node-llama-cpp 3.0.0-beta.3 → 3.0.0-beta.30

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 (690) hide show
  1. package/README.md +14 -11
  2. package/bins/linux-arm64/_nlcBuildMetadata.json +1 -0
  3. package/bins/linux-arm64/llama-addon.node +0 -0
  4. package/bins/linux-armv7l/_nlcBuildMetadata.json +1 -0
  5. package/bins/linux-armv7l/llama-addon.node +0 -0
  6. package/bins/linux-x64/_nlcBuildMetadata.json +1 -0
  7. package/bins/linux-x64/llama-addon.node +0 -0
  8. package/bins/linux-x64-cuda/_nlcBuildMetadata.json +1 -0
  9. package/bins/linux-x64-cuda/llama-addon.node +0 -0
  10. package/bins/linux-x64-vulkan/_nlcBuildMetadata.json +1 -0
  11. package/bins/linux-x64-vulkan/llama-addon.node +0 -0
  12. package/bins/mac-arm64-metal/_nlcBuildMetadata.json +1 -0
  13. package/bins/mac-arm64-metal/default.metallib +0 -0
  14. package/bins/mac-arm64-metal/llama-addon.node +0 -0
  15. package/bins/mac-x64/_nlcBuildMetadata.json +1 -0
  16. package/bins/mac-x64/llama-addon.node +0 -0
  17. package/bins/win-arm64/_nlcBuildMetadata.json +1 -0
  18. package/bins/win-arm64/llama-addon.exp +0 -0
  19. package/bins/win-arm64/llama-addon.lib +0 -0
  20. package/bins/win-arm64/llama-addon.node +0 -0
  21. package/bins/win-x64/_nlcBuildMetadata.json +1 -0
  22. package/bins/win-x64/llama-addon.exp +0 -0
  23. package/bins/win-x64/llama-addon.lib +0 -0
  24. package/bins/win-x64/llama-addon.node +0 -0
  25. package/bins/win-x64-cuda/_nlcBuildMetadata.json +1 -0
  26. package/bins/win-x64-cuda/llama-addon.exp +0 -0
  27. package/bins/win-x64-cuda/llama-addon.lib +0 -0
  28. package/bins/win-x64-cuda/llama-addon.node +0 -0
  29. package/bins/win-x64-vulkan/_nlcBuildMetadata.json +1 -0
  30. package/bins/win-x64-vulkan/llama-addon.exp +0 -0
  31. package/bins/win-x64-vulkan/llama-addon.lib +0 -0
  32. package/bins/win-x64-vulkan/llama-addon.node +0 -0
  33. package/dist/ChatWrapper.d.ts +8 -39
  34. package/dist/ChatWrapper.js +115 -72
  35. package/dist/ChatWrapper.js.map +1 -1
  36. package/dist/apiDocsIndex.d.ts +1 -0
  37. package/dist/apiDocsIndex.js +7 -0
  38. package/dist/apiDocsIndex.js.map +1 -0
  39. package/dist/{utils/getBin.d.ts → bindings/AddonTypes.d.ts} +54 -8
  40. package/dist/bindings/AddonTypes.js +2 -0
  41. package/dist/bindings/AddonTypes.js.map +1 -0
  42. package/dist/bindings/Llama.d.ts +47 -0
  43. package/dist/bindings/Llama.js +353 -0
  44. package/dist/bindings/Llama.js.map +1 -0
  45. package/dist/bindings/consts.d.ts +2 -0
  46. package/dist/bindings/consts.js +11 -0
  47. package/dist/bindings/consts.js.map +1 -0
  48. package/dist/bindings/getLlama.d.ts +148 -0
  49. package/dist/bindings/getLlama.js +401 -0
  50. package/dist/bindings/getLlama.js.map +1 -0
  51. package/dist/bindings/types.d.ts +56 -0
  52. package/dist/bindings/types.js +77 -0
  53. package/dist/bindings/types.js.map +1 -0
  54. package/dist/bindings/utils/MemoryOrchestrator.d.ts +21 -0
  55. package/dist/bindings/utils/MemoryOrchestrator.js +49 -0
  56. package/dist/bindings/utils/MemoryOrchestrator.js.map +1 -0
  57. package/dist/bindings/utils/NoBinaryFoundError.d.ts +2 -0
  58. package/dist/bindings/utils/NoBinaryFoundError.js +7 -0
  59. package/dist/bindings/utils/NoBinaryFoundError.js.map +1 -0
  60. package/dist/bindings/utils/asyncEvery.d.ts +5 -0
  61. package/dist/bindings/utils/asyncEvery.js +15 -0
  62. package/dist/bindings/utils/asyncEvery.js.map +1 -0
  63. package/dist/bindings/utils/asyncSome.d.ts +5 -0
  64. package/dist/bindings/utils/asyncSome.js +27 -0
  65. package/dist/bindings/utils/asyncSome.js.map +1 -0
  66. package/dist/{utils → bindings/utils}/binariesGithubRelease.js +1 -1
  67. package/dist/bindings/utils/binariesGithubRelease.js.map +1 -0
  68. package/dist/bindings/utils/clearAllLocalBuilds.d.ts +1 -0
  69. package/dist/bindings/utils/clearAllLocalBuilds.js +47 -0
  70. package/dist/bindings/utils/clearAllLocalBuilds.js.map +1 -0
  71. package/dist/bindings/utils/cloneLlamaCppRepo.d.ts +11 -0
  72. package/dist/bindings/utils/cloneLlamaCppRepo.js +166 -0
  73. package/dist/bindings/utils/cloneLlamaCppRepo.js.map +1 -0
  74. package/dist/bindings/utils/compileLLamaCpp.d.ts +17 -0
  75. package/dist/bindings/utils/compileLLamaCpp.js +226 -0
  76. package/dist/bindings/utils/compileLLamaCpp.js.map +1 -0
  77. package/dist/bindings/utils/detectAvailableComputeLayers.d.ts +14 -0
  78. package/dist/bindings/utils/detectAvailableComputeLayers.js +305 -0
  79. package/dist/bindings/utils/detectAvailableComputeLayers.js.map +1 -0
  80. package/dist/bindings/utils/detectGlibc.d.ts +4 -0
  81. package/dist/bindings/utils/detectGlibc.js +46 -0
  82. package/dist/bindings/utils/detectGlibc.js.map +1 -0
  83. package/dist/bindings/utils/getBestComputeLayersAvailable.d.ts +10 -0
  84. package/dist/bindings/utils/getBestComputeLayersAvailable.js +29 -0
  85. package/dist/bindings/utils/getBestComputeLayersAvailable.js.map +1 -0
  86. package/dist/bindings/utils/getBuildFolderNameForBuildOptions.d.ts +5 -0
  87. package/dist/bindings/utils/getBuildFolderNameForBuildOptions.js +93 -0
  88. package/dist/bindings/utils/getBuildFolderNameForBuildOptions.js.map +1 -0
  89. package/dist/bindings/utils/getCanUsePrebuiltBinaries.d.ts +1 -0
  90. package/dist/bindings/utils/getCanUsePrebuiltBinaries.js +8 -0
  91. package/dist/bindings/utils/getCanUsePrebuiltBinaries.js.map +1 -0
  92. package/dist/bindings/utils/getExampleUsageCodeOfGetLlama.d.ts +2 -0
  93. package/dist/bindings/utils/getExampleUsageCodeOfGetLlama.js +21 -0
  94. package/dist/bindings/utils/getExampleUsageCodeOfGetLlama.js.map +1 -0
  95. package/dist/bindings/utils/getGpuTypesToUseForOption.d.ts +12 -0
  96. package/dist/bindings/utils/getGpuTypesToUseForOption.js +30 -0
  97. package/dist/bindings/utils/getGpuTypesToUseForOption.js.map +1 -0
  98. package/dist/bindings/utils/getLinuxDistroInfo.d.ts +9 -0
  99. package/dist/bindings/utils/getLinuxDistroInfo.js +46 -0
  100. package/dist/bindings/utils/getLinuxDistroInfo.js.map +1 -0
  101. package/dist/bindings/utils/getLlamaWithoutBackend.d.ts +5 -0
  102. package/dist/bindings/utils/getLlamaWithoutBackend.js +27 -0
  103. package/dist/bindings/utils/getLlamaWithoutBackend.js.map +1 -0
  104. package/dist/bindings/utils/getPlatform.d.ts +2 -0
  105. package/dist/bindings/utils/getPlatform.js +15 -0
  106. package/dist/bindings/utils/getPlatform.js.map +1 -0
  107. package/dist/bindings/utils/getPlatformInfo.d.ts +5 -0
  108. package/dist/bindings/utils/getPlatformInfo.js +28 -0
  109. package/dist/bindings/utils/getPlatformInfo.js.map +1 -0
  110. package/dist/bindings/utils/hasBuildingFromSourceDependenciesInstalled.d.ts +3 -0
  111. package/dist/bindings/utils/hasBuildingFromSourceDependenciesInstalled.js +27 -0
  112. package/dist/bindings/utils/hasBuildingFromSourceDependenciesInstalled.js.map +1 -0
  113. package/dist/bindings/utils/hasFileInPath.d.ts +2 -0
  114. package/dist/bindings/utils/hasFileInPath.js +34 -0
  115. package/dist/bindings/utils/hasFileInPath.js.map +1 -0
  116. package/dist/bindings/utils/lastBuildInfo.d.ts +6 -0
  117. package/dist/bindings/utils/lastBuildInfo.js +17 -0
  118. package/dist/bindings/utils/lastBuildInfo.js.map +1 -0
  119. package/dist/bindings/utils/logBinaryUsageExampleToConsole.d.ts +2 -0
  120. package/dist/bindings/utils/logBinaryUsageExampleToConsole.js +22 -0
  121. package/dist/bindings/utils/logBinaryUsageExampleToConsole.js.map +1 -0
  122. package/dist/bindings/utils/logDistroInstallInstruction.d.ts +14 -0
  123. package/dist/bindings/utils/logDistroInstallInstruction.js +48 -0
  124. package/dist/bindings/utils/logDistroInstallInstruction.js.map +1 -0
  125. package/dist/bindings/utils/resolveCustomCmakeOptions.d.ts +1 -0
  126. package/dist/bindings/utils/resolveCustomCmakeOptions.js +47 -0
  127. package/dist/bindings/utils/resolveCustomCmakeOptions.js.map +1 -0
  128. package/dist/bindings/utils/testBindingBinary.d.ts +1 -0
  129. package/dist/bindings/utils/testBindingBinary.js +100 -0
  130. package/dist/bindings/utils/testBindingBinary.js.map +1 -0
  131. package/dist/bindings/utils/testCmakeBinary.d.ts +6 -0
  132. package/dist/bindings/utils/testCmakeBinary.js +32 -0
  133. package/dist/bindings/utils/testCmakeBinary.js.map +1 -0
  134. package/dist/chatWrappers/AlpacaChatWrapper.d.ts +2 -1
  135. package/dist/chatWrappers/AlpacaChatWrapper.js +9 -2
  136. package/dist/chatWrappers/AlpacaChatWrapper.js.map +1 -1
  137. package/dist/chatWrappers/ChatMLChatWrapper.d.ts +2 -9
  138. package/dist/chatWrappers/ChatMLChatWrapper.js +23 -21
  139. package/dist/chatWrappers/ChatMLChatWrapper.js.map +1 -1
  140. package/dist/chatWrappers/FalconChatWrapper.d.ts +4 -10
  141. package/dist/chatWrappers/FalconChatWrapper.js +38 -21
  142. package/dist/chatWrappers/FalconChatWrapper.js.map +1 -1
  143. package/dist/chatWrappers/FunctionaryChatWrapper.d.ts +8 -32
  144. package/dist/chatWrappers/FunctionaryChatWrapper.js +323 -118
  145. package/dist/chatWrappers/FunctionaryChatWrapper.js.map +1 -1
  146. package/dist/chatWrappers/GemmaChatWrapper.d.ts +7 -0
  147. package/dist/chatWrappers/GemmaChatWrapper.js +96 -0
  148. package/dist/chatWrappers/GemmaChatWrapper.js.map +1 -0
  149. package/dist/chatWrappers/GeneralChatWrapper.d.ts +4 -10
  150. package/dist/chatWrappers/GeneralChatWrapper.js +45 -22
  151. package/dist/chatWrappers/GeneralChatWrapper.js.map +1 -1
  152. package/dist/chatWrappers/Llama2ChatWrapper.d.ts +12 -0
  153. package/dist/chatWrappers/{LlamaChatWrapper.js → Llama2ChatWrapper.js} +38 -20
  154. package/dist/chatWrappers/Llama2ChatWrapper.js.map +1 -0
  155. package/dist/chatWrappers/Llama3ChatWrapper.d.ts +16 -0
  156. package/dist/chatWrappers/Llama3ChatWrapper.js +174 -0
  157. package/dist/chatWrappers/Llama3ChatWrapper.js.map +1 -0
  158. package/dist/chatWrappers/generic/JinjaTemplateChatWrapper.d.ts +67 -0
  159. package/dist/chatWrappers/generic/JinjaTemplateChatWrapper.js +371 -0
  160. package/dist/chatWrappers/generic/JinjaTemplateChatWrapper.js.map +1 -0
  161. package/dist/chatWrappers/generic/TemplateChatWrapper.d.ts +54 -0
  162. package/dist/chatWrappers/generic/TemplateChatWrapper.js +200 -0
  163. package/dist/chatWrappers/generic/TemplateChatWrapper.js.map +1 -0
  164. package/dist/chatWrappers/generic/utils/chatHistoryFunctionCallMessageTemplate.d.ts +23 -0
  165. package/dist/chatWrappers/generic/utils/chatHistoryFunctionCallMessageTemplate.js +45 -0
  166. package/dist/chatWrappers/generic/utils/chatHistoryFunctionCallMessageTemplate.js.map +1 -0
  167. package/dist/chatWrappers/utils/ChatModelFunctionsDocumentationGenerator.d.ts +42 -0
  168. package/dist/chatWrappers/utils/ChatModelFunctionsDocumentationGenerator.js +82 -0
  169. package/dist/chatWrappers/utils/ChatModelFunctionsDocumentationGenerator.js.map +1 -0
  170. package/dist/chatWrappers/utils/isJinjaTemplateEquivalentToSpecializedChatWrapper.d.ts +4 -0
  171. package/dist/chatWrappers/utils/isJinjaTemplateEquivalentToSpecializedChatWrapper.js +210 -0
  172. package/dist/chatWrappers/utils/isJinjaTemplateEquivalentToSpecializedChatWrapper.js.map +1 -0
  173. package/dist/chatWrappers/utils/resolveChatWrapper.d.ts +69 -0
  174. package/dist/chatWrappers/utils/resolveChatWrapper.js +243 -0
