.cpuRam - ramState.total) / ramState.total))));
157
+ const bonusContextSizePoints = 10 * Math.min(1, Math.max(0, resolvedContextSize - contextSize) / contextSize);
158
+ const compatibilityScore = canUseGpu
159
+ ? ((gpuLayersPoints + allLayersAreOffloadedPoints + contextSizePoints + ramUsageFitsInRamPoints) /
160
+ (rankPoints.gpuLayers + rankPoints.allLayersAreOffloaded + rankPoints.contextSize + rankPoints.ramUsageFitsInRam))
161
+ : ((contextSizePoints + ramUsageFitsInRamPoints + scoreModelSizeForCpuOnlyUsage(this._ggufInsights.modelSize)) /
162
+ (rankPoints.contextSize + rankPoints.ramUsageFitsInRam + rankPoints.cpuOnlySmallModelSize));
163
+ const bonusScore = bonusContextSizePoints / rankPoints.bonusContextSize;
164
+ return {
165
+ compatibilityScore,
166
+ bonusScore,
167
+ totalScore: compatibilityScore + bonusScore,
168
+ resolvedValues: {
169
+ gpuLayers: resolvedGpuLayers,
170
+ contextSize: resolvedContextSize,
171
+ modelRamUsage: estimatedModelResourceUsage.cpuRam,
172
+ contextRamUsage: estimatedContextResourceUsage.cpuRam,
173
+ totalRamUsage: estimatedModelResourceUsage.cpuRam + estimatedContextResourceUsage.cpuRam,
174
+ modelVramUsage: estimatedModelResourceUsage.gpuVram,
175
+ contextVramUsage: estimatedContextResourceUsage.gpuVram,
176
+ totalVramUsage: estimatedModelResourceUsage.gpuVram + estimatedContextResourceUsage.gpuVram
177
+ }
178
+ };
179
+ }
180
+ async resolveModelGpuLayers(gpuLayers, { ignoreMemorySafetyChecks = false, getVramState = (() => this._ggufInsights._llama._vramOrchestrator.getMemoryState()), llamaVramPaddingSize = this._ggufInsights._llama.vramPaddingSize, llamaGpu = this._ggufInsights._llama.gpu, llamaSupportsGpuOffloading = this._ggufInsights._llama.supportsGpuOffloading, defaultContextFlashAttention = false } = {}) {
181
+ return resolveModelGpuLayersOption(gpuLayers, {
182
+ ggufInsights: this._ggufInsights,
183
+ ignoreMemorySafetyChecks,
184
+ getVramState,
185
+ llamaVramPaddingSize,
186
+ llamaGpu,
187
+ llamaSupportsGpuOffloading,
188
+ defaultContextFlashAttention
189
+ });
190
+ }
191
+ async resolveContextContextSize(contextSize, { modelGpuLayers, batchSize, modelTrainContextSize, flashAttention = false, getVramState = (() => this._ggufInsights._llama._vramOrchestrator.getMemoryState()), llamaGpu = this._ggufInsights._llama.gpu, ignoreMemorySafetyChecks = false, isEmbeddingContext = false, sequences = getDefaultContextSequences() }) {
192
+ return await resolveContextContextSizeOption({
193
+ contextSize,
194
+ batchSize,
195
+ sequences,
196
+ modelFileInsights: this._ggufInsights,
197
