node-llama-cpp 3.0.0-beta.15 → 3.0.0-beta.17
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.
- package/dist/ChatWrapper.d.ts +1 -15
- package/dist/ChatWrapper.js +19 -29
- package/dist/ChatWrapper.js.map +1 -1
- package/dist/bindings/AddonTypes.d.ts +5 -1
- package/dist/bindings/Llama.d.ts +7 -0
- package/dist/bindings/Llama.js +12 -1
- package/dist/bindings/Llama.js.map +1 -1
- package/dist/bindings/getLlama.d.ts +2 -2
- package/dist/bindings/getLlama.js +1 -1
- package/dist/bindings/getLlama.js.map +1 -1
- package/dist/bindings/utils/cloneLlamaCppRepo.js +23 -23
- package/dist/bindings/utils/cloneLlamaCppRepo.js.map +1 -1
- package/dist/chatWrappers/FunctionaryChatWrapper.js +39 -40
- package/dist/chatWrappers/FunctionaryChatWrapper.js.map +1 -1
- package/dist/chatWrappers/{LlamaChatWrapper.d.ts → Llama2ChatWrapper.d.ts} +1 -1
- package/dist/chatWrappers/{LlamaChatWrapper.js → Llama2ChatWrapper.js} +3 -3
- package/dist/chatWrappers/Llama2ChatWrapper.js.map +1 -0
- package/dist/chatWrappers/Llama3ChatWrapper.d.ts +31 -0
- package/dist/chatWrappers/Llama3ChatWrapper.js +129 -0
- package/dist/chatWrappers/Llama3ChatWrapper.js.map +1 -0
- package/dist/chatWrappers/generic/JinjaTemplateChatWrapper.d.ts +2 -2
- package/dist/chatWrappers/generic/JinjaTemplateChatWrapper.js +6 -2
- package/dist/chatWrappers/generic/JinjaTemplateChatWrapper.js.map +1 -1
- package/dist/chatWrappers/generic/TemplateChatWrapper.d.ts +2 -2
- package/dist/chatWrappers/generic/TemplateChatWrapper.js.map +1 -1
- package/dist/chatWrappers/utils/ChatModelFunctionsDocumentationGenerator.d.ts +42 -0
- package/dist/chatWrappers/utils/ChatModelFunctionsDocumentationGenerator.js +82 -0
- package/dist/chatWrappers/utils/ChatModelFunctionsDocumentationGenerator.js.map +1 -0
- package/dist/chatWrappers/utils/isJinjaTemplateEquivalentToSpecializedChatWrapper.js +1 -1
- package/dist/chatWrappers/utils/isJinjaTemplateEquivalentToSpecializedChatWrapper.js.map +1 -1
- package/dist/chatWrappers/utils/resolveChatWrapper.d.ts +6 -4
- package/dist/chatWrappers/utils/resolveChatWrapper.js +26 -18
- package/dist/chatWrappers/utils/resolveChatWrapper.js.map +1 -1
- package/dist/cli/commands/BuildCommand.js +1 -1
- package/dist/cli/commands/BuildCommand.js.map +1 -1
- package/dist/cli/commands/ChatCommand.d.ts +4 -1
- package/dist/cli/commands/ChatCommand.js +93 -81
- package/dist/cli/commands/ChatCommand.js.map +1 -1
- package/dist/cli/commands/CompleteCommand.d.ts +4 -1
- package/dist/cli/commands/CompleteCommand.js +82 -61
- package/dist/cli/commands/CompleteCommand.js.map +1 -1
- package/dist/cli/commands/DebugCommand.js +4 -4
- package/dist/cli/commands/DownloadCommand.js +3 -4
- package/dist/cli/commands/DownloadCommand.js.map +1 -1
- package/dist/cli/commands/InfillCommand.d.ts +4 -1
- package/dist/cli/commands/InfillCommand.js +84 -65
- package/dist/cli/commands/InfillCommand.js.map +1 -1
- package/dist/cli/commands/inspect/commands/InspectGgufCommand.d.ts +2 -1
- package/dist/cli/commands/inspect/commands/InspectGgufCommand.js +20 -7
- package/dist/cli/commands/inspect/commands/InspectGgufCommand.js.map +1 -1
- package/dist/cli/commands/inspect/commands/InspectGpuCommand.js +42 -4
- package/dist/cli/commands/inspect/commands/InspectGpuCommand.js.map +1 -1
- package/dist/cli/commands/inspect/commands/InspectMeasureCommand.d.ts +4 -1
- package/dist/cli/commands/inspect/commands/InspectMeasureCommand.js +69 -36
- package/dist/cli/commands/inspect/commands/InspectMeasureCommand.js.map +1 -1
- package/dist/cli/recommendedModels.d.ts +2 -0
- package/dist/cli/recommendedModels.js +350 -0
- package/dist/cli/recommendedModels.js.map +1 -0
- package/dist/cli/utils/ConsoleInteraction.d.ts +23 -0
- package/dist/cli/utils/ConsoleInteraction.js +122 -0
- package/dist/cli/utils/ConsoleInteraction.js.map +1 -0
- package/dist/cli/utils/ConsoleTable.js +5 -5
- package/dist/cli/utils/basicChooseFromListConsoleInteraction.d.ts +13 -0
- package/dist/cli/utils/basicChooseFromListConsoleInteraction.js +111 -0
- package/dist/cli/utils/basicChooseFromListConsoleInteraction.js.map +1 -0
- package/dist/cli/utils/consolePromptQuestion.d.ts +5 -0
- package/dist/cli/utils/consolePromptQuestion.js +80 -0
- package/dist/cli/utils/consolePromptQuestion.js.map +1 -0
- package/dist/cli/utils/getReadablePath.d.ts +1 -0
- package/dist/cli/utils/getReadablePath.js +14 -0
- package/dist/cli/utils/getReadablePath.js.map +1 -0
- package/dist/cli/utils/printCommonInfoLines.js +5 -4
- package/dist/cli/utils/printCommonInfoLines.js.map +1 -1
- package/dist/cli/utils/printInfoLine.d.ts +4 -2