  175. package/dist/chatWrappers/utils/resolveChatWrapper.js.map +1 -0
  176. package/dist/cli/cli.js +21 -7
  177. package/dist/cli/cli.js.map +1 -1
  178. package/dist/cli/commands/BuildCommand.d.ts +11 -4
  179. package/dist/cli/commands/BuildCommand.js +114 -41
  180. package/dist/cli/commands/BuildCommand.js.map +1 -1
  181. package/dist/cli/commands/ChatCommand.d.ts +18 -6
  182. package/dist/cli/commands/ChatCommand.js +299 -143
  183. package/dist/cli/commands/ChatCommand.js.map +1 -1
  184. package/dist/cli/commands/ClearCommand.d.ts +1 -1
  185. package/dist/cli/commands/ClearCommand.js +11 -12
  186. package/dist/cli/commands/ClearCommand.js.map +1 -1
  187. package/dist/cli/commands/CompleteCommand.d.ts +29 -0
  188. package/dist/cli/commands/CompleteCommand.js +365 -0
  189. package/dist/cli/commands/CompleteCommand.js.map +1 -0
  190. package/dist/cli/commands/DebugCommand.d.ts +7 -0
  191. package/dist/cli/commands/DebugCommand.js +54 -0
  192. package/dist/cli/commands/DebugCommand.js.map +1 -0
  193. package/dist/cli/commands/DownloadCommand.d.ts +7 -4
  194. package/dist/cli/commands/DownloadCommand.js +121 -70
  195. package/dist/cli/commands/DownloadCommand.js.map +1 -1
  196. package/dist/cli/commands/InfillCommand.d.ts +31 -0
  197. package/dist/cli/commands/InfillCommand.js +401 -0
  198. package/dist/cli/commands/InfillCommand.js.map +1 -0
  199. package/dist/cli/commands/InitCommand.d.ts +11 -0
  200. package/dist/cli/commands/InitCommand.js +195 -0
  201. package/dist/cli/commands/InitCommand.js.map +1 -0
  202. package/dist/cli/commands/OnPostInstallCommand.js +9 -10
  203. package/dist/cli/commands/OnPostInstallCommand.js.map +1 -1
  204. package/dist/cli/commands/PullCommand.d.ts +12 -0
  205. package/dist/cli/commands/PullCommand.js +117 -0
  206. package/dist/cli/commands/PullCommand.js.map +1 -0
  207. package/dist/cli/commands/inspect/InspectCommand.d.ts +4 -0
  208. package/dist/cli/commands/inspect/InspectCommand.js +19 -0
  209. package/dist/cli/commands/inspect/InspectCommand.js.map +1 -0
  210. package/dist/cli/commands/inspect/commands/InspectGgufCommand.d.ts +12 -0
  211. package/dist/cli/commands/inspect/commands/InspectGgufCommand.js +136 -0
  212. package/dist/cli/commands/inspect/commands/InspectGgufCommand.js.map +1 -0
  213. package/dist/cli/commands/inspect/commands/InspectGpuCommand.d.ts +4 -0
  214. package/dist/cli/commands/inspect/commands/InspectGpuCommand.js +164 -0
  215. package/dist/cli/commands/inspect/commands/InspectGpuCommand.js.map +1 -0
  216. package/dist/cli/commands/inspect/commands/InspectMeasureCommand.d.ts +17 -0
  217. package/dist/cli/commands/inspect/commands/InspectMeasureCommand.js +613 -0
  218. package/dist/cli/commands/inspect/commands/InspectMeasureCommand.js.map +1 -0
  219. package/dist/cli/projectTemplates.d.ts +7 -0
  220. package/dist/cli/projectTemplates.js +10 -0
  221. package/dist/cli/projectTemplates.js.map +1 -0
  222. package/dist/cli/recommendedModels.d.ts +2 -0
  223. package/dist/cli/recommendedModels.js +363 -0
  224. package/dist/cli/recommendedModels.js.map +1 -0
  225. package/dist/cli/startCreateCli.d.ts +2 -0
  226. package/dist/cli/startCreateCli.js +26 -0
  227. package/dist/cli/startCreateCli.js.map +1 -0
  228. package/dist/cli/utils/ConsoleInteraction.d.ts +23 -0
  229. package/dist/cli/utils/ConsoleInteraction.js +122 -0
  230. package/dist/cli/utils/ConsoleInteraction.js.map +1 -0
  231. package/dist/cli/utils/ConsoleTable.d.ts +23 -0
  232. package/dist/cli/utils/ConsoleTable.js +86 -0
  233. package/dist/cli/utils/ConsoleTable.js.map +1 -0
  234. package/dist/cli/utils/basicChooseFromListConsoleInteraction.d.ts +13 -0
  235. package/dist/cli/utils/basicChooseFromListConsoleInteraction.js +111 -0
  236. package/dist/cli/utils/basicChooseFromListConsoleInteraction.js.map +1 -0
  237. package/dist/cli/utils/consolePromptQuestion.d.ts +6 -0
  238. package/dist/cli/utils/consolePromptQuestion.js +82 -0
  239. package/dist/cli/utils/consolePromptQuestion.js.map +1 -0
  240. package/dist/cli/utils/getReadablePath.d.ts +1 -0
  241. package/dist/cli/utils/getReadablePath.js +14 -0
  242. package/dist/cli/utils/getReadablePath.js.map +1 -0
  243. package/dist/cli/utils/interactivelyAskForModel.d.ts +7 -0
  244. package/dist/cli/utils/interactivelyAskForModel.js +451 -0
  245. package/dist/cli/utils/interactivelyAskForModel.js.map +1 -0
  246. package/dist/cli/utils/logUsedGpuTypeOption.d.ts +2 -0
  247. package/dist/cli/utils/logUsedGpuTypeOption.js +9 -0
  248. package/dist/cli/utils/logUsedGpuTypeOption.js.map +1 -0
  249. package/dist/cli/utils/printCommonInfoLines.d.ts +9 -0
  250. package/dist/cli/utils/printCommonInfoLines.js +75 -0
  251. package/dist/cli/utils/printCommonInfoLines.js.map +1 -0
  252. package/dist/cli/utils/printInfoLine.d.ts +12 -0
  253. package/dist/cli/utils/printInfoLine.js +54 -0
  254. package/dist/cli/utils/printInfoLine.js.map +1 -0
  255. package/dist/cli/utils/projectTemplates.d.ts +19 -0
  256. package/dist/cli/utils/projectTemplates.js +47 -0
  257. package/dist/cli/utils/projectTemplates.js.map +1 -0
  258. package/dist/cli/utils/resolveCommandGgufPath.d.ts +4 -0
  259. package/dist/cli/utils/resolveCommandGgufPath.js +71 -0
  260. package/dist/cli/utils/resolveCommandGgufPath.js.map +1 -0
  261. package/dist/cli/utils/resolveHeaderFlag.d.ts +1 -0
  262. package/dist/cli/utils/resolveHeaderFlag.js +21 -0
  263. package/dist/cli/utils/resolveHeaderFlag.js.map +1 -0
  264. package/dist/cli/utils/resolveModelRecommendationFileOptions.d.ts +19 -0
  265. package/dist/cli/utils/resolveModelRecommendationFileOptions.js +7 -0
  266. package/dist/cli/utils/resolveModelRecommendationFileOptions.js.map +1 -0
  267. package/dist/cli/utils/splitAnsiToLines.d.ts +1 -0
  268. package/dist/cli/utils/splitAnsiToLines.js +32 -0
  269. package/dist/cli/utils/splitAnsiToLines.js.map +1 -0
  270. package/dist/cli/utils/withCliCommandDescriptionDocsUrl.d.ts +2 -0
  271. package/dist/cli/utils/withCliCommandDescriptionDocsUrl.js +23 -0
  272. package/dist/cli/utils/withCliCommandDescriptionDocsUrl.js.map +1 -0
  273. package/dist/commands.d.ts +1 -0
  274. package/dist/commands.js +3 -0
  275. package/dist/commands.js.map +1 -1
  276. package/dist/config.d.ts +38 -5
  277. package/dist/config.js +61 -16
  278. package/dist/config.js.map +1 -1
  279. package/dist/consts.d.ts +3 -0
  280. package/dist/consts.js +10 -0
  281. package/dist/consts.js.map +1 -0
  282. package/dist/{llamaEvaluator → evaluator}/LlamaChat/LlamaChat.d.ts +112 -39
  283. package/dist/evaluator/LlamaChat/LlamaChat.js +1512 -0
  284. package/dist/evaluator/LlamaChat/LlamaChat.js.map +1 -0
  285. package/dist/evaluator/LlamaChat/utils/FunctionCallNameGrammar.d.ts +11 -0
  286. package/dist/evaluator/LlamaChat/utils/FunctionCallNameGrammar.js +55 -0
  287. package/dist/evaluator/LlamaChat/utils/FunctionCallNameGrammar.js.map +1 -0
  288. package/dist/evaluator/LlamaChat/utils/FunctionCallParamsGrammar.d.ts +16 -0
  289. package/dist/evaluator/LlamaChat/utils/FunctionCallParamsGrammar.js +45 -0
  290. package/dist/evaluator/LlamaChat/utils/FunctionCallParamsGrammar.js.map +1 -0
  291. package/dist/evaluator/LlamaChat/utils/LlamaFunctionCallValidationError.d.ts +8 -0
  292. package/dist/evaluator/LlamaChat/utils/LlamaFunctionCallValidationError.js +12 -0
  293. package/dist/evaluator/LlamaChat/utils/LlamaFunctionCallValidationError.js.map +1 -0
  294. package/dist/{llamaEvaluator → evaluator}/LlamaChat/utils/contextShiftStrategies/eraseFirstResponseAndKeepFirstSystemChatContextShiftStrategy.js +42 -16
  295. package/dist/evaluator/LlamaChat/utils/contextShiftStrategies/eraseFirstResponseAndKeepFirstSystemChatContextShiftStrategy.js.map +1 -0
  296. package/dist/evaluator/LlamaChatSession/LlamaChatSession.d.ts +288 -0
  297. package/dist/evaluator/LlamaChatSession/LlamaChatSession.js +419 -0
  298. package/dist/evaluator/LlamaChatSession/LlamaChatSession.js.map +1 -0
  299. package/dist/evaluator/LlamaChatSession/utils/LlamaChatSessionPromptCompletionEngine.d.ts +39 -0
  300. package/dist/evaluator/LlamaChatSession/utils/LlamaChatSessionPromptCompletionEngine.js +186 -0
  301. package/dist/evaluator/LlamaChatSession/utils/LlamaChatSessionPromptCompletionEngine.js.map +1 -0
  302. package/dist/{llamaEvaluator → evaluator}/LlamaChatSession/utils/defineChatSessionFunction.d.ts +3 -0
  303. package/dist/{llamaEvaluator → evaluator}/LlamaChatSession/utils/defineChatSessionFunction.js +3 -0
  304. package/dist/evaluator/LlamaChatSession/utils/defineChatSessionFunction.js.map +1 -0
  305. package/dist/evaluator/LlamaCompletion.d.ts +143 -0
  306. package/dist/evaluator/LlamaCompletion.js +418 -0
  307. package/dist/evaluator/LlamaCompletion.js.map +1 -0
  308. package/dist/{llamaEvaluator → evaluator}/LlamaContext/LlamaContext.d.ts +41 -21
  309. package/dist/{llamaEvaluator → evaluator}/LlamaContext/LlamaContext.js +270 -81
  310. package/dist/evaluator/LlamaContext/LlamaContext.js.map +1 -0
  311. package/dist/evaluator/LlamaContext/types.d.ts +140 -0
  312. package/dist/evaluator/LlamaContext/types.js.map +1 -0
  313. package/dist/evaluator/LlamaContext/utils/batchItemsPrioritizationStrategies/firstInFirstOutStrategy.js.map +1 -0
  314. package/dist/evaluator/LlamaContext/utils/batchItemsPrioritizationStrategies/maximumParallelismStrategy.js.map +1 -0
  315. package/dist/evaluator/LlamaContext/utils/resolveBatchItemsPrioritizationStrategy.d.ts +2 -0
  316. package/dist/{llamaEvaluator/LlamaContext/utils/resolveBatchItemsPrioritizingStrategy.js → evaluator/LlamaContext/utils/resolveBatchItemsPrioritizationStrategy.js} +4 -4
  317. package/dist/evaluator/LlamaContext/utils/resolveBatchItemsPrioritizationStrategy.js.map +1 -0
  318. package/dist/evaluator/LlamaEmbeddingContext.d.ts +51 -0
  319. package/dist/evaluator/LlamaEmbeddingContext.js +73 -0
  320. package/dist/evaluator/LlamaEmbeddingContext.js.map +1 -0
  321. package/dist/{llamaEvaluator → evaluator}/LlamaGrammar.d.ts +10 -7
  322. package/dist/{llamaEvaluator → evaluator}/LlamaGrammar.js +14 -11
  323. package/dist/evaluator/LlamaGrammar.js.map +1 -0
  324. package/dist/{llamaEvaluator → evaluator}/LlamaGrammarEvaluationState.js +4 -4
  325. package/dist/evaluator/LlamaGrammarEvaluationState.js.map +1 -0
  326. package/dist/{llamaEvaluator → evaluator}/LlamaJsonSchemaGrammar.d.ts +2 -1
  327. package/dist/{llamaEvaluator → evaluator}/LlamaJsonSchemaGrammar.js +3 -3
  328. package/dist/evaluator/LlamaJsonSchemaGrammar.js.map +1 -0
  329. package/dist/evaluator/LlamaModel/LlamaModel.d.ts +236 -0
  330. package/dist/evaluator/LlamaModel/LlamaModel.js +679 -0
  331. package/dist/evaluator/LlamaModel/LlamaModel.js.map +1 -0
  332. package/dist/evaluator/LlamaModel/utils/TokenAttributes.d.ts +29 -0
  333. package/dist/evaluator/LlamaModel/utils/TokenAttributes.js +65 -0
  334. package/dist/evaluator/LlamaModel/utils/TokenAttributes.js.map +1 -0
  335. package/dist/evaluator/TokenBias.d.ts +22 -0
  336. package/dist/evaluator/TokenBias.js +33 -0
  337. package/dist/evaluator/TokenBias.js.map +1 -0
  338. package/dist/evaluator/TokenMeter.d.ts +54 -0
  339. package/dist/evaluator/TokenMeter.js +86 -0