+ modelGpuLayers,
198
+ modelTrainContextSize,
199
+ flashAttention,
200
+ getVramState,
201
+ llamaGpu,
202
+ ignoreMemorySafetyChecks,
203
+ isEmbeddingContext
204
+ });
205
+ }
206
+ /** @internal */
207
+ static _create(ggufInsights) {
208
+ return new GgufInsightsConfigurationResolver(ggufInsights);
209
+ }
210
+ }
211
+ function scoreModelSizeForCpuOnlyUsage(modelSize) {
212
+ const s1GB = Math.pow(1024, 3);
213
+ return 60 - scoreLevels(modelSize, [{
214
+ start: s1GB,
215
+ end: s1GB * 2.5,
216
+ points: 40
217
+ }, {
218
+ start: s1GB * 2.5,
219
+ end: s1GB * 4,
220
+ points: 15
221
+ }, {
222
+ start: s1GB * 4,
223
+ points: 5
224
+ }]);
225
+ }
226
+ //# sourceMappingURL=GgufInsightsConfigurationResolver.js.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"GgufInsightsConfigurationResolver.js","sourceRoot":"","sources":["../../../src/gguf/insights/GgufInsightsConfigurationResolver.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,MAAM,IAAI,CAAC;AAIpB,OAAO,EAAC,0BAA0B,EAAC,MAAM,8CAA8C,CAAC;AACxF,OAAO,EAAC,2BAA2B,EAAC,MAAM,wCAAwC,CAAC;AACnF,OAAO,EAAC,+BAA+B,EAAC,MAAM,4CAA4C,CAAC;AAC3F,OAAO,EAAC,WAAW,EAAC,MAAM,wBAAwB,CAAC;AAGnD,MAAM,CAAC,MAAM,4CAA4C,GAAG,IAAI,CAAC;AAGjE,MAAM,OAAO,iCAAiC;IAC1C,gBAAgB,CAAkB,aAAa,CAAe;IAE9D,YAAoB,YAA0B;QAC1C,IAAI,CAAC,aAAa,GAAG,YAAY,CAAC;IACtC,CAAC;IAED,IAAW,YAAY;QACnB,OAAO,IAAI,CAAC,aAAa,CAAC;IAC9B,CAAC;IAED;;;;;;;;;OASG;IACI,KAAK,CAAC,qBAAqB,CAAC,EAC/B,eAAe,EACf,iBAAiB,EACjB,gBAAgB,GAAG,KAAK,EACxB,cAAc,GAAG,KAAK,KAMtB,EAAE,EAAE,EACJ,YAAY,GAAG,CAAC,GAAG,EAAE,CAAC,IAAI,CAAC,aAAa,CAAC,MAAM,CAAC,iBAAiB,CAAC,cAAc,EAAE,CAAC,EACnF,WAAW,GAAG,CAAC,KAAK,IAAI,EAAE,CAAC,CAAC,EAAC,KAAK,EAAE,EAAE,CAAC,QAAQ,EAAE,EAAE,IAAI,EAAE,EAAE,CAAC,OAAO,EAAE,EAAC,CAAC,CAAC,EACxE,oBAAoB,GAAG,IAAI,CAAC,aAAa,CAAC,MAAM,CAAC,eAAe,EAChE,QAAQ,GAAG,IAAI,CAAC,aAAa,CAAC,MAAM,CAAC,GAAG,EACxC,0BAA0B,GAAG,IAAI,CAAC,aAAa,CAAC,MAAM,CAAC,qBAAqB,KAO5E,EAAE;QACF,MAAM,kBAAkB,GAAG,MAAM,IAAI,CAAC,oCAAoC,CAAC;YACvE,cAAc;YACd,WAAW,EAAE,iBAAiB;YAC9B,gBAAgB;SACnB,EAAE;YACC,YAAY;YACZ,WAAW;YACX,oBAAoB;YACpB,QAAQ;YACR,0BAA0B;SAC7B,CAAC,CAAC;QAEH,IAAI,iBAAiB,IAAI,IAAI,IAAI,eAAe,IAAI,IAAI,EAAE,CAAC;YACvD,MAAM,SAAS,GAAG,MAAM,YAAY,EAAE,CAAC;YACvC,MAAM,iBAAiB,GAAG,MAAM,IAAI,CAAC,qBAAqB,CACtD,eAAe,IAAI,IAAI;gBACnB,CAAC,CAAC;oBACE,UAAU,EAAE;wBACR,WAAW,EAAE,iBAAiB;wBAC9B,gBAAgB;qBACnB;iBACJ;gBACD,CAAC,CAAC,eAAe,EACrB;gBACI,YAAY,EAAE,KAAK,IAAI,EAAE,CAAC,SAAS;gBACnC,4BAA4B,EAAE,cAAc;gBAC5C,wBAAwB,EAAE,eAAe,IAAI,IAAI;gBACjD,QAAQ;gBACR,0BAA0B;gBAC1B,oBAAoB;aACvB,CACJ,CAAC;YACF,MAAM,2BAA2B,GAAG,IAAI,CAAC,aAAa,CAAC,iCAAiC,CAAC;gBACrF,SAAS,EAAE,iBAAiB;aAC/B,CAAC,CAAC;YAEH,MAAM,mBAAmB