- package/dist/cli/utils/printInfoLine.js +15 -6
- package/dist/cli/utils/printInfoLine.js.map +1 -1
- package/dist/cli/utils/resolveCommandGgufPath.d.ts +2 -1
- package/dist/cli/utils/resolveCommandGgufPath.js +551 -2
- package/dist/cli/utils/resolveCommandGgufPath.js.map +1 -1
- package/dist/cli/utils/resolveHeaderFlag.d.ts +1 -0
- package/dist/cli/utils/resolveHeaderFlag.js +21 -0
- package/dist/cli/utils/resolveHeaderFlag.js.map +1 -0
- package/dist/cli/utils/resolveModelRecommendationFileOptions.d.ts +19 -0
- package/dist/cli/utils/resolveModelRecommendationFileOptions.js +10 -0
- package/dist/cli/utils/resolveModelRecommendationFileOptions.js.map +1 -0
- package/dist/cli/utils/splitAnsiToLines.d.ts +1 -0
- package/dist/cli/utils/splitAnsiToLines.js +17 -0
- package/dist/cli/utils/splitAnsiToLines.js.map +1 -0
- package/dist/config.d.ts +3 -0
- package/dist/config.js +5 -2
- package/dist/config.js.map +1 -1
- package/dist/consts.d.ts +2 -0
- package/dist/consts.js +8 -0
- package/dist/consts.js.map +1 -1
- package/dist/evaluator/LlamaChat/LlamaChat.d.ts +9 -2
- package/dist/evaluator/LlamaChat/LlamaChat.js +58 -22
- package/dist/evaluator/LlamaChat/LlamaChat.js.map +1 -1
- package/dist/evaluator/LlamaChat/utils/FunctionCallGrammar.js +3 -1
- package/dist/evaluator/LlamaChat/utils/FunctionCallGrammar.js.map +1 -1
- package/dist/evaluator/LlamaChatSession/LlamaChatSession.d.ts +10 -3
- package/dist/evaluator/LlamaChatSession/LlamaChatSession.js +5 -3
- package/dist/evaluator/LlamaChatSession/LlamaChatSession.js.map +1 -1
- package/dist/evaluator/LlamaCompletion.d.ts +11 -4
- package/dist/evaluator/LlamaCompletion.js +16 -15
- package/dist/evaluator/LlamaCompletion.js.map +1 -1
- package/dist/evaluator/LlamaContext/LlamaContext.d.ts +13 -23
- package/dist/evaluator/LlamaContext/LlamaContext.js +47 -60
- package/dist/evaluator/LlamaContext/LlamaContext.js.map +1 -1
- package/dist/evaluator/LlamaEmbeddingContext.js +1 -1
- package/dist/evaluator/LlamaEmbeddingContext.js.map +1 -1
- package/dist/evaluator/LlamaGrammar.d.ts +3 -2
- package/dist/evaluator/LlamaGrammar.js +3 -2
- package/dist/evaluator/LlamaGrammar.js.map +1 -1
- package/dist/evaluator/LlamaModel.d.ts +21 -18
- package/dist/evaluator/LlamaModel.js +34 -205
- package/dist/evaluator/LlamaModel.js.map +1 -1
- package/dist/evaluator/TokenBias.d.ts +22 -0
- package/dist/evaluator/TokenBias.js +33 -0
- package/dist/evaluator/TokenBias.js.map +1 -0
- package/dist/gguf/{GgufInsights.d.ts → insights/GgufInsights.d.ts} +5 -3
- package/dist/gguf/{GgufInsights.js → insights/GgufInsights.js} +46 -35
- package/dist/gguf/insights/GgufInsights.js.map +1 -0
- package/dist/gguf/insights/GgufInsightsConfigurationResolver.d.ts +87 -0
- package/dist/gguf/insights/GgufInsightsConfigurationResolver.js +136 -0
- package/dist/gguf/insights/GgufInsightsConfigurationResolver.js.map +1 -0
- package/dist/gguf/insights/utils/resolveContextContextSizeOption.d.ts +18 -0
- package/dist/gguf/insights/utils/resolveContextContextSizeOption.js +52 -0
- package/dist/gguf/insights/utils/resolveContextContextSizeOption.js.map +1 -0
- package/dist/gguf/insights/utils/resolveModelGpuLayersOption.d.ts +14 -0
- package/dist/gguf/insights/utils/resolveModelGpuLayersOption.js +177 -0
- package/dist/gguf/insights/utils/resolveModelGpuLayersOption.js.map +1 -0
- package/dist/gguf/insights/utils/scoreLevels.d.ts +5 -0
- package/dist/gguf/insights/utils/scoreLevels.js +16 -0
- package/dist/gguf/insights/utils/scoreLevels.js.map +1 -0
- package/dist/gguf/readGgufFileInfo.js +2 -1
- package/dist/gguf/readGgufFileInfo.js.map +1 -1
- package/dist/gguf/types/GgufMetadataTypes.d.ts +4 -0
- package/dist/gguf/types/GgufMetadataTypes.js.map +1 -1
- package/dist/index.d.ts +7 -4
- package/dist/index.js +5 -2
- package/dist/index.js.map +1 -1
- package/dist/types.d.ts +21 -1
- package/dist/types.js.map +1 -1
- package/dist/utils/LlamaText.d.ts +2 -1
- package/dist/utils/LlamaText.js +20 -0
- package/dist/utils/LlamaText.js.map +1 -1
- package/dist/utils/StopGenerationDetector.js +19 -18
- package/dist/utils/StopGenerationDetector.js.map +1 -1
- package/dist/utils/TokenStreamRegulator.d.ts +4 -2
- package/dist/utils/TokenStreamRegulator.js +22 -4
- package/dist/utils/TokenStreamRegulator.js.map +1 -1
- package/dist/utils/getConsoleLogPrefix.js +1 -1
- package/dist/utils/getReadableContextSize.d.ts +1 -0
- package/dist/utils/getReadableContextSize.js +7 -0
- package/dist/utils/getReadableContextSize.js.map +1 -0
- package/dist/utils/getTypeScriptTypeStringForGbnfJsonSchema.js +15 -11
- package/dist/utils/getTypeScriptTypeStringForGbnfJsonSchema.js.map +1 -1