  340. package/dist/evaluator/TokenMeter.js.map +1 -0
  341. package/dist/gguf/consts.d.ts +3 -0
  342. package/dist/gguf/consts.js +8 -0
  343. package/dist/gguf/consts.js.map +1 -0
  344. package/dist/gguf/errors/InvalidGgufMagicError.d.ts +3 -0
  345. package/dist/gguf/errors/InvalidGgufMagicError.js +6 -0
  346. package/dist/gguf/errors/InvalidGgufMagicError.js.map +1 -0
  347. package/dist/gguf/errors/UnsupportedGgufValueTypeError.d.ts +4 -0
  348. package/dist/gguf/errors/UnsupportedGgufValueTypeError.js +9 -0
  349. package/dist/gguf/errors/UnsupportedGgufValueTypeError.js.map +1 -0
  350. package/dist/gguf/fileReaders/GgufFileReader.d.ts +33 -0
  351. package/dist/gguf/fileReaders/GgufFileReader.js +76 -0
  352. package/dist/gguf/fileReaders/GgufFileReader.js.map +1 -0
  353. package/dist/gguf/fileReaders/GgufFsFileReader.d.ts +17 -0
  354. package/dist/gguf/fileReaders/GgufFsFileReader.js +46 -0
  355. package/dist/gguf/fileReaders/GgufFsFileReader.js.map +1 -0
  356. package/dist/gguf/fileReaders/GgufNetworkFetchFileReader.d.ts +22 -0
  357. package/dist/gguf/fileReaders/GgufNetworkFetchFileReader.js +63 -0
  358. package/dist/gguf/fileReaders/GgufNetworkFetchFileReader.js.map +1 -0
  359. package/dist/gguf/insights/GgufInsights.d.ts +48 -0
  360. package/dist/gguf/insights/GgufInsights.js +381 -0
  361. package/dist/gguf/insights/GgufInsights.js.map +1 -0
  362. package/dist/gguf/insights/GgufInsightsConfigurationResolver.d.ts +87 -0
  363. package/dist/gguf/insights/GgufInsightsConfigurationResolver.js +141 -0
  364. package/dist/gguf/insights/GgufInsightsConfigurationResolver.js.map +1 -0
  365. package/dist/gguf/insights/utils/resolveContextContextSizeOption.d.ts +18 -0
  366. package/dist/gguf/insights/utils/resolveContextContextSizeOption.js +76 -0
  367. package/dist/gguf/insights/utils/resolveContextContextSizeOption.js.map +1 -0
  368. package/dist/gguf/insights/utils/resolveModelGpuLayersOption.d.ts +14 -0
  369. package/dist/gguf/insights/utils/resolveModelGpuLayersOption.js +177 -0
  370. package/dist/gguf/insights/utils/resolveModelGpuLayersOption.js.map +1 -0
  371. package/dist/gguf/insights/utils/scoreLevels.d.ts +5 -0
  372. package/dist/gguf/insights/utils/scoreLevels.js +16 -0
  373. package/dist/gguf/insights/utils/scoreLevels.js.map +1 -0
  374. package/dist/gguf/parser/GgufV2Parser.d.ts +19 -0
  375. package/dist/gguf/parser/GgufV2Parser.js +115 -0
  376. package/dist/gguf/parser/GgufV2Parser.js.map +1 -0
  377. package/dist/gguf/parser/GgufV3Parser.d.ts +3 -0
  378. package/dist/gguf/parser/GgufV3Parser.js +4 -0
  379. package/dist/gguf/parser/GgufV3Parser.js.map +1 -0
  380. package/dist/gguf/parser/parseGguf.d.ts +8 -0
  381. package/dist/gguf/parser/parseGguf.js +63 -0
  382. package/dist/gguf/parser/parseGguf.js.map +1 -0
  383. package/dist/gguf/readGgufFileInfo.d.ts +33 -0
  384. package/dist/gguf/readGgufFileInfo.js +66 -0
  385. package/dist/gguf/readGgufFileInfo.js.map +1 -0
  386. package/dist/gguf/types/GgufFileInfoTypes.d.ts +84 -0
  387. package/dist/gguf/types/GgufFileInfoTypes.js +18 -0
  388. package/dist/gguf/types/GgufFileInfoTypes.js.map +1 -0
  389. package/dist/gguf/types/GgufMetadataTypes.d.ts +335 -0
  390. package/dist/gguf/types/GgufMetadataTypes.js +86 -0
  391. package/dist/gguf/types/GgufMetadataTypes.js.map +1 -0
  392. package/dist/gguf/types/GgufTensorInfoTypes.d.ts +37 -0
  393. package/dist/gguf/types/GgufTensorInfoTypes.js +33 -0
  394. package/dist/gguf/types/GgufTensorInfoTypes.js.map +1 -0
  395. package/dist/gguf/utils/GgufReadOffset.d.ts +6 -0
  396. package/dist/gguf/utils/GgufReadOffset.js +18 -0
  397. package/dist/gguf/utils/GgufReadOffset.js.map +1 -0
  398. package/dist/gguf/utils/convertMetadataKeyValueRecordToNestedObject.d.ts +5 -0
  399. package/dist/gguf/utils/convertMetadataKeyValueRecordToNestedObject.js +38 -0
  400. package/dist/gguf/utils/convertMetadataKeyValueRecordToNestedObject.js.map +1 -0
  401. package/dist/gguf/utils/getGgufFileTypeName.d.ts +4 -0
  402. package/dist/gguf/utils/getGgufFileTypeName.js +13 -0
  403. package/dist/gguf/utils/getGgufFileTypeName.js.map +1 -0
  404. package/dist/gguf/utils/getGgufMetadataArchitectureData.d.ts +3 -0
  405. package/dist/gguf/utils/getGgufMetadataArchitectureData.js +4 -0
  406. package/dist/gguf/utils/getGgufMetadataArchitectureData.js.map +1 -0
  407. package/dist/gguf/utils/normalizeGgufDownloadUrl.d.ts +1 -0
  408. package/dist/gguf/utils/normalizeGgufDownloadUrl.js +16 -0
  409. package/dist/gguf/utils/normalizeGgufDownloadUrl.js.map +1 -0
  410. package/dist/gguf/utils/resolveBinarySplitGgufPartUrls.d.ts +2 -0
  411. package/dist/gguf/utils/resolveBinarySplitGgufPartUrls.js +39 -0
  412. package/dist/gguf/utils/resolveBinarySplitGgufPartUrls.js.map +1 -0
  413. package/dist/gguf/utils/resolveSplitGgufParts.d.ts +7 -0
  414. package/dist/gguf/utils/resolveSplitGgufParts.js +55 -0
  415. package/dist/gguf/utils/resolveSplitGgufParts.js.map +1 -0
  416. package/dist/index.d.ts +41 -18
  417. package/dist/index.js +36 -15
  418. package/dist/index.js.map +1 -1
  419. package/dist/state.d.ts +4 -0
  420. package/dist/state.js +14 -0
  421. package/dist/state.js.map +1 -1
  422. package/dist/types.d.ts +116 -5
  423. package/dist/types.js.map +1 -1
  424. package/dist/utils/DisposeGuard.d.ts +13 -0
  425. package/dist/utils/DisposeGuard.js +120 -0
  426. package/dist/utils/DisposeGuard.js.map +1 -0
  427. package/dist/utils/InsufficientMemoryError.d.ts +3 -0
  428. package/dist/utils/InsufficientMemoryError.js +6 -0
  429. package/dist/utils/InsufficientMemoryError.js.map +1 -0
  430. package/dist/utils/LlamaText.d.ts +70 -26
  431. package/dist/utils/LlamaText.js +469 -157
  432. package/dist/utils/LlamaText.js.map +1 -1
  433. package/dist/utils/LruCache.d.ts +12 -0
  434. package/dist/utils/LruCache.js +44 -0
  435. package/dist/utils/LruCache.js.map +1 -0
  436. package/dist/utils/ReplHistory.js.map +1 -1
  437. package/dist/utils/StopGenerationDetector.d.ts +25 -9
  438. package/dist/utils/StopGenerationDetector.js +93 -22
  439. package/dist/utils/StopGenerationDetector.js.map +1 -1
  440. package/dist/utils/TokenStreamRegulator.d.ts +9 -4
  441. package/dist/utils/TokenStreamRegulator.js +81 -8
  442. package/dist/utils/TokenStreamRegulator.js.map +1 -1
  443. package/dist/utils/UnsupportedError.d.ts +2 -0
  444. package/dist/utils/UnsupportedError.js +7 -0
  445. package/dist/utils/UnsupportedError.js.map +1 -0
  446. package/dist/utils/appendUserMessageToChatHistory.js.map +1 -1
  447. package/dist/utils/clearTempFolder.js.map +1 -1
  448. package/dist/utils/cmake.js +38 -20
  449. package/dist/utils/cmake.js.map +1 -1
  450. package/dist/utils/createModelDownloader.d.ts +108 -0
  451. package/dist/utils/createModelDownloader.js +231 -0
  452. package/dist/utils/createModelDownloader.js.map +1 -0
  453. package/dist/utils/findBestOption.d.ts +4 -0
  454. package/dist/utils/findBestOption.js +15 -0
  455. package/dist/utils/findBestOption.js.map +1 -0
  456. package/dist/utils/findCharacterRemovalCountToFitChatHistoryInContext.d.ts +1 -0
  457. package/dist/utils/findCharacterRemovalCountToFitChatHistoryInContext.js +23 -12
  458. package/dist/utils/findCharacterRemovalCountToFitChatHistoryInContext.js.map +1 -1
  459. package/dist/utils/gbnfJson/GbnfGrammarGenerator.js.map +1 -1
  460. package/dist/utils/gbnfJson/getGbnfGrammarForGbnfJsonSchema.d.ts +5 -0
  461. package/dist/utils/gbnfJson/getGbnfGrammarForGbnfJsonSchema.js +11 -0
  462. package/dist/utils/gbnfJson/getGbnfGrammarForGbnfJsonSchema.js.map +1 -0
  463. package/dist/utils/gbnfJson/terminals/GbnfArray.d.ts +3 -1
  464. package/dist/utils/gbnfJson/terminals/GbnfArray.js +10 -5
  465. package/dist/utils/gbnfJson/terminals/GbnfArray.js.map +1 -1
  466. package/dist/utils/gbnfJson/terminals/GbnfBoolean.d.ts +1 -1
  467. package/dist/utils/gbnfJson/terminals/GbnfBoolean.js.map +1 -1
  468. package/dist/utils/gbnfJson/terminals/GbnfBooleanValue.js.map +1 -1
  469. package/dist/utils/gbnfJson/terminals/GbnfGrammar.js.map +1 -1
  470. package/dist/utils/gbnfJson/terminals/GbnfNull.d.ts +1 -1
  471. package/dist/utils/gbnfJson/terminals/GbnfNull.js.map +1 -1
  472. package/dist/utils/gbnfJson/terminals/GbnfNumber.d.ts +1 -1
  473. package/dist/utils/gbnfJson/terminals/GbnfNumber.js.map +1 -1
  474. package/dist/utils/gbnfJson/terminals/GbnfNumberValue.js.map +1 -1
  475. package/dist/utils/gbnfJson/terminals/GbnfObjectMap.d.ts +3 -1
  476. package/dist/utils/gbnfJson/terminals/GbnfObjectMap.js +9 -4
  477. package/dist/utils/gbnfJson/terminals/GbnfObjectMap.js.map +1 -1
  478. package/dist/utils/gbnfJson/terminals/GbnfOr.js.map +1 -1
  479. package/dist/utils/gbnfJson/terminals/GbnfRepetition.d.ts +9 -0
  480. package/dist/utils/gbnfJson/terminals/GbnfRepetition.js +37 -0
  481. package/dist/utils/gbnfJson/terminals/GbnfRepetition.js.map +1 -0
  482. package/dist/utils/gbnfJson/terminals/GbnfString.d.ts +1 -1
  483. package/dist/utils/gbnfJson/terminals/GbnfString.js +23 -5
  484. package/dist/utils/gbnfJson/terminals/GbnfString.js.map +1 -1
  485. package/dist/utils/gbnfJson/terminals/GbnfStringValue.js.map +1 -1
  486. package/dist/utils/gbnfJson/terminals/GbnfVerbatimText.js.map +1 -1
  487. package/dist/utils/gbnfJson/terminals/GbnfWhitespace.d.ts +7 -4
  488. package/dist/utils/gbnfJson/terminals/GbnfWhitespace.js +37 -9
  489. package/dist/utils/gbnfJson/terminals/GbnfWhitespace.js.map +1 -1
  490. package/dist/utils/gbnfJson/terminals/gbnfConsts.d.ts +5 -4
  491. package/dist/utils/gbnfJson/terminals/gbnfConsts.js +14 -3
  492. package/dist/utils/gbnfJson/terminals/gbnfConsts.js.map +1 -1
  493. package/dist/utils/gbnfJson/types.d.ts +3 -0
  494. package/dist/utils/gbnfJson/types.js.map +1 -1
  495. package/dist/utils/gbnfJson/utils/GbnfJsonScopeState.d.ts +10 -0
  496. package/dist/utils/gbnfJson/utils/GbnfJsonScopeState.js +15 -0
  497. package/dist/utils/gbnfJson/utils/GbnfJsonScopeState.js.map +1 -0
  498. package/dist/utils/gbnfJson/utils/getGbnfJsonTerminalForGbnfJsonSchema.d.ts +2 -1
  499. package/dist/utils/gbnfJson/utils/getGbnfJsonTerminalForGbnfJsonSchema.js +6 -5
  500. package/dist/utils/gbnfJson/utils/getGbnfJsonTerminalForGbnfJsonSchema.js.map +1 -1
  501. package/dist/utils/gbnfJson/utils/validateObjectAgainstGbnfSchema.js +2 -2
  502. package/dist/utils/gbnfJson/utils/validateObjectAgainstGbnfSchema.js.map +1 -1
  503. package/dist/utils/getBuildDefaults.d.ts +1 -2
  504. package/dist/utils/getBuildDefaults.js +2 -3
  505. package/dist/utils/getBuildDefaults.js.map +1 -1
  506. package/dist/utils/getConsoleLogPrefix.d.ts +1 -0
  507. package/dist/utils/getConsoleLogPrefix.js +10 -0
  508. package/dist/utils/getConsoleLogPrefix.js.map +1 -0
  509. package/dist/utils/getGrammarsFolder.d.ts +2 -1
  510. package/dist/utils/getGrammarsFolder.js +8 -7
  511. package/dist/utils/getGrammarsFolder.js.map +1 -1
  512. package/dist/utils/getModuleVersion.d.ts +1 -0
  513. package/dist/utils/getModuleVersion.js +13 -0
  514. package/dist/utils/getModuleVersion.js.map +1 -0
  515. package/dist/utils/getQueuedTokensBeforeStopTrigger.d.ts +6 -0
  516. package/dist/utils/getQueuedTokensBeforeStopTrigger.js +22 -0
  517. package/dist/utils/getQueuedTokensBeforeStopTrigger.js.map +1 -0