,GAAG,MAAM,IAAI,CAAC,aAAa,CAAC,qBAAqB,CAAC,yBAAyB,CAAC,iBAAiB,IAAI,MAAM,EAAE;gBAC9H,YAAY,EAAE,KAAK,IAAI,EAAE,CAAC,CAAC;oBACvB,KAAK,EAAE,SAAS,CAAC,KAAK;oBACtB,IAAI,EAAE,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,SAAS,CAAC,IAAI,GAAG,2BAA2B,CAAC,OAAO,CAAC;iBAC1E,CAAC;gBACF,kBAAkB,EAAE,gBAAgB;gBACpC,cAAc,EAAE,iBAAiB;gBACjC,qBAAqB,EAAE,IAAI,CAAC,aAAa,CAAC,gBAAgB,IAAI,4CAA4C;gBAC1G,cAAc;gBACd,wBAAwB,EAAE,iBAAiB,IAAI,IAAI;gBACnD,QAAQ;aACX,CAAC,CAAC;YACH,MAAM,6BAA6B,GAAG,IAAI,CAAC,aAAa,CAAC,mCAAmC,CAAC;gBACzF,WAAW,EAAE,mBAAmB;gBAChC,kBAAkB,EAAE,gBAAgB;gBACpC,cAAc,EAAE,iBAAiB;gBACjC,cAAc;aACjB,CAAC,CAAC;YAEH,kBAAkB,CAAC,cAAc,GAAG;gBAChC,SAAS,EAAE,iBAAiB;gBAC5B,WAAW,EAAE,mBAAmB;gBAEhC,aAAa,EAAE,2BAA2B,CAAC,MAAM;gBACjD,eAAe,EAAE,6BAA6B,CAAC,MAAM;gBACrD,aAAa,EAAE,2BAA2B,CAAC,MAAM,GAAG,6BAA6B,CAAC,MAAM;gBAExF,cAAc,EAAE,2BAA2B,CAAC,OAAO;gBACnD,gBAAgB,EAAE,6BAA6B,CAAC,OAAO;gBACvD,cAAc,EAAE,2BAA2B,CAAC,OAAO,GAAG,6BAA6B,CAAC,OAAO;aAC9F,CAAC;YAEF,IAAI,kBAAkB,CAAC,cAAc,CAAC,cAAc,GAAG,SAAS,CAAC,KAAK,EAAE,CAAC;gBACrE,kBAAkB,CAAC,kBAAkB,GAAG,CAAC,CAAC;gBAC1C,kBAAkB,CAAC,UAAU,GAAG,CAAC,CAAC;gBAClC,kBAAkB,CAAC,UAAU,GAAG,CAAC,CAAC;YACtC,CAAC;QACL,CAAC;QAED,OAAO,kBAAkB,CAAC;IAC9B,CAAC;IAED;;;;;;;;;;;;OAYG;IACI,KAAK,CAAC,oCAAoC,CAAC,EAC9C,WAAW,GAAG,IAAI,CAAC,GAAG,CAAC,IAAI,EAAE,IAAI,CAAC,aAAa,CAAC,gBAAgB,IAAI,IAAI,CAAC,EACzE,gBAAgB,GAAG,KAAK,EACxB,cAAc,GAAG,KAAK,KAKtB,EAAE,EAAE,EACJ,YAAY,GAAG,CAAC,GAAG,EAAE,CAAC,IAAI,CAAC,aAAa,CAAC,MAAM,CAAC,iBAAiB,CAAC,cAAc,EAAE,CAAC,EACnF,WAAW,GAAG,CAAC,KAAK,IAAI,EAAE,CAAC,CAAC,EAAC,KAAK,EAAE,EAAE,CAAC,QAAQ,EAAE,EAAE,IAAI,EAAE,EAAE,CAAC,OAAO,EAAE,EAAC,CAAC,CAAC,EACxE,oBAAoB,GAAG,IAAI,CAAC,aAAa,CAAC,MAAM,CAAC,eAAe,EAChE,QAAQ,GAAG,IAAI,CAAC,aAAa,CAAC,MAAM,CAAC,GAAG,EACxC,0BAA0B,GAAG,IAAI,CAAC,aAAa,CAAC,MAAM,CAAC,qBAAqB,KAO5E,EAAE;QAiCF,MAAM,CACF,SAAS,EACT,QAAQ,CACX,GAAG,MAAM,OAAO,CAAC,GAAG,CAAC;YAClB,YAAY,EAAE;YACd,WAAW,EAAE;SAChB,CAAC,CAAC;QACH,MAAM,iBAAiB,GAAG,MAAM,IAAI,CAAC,qBAAqB,CACtD,gBAAgB;YACZ,CAAC,CAAC,EAAC,UAAU,EAAE,EAAC,gBAAgB,EAAE,IAAI,EAAC,EAAC;YACxC,CAAC,CAAC,MAAM,EACZ;YACI,YAAY,EAAE,KAAK,IAAI,EAAE,CAAC,SAAS;YACnC,oBAAoB;YACpB,QAAQ;YACR,0BAA0B;YAC1B,4BAA4B,EAAE,cAAc;SAC/C,CACJ,CAAC;QACF,MAAM,SAAS,GAAG,0BAA0B,IAAI,QAAQ,KAAK,KAAK,CAAC;QACnE,MAAM,2BAA2B,GAAG,IAAI,CAAC,aAAa,CAAC,iCAAiC,CAAC;YACrF,SAAS,EAAE,iBAAiB;SAC/B,CAAC,CAAC;QAEH,MAAM,mBAAmB,GAAG,MAAM,IAAI,CAAC,yBAAyB,CAAC,MAAM,EAAE;YACrE,YAAY,EAAE,KAAK,IAAI,EAAE,CAAC,CAAC;gBACvB,KAAK,EAAE,SAAS,CAAC,KAAK;gBACtB,IAAI,EAAE,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,SAAS,CAAC,IAAI,GAAG,2BAA2B,CAAC,OAAO,CAAC;aAC1E,CAAC;YACF,QAAQ;YACR,kBAAkB,EAAE,gBAAgB;YACpC,cAAc,EAAE,iBAAiB;YACjC,qBAAqB,EAAE,IAAI,CAAC,aAAa,CAAC,gBAAgB,IAAI,4CAA4C;YAC1G,cAAc;SACjB,CAAC,CAAC;QACH,MAAM,6BAA6B,GAAG,IAAI,CAAC,aAAa,CAAC,mCAAmC,CAAC;YACzF,WAAW,EAAE,mBAAmB;YAChC,kBAAkB,EAAE,gBAAgB;YACpC,cAAc,EAAE,iBAAiB;YACjC,cAAc;SACjB,CAAC,CAAC;QAEH,MAAM,UAAU,GAAG;YACf,SAAS,EAAE,EAAE;YACb,qBAAqB,EAAE,EAAE;YACzB,WAAW,EAAE,EAAE;YACf,iBAAiB,EAAE,EAAE;YACrB,qBAAqB,EAAE,EAAE,EAAE,sDAAsD;YACjF,gBAAgB,EAAE,EAAE;SACd,CAAC;QAEX,MAAM,eAAe,GAAG,UAAU,CAAC,SAAS,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,iBAAiB,GAAG,IAAI,CAAC,aAAa,CAAC,WAAW,CAAC,CAAC;QAC/G,MAAM,2BAA2B,GAAG,UAAU,CAAC,qBAAqB,GAAG,CACnE,iBAAiB,KAAK,IAAI,CAAC,aAAa,CAAC,WAAW,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAC/D,CAAC;QACF,MAAM,iBAAiB,GAAG,UAAU,CAAC,WAAW,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,mBAAmB,GAAG,WAAW,CAAC,CAAC;QAClG,MAAM,uBAAuB,GAAG,UAAU,CAAC,iBAAiB,GAAG,CAC3D,2BAA2B,CAAC,MAAM,IAAI,QAAQ,CAAC,IAAI;YAC/C,CAAC,CAAC,CAAC;YACH,CAAC,CAAC,2BAA2B,CAAC,MAAM,IAAI,QAAQ,CAAC,KAAK;gBAClD,CAAC,CAAC,GAAG;gBACL,CAAC,CAAC,CACE,GAAG,GAAG,IAAI,CAAC,GAAG,CACV,GAAG,EACH,GAAG,GAAG,CACF,CAAC,2BAA2B,CAAC,MAAM,GAAG,QAAQ,CAAC,KAAK,CAAC,GAAG,QAAQ,CAAC,KAAK,CACzE,CACJ,CACJ,CACZ,CAAC;QACF,MAAM,sBAAsB,GAAG,EAAE,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,mBAAmB,GAAG,WAAW,CAAC,GAAG,WAAW,CAAC,CAAC;QAE9G,MAAM,kBAAkB,GAAG,SAAS;YAChC,CAAC,CAAC,CACE,CAAC,eAAe,GAAG,2BAA2B,GAAG,iBAAiB,GAAG,uBAAuB,CAAC;gBAC7F,CAAC,UAAU,CAAC,SAAS,GAAG,UAAU,CAAC,qBAAqB,GAAG,UAAU,CAAC,WAAW,GAAG,UAAU,CAAC,iBAAiB,CAAC,CACpH;YACD,CAAC,CAAC,CACE,CAAC,iBAAiB,GAAG,uBAAuB,GAAG,6BAA6B,CAAC,IAAI,CAAC,aAAa,CAAC,SAAS,CAAC,CAAC;gBAC3G,CAAC,UAAU,CAAC,WAAW,GAAG,UAAU,CAAC,iBAAiB,GAAG,UAAU,CAAC,qBAAqB,CAAC,CAAC,CAAC;QACpG,MAAM,UAAU,GAAG,sBAAsB,GAAG,UAAU,CAAC,gBAAgB,CAAC;QAExE,OAAO;YACH,kBAAkB;YAClB,UAAU;YACV,UAAU,EAAE,kBAAkB,GAAG,UAAU;YAE3C,cAAc,EAAE;gBACZ,SAAS,EAAE,iBAAiB;gBAC5B,WAAW,EAAE,mBAAmB;gBAEhC,aAAa,EAAE,2BAA2B,CAAC,MAAM;gBACjD,eAAe,EAAE,6BAA6B,CAAC,MAAM;gBACrD,aAAa,EAAE,2BAA2B,CAAC,MAAM,GAAG,6BAA6B,CAAC,MAAM;gBAExF,cAAc,EAAE,2BAA2B,CAAC,OAAO;gBACnD,gBAAgB,EAAE,6BAA6B,CAAC,OAAO;gBACvD,cAAc,EAAE,2BAA2B,CAAC,OAAO,GAAG,6BAA6B,CAAC,OAAO;aAC9F;SACJ,CAAC;IACN,CAAC;IAEM,KAAK,CAAC,qBAAqB,CAAC,SAA0C,EAAE,EAC3E,wBAAwB,GAAG,KAAK,EAChC,YAAY,GAAG,CAAC,GAAG,EAAE,CAAC,IAAI,CAAC,aAAa,CAAC,MAAM,CAAC,iBAAiB,CAAC,cAAc,EAAE,CAAC,EACnF,oBAAoB,GAAG,IAAI,CAAC,aAAa,CAAC,MAAM,CAAC,eAAe,EAAE,QAAQ,GAAG,IAAI,CAAC,aAAa,CAAC,MAAM,CAAC,GAAG,EAC1G,0BAA0B,GAAG,IAAI,CAAC,aAAa,CAAC,MAAM,CAAC,qBAAqB,EAC5E,4BAA4B,GAAG,KAAK,KAIpC,EAAE;QACF,OAAO,2BAA2B,CAAC,SAAS,EAAE;YAC1C,YAAY,EAAE,IAAI,CAAC,aAAa;YAChC,wBAAwB;YACxB,YAAY;YACZ,oBAAoB;YACpB,QAAQ;YACR,0BAA0B;YAC1B,4BAA4B;SAC/B,CAAC,CAAC;IACP,CAAC;IAEM,KAAK,CAAC,yBAAyB,CAAC,WAA+C,EAAE,EACpF,cAAc,EACd,SAAS,EACT,qBAAqB,EACrB,cAAc,GAAG,KAAK,EACtB,YAAY,GAAG,CAAC,GAAG,EAAE,CAAC,IAAI,CAAC,aAAa,CAAC,MAAM,CAAC,iBAAiB,CAAC,cAAc,EAAE,CAAC,EACnF,QAAQ,GAAG,IAAI,CAAC,aAAa,CAAC,MAAM,CAAC,GAAG,EACxC,wBAAwB,GAAG,KAAK,EAChC,kBAAkB,GAAG,KAAK,EAC1B,SAAS,GAAG,0BAA0B,EAAE,EAW3C;QACG,OAAO,MAAM,+BAA+B,CAAC;YACzC,WAAW;YACX,SAAS;YACT,SAAS;YACT,iBAAiB,EAAE,IAAI,CAAC,aAAa;YACrC,cAAc;YACd,qBAAqB;YACrB,cAAc;YACd,YAAY;YACZ,QAAQ;YACR,wBAAwB;YACxB,kBAAkB;SACrB,CAAC,CAAC;IACP,CAAC;IAED,gBAAgB;IACT,MAAM,CAAC,OAAO,CAAC,YAA0B;QAC5C,OAAO,IAAI,iCAAiC,CAAC,YAAY,CAAC,CAAC;IAC/D,CAAC;CACJ;AAED,SAAS,6BAA6B,CAAC,SAAiB;IACpD,MAAM,IAAI,GAAG,IAAI,CAAC,GAAG,CAAC,IAAI,EAAE,CAAC,CAAC,CAAC;IAC/B,OAAO,EAAE,GAAG,WAAW,CAAC,SAAS,EAAE,CAAC;YAChC,KAAK,EAAE,IAAI;YACX,GAAG,EAAE,IAAI,GAAG,GAAG;YACf,MAAM,EAAE,EAAE;SACb,EAAE;YACC,KAAK,EAAE,IAAI,GAAG,GAAG;YACjB,GAAG,EAAE,IAAI,GAAG,CAAC;YACb,MAAM,EAAE,EAAE;SACb,EAAE;YACC,KAAK,EAAE,IAAI,GAAG,CAAC;YACf,MAAM,EAAE,CAAC;SACZ,CAAC,CAAC,CAAC;AACR,CAAC"}
@@ -0,0 +1,19 @@
1
+ import { LlamaContextOptions } from "../../../evaluator/LlamaContext/types.js";
2
+ import { GgufInsights } from "../GgufInsights.js";
3
+ import { BuildGpu } from "../../../bindings/types.js";
4
+ export declare function resolveContextContextSizeOption({ contextSize, batchSize, sequences, modelFileInsights, modelGpuLayers, modelTrainContextSize, flashAttention, getVramState, llamaGpu, ignoreMemorySafetyChecks, isEmbeddingContext }: {
5
+ contextSize?: LlamaContextOptions["contextSize"];
6
+ batchSize?: LlamaContextOptions["batchSize"];
7
+ sequences: number;
8
+ modelFileInsights: GgufInsights;
9
+ modelGpuLayers: number;
10
+ modelTrainContextSize: number;
11
+ flashAttention: boolean;
12
+ getVramState(): Promise<{
13
+ total: number;
14
+ free: number;
15
+ }>;
16
+ llamaGpu: BuildGpu;
17
+ ignoreMemorySafetyChecks?: boolean;
18
+ isEmbeddingContext?: boolean;
19
+ }): Promise<number>;
@@ -0,0 +1,78 @@
1
+ import { minAllowedContextSizeInCalculations } from "../../../config.js";
2
+ import { getDefaultContextBatchSize, getDefaultModelContextSize } from "../../../evaluator/LlamaContext/LlamaContext.js";
3
+ export async function resolveContextContextSizeOption({ contextSize, batchSize, sequences, modelFileInsights, modelGpuLayers, modelTrainContextSize, flashAttention, getVramState, llamaGpu, ignoreMemorySafetyChecks = false, isEmbeddingContext = false }) {
4
+ if (contextSize == null)
5
+ contextSize = "auto";
6
+ if (typeof contextSize === "number") {
7
+ const resolvedContextSize = Math.max(1, Math.floor(contextSize));