- package/dist/utils/isToken.d.ts +2 -0
- package/dist/utils/isToken.js +4 -0
- package/dist/utils/isToken.js.map +1 -0
- package/dist/utils/isUrl.d.ts +1 -0
- package/dist/utils/isUrl.js +15 -0
- package/dist/utils/isUrl.js.map +1 -0
- package/dist/utils/tokenizeInput.d.ts +1 -1
- package/dist/utils/tokenizeInput.js +3 -0
- package/dist/utils/tokenizeInput.js.map +1 -1
- package/dist/utils/utilTypes.d.ts +3 -0
- package/dist/utils/utilTypes.js +2 -0
- package/dist/utils/utilTypes.js.map +1 -0
- package/dist/utils/withOra.d.ts +1 -0
- package/dist/utils/withOra.js +13 -7
- package/dist/utils/withOra.js.map +1 -1
- package/dist/utils/withProgressLog.d.ts +23 -0
- package/dist/utils/withProgressLog.js +211 -0
- package/dist/utils/withProgressLog.js.map +1 -0
- package/dist/utils/withStatusLogs.js +1 -1
- package/dist/utils/withStatusLogs.js.map +1 -1
- package/llama/addon.cpp +90 -10
- package/llama/binariesGithubRelease.json +1 -1
- package/llama/gitRelease.bundle +0 -0
- package/llama/grammars/README.md +10 -0
- package/llama/llama.cpp.info.json +1 -1
- package/llamaBins/linux-arm64/_nlcBuildMetadata.json +1 -1
- package/llamaBins/linux-arm64/llama-addon.node +0 -0
- package/llamaBins/linux-armv7l/_nlcBuildMetadata.json +1 -1
- package/llamaBins/linux-armv7l/llama-addon.node +0 -0
- package/llamaBins/linux-x64/_nlcBuildMetadata.json +1 -1
- package/llamaBins/linux-x64/llama-addon.node +0 -0
- package/llamaBins/linux-x64-cuda/_nlcBuildMetadata.json +1 -1
- package/llamaBins/linux-x64-cuda/llama-addon.node +0 -0
- package/llamaBins/linux-x64-vulkan/_nlcBuildMetadata.json +1 -1
- package/llamaBins/linux-x64-vulkan/llama-addon.node +0 -0
- package/llamaBins/mac-arm64-metal/_nlcBuildMetadata.json +1 -1
- package/llamaBins/mac-arm64-metal/default.metallib +0 -0
- package/llamaBins/mac-arm64-metal/llama-addon.node +0 -0
- package/llamaBins/mac-x64/_nlcBuildMetadata.json +1 -1
- package/llamaBins/mac-x64/llama-addon.node +0 -0
- package/llamaBins/win-arm64/_nlcBuildMetadata.json +1 -1
- package/llamaBins/win-arm64/llama-addon.node +0 -0
- package/llamaBins/win-x64/_nlcBuildMetadata.json +1 -1
- package/llamaBins/win-x64/llama-addon.node +0 -0
- package/llamaBins/win-x64-cuda/_nlcBuildMetadata.json +1 -1
- package/llamaBins/win-x64-cuda/llama-addon.node +0 -0
- package/llamaBins/win-x64-vulkan/_nlcBuildMetadata.json +1 -1
- package/llamaBins/win-x64-vulkan/llama-addon.node +0 -0
- package/package.json +9 -7
- package/dist/chatWrappers/LlamaChatWrapper.js.map +0 -1
- package/dist/gguf/GgufInsights.js.map +0 -1
|
@@ -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,oCAAoC,CAAC,EACxC,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,GAAG,EAAE,CAAC,CAAC,EAAC,KAAK,EAAE,EAAE,CAAC,QAAQ,EAAE,EAAE,IAAI,EAAE,EAAE,CAAC,OAAO,EAAE,EAAC,CAAC,CAAC,EAClE,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,SAAS,GAAG,YAAY,EAAE,CAAC;QACjC,MAAM,QAAQ,GAAG,WAAW,EAAE,CAAC;QAC/B,MAAM,iBAAiB,GAAG,IAAI,CAAC,qBAAqB,CAAC,MAAM,EAAE;YACzD,YAAY,EAAE,GAAG,EAAE,CAAC,SAAS;YAC7B,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,IAAI,CAAC,yBAAyB,CAAC,MAAM,EAAE;YAC/D,YAAY,EAAE,GAAG,EAAE,CAAC,CAAC,EAAC,KAAK,EAAE,SAAS,CAAC,KAAK,EAAE,IAAI,EAAE,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,SAAS,CAAC,IAAI,GAAG,2BAA2B,CAAC,OAAO,CAAC,EAAC,CAAC;YACvH,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;YACzB,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,qBAAqB,CAAC,SAAyC,EAAE,EACpE,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,yBAAyB,CAAC,WAA+C,EAAE,EAC9E,cAAc,EACd,SAAS,EACT,SAAS,GAAG,0BAA0B,EAAE,EACxC,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,EAU7B;QACG,OAAO,+BAA+B,CAAC;YACnC,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(): {
|
|
12
|
+
total: number;
|
|
13
|
+
free: number;
|
|
14
|
+
};
|
|
15
|
+
llamaGpu: BuildGpu;
|
|
16
|
+
ignoreMemorySafetyChecks?: boolean;
|
|
17
|
+
isEmbeddingContext?: boolean;
|
|
18
|
+
}): number;
|
|
@@ -0,0 +1,52 @@
|
|
|
1
|
+
import { minAllowedContextSizeInCalculations } from "../../../config.js";
|
|
2
|
+
import { getDefaultContextBatchSize, getDefaultModelContextSize } from "../../../evaluator/LlamaContext/LlamaContext.js";
|
|
3
|
+
export 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 = 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 = 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
|
+
for (let testContextSize = maxContextSize; testContextSize >= minContextSize; testContextSize--) {
|
|
36
|
+
const contextVram = modelFileInsights.estimateContextResourceRequirements({
|
|
37
|
+
contextSize: testContextSize,
|
|
38
|
+