  518. package/dist/utils/getReadableContextSize.d.ts +1 -0
  519. package/dist/utils/getReadableContextSize.js +7 -0
  520. package/dist/utils/getReadableContextSize.js.map +1 -0
  521. package/dist/utils/getTypeScriptTypeStringForGbnfJsonSchema.js +15 -11
  522. package/dist/utils/getTypeScriptTypeStringForGbnfJsonSchema.js.map +1 -1
  523. package/dist/utils/gitReleaseBundles.js +73 -5
  524. package/dist/utils/gitReleaseBundles.js.map +1 -1
  525. package/dist/utils/hashString.d.ts +1 -0
  526. package/dist/utils/hashString.js +8 -0
  527. package/dist/utils/hashString.js.map +1 -0
  528. package/dist/utils/isLockfileActive.d.ts +4 -0
  529. package/dist/utils/isLockfileActive.js +12 -0
  530. package/dist/utils/isLockfileActive.js.map +1 -0
  531. package/dist/utils/isToken.d.ts +2 -0
  532. package/dist/utils/isToken.js +4 -0
  533. package/dist/utils/isToken.js.map +1 -0
  534. package/dist/utils/isUrl.d.ts +1 -0
  535. package/dist/utils/isUrl.js +15 -0
  536. package/dist/utils/isUrl.js.map +1 -0
  537. package/dist/utils/mergeUnionTypes.d.ts +10 -0
  538. package/dist/utils/mergeUnionTypes.js +2 -0
  539. package/dist/utils/mergeUnionTypes.js.map +1 -0
  540. package/dist/utils/parseModelFileName.d.ts +1 -0
  541. package/dist/utils/parseModelFileName.js +6 -1
  542. package/dist/utils/parseModelFileName.js.map +1 -1
  543. package/dist/utils/parseTextTemplate.d.ts +66 -0
  544. package/dist/utils/parseTextTemplate.js +116 -0
  545. package/dist/utils/parseTextTemplate.js.map +1 -0
  546. package/dist/utils/prettyPrintObject.d.ts +10 -0
  547. package/dist/utils/prettyPrintObject.js +84 -0
  548. package/dist/utils/prettyPrintObject.js.map +1 -0
  549. package/dist/utils/removeNullFields.d.ts +2 -1
  550. package/dist/utils/removeNullFields.js +8 -0
  551. package/dist/utils/removeNullFields.js.map +1 -1
  552. package/dist/utils/resolveGithubRelease.d.ts +2 -0
  553. package/dist/utils/resolveGithubRelease.js +36 -0
  554. package/dist/utils/resolveGithubRelease.js.map +1 -0
  555. package/dist/utils/runtime.d.ts +4 -0
  556. package/dist/utils/runtime.js +8 -0
  557. package/dist/utils/runtime.js.map +1 -0
  558. package/dist/utils/safeEventCallback.d.ts +6 -0
  559. package/dist/utils/safeEventCallback.js +29 -0
  560. package/dist/utils/safeEventCallback.js.map +1 -0
  561. package/dist/utils/spawnCommand.d.ts +11 -1
  562. package/dist/utils/spawnCommand.js +56 -6
  563. package/dist/utils/spawnCommand.js.map +1 -1
  564. package/dist/utils/tokenizeInput.d.ts +3 -0
  565. package/dist/utils/tokenizeInput.js +12 -0
  566. package/dist/utils/tokenizeInput.js.map +1 -0
  567. package/dist/utils/truncateTextAndRoundToWords.d.ts +2 -0
  568. package/dist/utils/truncateTextAndRoundToWords.js +30 -0
  569. package/dist/utils/truncateTextAndRoundToWords.js.map +1 -1
  570. package/dist/utils/utilTypes.d.ts +3 -0
  571. package/dist/utils/utilTypes.js +2 -0
  572. package/dist/utils/utilTypes.js.map +1 -0
  573. package/dist/utils/waitForLockfileRelease.d.ts +5 -0
  574. package/dist/utils/waitForLockfileRelease.js +20 -0
  575. package/dist/utils/waitForLockfileRelease.js.map +1 -0
  576. package/dist/utils/withLockfile.d.ts +7 -0
  577. package/dist/utils/withLockfile.js +44 -0
  578. package/dist/utils/withLockfile.js.map +1 -0
  579. package/dist/utils/withOra.d.ts +2 -0
  580. package/dist/utils/withOra.js +22 -6
  581. package/dist/utils/withOra.js.map +1 -1
  582. package/dist/utils/withProgressLog.d.ts +23 -0
  583. package/dist/utils/withProgressLog.js +211 -0
  584. package/dist/utils/withProgressLog.js.map +1 -0
  585. package/dist/utils/withStatusLogs.d.ts +2 -1
  586. package/dist/utils/withStatusLogs.js +12 -9
  587. package/dist/utils/withStatusLogs.js.map +1 -1
  588. package/dist/utils/wrapAbortSignal.d.ts +2 -0
  589. package/dist/utils/wrapAbortSignal.js +9 -0
  590. package/dist/utils/wrapAbortSignal.js.map +1 -0
  591. package/llama/.clang-format +1 -2
  592. package/llama/CMakeLists.txt +115 -4
  593. package/llama/addon.cpp +1300 -97
  594. package/llama/binariesGithubRelease.json +1 -1
  595. package/llama/gitRelease.bundle +0 -0
  596. package/llama/gpuInfo/cuda-gpu-info.cu +120 -0
  597. package/llama/gpuInfo/cuda-gpu-info.h +10 -0
  598. package/llama/gpuInfo/metal-gpu-info.h +8 -0
  599. package/llama/gpuInfo/metal-gpu-info.mm +30 -0
  600. package/llama/gpuInfo/vulkan-gpu-info.cpp +83 -0
  601. package/llama/gpuInfo/vulkan-gpu-info.h +9 -0
  602. package/llama/grammars/README.md +58 -5
  603. package/llama/grammars/json.gbnf +4 -4
  604. package/llama/grammars/json_arr.gbnf +4 -4
  605. package/llama/llama.cpp.info.json +4 -0
  606. package/llama/toolchains/win32.host-x64.target-arm64.cmake +41 -0
  607. package/package.json +78 -53
  608. package/templates/packed/electron-typescript-react.json +1 -0
  609. package/templates/packed/node-typescript.json +1 -0
  610. package/dist/AbortError.d.ts +0 -2
  611. package/dist/AbortError.js +0 -7
  612. package/dist/AbortError.js.map +0 -1
  613. package/dist/chatWrappers/LlamaChatWrapper.d.ts +0 -13
  614. package/dist/chatWrappers/LlamaChatWrapper.js.map +0 -1
  615. package/dist/chatWrappers/resolveChatWrapperBasedOnModel.d.ts +0 -13
  616. package/dist/chatWrappers/resolveChatWrapperBasedOnModel.js +0 -57
  617. package/dist/chatWrappers/resolveChatWrapperBasedOnModel.js.map +0 -1
  618. package/dist/llamaEvaluator/LlamaBins.d.ts +0 -18
  619. package/dist/llamaEvaluator/LlamaBins.js +0 -5
  620. package/dist/llamaEvaluator/LlamaBins.js.map +0 -1
  621. package/dist/llamaEvaluator/LlamaChat/LlamaChat.js +0 -704
  622. package/dist/llamaEvaluator/LlamaChat/LlamaChat.js.map +0 -1
  623. package/dist/llamaEvaluator/LlamaChat/utils/FunctionCallGrammar.d.ts +0 -21
  624. package/dist/llamaEvaluator/LlamaChat/utils/FunctionCallGrammar.js +0 -120
  625. package/dist/llamaEvaluator/LlamaChat/utils/FunctionCallGrammar.js.map +0 -1
  626. package/dist/llamaEvaluator/LlamaChat/utils/contextShiftStrategies/eraseFirstResponseAndKeepFirstSystemChatContextShiftStrategy.js.map +0 -1
  627. package/dist/llamaEvaluator/LlamaChatSession/LlamaChatSession.d.ts +0 -146
  628. package/dist/llamaEvaluator/LlamaChatSession/LlamaChatSession.js +0 -211
  629. package/dist/llamaEvaluator/LlamaChatSession/LlamaChatSession.js.map +0 -1
  630. package/dist/llamaEvaluator/LlamaChatSession/utils/defineChatSessionFunction.js.map +0 -1
  631. package/dist/llamaEvaluator/LlamaContext/LlamaContext.js.map +0 -1
  632. package/dist/llamaEvaluator/LlamaContext/types.d.ts +0 -82
  633. package/dist/llamaEvaluator/LlamaContext/types.js.map +0 -1
  634. package/dist/llamaEvaluator/LlamaContext/utils/batchItemsPrioritizingStrategies/firstInFirstOutStrategy.js.map +0 -1
  635. package/dist/llamaEvaluator/LlamaContext/utils/batchItemsPrioritizingStrategies/maximumParallelismStrategy.js.map +0 -1
  636. package/dist/llamaEvaluator/LlamaContext/utils/resolveBatchItemsPrioritizingStrategy.d.ts +0 -2
  637. package/dist/llamaEvaluator/LlamaContext/utils/resolveBatchItemsPrioritizingStrategy.js.map +0 -1
  638. package/dist/llamaEvaluator/LlamaEmbeddingContext.d.ts +0 -35
  639. package/dist/llamaEvaluator/LlamaEmbeddingContext.js +0 -73
  640. package/dist/llamaEvaluator/LlamaEmbeddingContext.js.map +0 -1
  641. package/dist/llamaEvaluator/LlamaGrammar.js.map +0 -1
  642. package/dist/llamaEvaluator/LlamaGrammarEvaluationState.js.map +0 -1
  643. package/dist/llamaEvaluator/LlamaJsonSchemaGrammar.js.map +0 -1
  644. package/dist/llamaEvaluator/LlamaModel.d.ts +0 -119
  645. package/dist/llamaEvaluator/LlamaModel.js +0 -322
  646. package/dist/llamaEvaluator/LlamaModel.js.map +0 -1
  647. package/dist/utils/binariesGithubRelease.js.map +0 -1
  648. package/dist/utils/clearLlamaBuild.d.ts +0 -1
  649. package/dist/utils/clearLlamaBuild.js +0 -12
  650. package/dist/utils/clearLlamaBuild.js.map +0 -1
  651. package/dist/utils/cloneLlamaCppRepo.d.ts +0 -2
  652. package/dist/utils/cloneLlamaCppRepo.js +0 -102
  653. package/dist/utils/cloneLlamaCppRepo.js.map +0 -1
  654. package/dist/utils/compileLLamaCpp.d.ts +0 -8
  655. package/dist/utils/compileLLamaCpp.js +0 -132
  656. package/dist/utils/compileLLamaCpp.js.map +0 -1
  657. package/dist/utils/getBin.js +0 -78
  658. package/dist/utils/getBin.js.map +0 -1
  659. package/dist/utils/getGbnfGrammarForGbnfJsonSchema.d.ts +0 -2
  660. package/dist/utils/getGbnfGrammarForGbnfJsonSchema.js +0 -9
  661. package/dist/utils/getGbnfGrammarForGbnfJsonSchema.js.map +0 -1
  662. package/dist/utils/getReleaseInfo.d.ts +0 -7
  663. package/dist/utils/getReleaseInfo.js +0 -30
  664. package/dist/utils/getReleaseInfo.js.map +0 -1
  665. package/dist/utils/parseModelTypeDescription.d.ts +0 -6
  666. package/dist/utils/parseModelTypeDescription.js +0 -9
  667. package/dist/utils/parseModelTypeDescription.js.map +0 -1
  668. package/dist/utils/resolveChatWrapper.d.ts +0 -4
  669. package/dist/utils/resolveChatWrapper.js +0 -16
  670. package/dist/utils/resolveChatWrapper.js.map +0 -1
  671. package/dist/utils/usedBinFlag.d.ts +0 -6
  672. package/dist/utils/usedBinFlag.js +0 -15
  673. package/dist/utils/usedBinFlag.js.map +0 -1
  674. package/llama/usedBin.json +0 -3
  675. package/llamaBins/linux-arm64/llama-addon.node +0 -0
  676. package/llamaBins/linux-armv7l/llama-addon.node +0 -0
  677. package/llamaBins/linux-x64/llama-addon.node +0 -0
  678. package/llamaBins/mac-arm64/llama-addon.node +0 -0
  679. package/llamaBins/mac-x64/llama-addon.node +0 -0
  680. package/llamaBins/win-x64/llama-addon.exp +0 -0
  681. package/llamaBins/win-x64/llama-addon.lib +0 -0
  682. package/llamaBins/win-x64/llama-addon.node +0 -0
  683. /package/dist/{utils → bindings/utils}/binariesGithubRelease.d.ts +0 -0
  684. /package/dist/{llamaEvaluator → evaluator}/LlamaChat/utils/contextShiftStrategies/eraseFirstResponseAndKeepFirstSystemChatContextShiftStrategy.d.ts +0 -0
  685. /package/dist/{llamaEvaluator → evaluator}/LlamaContext/types.js +0 -0
  686. /package/dist/{llamaEvaluator/LlamaContext/utils/batchItemsPrioritizingStrategies → evaluator/LlamaContext/utils/batchItemsPrioritizationStrategies}/firstInFirstOutStrategy.d.ts +0 -0
  687. /package/dist/{llamaEvaluator/LlamaContext/utils/batchItemsPrioritizingStrategies → evaluator/LlamaContext/utils/batchItemsPrioritizationStrategies}/firstInFirstOutStrategy.js +0 -0
  688. /package/dist/{llamaEvaluator/LlamaContext/utils/batchItemsPrioritizingStrategies → evaluator/LlamaContext/utils/batchItemsPrioritizationStrategies}/maximumParallelismStrategy.d.ts +0 -0
  689. /package/dist/{llamaEvaluator/LlamaContext/utils/batchItemsPrioritizingStrategies → evaluator/LlamaContext/utils/batchItemsPrioritizationStrategies}/maximumParallelismStrategy.js +0 -0
  690. /package/dist/{llamaEvaluator → evaluator}/LlamaGrammarEvaluationState.d.ts +0 -0
@@ -0,0 +1,141 @@
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 class GgufInsightsConfigurationResolver {
7
+ /** @internal */ _ggufInsights;
8
+ constructor(ggufInsights) {
9
+ this._ggufInsights = ggufInsights;
10
+ }
11
+ get ggufInsights() {
12
+ return this._ggufInsights;
13
+ }
14
+ /**
15
+ * Score the compatibility of the model configuration with the current GPU and VRAM state.