8
+ if (ignoreMemorySafetyChecks)
9
+ return resolvedContextSize;
10
+ const vramState = await getVramState();
11
+ const contextVram = modelFileInsights.estimateContextResourceRequirements({
12
+ contextSize: resolvedContextSize,
13
+ batchSize: batchSize ?? getDefaultContextBatchSize({ contextSize: resolvedContextSize, sequences }),
14
+ modelGpuLayers: modelGpuLayers,
15
+ sequences,
16
+ flashAttention,
17
+ isEmbeddingContext
18
+ }).gpuVram;
19
+ if (contextVram > vramState.free)
20
+ throw new Error(`The context size of ${resolvedContextSize}${sequences > 1 ? ` with ${sequences} sequences` : ""} is too large for the available VRAM`);
21
+ return resolvedContextSize;
22
+ }
23
+ else if (contextSize === "auto" || typeof contextSize === "object") {
24
+ if (llamaGpu === false)
25
+ return modelTrainContextSize;
26
+ const vramState = await getVramState();
27
+ if (vramState.total === 0)
28
+ return modelTrainContextSize;
29
+ const freeVram = vramState.free;
30
+ const maxContextSize = contextSize === "auto"
31
+ ? getDefaultModelContextSize({ trainContextSize: modelTrainContextSize })
32
+ : Math.min(contextSize.max ?? getDefaultModelContextSize({ trainContextSize: modelTrainContextSize }), getDefaultModelContextSize({ trainContextSize: modelTrainContextSize }));
33
+ const minContextSize = contextSize === "auto"
34
+ ? minAllowedContextSizeInCalculations
35
+ : Math.max(contextSize.min ?? minAllowedContextSizeInCalculations, minAllowedContextSizeInCalculations);
36
+ let highestCompatibleContextSize = null;
37
+ let step = -Math.max(1, Math.floor((maxContextSize - minContextSize) / 4));
38
+ for (let testContextSize = maxContextSize; testContextSize >= minContextSize && testContextSize <= maxContextSize;) {
39
+ const contextVram = modelFileInsights.estimateContextResourceRequirements({
40
+ contextSize: testContextSize,
41
+ batchSize: batchSize ?? getDefaultContextBatchSize({ contextSize: testContextSize, sequences }),
42
+ modelGpuLayers: modelGpuLayers,
43
+ sequences,
44
+ flashAttention,
45
+ isEmbeddingContext
46
+ }).gpuVram;
47
+ if (contextVram <= freeVram) {
48
+ if (highestCompatibleContextSize == null || testContextSize > highestCompatibleContextSize) {
49
+ highestCompatibleContextSize = testContextSize;
50
+ if (step === -1)
51
+ break;
52
+ else if (step < 0)
53
+ step = Math.max(1, Math.floor(-step / 2));
54
+ }
55
+ }
56
+ else if (step > 0)
57
+ step = -Math.max(1, Math.floor(step / 2));
58
+ if (testContextSize == minContextSize && step === -1)
59
+ break;
60
+ testContextSize += step;
61