batchSize: batchSize ?? getDefaultContextBatchSize({ contextSize: testContextSize, sequences }),
|
|
39
|
+
modelGpuLayers: modelGpuLayers,
|
|
40
|
+
sequences,
|
|
41
|
+
isEmbeddingContext
|
|
42
|
+
}).gpuVram;
|
|
43
|
+
if (contextVram <= freeVram)
|
|
44
|
+
return testContextSize;
|
|
45
|
+
}
|
|
46
|
+
if (ignoreMemorySafetyChecks)
|
|
47
|
+
return minContextSize;
|
|
48
|
+
throw new Error(`The available VRAM is too small to fit the context size of ${maxContextSize}${sequences > 1 ? ` with ${sequences} sequences` : ""}`);
|
|
49
|
+
}
|
|
50
|
+
throw new Error(`Invalid context size: "${contextSize}"`);
|
|
51
|
+
}
|
|
52
|
+
//# 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,UAAU,+BAA+B,CAAC,EAC5C,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;QACjC,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,YAAY,EAAE,CAAC;QACjC,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;KAC9B;SAAM,IAAI,WAAW,KAAK,MAAM,IAAI,OAAO,WAAW,KAAK,QAAQ,EAAE;QAClE,IAAI,QAAQ,KAAK,KAAK;YAClB,OAAO,qBAAqB,CAAC;QAEjC,MAAM,SAAS,GAAG,YAAY,EAAE,CAAC;QAEjC,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,KAAK,IAAI,eAAe,GAAG,cAAc,EAAE,eAAe,IAAI,cAAc,EAAE,eAAe,EAAE,EAAE;YAC7F,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;gBACvB,OAAO,eAAe,CAAC;SAC9B;QAED,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;KACzJ;IAED,MAAM,IAAI,KAAK,CAAC,0BAA0B,WAAW,GAAG,CAAC,CAAC;AAC9D,CAAC"}
|
|
@@ -0,0 +1,14 @@
|
|
|
1
|
+
import { LlamaModelOptions } from "../../../evaluator/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(): {
|
|
8
|
+
total: number;
|
|
9
|
+
free: number;
|
|
10
|
+
};
|
|
11
|
+
llamaVramPaddingSize: number;
|
|
12
|
+
llamaGpu: BuildGpu;
|
|
13
|
+
llamaSupportsGpuOffloading: boolean;
|
|
14
|
+
}): 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 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 = 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 = 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,UAAU,2BAA2B,CAAC,SAAyC,EAAE,EACnF,YAAY,EAAE,wBAAwB,GAAG,KAAK,EAAE,YAAY,EAAE,oBAAoB,EAClF,QAAQ,EAAE,0BAA0B,EAIvC;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;QACtD,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,YAAY,EAAE,CAAC;QACjC,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;KAC5B;SAAM,IAAI,SAAS,KAAK,MAAM,IAAI,OAAO,SAAS,KAAK,QAAQ,EAAE;QAC9D,IAAI,QAAQ,KAAK,KAAK;YAClB,OAAO,CAAC,CAAC;QAEb,MAAM,SAAS,GAAG,YAAY,EAAE,CAAC;QACjC,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;YAC5E,QAAQ,IAAI,oBAAoB,GAAG,sCAAsC,CAAC;YAE1E,IAAI,QAAQ,GAAG,CAAC;gBACZ,QAAQ,GAAG,CAAC,CAAC;SACpB;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;KACnC;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;gBACxD,MAAM;oBACF,SAAS,EAAE,MAAM;iBACpB,CAAC;aACL;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;QACtD,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;KACL;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;QACtG,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;KACT;IAED,OAAO,IAAI,CAAC;AAChB,CAAC"}
|
|
@@ -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;QACpC,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;KAC7D;IAED,OAAO,GAAG,CAAC;AACf,CAAC"}
|
|
@@ -1,3 +1,4 @@
|
|
|
1
|
+
import { isUrl } from "../utils/isUrl.js";
|
|
1
2
|
import { parseGguf } from "./parser/parseGguf.js";
|
|
2
3
|
import { GgufNetworkFetchFileReader } from "./fileReaders/GgufNetworkFetchFileReader.js";
|
|
3
4
|
import { GgufFsFileReader } from "./fileReaders/GgufFsFileReader.js";
|
|
@@ -9,7 +10,7 @@ import { normalizeGgufDownloadUrl } from "./utils/normalizeGgufDownloadUrl.js";
|
|
|
9
10
|
*/
|
|
10
11
|
export async function readGgufFileInfo(pathOrUrl, { readTensorInfo = true, sourceType, ignoreKeys = [], logWarnings = true, fetchRetryOptions = ggufDefaultFetchRetryOptions, fetchHeaders = {}, signal } = {}) {
|
|
11
12
|
function createFileReader() {
|
|
12
|
-
if (sourceType === "network" || (sourceType == null && (pathOrUrl
|
|
13
|
+
if (sourceType === "network" || (sourceType == null && isUrl(pathOrUrl))) {
|
|
13
14
|
return new GgufNetworkFetchFileReader({
|
|
14
15
|
url: normalizeGgufDownloadUrl(pathOrUrl),
|
|
15
16
|
retryOptions: fetchRetryOptions,
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"readGgufFileInfo.js","sourceRoot":"","sources":["../../src/gguf/readGgufFileInfo.ts"],"names":[],"mappings":"AACA,OAAO,EAAC,SAAS,EAAC,MAAM,uBAAuB,CAAC;AAChD,OAAO,EAAC,0BAA0B,EAAC,MAAM,6CAA6C,CAAC;AACvF,OAAO,EAAC,gBAAgB,EAAC,MAAM,mCAAmC,CAAC;AACnE,OAAO,EAAC,4BAA4B,EAAC,MAAM,aAAa,CAAC;AACzD,OAAO,EAAC,wBAAwB,EAAC,MAAM,qCAAqC,CAAC;AAG7E;;;GAGG;AACH,MAAM,CAAC,KAAK,UAAU,gBAAgB,CAAC,SAAiB,EAAE,EACtD,cAAc,GAAG,IAAI,EACrB,UAAU,EACV,UAAU,GAAG,EAAE,EACf,WAAW,GAAG,IAAI,EAClB,iBAAiB,GAAG,4BAA4B,EAChD,YAAY,GAAG,EAAE,EACjB,MAAM,KA8BN,EAAE;IACF,SAAS,gBAAgB;QACrB,IAAI,UAAU,KAAK,SAAS,IAAI,CAAC,UAAU,IAAI,IAAI,IAAI,