16
+ * Assumes a model is loaded with the default `"auto"` configurations.
17
+ * Scored based on the following criteria:
18
+ * - 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)
19
+ * - Whether all layers can be offloaded to the GPU (gives additional points)
20
+ * - Whether the resolved context size is at least as large as the specified `contextSize`
21
+ *
22
+ * IF the resolved context size is larger than the specified context size, for each multiplier of the specified `contextSize`
23
+ * that the resolved context size is larger by, 1 bonus point is given in the `bonusScore`.
24
+ *
25
+ * `contextSize` defaults to `4096` (if the model train context size is lower than this, the model train context size is used instead).
26
+ */
27
+ async scoreModelConfigurationCompatibility({ contextSize = Math.min(4096, this._ggufInsights.trainContextSize ?? 4096), embeddingContext = 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 } = {}) {
28
+ const defaultTrainContextSize = 4096;
29
+ const [vramState, ramState] = await Promise.all([
30
+ getVramState(),
31
+ getRamState()
32
+ ]);
33
+ const resolvedGpuLayers = await this.resolveModelGpuLayers("auto", {
34
+ getVramState: async () => vramState,
35
+ llamaVramPaddingSize,
36
+ llamaGpu,
37
+ llamaSupportsGpuOffloading
38
+ });
39
+ const canUseGpu = llamaSupportsGpuOffloading && llamaGpu !== false;
40
+ const estimatedModelResourceUsage = this._ggufInsights.estimateModelResourceRequirements({
41
+ gpuLayers: resolvedGpuLayers
42
+ });
43
+ const resolvedContextSize = await this.resolveContextContextSize("auto", {
44
+ getVramState: async () => ({
45
+ total: vramState.total,
46
+ free: Math.max(0, vramState.free - estimatedModelResourceUsage.gpuVram)
47
+ }),
48
+ llamaGpu,
49
+ isEmbeddingContext: embeddingContext,
50
+ modelGpuLayers: resolvedGpuLayers,
51
+ modelTrainContextSize: this._ggufInsights.trainContextSize ?? defaultTrainContextSize
52
+ });
53
+ const estimatedContextResourceUsage = this._ggufInsights.estimateContextResourceRequirements({
54
+ contextSize: resolvedContextSize,
55
+ isEmbeddingContext: embeddingContext,
56
+ modelGpuLayers: resolvedGpuLayers
57
+ });
58
+ const rankPoints = {
59
+ gpuLayers: 60,
60
+ allLayersAreOffloaded: 10,
61
+ contextSize: 30,
62
+ ramUsageFitsInRam: 10,
63
+ cpuOnlySmallModelSize: 60, // also defined inside `scoreModelSizeForCpuOnlyUsage`
64
+ bonusContextSize: 10
65
+ };
66
+ const gpuLayersPoints = rankPoints.gpuLayers * Math.min(1, resolvedGpuLayers / this._ggufInsights.totalLayers);
67
+ const allLayersAreOffloadedPoints = rankPoints.allLayersAreOffloaded * (resolvedGpuLayers === this._ggufInsights.totalLayers ? 1 : 0);
68
+ const contextSizePoints = rankPoints.contextSize * Math.min(1, resolvedContextSize / contextSize);
69
+ const ramUsageFitsInRamPoints = rankPoints.ramUsageFitsInRam * (estimatedModelResourceUsage.cpuRam <= ramState.free
70
+ ? 1
71
+ : estimatedModelResourceUsage.cpuRam <= ramState.total
72
+ ? 0.5
73
+ : (0.5 - Math.min(0.5, 0.5 * ((estimatedModelResourceUsage.cpuRam - ramState.total) / ramState.total))));
74
+ const bonusContextSizePoints = 10 * Math.min(1, Math.max(0, resolvedContextSize - contextSize) / contextSize);
75
+ const compatibilityScore = canUseGpu
76
+ ? ((gpuLayersPoints + allLayersAreOffloadedPoints + contextSizePoints + ramUsageFitsInRamPoints) /
77
+ (rankPoints.gpuLayers + rankPoints.allLayersAreOffloaded + rankPoints.contextSize + rankPoints.ramUsageFitsInRam))
78
+ : ((contextSizePoints + ramUsageFitsInRamPoints + scoreModelSizeForCpuOnlyUsage(this._ggufInsights.modelSize)) /
79
+ (rankPoints.contextSize + rankPoints.ramUsageFitsInRam + rankPoints.cpuOnlySmallModelSize));
80
+ const bonusScore = bonusContextSizePoints / rankPoints.bonusContextSize;
81
+ return {
82
+ compatibilityScore,
83
+ bonusScore,
84
+ totalScore: compatibilityScore + bonusScore,
85
+ resolvedValues: {
86
+ gpuLayers: resolvedGpuLayers,
87
+ contextSize: resolvedContextSize,
88
+ modelRamUsage: estimatedModelResourceUsage.cpuRam,
89
+ contextRamUsage: estimatedContextResourceUsage.cpuRam,
90
+ totalRamUsage: estimatedModelResourceUsage.cpuRam + estimatedContextResourceUsage.cpuRam,
91
+ modelVramUsage: estimatedModelResourceUsage.gpuVram,
92
+ contextVramUsage: estimatedContextResourceUsage.gpuVram,
93
+ totalVramUsage: estimatedModelResourceUsage.gpuVram + estimatedContextResourceUsage.gpuVram
94
+ }
95
+ };
96
+ }
97
+ 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 } = {}) {
98
+ return resolveModelGpuLayersOption(gpuLayers, {
99
+ ggufInsights: this._ggufInsights,
100
+ ignoreMemorySafetyChecks,
101
+ getVramState,
102
+ llamaVramPaddingSize,
103
+ llamaGpu,
104
+ llamaSupportsGpuOffloading
105
+ });
106
+ }
107
+ async resolveContextContextSize(contextSize, { modelGpuLayers, batchSize, modelTrainContextSize, getVramState = (() => this._ggufInsights._llama._vramOrchestrator.getMemoryState()), llamaGpu = this._ggufInsights._llama.gpu, ignoreMemorySafetyChecks = false, isEmbeddingContext = false, sequences = getDefaultContextSequences() }) {
108
+ return await resolveContextContextSizeOption({
109
+ contextSize,
110
+ batchSize,
111
+ sequences,
112
+ modelFileInsights: this._ggufInsights,
113
+ modelGpuLayers,
114
+ modelTrainContextSize,
115
+ getVramState,
116
+ llamaGpu,
117
+ ignoreMemorySafetyChecks,
118
+ isEmbeddingContext
119
+ });
120
+ }
121
+ /** @internal */
122
+ static _create(ggufInsights) {
123
+ return new GgufInsightsConfigurationResolver(ggufInsights);
124
+ }
125
+ }
126
+ function scoreModelSizeForCpuOnlyUsage(modelSize) {
127
+ const s1GB = Math.pow(1024, 3);
128
+ return 60 - scoreLevels(modelSize, [{
129
+ start: s1GB,
130
+ end: s1GB * 2.5,
131
+ points: 40
132
+ }, {
133
+ start: s1GB * 2.5,
134
+ end: s1GB * 4,
135
+ points: 15
136
+ }, {
137
+ start: s1GB * 4,
138
+ points: 5
139
+ }]);
140
+ }
141
+ //# 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;AAInD,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;;;;;;;;;;;;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,KAIxB,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,uBAAuB,GAAG,IAAI,CAAC;QAErC,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,CAAC,MAAM,EAAE;YAC/D,YAAY,EAAE,KAAK,IAAI,EAAE,CAAC,SAAS;YACnC,oBAAoB;YACpB,QAAQ;YACR,0BAA0B;SAC7B,CAAC,CAAC;QACH,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,uBAAuB;SACxF,CAAC,CAAC;QACH,MAAM,6BAA6B,GAAG,IAAI,CAAC,aAAa,CAAC,mCAAmC,CAAC;YACzF,WAAW,EAAE,mBAAmB;YAChC,kBAAkB,EAAE,gBAAgB;YACpC,cAAc,EAAE,iBAAiB;SACpC,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,SAAyC,EAAE,EAC1E,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,KAI5E,EAAE;QACF,OAAO,2BAA2B,CAAC,SAAS,EAAE;YAC1C,YAAY,EAAE,IAAI,CAAC,aAAa;YAChC,wBAAwB;YACxB,YAAY;YACZ,oBAAoB;YACpB,QAAQ;YACR,0BAA0B;SAC7B,CAAC,CAAC;IACP,CAAC;IAEM,KAAK,CAAC,yBAAyB,CAAC,WAA+C,EAAE,EACpF,cAAc,EACd,SAAS,EACT,qBAAqB,EACrB,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,EAU3C;QACG,OAAO,MAAM,+BAA+B,CAAC;YACzC,WAAW;YACX,SAAS;YACT,SAAS;YACT,iBAAiB,EAAE,IAAI,CAAC,aAAa;YACrC,cAAc;YACd,qBAAqB;YACrB,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 +1,18 @@
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, 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
+ getVramState(): Promise<{
12
+ total: number;
13
+ free: number;
14
+ }>;
15
+ llamaGpu: BuildGpu;
16
+ ignoreMemorySafetyChecks?: boolean;
17
+ isEmbeddingContext?: boolean;
18
+ }): Promise<number>;
@@ -0,0 +1,76 @@
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, 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
+ isEmbeddingContext
17
+ }).gpuVram;
18
+ if (contextVram > vramState.free)
19
+ throw new Error(`The context size of ${resolvedContextSize}${sequences > 1 ? ` with ${sequences} sequences` : ""} is too large for the available VRAM`);
20
+ return resolvedContextSize;
21
+ }
22
+ else if (contextSize === "auto" || typeof contextSize === "object") {
23
+ if (llamaGpu === false)
24
+ return modelTrainContextSize;
25
+ const vramState = await getVramState();
26
+ if (vramState.total === 0)
27
+ return modelTrainContextSize;
28
+ const freeVram = vramState.free;
29
+ const maxContextSize = contextSize === "auto"
30
+ ? getDefaultModelContextSize({ trainContextSize: modelTrainContextSize })
31
+ : Math.min(contextSize.max ?? getDefaultModelContextSize({ trainContextSize: modelTrainContextSize }), getDefaultModelContextSize({ trainContextSize: modelTrainContextSize }));
32
+ const minContextSize = contextSize === "auto"
33
+ ? minAllowedContextSizeInCalculations
34
+ : Math.max(contextSize.min ?? minAllowedContextSizeInCalculations, minAllowedContextSizeInCalculations);
35
+ let highestCompatibleContextSize = null;
36
+ let step = -Math.max(1, Math.floor((maxContextSize - minContextSize) / 4));
37
+ for (let testContextSize = maxContextSize; testContextSize >= minContextSize && testContextSize <= maxContextSize;) {
38
+ const contextVram = modelFileInsights.estimateContextResourceRequirements({
39
+ contextSize: testContextSize,
40
+ batchSize: batchSize ?? getDefaultContextBatchSize({ contextSize: testContextSize, sequences }),
41
+ modelGpuLayers: modelGpuLayers,
42
+ sequences,
43
+ isEmbeddingContext
44
+ }).gpuVram;
45
+ if (contextVram <= freeVram) {