+ if (testContextSize < minContextSize) {
62
+ testContextSize = minContextSize;
63
+ step = Math.max(1, Math.floor(Math.abs(step) / 2));
64
+ }
65
+ else if (testContextSize > maxContextSize) {
66
+ testContextSize = maxContextSize;
67
+ step = -Math.max(1, Math.floor(Math.abs(step) / 2));
68
+ }
69
+ }
70
+ if (highestCompatibleContextSize != null)
71
+ return highestCompatibleContextSize;
72
+ if (ignoreMemorySafetyChecks)
73
+ return minContextSize;
74
+ throw new Error(`The available VRAM is too small to fit the context size of ${maxContextSize}${sequences > 1 ? ` with ${sequences} sequences` : ""}`);
75
+ }
76
+ throw new Error(`Invalid context size: "${contextSize}"`);
77
+ }
78
+ //# sourceMappingURL=resolveContextContextSizeOption.js.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"resolveContextContextSizeOption.js","sourceRoot":"","sources":["../../../../src/gguf/insights/utils/resolveContextContextSizeOption.ts"],"names":[],"mappings":"AAGA,OAAO,EAAC,mCAAmC,EAAC,MAAM,oBAAoB,CAAC;AACvE,OAAO,EAAC,0BAA0B,EAAE,0BAA0B,EAAC,MAAM,iDAAiD,CAAC;AAEvH,MAAM,CAAC,KAAK,UAAU,+BAA+B,CAAC,EAClD,WAAW,EAAE,SAAS,EAAE,SAAS,EAAE,iBAAiB,EAAE,cAAc,EAAE,qBAAqB,EAAE,cAAc,EAAE,YAAY,EAAE,QAAQ,EACnI,wBAAwB,GAAG,KAAK,EAAE,kBAAkB,GAAG,KAAK,EAa/D;IACG,IAAI,WAAW,IAAI,IAAI;QACnB,WAAW,GAAG,MAAM,CAAC;IAEzB,IAAI,OAAO,WAAW,KAAK,QAAQ,EAAE,CAAC;QAClC,MAAM,mBAAmB,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,IAAI,CAAC,KAAK,CAAC,WAAW,CAAC,CAAC,CAAC;QAEjE,IAAI,wBAAwB;YACxB,OAAO,mBAAmB,CAAC;QAE/B,MAAM,SAAS,GAAG,MAAM,YAAY,EAAE,CAAC;QACvC,MAAM,WAAW,GAAG,iBAAiB,CAAC,mCAAmC,CAAC;YACtE,WAAW,EAAE,mBAAmB;YAChC,SAAS,EAAE,SAAS,IAAI,0BAA0B,CAAC,EAAC,WAAW,EAAE,mBAAmB,EAAE,SAAS,EAAC,CAAC;YACjG,cAAc,EAAE,cAAc;YAC9B,SAAS;YACT,cAAc;YACd,kBAAkB;SACrB,CAAC,CAAC,OAAO,CAAC;QAEX,IAAI,WAAW,GAAG,SAAS,CAAC,IAAI;YAC5B,MAAM,IAAI,KAAK,CAAC,uBAAuB,mBAAmB,GAAG,SAAS,GAAG,CAAC,CAAC,CAAC,CAAC,SAAS,SAAS,YAAY,CAAC,CAAC,CAAC,EAAE,sCAAsC,CAAC,CAAC;QAE5J,OAAO,mBAAmB,CAAC;IAC/B,CAAC;SAAM,IAAI,WAAW,KAAK,MAAM,IAAI,OAAO,WAAW,KAAK,QAAQ,EAAE,CAAC;QACnE,IAAI,QAAQ,KAAK,KAAK;YAClB,OAAO,qBAAqB,CAAC;QAEjC,MAAM,SAAS,GAAG,MAAM,YAAY,EAAE,CAAC;QAEvC,IAAI,SAAS,CAAC,KAAK,KAAK,CAAC;YACrB,OAAO,qBAAqB,CAAC;QAEjC,MAAM,QAAQ,GAAG,SAAS,CAAC,IAAI,CAAC;QAEhC,MAAM,cAAc,GAAG,WAAW,KAAK,MAAM;YACzC,CAAC,CAAC,0BAA0B,CAAC,EAAC,gBAAgB,EAAE,qBAAqB,EAAC,CAAC;YACvE,CAAC,CAAC,IAAI,CAAC,GAAG,CACN,WAAW,CAAC,GAAG,IAAI,0BAA0B,CAAC,EAAC,gBAAgB,EAAE,qBAAqB,EAAC,CAAC,EACxF,0BAA0B,CAAC,EAAC,gBAAgB,EAAE,qBAAqB,EAAC,CAAC,CACxE,CAAC;QAEN,MAAM,cAAc,GAAG,WAAW,KAAK,MAAM;YACzC,CAAC,CAAC,mCAAmC;YACrC,CAAC,CAAC,IAAI,CAAC,GAAG,CACN,WAAW,CAAC,GAAG,IAAI,mCAAmC,EACtD,mCAAmC,CACtC,