|
|
1
|
+
{"version":3,"file":"readGgufFileInfo.js","sourceRoot":"","sources":["../../src/gguf/readGgufFileInfo.ts"],"names":[],"mappings":"AACA,OAAO,EAAC,KAAK,EAAC,MAAM,mBAAmB,CAAC;AACxC,OAAO,EAAC,SAAS,EAAC,MAAM,uBAAuB,CAAC;AAChD,OAAO,EAAC,0BAA0B,EAAC,MAAM,6CAA6C,CAAC;AACvF,OAAO,EAAC,gBAAgB,EAAC,MAAM,mCAAmC,CAAC;AACnE,OAAO,EAAC,4BAA4B,EAAC,MAAM,aAAa,CAAC;AACzD,OAAO,EAAC,wBAAwB,EAAC,MAAM,qCAAqC,CAAC;AAG7E;;;GAGG;AACH,MAAM,CAAC,KAAK,UAAU,gBAAgB,CAAC,SAAiB,EAAE,EACtD,cAAc,GAAG,IAAI,EACrB,UAAU,EACV,UAAU,GAAG,EAAE,EACf,WAAW,GAAG,IAAI,EAClB,iBAAiB,GAAG,4BAA4B,EAChD,YAAY,GAAG,EAAE,EACjB,MAAM,KA8BN,EAAE;IACF,SAAS,gBAAgB;QACrB,IAAI,UAAU,KAAK,SAAS,IAAI,CAAC,UAAU,IAAI,IAAI,IAAI,KAAK,CAAC,SAAS,CAAC,CAAC,EAAE;YACtE,OAAO,IAAI,0BAA0B,CAAC;gBAClC,GAAG,EAAE,wBAAwB,CAAC,SAAS,CAAC;gBACxC,YAAY,EAAE,iBAAiB;gBAC/B,OAAO,EAAE,YAAY;gBACrB,MAAM;aACT,CAAC,CAAC;SACN;aAAM,IAAI,UAAU,KAAK,YAAY,IAAI,UAAU,IAAI,IAAI,EAAE;YAC1D,OAAO,IAAI,gBAAgB,CAAC;gBACxB,QAAQ,EAAE,SAAS;gBACnB,MAAM;aACT,CAAC,CAAC;SACN;QAED,KAAK,CAAC,UAA0B,CAAC,CAAC;QAClC,MAAM,IAAI,KAAK,CAAC,2BAA2B,UAAU,EAAE,CAAC,CAAC;IAC7D,CAAC;IAED,MAAM,UAAU,GAAG,gBAAgB,EAAE,CAAC;IACtC,OAAO,MAAM,SAAS,CAAC;QACnB,UAAU;QACV,UAAU;QACV,cAAc;QACd,WAAW;KACd,CAAC,CAAC;AACP,CAAC"}
|
|
@@ -166,6 +166,10 @@ export type GgufMetadataTokenizer = {
|
|
|
166
166
|
readonly add_eos_token?: boolean;
|
|
167
167
|
readonly add_space_prefix?: boolean;
|
|
168
168
|
readonly added_tokens?: readonly string[];
|
|
169
|
+
readonly prefix_token_id?: number;
|
|
170
|
+
readonly suffix_token_id?: number;
|
|
171
|
+
readonly middle_token_id?: number;
|
|
172
|
+
readonly eot_token_id?: number;
|
|
169
173
|
};
|
|
170
174
|
readonly huggingface?: {
|
|
171
175
|
readonly json?: string;
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"GgufMetadataTypes.js","sourceRoot":"","sources":["../../../src/gguf/types/GgufMetadataTypes.ts"],"names":[],"mappings":"AAAA,MAAM,CAAN,IAAkB,oBA4BjB;AA5BD,WAAkB,oBAAoB;IAClC,uCAAe,CAAA;IACf,yCAAiB,CAAA;IACjB,qCAAa,CAAA;IACb,qCAAa,CAAA;IACb,2CAAmB,CAAA;IACnB,mCAAW,CAAA;IACX,6CAAqB,CAAA;IACrB,+CAAuB,CAAA;IACvB,+CAAuB,CAAA;IACvB,yCAAiB,CAAA;IACjB,qCAAa,CAAA;IACb,gDAAwB,CAAA;IACxB,uCAAe,CAAA;IACf,6CAAqB,CAAA;IACrB,qCAAa,CAAA;IACb,uCAAe,CAAA;IACf,qCAAa,CAAA;IACb,uCAAe,CAAA;IACf,+CAAuB,CAAA;IACvB,uCAAe,CAAA;IACf,+CAAuB,CAAA;IACvB,2CAAmB,CAAA;IACnB,uCAAe,CAAA;IACf,iDAAyB,CAAA;IACzB,uCAAe,CAAA;IACf,8CAAsB,CAAA;IACtB,qCAAa,CAAA;AACjB,CAAC,EA5BiB,oBAAoB,KAApB,oBAAoB,QA4BrC;AA+BD,yEAAyE;AACzE,MAAM,CAAN,IAAY,YAgCX;AAhCD,WAAY,YAAY;IACpB,qDAAW,CAAA;IACX,2DAAc,CAAA;IACd,6DAAe,CAAA;IACf,6DAAe,CAAA;IACf,+EAAwB,CAAA;IACxB,6DAAe,CAAA;IACf,6DAAe,CAAA;IACf,6DAAe,CAAA;IACf,6DAAe,CAAA;IACf,6DAAe,CAAA;IACf,8DAAgB,CAAA;IAChB,kEAAkB,CAAA;IAClB,kEAAkB,CAAA;IAClB,kEAAkB,CAAA;IAClB,kEAAkB,CAAA;IAClB,kEAAkB,CAAA;IAClB,kEAAkB,CAAA;IAClB,kEAAkB,CAAA;IAClB,8DAAgB,CAAA;IAChB,oEAAmB,CAAA;IACnB,kEAAkB,CAAA;IAClB,kEAAkB,CAAA;IAClB,kEAAkB,CAAA;IAClB,oEAAmB,CAAA;IACnB,gEAAiB,CAAA;IACjB,kEAAkB,CAAA;IAClB,gEAAiB,CAAA;IACjB,gEAAiB,CAAA;IACjB,gEAAiB,CAAA;IACjB,gEAAiB,CAAA;IACjB,kEAAkB,CAAA;AACtB,CAAC,EAhCW,YAAY,KAAZ,YAAY,QAgCvB;AAyED,MAAM,CAAN,IAAkB,8BAQjB;AARD,WAAkB,8BAA8B;IAC5C,6FAAa,CAAA;IACb,uFAAU,CAAA;IACV,yFAAW,CAAA;IACX,yFAAW,CAAA;IACX,iGAAe,CAAA;IACf,uFAAU,CAAA;IACV,mFAAQ,CAAA;AACZ,CAAC,EARiB,8BAA8B,KAA9B,8BAA8B,QAQ/C;
|
|
1
|
+
{"version":3,"file":"GgufMetadataTypes.js","sourceRoot":"","sources":["../../../src/gguf/types/GgufMetadataTypes.ts"],"names":[],"mappings":"AAAA,MAAM,CAAN,IAAkB,oBA4BjB;AA5BD,WAAkB,oBAAoB;IAClC,uCAAe,CAAA;IACf,yCAAiB,CAAA;IACjB,qCAAa,CAAA;IACb,qCAAa,CAAA;IACb,2CAAmB,CAAA;IACnB,mCAAW,CAAA;IACX,6CAAqB,CAAA;IACrB,+CAAuB,CAAA;IACvB,+CAAuB,CAAA;IACvB,yCAAiB,CAAA;IACjB,qCAAa,CAAA;IACb,gDAAwB,CAAA;IACxB,uCAAe,CAAA;IACf,6CAAqB,CAAA;IACrB,qCAAa,CAAA;IACb,uCAAe,CAAA;IACf,qCAAa,CAAA;IACb,uCAAe,CAAA;IACf,+CAAuB,CAAA;IACvB,uCAAe,CAAA;IACf,+CAAuB,CAAA;IACvB,2CAAmB,CAAA;IACnB,uCAAe,CAAA;IACf,iDAAyB,CAAA;IACzB,uCAAe,CAAA;IACf,8CAAsB,CAAA;IACtB,qCAAa,CAAA;AACjB,CAAC,EA5BiB,oBAAoB,KAApB,oBAAoB,QA4BrC;AA+BD,yEAAyE;AACzE,MAAM,CAAN,IAAY,YAgCX;AAhCD,WAAY,YAAY;IACpB,qDAAW,CAAA;IACX,2DAAc,CAAA;IACd,6DAAe,CAAA;IACf,6DAAe,CAAA;IACf,+EAAwB,CAAA;IACxB,6DAAe,CAAA;IACf,6DAAe,CAAA;IACf,6DAAe,CAAA;IACf,6DAAe,CAAA;IACf,6DAAe,CAAA;IACf,8DAAgB,CAAA;IAChB,kEAAkB,CAAA;IAClB,kEAAkB,CAAA;IAClB,kEAAkB,CAAA;IAClB,kEAAkB,CAAA;IAClB,kEAAkB,CAAA;IAClB,kEAAkB,CAAA;IAClB,kEAAkB,CAAA;IAClB,8DAAgB,CAAA;IAChB,oEAAmB,CAAA;IACnB,kEAAkB,CAAA;IAClB,kEAAkB,CAAA;IAClB,kEAAkB,CAAA;IAClB,oEAAmB,CAAA;IACnB,gEAAiB,CAAA;IACjB,kEAAkB,CAAA;IAClB,gEAAiB,CAAA;IACjB,gEAAiB,CAAA;IACjB,gEAAiB,CAAA;IACjB,gEAAiB,CAAA;IACjB,kEAAkB,CAAA;AACtB,CAAC,EAhCW,YAAY,KAAZ,YAAY,QAgCvB;AAyED,MAAM,CAAN,IAAkB,8BAQjB;AARD,WAAkB,8BAA8B;IAC5C,6FAAa,CAAA;IACb,uFAAU,CAAA;IACV,yFAAW,CAAA;IACX,yFAAW,CAAA;IACX,iGAAe,CAAA;IACf,uFAAU,CAAA;IACV,mFAAQ,CAAA;AACZ,CAAC,EARiB,8BAA8B,KAA9B,8BAA8B,QAQ/C;AA8BD,MAAM,CAAN,IAAkB,mCAKjB;AALD,WAAkB,mCAAmC;IACjD,4GAAgB,CAAA;IAChB,6FAAQ,CAAA;IACR,6FAAQ,CAAA;IACR,2FAAO,CAAA;AACX,CAAC,EALiB,mCAAmC,KAAnC,mCAAmC,QAKpD;AAuND,MAAM,UAAU,gCAAgC,CAC5C,QAAsB,EAAE,IAAO;IAE/B,OAAO,QAAQ,EAAE,OAAO,EAAE,YAAY,KAAK,IAAI,CAAC;AACpD,CAAC"}