46
+ if (highestCompatibleContextSize == null || testContextSize > highestCompatibleContextSize) {
47
+ highestCompatibleContextSize = testContextSize;
48
+ if (step === -1)
49
+ break;
50
+ else if (step < 0)
51
+ step = Math.max(1, Math.floor(-step / 2));
52
+ }
53
+ }
54
+ else if (step > 0)
55
+ step = -Math.max(1, Math.floor(step / 2));
56
+ if (testContextSize == minContextSize && step === -1)
57
+ break;
58
+ testContextSize += step;
59
+ if (testContextSize < minContextSize) {
60
+ testContextSize = minContextSize;
61
+ step = Math.max(1, Math.floor(Math.abs(step) / 2));
62
+ }
63
+ else if (testContextSize > maxContextSize) {
64
+ testContextSize = maxContextSize;
65
+ step = -Math.max(1, Math.floor(Math.abs(step) / 2));
66
+ }
67
+ }
68
+ if (highestCompatibleContextSize != null)
69
+ return highestCompatibleContextSize;
70
+ if (ignoreMemorySafetyChecks)
71
+ return minContextSize;
72
+ throw new Error(`The available VRAM is too small to fit the context size of ${maxContextSize}${sequences > 1 ? ` with ${sequences} sequences` : ""}`);
73
+ }
74
+ throw new Error(`Invalid context size: "${contextSize}"`);
75
+ }
76
+ //# 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,YAAY,EAAE,QAAQ,EACnH,wBAAwB,GAAG,KAAK,EAAE,kBAAkB,GAAG,KAAK,EAY/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,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,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 +1,14 @@
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 }: {
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
+ }): Promise<number>;
@@ -0,0 +1,177 @@
1
+ import { InsufficientMemoryError } from "../../../utils/InsufficientMemoryError.js";
2
+ import { findBestOption } from "../../../utils/findBestOption.js";
3
+ import { getDefaultContextBatchSize, getDefaultModelContextSize } from "../../../evaluator/LlamaContext/LlamaContext.js";
4
+ import { minAllowedContextSizeInCalculations } from "../../../config.js";
5
+ import { scoreLevels } from "./scoreLevels.js";
6
+ const fitContextExtraMemoryPaddingPercentage = 0.5;
7
+ export async function resolveModelGpuLayersOption(gpuLayers, { ggufInsights, ignoreMemorySafetyChecks = false, getVramState, llamaVramPaddingSize, llamaGpu, llamaSupportsGpuOffloading }) {
8
+ if (gpuLayers == null)
9
+ gpuLayers = "auto";
10
+ if (!llamaSupportsGpuOffloading)
11
+ return 0;
12
+ if (gpuLayers === "max" || typeof gpuLayers === "number") {
13
+ const resolvedGpuLayers = typeof gpuLayers === "number"
14
+ ? Math.max(0, Math.min(ggufInsights.totalLayers, gpuLayers))
15
+ : ggufInsights.totalLayers;
16
+ if (ignoreMemorySafetyChecks)
17
+ return resolvedGpuLayers;
18
+ const vramState = await getVramState();
19
+ const maxLayersRequirements = getVramRequiredForGpuLayers({
20
+ gpuLayers: resolvedGpuLayers,
21
+ ggufInsights,
22
+ currentVram: vramState.free
23
+ });
24
+ if (maxLayersRequirements == null)
25
+ throw new InsufficientMemoryError("Not enough VRAM to fit the model with the specified settings");
26
+ return resolvedGpuLayers;
27
+ }
28
+ else if (gpuLayers === "auto" || typeof gpuLayers === "object") {
29
+ if (llamaGpu === false)
30
+ return 0;
31
+ const vramState = await getVramState();
32
+ if (vramState.total === 0)
33
+ return 0;
34
+ let freeVram = vramState.free;
35
+ if (typeof gpuLayers === "object" && gpuLayers.fitContext?.contextSize != null) {
36
+ freeVram -= llamaVramPaddingSize * fitContextExtraMemoryPaddingPercentage;
37
+ if (freeVram < 0)
38
+ freeVram = 0;
39
+ }
40
+ const bestGpuLayersOption = getBestGpuLayersForFreeVram({
41
+ ggufInsights,
42
+ freeVram,
43
+ fitContext: typeof gpuLayers === "object"
44
+ ? gpuLayers.fitContext
45
+ : undefined,
46
+ minGpuLayers: typeof gpuLayers === "object"
47
+ ? gpuLayers.min
48
+ : undefined,
49
+ maxGpuLayers: typeof gpuLayers === "object"
50
+ ? gpuLayers.max
51
+ : undefined
52
+ });
53
+ const hasGpuLayersRequirements = typeof gpuLayers === "object" &&
54
+ (gpuLayers.min != null || gpuLayers.max != null || gpuLayers.fitContext?.contextSize != null);
55
+ if (!ignoreMemorySafetyChecks && bestGpuLayersOption == null && hasGpuLayersRequirements)
56
+ throw new InsufficientMemoryError("Not enough VRAM to fit the model with the specified settings");
57
+ return bestGpuLayersOption ?? 0;
58
+ }
59
+ throw new Error(`Invalid gpuLayers value: ${gpuLayers}`);
60
+ }
61
+ function getBestGpuLayersForFreeVram({ ggufInsights, freeVram, fitContext, minGpuLayers, maxGpuLayers }) {
62
+ return findBestOption({
63
+ *generator() {
64
+ const minLayers = Math.floor(Math.max(0, minGpuLayers ?? 0));
65
+ const maxLayers = Math.floor(Math.min(ggufInsights.totalLayers, maxGpuLayers ?? ggufInsights.totalLayers));
66
+ for (let layers = maxLayers; layers >= minLayers; layers--) {
67
+ yield {
68
+ gpuLayers: layers
69
+ };
70
+ }
71
+ },
72
+ score(option) {
73
+ const layersRequirements = getVramRequiredForGpuLayers({
74
+ gpuLayers: option.gpuLayers,
75
+ ggufInsights,
76
+ currentVram: freeVram,
77
+ fitContext
78
+ });
79
+ if (layersRequirements == null)
80
+ return null;
81
+ return scoreGpuLayersAndContextCombination({ gpuLayers: option.gpuLayers, contextSize: layersRequirements.contextSize }, {
82
+ totalGpuLayers: ggufInsights.totalLayers,
83
+ trainContextSize: getDefaultModelContextSize({ trainContextSize: ggufInsights.trainContextSize })
84
+ });
85
+ }
86
+ })?.gpuLayers ?? null;
87
+ }
88
+ function scoreGpuLayersAndContextCombination({ gpuLayers, contextSize }, { totalGpuLayers, trainContextSize }) {
89
+ function scoreGpuLayers() {
90
+ return scoreLevels(gpuLayers, [{
91
+ start: 0,
92
+ points: 4
93
+ }, {
94
+ start: 1,
95
+ points: 26
96
+ }, {
97
+ start: totalGpuLayers,
98
+ points: 14,
99
+ end: totalGpuLayers
100
+ }]);
101
+ }
102
+ function scoreContextSize() {
103
+ const gpuLayersPercentage = gpuLayers / totalGpuLayers;
104
+ return scoreLevels(contextSize, [{
105
+ start: 0,
106
+ points: 2
107
+ }, {
108
+ start: 1024,
109
+ points: 4
110
+ }, {
111
+ start: 2048,
112
+ points: gpuLayersPercentage < 0.1 ? 1 : 8
113
+ }, {
114
+ start: 4096,
115
+ points: gpuLayersPercentage < 0.3 ? 4 : 16
116
+ }, {
117
+ start: 8192,
118
+ points: gpuLayersPercentage < 0.6 ? 1 : 8,
119
+ end: Math.max(trainContextSize, 16384)
120
+ }]);
121
+ }
122
+ return scoreGpuLayers() + scoreContextSize();
123
+ }
124
+ function getVramRequiredForGpuLayers({ gpuLayers, ggufInsights, currentVram, fitContext }) {
125
+ const modelVram = ggufInsights.estimateModelResourceRequirements({ gpuLayers }).gpuVram;
126
+ if (modelVram > currentVram)
127
+ return null;
128
+ if (fitContext != null && fitContext.contextSize != null) {
129
+ const contextVram = ggufInsights.estimateContextResourceRequirements({
130
+ contextSize: fitContext.contextSize,
131
+ batchSize: getDefaultContextBatchSize({ contextSize: fitContext.contextSize, sequences: 1 }),
132
+ modelGpuLayers: gpuLayers,
133
+ sequences: 1,
134
+ isEmbeddingContext: fitContext.embeddingContext ?? false
135
+ }).gpuVram;
136
+ const totalVram = modelVram + contextVram;
137
+ if (totalVram > currentVram)
138
+ return null;
139
+ return {
140
+ contextSize: fitContext.contextSize,
141
+ contextVram,
142
+ totalVram
143
+ };
144
+ }
145
+ const maxContext = findMaxPossibleContextSizeForVram({
146
+ gpuLayers,
147
+ ggufInsights,
148
+ vram: currentVram - modelVram,
149
+ isEmbeddingContext: fitContext?.embeddingContext ?? false
150
+ });
151
+ if (maxContext == null || modelVram + maxContext.vram > currentVram)
152
+ return null;
153
+ return {
154
+ contextSize: maxContext.contextSize,
155
+ contextVram: maxContext.vram,
156
+ totalVram: modelVram + maxContext.vram
157
+ };
158
+ }
159
+ function findMaxPossibleContextSizeForVram({ gpuLayers, ggufInsights, vram, isEmbeddingContext }) {
160
+ const maxContextSize = getDefaultModelContextSize({ trainContextSize: ggufInsights.trainContextSize });
161
+ for (let contextSize = maxContextSize; contextSize >= minAllowedContextSizeInCalculations; contextSize--) {
162
+ const contextVram = ggufInsights.estimateContextResourceRequirements({
163
+ contextSize,
164
+ batchSize: getDefaultContextBatchSize({ contextSize, sequences: 1 }),
165
+ modelGpuLayers: gpuLayers,
166
+ sequences: 1,
167
+ isEmbeddingContext
168
+ }).gpuVram;
169
+ if (contextVram <= vram)
170
+ return {
171
+ contextSize,
172
+ vram: contextVram
173
+ };
174
+ }
175
+ return null;
176
+ }
177