CAAC;QAEN,IAAI,4BAA4B,GAAkB,IAAI,CAAC;QACvD,IAAI,IAAI,GAAG,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,IAAI,CAAC,KAAK,CAAC,CAAC,cAAc,GAAG,cAAc,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC;QAC3E,KAAK,IAAI,eAAe,GAAG,cAAc,EAAE,eAAe,IAAI,cAAc,IAAI,eAAe,IAAI,cAAc,GAAG,CAAC;YACjH,MAAM,WAAW,GAAG,iBAAiB,CAAC,mCAAmC,CAAC;gBACtE,WAAW,EAAE,eAAe;gBAC5B,SAAS,EAAE,SAAS,IAAI,0BAA0B,CAAC,EAAC,WAAW,EAAE,eAAe,EAAE,SAAS,EAAC,CAAC;gBAC7F,cAAc,EAAE,cAAc;gBAC9B,SAAS;gBACT,cAAc;gBACd,kBAAkB;aACrB,CAAC,CAAC,OAAO,CAAC;YAEX,IAAI,WAAW,IAAI,QAAQ,EAAE,CAAC;gBAC1B,IAAI,4BAA4B,IAAI,IAAI,IAAI,eAAe,GAAG,4BAA4B,EAAE,CAAC;oBACzF,4BAA4B,GAAG,eAAe,CAAC;oBAE/C,IAAI,IAAI,KAAK,CAAC,CAAC;wBACX,MAAM;yBACL,IAAI,IAAI,GAAG,CAAC;wBACb,IAAI,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,IAAI,CAAC,KAAK,CAAC,CAAC,IAAI,GAAG,CAAC,CAAC,CAAC,CAAC;gBAClD,CAAC;YACL,CAAC;iBAAM,IAAI,IAAI,GAAG,CAAC;gBACf,IAAI,GAAG,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,IAAI,CAAC,KAAK,CAAC,IAAI,GAAG,CAAC,CAAC,CAAC,CAAC;YAE9C,IAAI,eAAe,IAAI,cAAc,IAAI,IAAI,KAAK,CAAC,CAAC;gBAChD,MAAM;YAEV,eAAe,IAAI,IAAI,CAAC;YACxB,IAAI,eAAe,GAAG,cAAc,EAAE,CAAC;gBACnC,eAAe,GAAG,cAAc,CAAC;gBACjC,IAAI,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,GAAG,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC;YACvD,CAAC;iBAAM,IAAI,eAAe,GAAG,cAAc,EAAE,CAAC;gBAC1C,eAAe,GAAG,cAAc,CAAC;gBACjC,IAAI,GAAG,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,GAAG,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC;YACxD,CAAC;QACL,CAAC;QAED,IAAI,4BAA4B,IAAI,IAAI;YACpC,OAAO,4BAA4B,CAAC;QAExC,IAAI,wBAAwB;YACxB,OAAO,cAAc,CAAC;QAE1B,MAAM,IAAI,KAAK,CAAC,8DAA8D,cAAc,GAAG,SAAS,GAAG,CAAC,CAAC,CAAC,CAAC,SAAS,SAAS,YAAY,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC;IAC1J,CAAC;IAED,MAAM,IAAI,KAAK,CAAC,0BAA0B,WAAW,GAAG,CAAC,CAAC;AAC9D,CAAC"}
@@ -0,0 +1,15 @@
1
+ import { LlamaModelOptions } from "../../../evaluator/LlamaModel/LlamaModel.js";
2
+ import { BuildGpu } from "../../../bindings/types.js";
3
+ import type { GgufInsights } from "../GgufInsights.js";
4
+ export declare function resolveModelGpuLayersOption(gpuLayers: LlamaModelOptions["gpuLayers"], { ggufInsights, ignoreMemorySafetyChecks, getVramState, llamaVramPaddingSize, llamaGpu, llamaSupportsGpuOffloading, defaultContextFlashAttention }: {
5
+ ggufInsights: GgufInsights;
6
+ ignoreMemorySafetyChecks?: boolean;
7
+ getVramState(): Promise<{
8
+ total: number;
9
+ free: number;
10
+ }>;
11
+ llamaVramPaddingSize: number;
12
+ llamaGpu: BuildGpu;
13
+ llamaSupportsGpuOffloading: boolean;
14
+ defaultContextFlashAttention: boolean;
15
+ }): Promise<number>;