|
package/dist/index.d.ts
CHANGED
|
@@ -11,6 +11,7 @@ import { LlamaGrammarEvaluationState, LlamaGrammarEvaluationStateOptions } from
|
|
|
11
11
|
import { LlamaContext, LlamaContextSequence } from "./evaluator/LlamaContext/LlamaContext.js";
|
|
12
12
|
import { LlamaEmbeddingContext, type LlamaEmbeddingContextOptions, LlamaEmbedding, type LlamaEmbeddingJSON } from "./evaluator/LlamaEmbeddingContext.js";
|
|
13
13
|
import { type LlamaContextOptions, type BatchingOptions, type LlamaContextSequenceRepeatPenalty, type CustomBatchingDispatchSchedule, type CustomBatchingPrioritizationStrategy, type BatchItem, type PrioritizedBatchItem, type ContextShiftOptions, type ContextTokensDeleteRange, type EvaluationPriority } from "./evaluator/LlamaContext/types.js";
|
|
14
|
+
import { TokenBias } from "./evaluator/TokenBias.js";
|
|
14
15
|
import { LlamaChatSession, type LlamaChatSessionOptions, type LlamaChatSessionContextShiftOptions, type LLamaChatPromptOptions, type LlamaChatSessionRepeatPenalty } from "./evaluator/LlamaChatSession/LlamaChatSession.js";
|
|
15
16
|
import { defineChatSessionFunction } from "./evaluator/LlamaChatSession/utils/defineChatSessionFunction.js";
|
|
16
17
|
import { LlamaChat, type LlamaChatOptions, type LLamaChatGenerateResponseOptions, type LLamaChatContextShiftOptions, type LlamaChatResponse, type LlamaChatResponseFunctionCall } from "./evaluator/LlamaChat/LlamaChat.js";
|
|
@@ -18,9 +19,10 @@ import { LlamaCompletion, type LlamaCompletionOptions, type LlamaCompletionGener
|
|
|
18
19
|
import { TokenMeter, type TokenMeterState } from "./evaluator/TokenMeter.js";
|
|
19
20
|
import { UnsupportedError } from "./utils/UnsupportedError.js";
|
|
20
21
|
import { InsufficientMemoryError } from "./utils/InsufficientMemoryError.js";
|
|
21
|
-
import { ChatWrapper
|
|
22
|
+
import { ChatWrapper } from "./ChatWrapper.js";
|
|
22
23
|
import { EmptyChatWrapper } from "./chatWrappers/EmptyChatWrapper.js";
|
|
23
|
-
import {
|
|
24
|
+
import { Llama3ChatWrapper } from "./chatWrappers/Llama3ChatWrapper.js";
|
|
25
|
+
import { Llama2ChatWrapper } from "./chatWrappers/Llama2ChatWrapper.js";
|
|
24
26
|
import { GeneralChatWrapper } from "./chatWrappers/GeneralChatWrapper.js";
|
|
25
27
|
import { ChatMLChatWrapper } from "./chatWrappers/ChatMLChatWrapper.js";
|
|
26
28
|
import { FalconChatWrapper } from "./chatWrappers/FalconChatWrapper.js";
|
|
@@ -30,13 +32,14 @@ import { GemmaChatWrapper } from "./chatWrappers/GemmaChatWrapper.js";
|
|
|
30
32
|
import { TemplateChatWrapper, type TemplateChatWrapperOptions } from "./chatWrappers/generic/TemplateChatWrapper.js";
|
|
31
33
|
import { JinjaTemplateChatWrapper, type JinjaTemplateChatWrapperOptions } from "./chatWrappers/generic/JinjaTemplateChatWrapper.js";
|
|
32
34
|
import { resolvableChatWrapperTypeNames, type ResolvableChatWrapperTypeName, specializedChatWrapperTypeNames, type SpecializedChatWrapperTypeName, templateChatWrapperTypeNames, type TemplateChatWrapperTypeName, resolveChatWrapper, type ResolveChatWrapperOptions } from "./chatWrappers/utils/resolveChatWrapper.js";
|
|
35
|
+
import { ChatModelFunctionsDocumentationGenerator } from "./chatWrappers/utils/ChatModelFunctionsDocumentationGenerator.js";
|
|
33
36
|
import { LlamaText, SpecialTokensText, SpecialToken, isLlamaText, tokenizeText, type LlamaTextJSON, type LlamaTextJSONValue, type LlamaTextSpecialTokensTextJSON, type LlamaTextSpecialTokenJSON } from "./utils/LlamaText.js";
|
|
34
37
|
import { appendUserMessageToChatHistory } from "./utils/appendUserMessageToChatHistory.js";
|
|
35
38
|
import { getModuleVersion } from "./utils/getModuleVersion.js";
|
|
36
39
|
import { readGgufFileInfo } from "./gguf/readGgufFileInfo.js";
|
|
37
|
-
import { type ChatHistoryItem, type ChatModelFunctionCall, type ChatModelFunctions, type ChatModelResponse, type ChatSessionModelFunction, type ChatSessionModelFunctions, type ChatSystemMessage, type ChatUserMessage, type Token, isChatModelResponseFunctionCall, type LLamaContextualRepeatPenalty } from "./types.js";