+ //# sourceMappingURL=resolveModelGpuLayersOption.js.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"resolveModelGpuLayersOption.js","sourceRoot":"","sources":["../../../../src/gguf/insights/utils/resolveModelGpuLayersOption.ts"],"names":[],"mappings":"AAEA,OAAO,EAAC,uBAAuB,EAAC,MAAM,2CAA2C,CAAC;AAClF,OAAO,EAAC,cAAc,EAAC,MAAM,kCAAkC,CAAC;AAChE,OAAO,EAAC,0BAA0B,EAAE,0BAA0B,EAAC,MAAM,iDAAiD,CAAC;AACvH,OAAO,EAAC,mCAAmC,EAAC,MAAM,oBAAoB,CAAC;AACvE,OAAO,EAAC,WAAW,EAAC,MAAM,kBAAkB,CAAC;AAG7C,MAAM,sCAAsC,GAAG,GAAG,CAAC;AAEnD,MAAM,CAAC,KAAK,UAAU,2BAA2B,CAAC,SAAyC,EAAE,EACzF,YAAY,EAAE,wBAAwB,GAAG,KAAK,EAAE,YAAY,EAAE,oBAAoB,EAClF,QAAQ,EAAE,0BAA0B,EAKvC;IACG,IAAI,SAAS,IAAI,IAAI;QACjB,SAAS,GAAG,MAAM,CAAC;IAEvB,IAAI,CAAC,0BAA0B;QAC3B,OAAO,CAAC,CAAC;IAEb,IAAI,SAAS,KAAK,KAAK,IAAI,OAAO,SAAS,KAAK,QAAQ,EAAE,CAAC;QACvD,MAAM,iBAAiB,GAAG,OAAO,SAAS,KAAK,QAAQ;YACnD,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,IAAI,CAAC,GAAG,CAAC,YAAY,CAAC,WAAW,EAAE,SAAS,CAAC,CAAC;YAC5D,CAAC,CAAC,YAAY,CAAC,WAAW,CAAC;QAE/B,IAAI,wBAAwB;YACxB,OAAO,iBAAiB,CAAC;QAE7B,MAAM,SAAS,GAAG,MAAM,YAAY,EAAE,CAAC;QACvC,MAAM,qBAAqB,GAAG,2BAA2B,CAAC;YACtD,SAAS,EAAE,iBAAiB;YAC5B,YAAY;YACZ,WAAW,EAAE,SAAS,CAAC,IAAI;SAC9B,CAAC,CAAC;QAEH,IAAI,qBAAqB,IAAI,IAAI;YAC7B,MAAM,IAAI,uBAAuB,CAAC,8DAA8D,CAAC,CAAC;QAEtG,OAAO,iBAAiB,CAAC;IAC7B,CAAC;SAAM,IAAI,SAAS,KAAK,MAAM,IAAI,OAAO,SAAS,KAAK,QAAQ,EAAE,CAAC;QAC/D,IAAI,QAAQ,KAAK,KAAK;YAClB,OAAO,CAAC,CAAC;QAEb,MAAM,SAAS,GAAG,MAAM,YAAY,EAAE,CAAC;QACvC,IAAI,SAAS,CAAC,KAAK,KAAK,CAAC;YACrB,OAAO,CAAC,CAAC;QAEb,IAAI,QAAQ,GAAG,SAAS,CAAC,IAAI,CAAC;QAC9B,IAAI,OAAO,SAAS,KAAK,QAAQ,IAAI,SAAS,CAAC,UAAU,EAAE,WAAW,IAAI,IAAI,EAAE,CAAC;YAC7E,QAAQ,IAAI,oBAAoB,GAAG,sCAAsC,CAAC;YAE1E,IAAI,QAAQ,GAAG,CAAC;gBACZ,QAAQ,GAAG,CAAC,CAAC;QACrB,CAAC;QAED,MAAM,mBAAmB,GAAG,2BAA2B,CAAC;YACpD,YAAY;YACZ,QAAQ;YACR,UAAU,EAAE,OAAO,SAAS,KAAK,QAAQ;gBACrC,CAAC,CAAC,SAAS,CAAC,UAAU;gBACtB,CAAC,CAAC,SAAS;YACf,YAAY,EAAE,OAAO,SAAS,KAAK,QAAQ;gBACvC,CAAC,CAAC,SAAS,CAAC,GAAG;gBACf,CAAC,CAAC,SAAS;YACf,YAAY,EAAE,OAAO,SAAS,KAAK,QAAQ;gBACvC,CAAC,CAAC,SAAS,CAAC,GAAG;gBACf,CAAC,CAAC,SAAS;SAClB,CAAC,CAAC;QAEH,MAAM,wBAAwB,GAAG,OAAO,SAAS,KAAK,QAAQ;YAC1D,CAAC,SAAS,CAAC,GAAG,IAAI,IAAI,IAAI,SAAS,CAAC,GAAG,IAAI,IAAI,IAAI,SAAS,CAAC,UAAU,EAAE,WAAW,IAAI,IAAI,CAAC,CAAC;QAElG,IAAI,CAAC,wBAAwB,IAAI,mBAAmB,IAAI,IAAI,IAAI,wBAAwB;YACpF,MAAM,IAAI,uBAAuB,CAAC,8DAA8D,CAAC,CAAC;QAEtG,OAAO,mBAAmB,IAAI,CAAC,CAAC;IACpC,CAAC;IAED,MAAM,IAAI,KAAK,CAAC,4BAA4B,SAAS,EAAE,CAAC,CAAC;AAC7D,CAAC;AAED,SAAS,2BAA2B,CAAC,EACjC,YAAY,EACZ,QAAQ,EACR,UAAU,EACV,YAAY,EACZ,YAAY,EAOf;IACG,OAAO,cAAc,CAAC;QAClB,CAAC,SAAS;YACN,MAAM,SAAS,GAAG,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,YAAY,IAAI,CAAC,CAAC,CAAC,CAAC;YAC7D,MAAM,SAAS,GAAG,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,GAAG,CAAC,YAAY,CAAC,WAAW,EAAE,YAAY,IAAI,YAAY,CAAC,WAAW,CAAC,CAAC,CAAC;YAE3G,KAAK,IAAI,MAAM,GAAG,SAAS,EAAE,MAAM,IAAI,SAAS,EAAE,MAAM,EAAE,EAAE,CAAC;gBACzD,MAAM;oBACF,SAAS,EAAE,MAAM;iBACpB,CAAC;YACN,CAAC;QACL,CAAC;QACD,KAAK,CAAC,MAAM;YACR,MAAM,kBAAkB,GAAG,2BAA2B,CAAC;gBACnD,SAAS,EAAE,MAAM,CAAC,SAAS;gBAC3B,YAAY;gBACZ,WAAW,EAAE,QAAQ;gBACrB,UAAU;aACb,CAAC,CAAC;YAEH,IAAI,kBAAkB,IAAI,IAAI;gBAC1B,OAAO,IAAI,CAAC;YAEhB,OAAO,mCAAmC,CAAC,EAAC,SAAS,EAAE,MAAM,CAAC,SAAS,EAAE,WAAW,EAAE,kBAAkB,CAAC,WAAW,EAAC,EAAE;gBACnH,cAAc,EAAE,YAAY,CAAC,WAAW;gBACxC,gBAAgB,EAAE,0BAA0B,CAAC,EAAC,gBAAgB,EAAE,YAAY,CAAC,gBAAgB,EAAC,CAAC;aAClG,CAAC,CAAC;QACP,CAAC;KACJ,CAAC,EAAE,SAAS,IAAI,IAAI,CAAC;AAC1B,CAAC;AAED,SAAS,mCAAmC,CAAC,EAAC,SAAS,EAAE,WAAW,EAA2C,EAAE,EAC7G,cAAc,EAAE,gBAAgB,EAGnC;IACG,SAAS,cAAc;QACnB,OAAO,WAAW,CAAC,SAAS,EAAE,CAAC;gBAC3B,KAAK,EAAE,CAAC;gBACR,MAAM,EAAE,CAAC;aACZ,EAAE;gBACC,KAAK,EAAE,CAAC;gBACR,MAAM,EAAE,EAAE;aACb,EAAE;gBACC,KAAK,EAAE,cAAc;gBACrB,MAAM,EAAE,EAAE;gBACV,GAAG,EAAE,cAAc;aACtB,CAAC,CAAC,CAAC;IACR,CAAC;IAED,SAAS,gBAAgB;QACrB,MAAM,mBAAmB,GAAG,SAAS,GAAG,cAAc,CAAC;QAEvD,OAAO,WAAW,CAAC,WAAW,EAAE,CAAC;gBAC7B,KAAK,EAAE,CAAC;gBACR,MAAM,EAAE,CAAC;aACZ,EAAE;gBACC,KAAK,EAAE,IAAI;gBACX,MAAM,EAAE,CAAC;aACZ,EAAE;gBACC,KAAK,EAAE,IAAI;gBACX,MAAM,EAAE,mBAAmB,GAAG,GAAG,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;aAC5C,EAAE;gBACC,KAAK,EAAE,IAAI;gBACX,MAAM,EAAE,mBAAmB,GAAG,GAAG,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE;aAC7C,EAAE;gBACC,KAAK,EAAE,IAAI;gBACX,MAAM,EAAE,mBAAmB,GAAG,GAAG,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;gBACzC,GAAG,EAAE,IAAI,CAAC,GAAG,CAAC,gBAAgB,EAAE,KAAK,CAAC;aACzC,CAAC,CAAC,CAAC;IACR,CAAC;IAED,OAAO,cAAc,EAAE,GAAG,gBAAgB,EAAE,CAAC;AACjD,CAAC;AAED,SAAS,2BAA2B,CAAC,EACjC,SAAS,EAAE,YAAY,EAAE,WAAW,EAAE,UAAU,EAGnD;IACG,MAAM,SAAS,GAAG,YAAY,CAAC,iCAAiC,CAAC,EAAC,SAAS,EAAC,CAAC,CAAC,OAAO,CAAC;IAEtF,IAAI,SAAS,GAAG,WAAW;QACvB,OAAO,IAAI,CAAC;IAEhB,IAAI,UAAU,IAAI,IAAI,IAAI,UAAU,CAAC,WAAW,IAAI,IAAI,EAAE,CAAC;QACvD,MAAM,WAAW,GAAG,YAAY,CAAC,mCAAmC,CAAC;YACjE,WAAW,EAAE,UAAU,CAAC,WAAW;YACnC,SAAS,EAAE,0BAA0B,CAAC,EAAC,WAAW,EAAE,UAAU,CAAC,WAAW,EAAE,SAAS,EAAE,CAAC,EAAC,CAAC;YAC1F,cAAc,EAAE,SAAS;YACzB,SAAS,EAAE,CAAC;YACZ,kBAAkB,EAAE,UAAU,CAAC,gBAAgB,IAAI,KAAK;SAC3D,CAAC,CAAC,OAAO,CAAC;QAEX,MAAM,SAAS,GAAG,SAAS,GAAG,WAAW,CAAC;QAC1C,IAAI,SAAS,GAAG,WAAW;YACvB,OAAO,IAAI,CAAC;QAEhB,OAAO;YACH,WAAW,EAAE,UAAU,CAAC,WAAW;YACnC,WAAW;YACX,SAAS;SACZ,CAAC;IACN,CAAC;IAED,MAAM,UAAU,GAAG,iCAAiC,CAAC;QACjD,SAAS;QACT,YAAY;QACZ,IAAI,EAAE,WAAW,GAAG,SAAS;QAC7B,kBAAkB,EAAE,UAAU,EAAE,gBAAgB,IAAI,KAAK;KAC5D,CAAC,CAAC;IAEH,IAAI,UAAU,IAAI,IAAI,IAAI,SAAS,GAAG,UAAU,CAAC,IAAI,GAAG,WAAW;QAC/D,OAAO,IAAI,CAAC;IAEhB,OAAO;QACH,WAAW,EAAE,UAAU,CAAC,WAAW;QACnC,WAAW,EAAE,UAAU,CAAC,IAAI;QAC5B,SAAS,EAAE,SAAS,GAAG,UAAU,CAAC,IAAI;KACzC,CAAC;AACN,CAAC;AAED,SAAS,iCAAiC,CAAC,EAAC,SAAS,EAAE,YAAY,EAAE,IAAI,EAAE,kBAAkB,EAE5F;IACG,MAAM,cAAc,GAAG,0BAA0B,CAAC,EAAC,gBAAgB,EAAE,YAAY,CAAC,gBAAgB,EAAC,CAAC,CAAC;IAErG,KAAK,IAAI,WAAW,GAAG,cAAc,EAAE,WAAW,IAAI,mCAAmC,EAAE,WAAW,EAAE,EAAE,CAAC;QACvG,MAAM,WAAW,GAAG,YAAY,CAAC,mCAAmC,CAAC;YACjE,WAAW;YACX,SAAS,EAAE,0BAA0B,CAAC,EAAC,WAAW,EAAE,SAAS,EAAE,CAAC,EAAC,CAAC;YAClE,cAAc,EAAE,SAAS;YACzB,SAAS,EAAE,CAAC;YACZ,kBAAkB;SACrB,CAAC,CAAC,OAAO,CAAC;QAEX,IAAI,WAAW,IAAI,IAAI;YACnB,OAAO;gBACH,WAAW;gBACX,IAAI,EAAE,WAAW;aACpB,CAAC;IACV,CAAC;IAED,OAAO,IAAI,CAAC;AAChB,CAAC"}
@@ -0,0 +1,5 @@
1
+ export declare function scoreLevels(num: number, levels: {
2
+ start: number;
3
+ end?: number;
4
+ points: number;
5
+ }[]): number;
@@ -0,0 +1,16 @@
1
+ export function scoreLevels(num, levels) {
2
+ let res = 0;
3
+ for (let i = 0; i < levels.length; i++) {
4
+ const level = levels[i];
5
+ const start = level.start;
6
+ const end = level.end ?? levels[i + 1]?.start ?? Math.max(start, num);
7
+ if (num < start)
8
+ break;
9
+ else if (num >= end)
10
+ res += level.points;
11
+ else
12
+ res += level.points * ((num - start) / (end - start));
13
+ }
14
+ return res;
15
+ }
16
+ //# sourceMappingURL=scoreLevels.js.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"scoreLevels.js","sourceRoot":"","sources":["../../../../src/gguf/insights/utils/scoreLevels.ts"],"names":[],"mappings":"AAAA,MAAM,UAAU,WAAW,CAAC,GAAW,EAAE,MAAyD;IAC9F,IAAI,GAAG,GAAG,CAAC,CAAC;IAEZ,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,MAAM,CAAC,MAAM,EAAE,CAAC,EAAE,EAAE,CAAC;QACrC,MAAM,KAAK,GAAG,MAAM,CAAC,CAAC,CAAC,CAAC;QACxB,MAAM,KAAK,GAAG,KAAK,CAAC,KAAK,CAAC;QAC1B,MAAM,GAAG,GAAG,KAAK,CAAC,GAAG,IAAI,MAAM,CAAC,CAAC,GAAG,CAAC,CAAC,EAAE,KAAK,IAAI,IAAI,CAAC,GAAG,CAAC,KAAK,EAAE,GAAG,CAAC,CAAC;QAEtE,IAAI,GAAG,GAAG,KAAK;YACX,MAAM;aACL,IAAI,GAAG,IAAI,GAAG;YACf,GAAG,IAAI,KAAK,CAAC,MAAM,CAAC;;YAEpB,GAAG,IAAI,KAAK,CAAC,MAAM,GAAG,CAAC,CAAC,GAAG,GAAG,KAAK,CAAC,GAAG,CAAC,GAAG,GAAG,KAAK,CAAC,CAAC,CAAC;IAC9D,CAAC;IAED,OAAO,GAAG,CAAC;AACf,CAAC"}
@@ -0,0 +1,19 @@
1
+ import { GgufReadOffset } from "../utils/GgufReadOffset.js";
2
+ import { GgufValueType, GgufVersionParserOptions, GgufVersionParserResult, MetadataKeyValueRecord, MetadataValue } from "../types/GgufFileInfoTypes.js";
3
+ export declare class GgufV2Parser {
4
+ private readonly _fileReader;
5
+ private readonly _shouldReadTensorInfo;
6
+ private readonly _ignoreKeys;
7
+ private readonly _readOffset;
8
+ private readonly _logWarnings;
9
+ constructor({ fileReader, readTensorInfo, ignoreKeys, readOffset, logWarnings }: GgufVersionParserOptions);
10
+ parse(): Promise<GgufVersionParserResult>;
11
+ protected _readGgufValue(type: GgufValueType, offset: number | GgufReadOffset): Promise<MetadataValue>;
12
+ protected _readStringValue(offset: number | GgufReadOffset): Promise<string>;
13
+ protected _readRawHeader(readOffset: GgufReadOffset): Promise<{
14
+ tensorCount: number | bigint;
15
+ metadata: MetadataKeyValueRecord;
16
+ headerSize: number;
17
+ }>;
18
+ private _readTensorInfo;
19
+ }
@@ -0,0 +1,115 @@
1
+ import { GgufFileReader, valueTypeToBytesToRead } from "../fileReaders/GgufFileReader.js";
2
+ import { GgufReadOffset } from "../utils/GgufReadOffset.js";
3
+ import { UnsupportedGgufValueTypeError } from "../errors/UnsupportedGgufValueTypeError.js";
4
+ import { GgufValueType } from "../types/GgufFileInfoTypes.js";
5
+ import { convertMetadataKeyValueRecordToNestedObject } from "../utils/convertMetadataKeyValueRecordToNestedObject.js";
6
+ export class GgufV2Parser {
7
+ _fileReader;
8
+ _shouldReadTensorInfo;
9
+ _ignoreKeys;
10
+ _readOffset;
11
+ _logWarnings;
12
+ constructor({ fileReader, readTensorInfo = true, ignoreKeys = [], readOffset, logWarnings }) {
13
+ this._fileReader = fileReader;
14
+ this._shouldReadTensorInfo = readTensorInfo;
15
+ this._ignoreKeys = ignoreKeys;
16
+ this._readOffset = readOffset;
17
+ this._logWarnings = logWarnings;
18
+ }
19
+ async parse() {
20
+ const readOffset = this._readOffset;
21
+ const initialOffset = readOffset.offset;
22
+ const headerReadResult = await this._readRawHeader(readOffset);
23
+ const tensorReadResult = this._shouldReadTensorInfo
24
+ ? await this._readTensorInfo(headerReadResult.tensorCount, readOffset)
25
+ : null;
26
+ const metadata = convertMetadataKeyValueRecordToNestedObject(headerReadResult.metadata, {
27
+ logOverrideWarnings: this._logWarnings,
28
+ ignoreKeys: this._ignoreKeys
29
+ });
30
+ return {
31
+ tensorCount: headerReadResult.tensorCount,
32
+ metadata: metadata,
33
+ tensorInfo: tensorReadResult?.tensorInfo,