|
|
40
|
+
import { type ChatHistoryItem, type ChatModelFunctionCall, type ChatModelFunctions, type ChatModelResponse, type ChatSessionModelFunction, type ChatSessionModelFunctions, type ChatSystemMessage, type ChatUserMessage, type Token, isChatModelResponseFunctionCall, type LLamaContextualRepeatPenalty, type ChatWrapperSettings } from "./types.js";
|
|
38
41
|
import { type GbnfJsonArraySchema, type GbnfJsonBasicSchema, type GbnfJsonConstSchema, type GbnfJsonEnumSchema, type GbnfJsonObjectSchema, type GbnfJsonOneOfSchema, type GbnfJsonSchema, type GbnfJsonSchemaImmutableType, type GbnfJsonSchemaToType } from "./utils/gbnfJson/types.js";
|
|
39
42
|
import { type GgufFileInfo } from "./gguf/types/GgufFileInfoTypes.js";
|
|
40
43
|
import { type GgufMetadata, type GgufMetadataLlmToType, GgufArchitectureType, GgufFileType, GgufMetadataTokenizerTokenType, GgufMetadataArchitecturePoolingType, type GgufMetadataGeneral, type GgufMetadataTokenizer, type GgufMetadataDefaultArchitectureType, type GgufMetadataLlmLLaMA, type GgufMetadataMPT, type GgufMetadataGPTNeoX, type GgufMetadataGPTJ, type GgufMetadataGPT2, type GgufMetadataBloom, type GgufMetadataFalcon, type GgufMetadataMamba, type GgufMetadataRWKV, isGgufMetadataOfArchitectureType } from "./gguf/types/GgufMetadataTypes.js";
|
|
41
44
|
import { GgmlType, type GgufTensorInfo } from "./gguf/types/GgufTensorInfoTypes.js";
|
|
42
|
-
export { Llama, getLlama, type LlamaOptions, LlamaLogLevel, NoBinaryFoundError, LlamaModel, LlamaModelTokens, LlamaModelInfillTokens, type LlamaModelOptions, LlamaGrammar, type LlamaGrammarOptions, LlamaJsonSchemaGrammar, LlamaJsonSchemaValidationError, LlamaGrammarEvaluationState, type LlamaGrammarEvaluationStateOptions, LlamaContext, LlamaContextSequence, type LlamaContextOptions, type BatchingOptions, type CustomBatchingDispatchSchedule, type CustomBatchingPrioritizationStrategy, type BatchItem, type PrioritizedBatchItem, type ContextShiftOptions, type ContextTokensDeleteRange, type EvaluationPriority, type LlamaContextSequenceRepeatPenalty, LlamaEmbeddingContext, type LlamaEmbeddingContextOptions, LlamaEmbedding, type LlamaEmbeddingJSON, LlamaChatSession, defineChatSessionFunction, type LlamaChatSessionOptions, type LlamaChatSessionContextShiftOptions, type LLamaChatPromptOptions, type LlamaChatSessionRepeatPenalty, LlamaChat, type LlamaChatOptions, type LLamaChatGenerateResponseOptions, type LLamaChatContextShiftOptions, type LLamaContextualRepeatPenalty, type LlamaChatResponse, type LlamaChatResponseFunctionCall, LlamaCompletion, type LlamaCompletionOptions, type LlamaCompletionGenerationOptions, type LlamaInfillGenerationOptions, TokenMeter, type TokenMeterState, UnsupportedError, InsufficientMemoryError, DisposedError, ChatWrapper, type ChatWrapperSettings, EmptyChatWrapper,
|
|
45
|
+
export { Llama, getLlama, type LlamaOptions, LlamaLogLevel, NoBinaryFoundError, LlamaModel, LlamaModelTokens, LlamaModelInfillTokens, type LlamaModelOptions, LlamaGrammar, type LlamaGrammarOptions, LlamaJsonSchemaGrammar, LlamaJsonSchemaValidationError, LlamaGrammarEvaluationState, type LlamaGrammarEvaluationStateOptions, LlamaContext, LlamaContextSequence, type LlamaContextOptions, type BatchingOptions, type CustomBatchingDispatchSchedule, type CustomBatchingPrioritizationStrategy, type BatchItem, type PrioritizedBatchItem, type ContextShiftOptions, type ContextTokensDeleteRange, type EvaluationPriority, type LlamaContextSequenceRepeatPenalty, TokenBias, LlamaEmbeddingContext, type LlamaEmbeddingContextOptions, LlamaEmbedding, type LlamaEmbeddingJSON, LlamaChatSession, defineChatSessionFunction, type LlamaChatSessionOptions, type LlamaChatSessionContextShiftOptions, type LLamaChatPromptOptions, type LlamaChatSessionRepeatPenalty, LlamaChat, type LlamaChatOptions, type LLamaChatGenerateResponseOptions, type LLamaChatContextShiftOptions, type LLamaContextualRepeatPenalty, type LlamaChatResponse, type LlamaChatResponseFunctionCall, LlamaCompletion, type LlamaCompletionOptions, type LlamaCompletionGenerationOptions, type LlamaInfillGenerationOptions, TokenMeter, type TokenMeterState, UnsupportedError, InsufficientMemoryError, DisposedError, ChatWrapper, type ChatWrapperSettings, EmptyChatWrapper, Llama3ChatWrapper, Llama2ChatWrapper, GeneralChatWrapper, ChatMLChatWrapper, FalconChatWrapper, AlpacaChatWrapper, FunctionaryChatWrapper, GemmaChatWrapper, TemplateChatWrapper, type TemplateChatWrapperOptions, JinjaTemplateChatWrapper, type