34
+ metadataSize: headerReadResult.headerSize + initialOffset,
35
+ tensorInfoSize: tensorReadResult?.tensorInfoSize
36
+ };
37
+ }
38
+ async _readGgufValue(type, offset) {
39
+ const readOffset = GgufReadOffset.resolveReadOffset(offset);
40
+ switch (type) {
41
+ case GgufValueType.Uint8: return await this._fileReader.readUint8(readOffset);
42
+ case GgufValueType.Int8: return await this._fileReader.readInt8(readOffset);
43
+ case GgufValueType.Uint16: return await this._fileReader.readUint16(readOffset);
44
+ case GgufValueType.Int16: return await this._fileReader.readInt16(readOffset);
45
+ case GgufValueType.Uint32: return await this._fileReader.readUint32(readOffset);
46
+ case GgufValueType.Int32: return await this._fileReader.readInt32(readOffset);
47
+ case GgufValueType.Float32: return await this._fileReader.readFloat32(readOffset);
48
+ case GgufValueType.Bool: return await this._fileReader.readBool(readOffset);
49
+ case GgufValueType.String: return await this._readStringValue(readOffset);
50
+ case GgufValueType.Uint64: return await this._fileReader.readUint64(readOffset);
51
+ case GgufValueType.Int64: return await this._fileReader.readInt64(readOffset);
52
+ case GgufValueType.Float64: return await this._fileReader.readFloat64(readOffset);
53
+ }
54
+ if (type === GgufValueType.Array) {
55
+ const arrayType = await this._fileReader.readUint32(readOffset);
56
+ const arrayLength = await this._fileReader.readUint64(readOffset);
57
+ const arrayValues = [];
58
+ for (let i = 0; i < arrayLength; i++) {
59
+ const value = await this._readGgufValue(arrayType, readOffset);
60
+ arrayValues.push(value);
61
+ }
62
+ return arrayValues;
63
+ }
64
+ throw new UnsupportedGgufValueTypeError(type);
65
+ }
66
+ async _readStringValue(offset) {
67
+ const readOffset = GgufReadOffset.resolveReadOffset(offset);
68
+ const length = Number(await this._fileReader.readUint64(readOffset));
69
+ const readLength = valueTypeToBytesToRead.uint8 * length;
70
+ const stringBytes = await this._fileReader.readByteRange(readOffset, readLength);
71
+ return String.fromCharCode(...stringBytes);
72
+ }
73
+ async _readRawHeader(readOffset) {
74
+ const initialOffset = readOffset.offset;
75
+ const tensorCount = await this._fileReader.readUint64(readOffset);
76
+ const metadataKVCount = Number(await this._fileReader.readUint64(readOffset));
77
+ const metadata = {};
78
+ for (let i = 0; i < metadataKVCount; i++) {
79
+ const keyResult = await this._readStringValue(readOffset);
80
+ const valueType = await this._fileReader.readUint32(readOffset);
81
+ metadata[keyResult] = await this._readGgufValue(valueType, readOffset);
82
+ }
83
+ return {
84
+ tensorCount: GgufFileReader.castNumberIfSafe(tensorCount),
85
+ metadata: metadata,
86
+ headerSize: readOffset.offset - initialOffset
87
+ };
88
+ }
89
+ async _readTensorInfo(tensorCount, readOffset) {
90
+ const initialOffset = readOffset.offset;
91
+ const tensorInfo = [];
92
+ for (let i = 0n; i < BigInt(tensorCount); i++) {
93
+ const name = await this._readStringValue(readOffset);
94
+ const dimensionsNumber = await this._fileReader.readUint32(readOffset);
95
+ const dimensions = [];
96
+ for (let i = 0; i < dimensionsNumber; i++) {
97
+ const dimension = await this._fileReader.readUint64(readOffset);
98
+ dimensions.push(GgufFileReader.castNumberIfSafe(dimension));
99
+ }
100
+ const ggmlType = await this._fileReader.readUint32(readOffset);
101
+ const offset = await this._fileReader.readUint64(readOffset);
102
+ tensorInfo.push({
103
+ name,
104
+ dimensions,
105
+ ggmlType: ggmlType,
106
+ offset: GgufFileReader.castNumberIfSafe(offset)
107
+ });
108
+ }
109
+ return {
110
+ tensorInfo,
111
+ tensorInfoSize: readOffset.offset - initialOffset
112
+ };
113
+ }
114
+ }
115
+ //# sourceMappingURL=GgufV2Parser.js.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"GgufV2Parser.js","sourceRoot":"","sources":["../../../src/gguf/parser/GgufV2Parser.ts"],"names":[],"mappings":"AAAA,OAAO,EAAC,cAAc,EAAE,sBAAsB,EAAC,MAAM,kCAAkC,CAAC;AACxF,OAAO,EAAC,cAAc,EAAC,MAAM,4BAA4B,CAAC;AAC1D,OAAO,EAAC,6BAA6B,EAAC,MAAM,4CAA4C,CAAC;AACzF,OAAO,EACH,aAAa,EAChB,MAAM,+BAA+B,CAAC;AAGvC,OAAO,EAAC,2CAA2C,EAAC,MAAM,yDAAyD,CAAC;AAEpH,MAAM,OAAO,YAAY;IACJ,WAAW,CAAiB;IAC5B,qBAAqB,CAAU;IAC/B,WAAW,CAAW;IACtB,WAAW,CAAiB;IAC5B,YAAY,CAAU;IAEvC,YAAmB,EAAC,UAAU,EAAE,cAAc,GAAG,IAAI,EAAE,UAAU,GAAG,EAAE,EAAE,UAAU,EAAE,WAAW,EAA2B;QACtH,IAAI,CAAC,WAAW,GAAG,UAAU,CAAC;QAC9B,IAAI,CAAC,qBAAqB,GAAG,cAAc,CAAC;QAC5C,IAAI,CAAC,WAAW,GAAG,UAAU,CAAC;QAC9B,IAAI,CAAC,WAAW,GAAG,UAAU,CAAC;QAC9B,IAAI,CAAC,YAAY,GAAG,WAAW,CAAC;IACpC,CAAC;IAEM,KAAK,CAAC,KAAK;QACd,MAAM,UAAU,GAAG,IAAI,CAAC,WAAW,CAAC;QACpC,MAAM,aAAa,GAAG,UAAU,CAAC,MAAM,CAAC;QAExC,MAAM,gBAAgB,GAAG,MAAM,IAAI,CAAC,cAAc,CAAC,UAAU,CAAC,CAAC;QAC/D,MAAM,gBAAgB,GAAG,IAAI,CAAC,qBAAqB;YAC/C,CAAC,CAAC,MAAM,IAAI,CAAC,eAAe,CAAC,gBAAgB,CAAC,WAAW,EAAE,UAAU,CAAC;YACtE,CAAC,CAAC,IAAI,CAAC;QACX,MAAM,QAAQ,GAAG,2CAA2C,CAAC,gBAAgB,CAAC,QAAQ,EAAE;YACpF,mBAAmB,EAAE,IAAI,CAAC,YAAY;YACtC,UAAU,EAAE,IAAI,CAAC,WAAW;SAC/B,CAAC,CAAC;QAEH,OAAO;YACH,WAAW,EAAE,gBAAgB,CAAC,WAAW;YACzC,QAAQ,EAAE,QAA+B;YACzC,UAAU,EAAE,gBAAgB,EAAE,UAAU;YACxC,YAAY,EAAE,gBAAgB,CAAC,UAAU,GAAG,aAAa;YACzD,cAAc,EAAE,gBAAgB,EAAE,cAAc;SACnD,CAAC;IACN,CAAC;IAES,KAAK,CAAC,cAAc,CAAC,IAAmB,EAAE,MAA+B;QAC/E,MAAM,UAAU,GAAG,cAAc,CAAC,iBAAiB,CAAC,MAAM,CAAC,CAAC;QAE5D,QAAQ,IAAI,EAAE,CAAC;YACX,KAAK,aAAa,CAAC,KAAK,CAAC,CAAC,OAAO,MAAM,IAAI,CAAC,WAAW,CAAC,SAAS,CAAC,UAAU,CAAC,CAAC;YAC9E,KAAK,aAAa,CAAC,IAAI,CAAC,CAAC,OAAO,MAAM,IAAI,CAAC,WAAW,CAAC,QAAQ,CAAC,UAAU,CAAC,CAAC;YAC5E,KAAK,aAAa,CAAC,MAAM,CAAC,CAAC,OAAO,MAAM,IAAI,CAAC,WAAW,CAAC,UAAU,CAAC,UAAU,CAAC,CAAC;YAChF,KAAK,aAAa,CAAC,KAAK,CAAC,CAAC,OAAO,MAAM,IAAI,CAAC,WAAW,CAAC,SAAS,CAAC,UAAU,CAAC,CAAC;YAC9E,KAAK,aAAa,CAAC,MAAM,CAAC,CAAC,OAAO,MAAM,IAAI,CAAC,WAAW,CAAC,UAAU,CAAC,UAAU,CAAC,CAAC;YAChF,KAAK,aAAa,CAAC,KAAK,CAAC,CAAC,OAAO,MAAM,IAAI,CAAC,WAAW,CAAC,SAAS,CAAC,UAAU,CAAC,CAAC;YAC9E,KAAK,aAAa,CAAC,OAAO,CAAC,CAAC,OAAO,MAAM,IAAI,CAAC,WAAW,CAAC,WAAW,CAAC,UAAU,CAAC,CAAC;YAClF,KAAK,aAAa,CAAC,IAAI,CAAC,CAAC,OAAO,MAAM,IAAI,CAAC,WAAW,CAAC,QAAQ,CAAC,UAAU,CAAC,CAAC;YAC5E,KAAK,aAAa,CAAC,MAAM,CAAC,CAAC,OAAO,MAAM,IAAI,CAAC,gBAAgB,CAAC,UAAU,CAAC,CAAC;YAC1E,KAAK,aAAa,CAAC,MAAM,CAAC,CAAC,OAAO,MAAM,IAAI,CAAC,WAAW,CAAC,UAAU,CAAC,UAAU,CAAC,CAAC;YAChF,KAAK,aAAa,CAAC,KAAK,CAAC,CAAC,OAAO,MAAM,IAAI,CAAC,WAAW,CAAC,SAAS,CAAC,UAAU,CAAC,CAAC;YAC9E,KAAK,aAAa,CAAC,OAAO,CAAC,CAAC,OAAO,MAAM,IAAI,CAAC,WAAW,CAAC,WAAW,CAAC,UAAU,CAAC,CAAC;QACtF,CAAC;QAED,IAAI,IAAI,KAAK,aAAa,CAAC,KAAK,EAAE,CAAC;YAC/B,MAAM,SAAS,GAAG,MAAM,IAAI,CAAC,WAAW,CAAC,UAAU,CAAC,UAAU,CAAC,CAAC;YAChE,MAAM,WAAW,GAAG,MAAM,IAAI,CAAC,WAAW,CAAC,UAAU,CAAC,UAAU,CAAC,CAAC;YAElE,MAAM,WAAW,GAAoB,EAAE,CAAC;YACxC,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,WAAW,EAAE,CAAC,EAAE,EAAE,CAAC;gBACnC,MAAM,KAAK,GAAG,MAAM,IAAI,CAAC,cAAc,CAAC,SAAS,EAAE,UAAU,CAAC,CAAC;gBAC/D,WAAW,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;YAC5B,CAAC;YACD,OAAO,WAAW,CAAC;QACvB,CAAC;QAED,MAAM,IAAI,6BAA6B,CAAC,IAAI,CAAC,CAAC;IAClD,CAAC;IAES,KAAK,CAAC,gBAAgB,CAAC,MAA+B;QAC5D,MAAM,UAAU,GAAG,cAAc,CAAC,iBAAiB,CAAC,MAAM,CAAC,CAAC;QAC5D,MAAM,MAAM,GAAG,MAAM,CAAC,MAAM,IAAI,CAAC,WAAW,CAAC,UAAU,CAAC,UAAU,CAAC,CAAC,CAAC;QAErE,MAAM,UAAU,GAAG,sBAAsB,CAAC,KAAK,GAAG,MAAM,CAAC;QACzD,MAAM,WAAW,GAAG,MAAM,IAAI,CAAC,WAAW,CAAC,aAAa,CAAC,UAAU,EAAE,UAAU,CAAC,CAAC;QAEjF,OAAO,MAAM,CAAC,YAAY,CAAC,GAAG,WAAW,CAAC,CAAC;IAC/C,CAAC;IAES,KAAK,CAAC,cAAc,CAAC,UAA0B;QACrD,MAAM,aAAa,GAAG,UAAU,CAAC,MAAM,CAAC;QAExC,MAAM,WAAW,GAAG,MAAM,IAAI,CAAC,WAAW,CAAC,UAAU,CAAC,UAAU,CAAC,CAAC;QAClE,MAAM,eAAe,GAAG,MAAM,CAAC,MAAM,IAAI,CAAC,WAAW,CAAC,UAAU,CAAC,UAAU,CAAC,CAAC,CAAC;QAE9E,MAAM,QAAQ,GAA2B,EAAE,CAAC;QAE5C,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,eAAe,EAAE,CAAC,EAAE,EAAE,CAAC;YACvC,MAAM,SAAS,GAAG,MAAM,IAAI,CAAC,gBAAgB,CAAC,UAAU,CAAC,CAAC;YAC1D,MAAM,SAAS,GAAG,MAAM,IAAI,CAAC,WAAW,CAAC,UAAU,CAAC,UAAU,CAAC,CAAC;YAChE,QAAQ,CAAC,SAAS,CAAC,GAAG,MAAM,IAAI,CAAC,cAAc,CAAC,SAAS,EAAE,UAAU,CAAC,CAAC;QAC3E,CAAC;QAED,OAAO;YACH,WAAW,EAAE,cAAc,CAAC,gBAAgB,CAAC,WAAW,CAAC;YACzD,QAAQ,EAAE,QAAQ;YAClB,UAAU,EAAE,UAAU,CAAC,MAAM,GAAG,aAAa;SAChD,CAAC;IACN,CAAC;IAEO,KAAK,CAAC,eAAe,CAAC,WAA4B,EAAE,UAA0B;QAClF,MAAM,aAAa,GAAG,UAAU,CAAC,MAAM,CAAC;QACxC,MAAM,UAAU,GAAqB,EAAE,CAAC;QAExC,KAAK,IAAI,CAAC,GAAG,EAAE,EAAE,CAAC,GAAG,MAAM,CAAC,WAAW,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC;YAC5C,MAAM,IAAI,GAAG,MAAM,IAAI,CAAC,gBAAgB,CAAC,UAAU,CAAC,CAAC;YACrD,MAAM,gBAAgB,GAAG,MAAM,IAAI,CAAC,WAAW,CAAC,UAAU,CAAC,UAAU,CAAC,CAAC;YACvE,MAAM,UAAU,GAAwB,EAAE,CAAC;YAE3C,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,gBAAgB,EAAE,CAAC,EAAE,EAAE,CAAC;gBACxC,MAAM,SAAS,GAAG,MAAM,IAAI,CAAC,WAAW,CAAC,UAAU,CAAC,UAAU,CAAC,CAAC;gBAChE,UAAU,CAAC,IAAI,CAAC,cAAc,CAAC,gBAAgB,CAAC,SAAS,CAAC,CAAC,CAAC;YAChE,CAAC;YAED,MAAM,QAAQ,GAAG,MAAM,IAAI,CAAC,WAAW,CAAC,UAAU,CAAC,UAAU,CAAC,CAAC;YAC/D,MAAM,MAAM,GAAG,MAAM,IAAI,CAAC,WAAW,CAAC,UAAU,CAAC,UAAU,CAAC,CAAC;YAE7D,UAAU,CAAC,IAAI,CAAC;gBACZ,IAAI;gBACJ,UAAU;gBACV,QAAQ,EAAE,QAAoB;gBAC9B,MAAM,EAAE,cAAc,CAAC,gBAAgB,CAAC,MAAM,CAAC;aAClD,CAAC,CAAC;QACP,CAAC;QAED,OAAO;YACH,UAAU;YACV,cAAc,EAAE,UAAU,CAAC,MAAM,GAAG,aAAa;SACpD,CAAC;IACN,CAAC;CACJ"}
@@ -0,0 +1,3 @@
1
+ import { GgufV2Parser } from "./GgufV2Parser.js";
2
+ export declare class GgufV3Parser extends GgufV2Parser {
3
+ }