JinjaTemplateChatWrapperOptions, resolveChatWrapper, type ResolveChatWrapperOptions, resolvableChatWrapperTypeNames, type ResolvableChatWrapperTypeName, specializedChatWrapperTypeNames, type SpecializedChatWrapperTypeName, templateChatWrapperTypeNames, type TemplateChatWrapperTypeName, ChatModelFunctionsDocumentationGenerator, LlamaText, SpecialTokensText, SpecialToken, isLlamaText, tokenizeText, type LlamaTextJSON, type LlamaTextJSONValue, type LlamaTextSpecialTokensTextJSON, type LlamaTextSpecialTokenJSON, appendUserMessageToChatHistory, getModuleVersion, type ChatHistoryItem, type ChatModelFunctionCall, type ChatModelFunctions, type ChatModelResponse, type ChatSessionModelFunction, type ChatSessionModelFunctions, type ChatSystemMessage, type ChatUserMessage, type Token, isChatModelResponseFunctionCall, type GbnfJsonSchema, type GbnfJsonSchemaToType, type GbnfJsonSchemaImmutableType, type GbnfJsonBasicSchema, type GbnfJsonConstSchema, type GbnfJsonEnumSchema, type GbnfJsonOneOfSchema, type GbnfJsonObjectSchema, type GbnfJsonArraySchema, LlamaVocabularyType, LlamaLogLevelGreaterThan, LlamaLogLevelGreaterThanOrEqual, readGgufFileInfo, type GgufFileInfo, type GgufMetadata, type GgufTensorInfo, type GgufMetadataLlmToType, GgufArchitectureType, GgufFileType, GgufMetadataTokenizerTokenType, GgufMetadataArchitecturePoolingType, type GgufMetadataGeneral, type GgufMetadataTokenizer, type GgufMetadataDefaultArchitectureType, type GgufMetadataLlmLLaMA, type GgufMetadataMPT, type GgufMetadataGPTNeoX, type GgufMetadataGPTJ, type GgufMetadataGPT2, type GgufMetadataBloom, type GgufMetadataFalcon, type GgufMetadataMamba, type GgufMetadataRWKV, GgmlType, isGgufMetadataOfArchitectureType };
|
package/dist/index.js
CHANGED
|
@@ -10,6 +10,7 @@ import { LlamaJsonSchemaValidationError } from "./utils/gbnfJson/utils/validateO
|
|
|
10
10
|
import { LlamaGrammarEvaluationState } from "./evaluator/LlamaGrammarEvaluationState.js";
|
|
11
11
|
import { LlamaContext, LlamaContextSequence } from "./evaluator/LlamaContext/LlamaContext.js";
|
|
12
12
|
import { LlamaEmbeddingContext, LlamaEmbedding } from "./evaluator/LlamaEmbeddingContext.js";
|
|
13
|
+
import { TokenBias } from "./evaluator/TokenBias.js";
|
|
13
14
|
import { LlamaChatSession } from "./evaluator/LlamaChatSession/LlamaChatSession.js";
|
|
14
15
|
import { defineChatSessionFunction } from "./evaluator/LlamaChatSession/utils/defineChatSessionFunction.js";
|
|
15
16
|
import { LlamaChat } from "./evaluator/LlamaChat/LlamaChat.js";
|
|
@@ -19,7 +20,8 @@ import { UnsupportedError } from "./utils/UnsupportedError.js";
|
|
|
19
20
|
import { InsufficientMemoryError } from "./utils/InsufficientMemoryError.js";
|
|
20
21
|
import { ChatWrapper } from "./ChatWrapper.js";
|
|
21
22
|
import { EmptyChatWrapper } from "./chatWrappers/EmptyChatWrapper.js";
|
|
22
|
-
import {
|
|
23
|
+
import { Llama3ChatWrapper } from "./chatWrappers/Llama3ChatWrapper.js";
|
|
24
|
+
import { Llama2ChatWrapper } from "./chatWrappers/Llama2ChatWrapper.js";
|
|
23
25
|
import { GeneralChatWrapper } from "./chatWrappers/GeneralChatWrapper.js";
|
|
24
26
|
import { ChatMLChatWrapper } from "./chatWrappers/ChatMLChatWrapper.js";
|
|
25
27
|
import { FalconChatWrapper } from "./chatWrappers/FalconChatWrapper.js";
|
|
@@ -29,6 +31,7 @@ import { GemmaChatWrapper } from "./chatWrappers/GemmaChatWrapper.js";
|
|
|
29
31
|
import { TemplateChatWrapper } from "./chatWrappers/generic/TemplateChatWrapper.js";
|
|
30
32
|
import { JinjaTemplateChatWrapper } from "./chatWrappers/generic/JinjaTemplateChatWrapper.js";
|
|
31
33
|
import { resolvableChatWrapperTypeNames, specializedChatWrapperTypeNames, templateChatWrapperTypeNames, resolveChatWrapper } from "./chatWrappers/utils/resolveChatWrapper.js";
|
|
34
|
+
import { ChatModelFunctionsDocumentationGenerator } from "./chatWrappers/utils/ChatModelFunctionsDocumentationGenerator.js";
|
|
32
35
|
import { LlamaText, SpecialTokensText, SpecialToken, isLlamaText, tokenizeText } from "./utils/LlamaText.js";
|
|
33
36
|
import { appendUserMessageToChatHistory } from "./utils/appendUserMessageToChatHistory.js";
|
|
34
37
|
import { getModuleVersion } from "./utils/getModuleVersion.js";
|
|
@@ -36,5 +39,5 @@ import { readGgufFileInfo } from "./gguf/readGgufFileInfo.js";
|
|
|
36
39
|
import { isChatModelResponseFunctionCall } from "./types.js";
|
|
37
40
|
import { GgufArchitectureType, GgufFileType, GgufMetadataTokenizerTokenType, GgufMetadataArchitecturePoolingType, isGgufMetadataOfArchitectureType } from "./gguf/types/GgufMetadataTypes.js";
|
|
38
41
|
import { GgmlType } from "./gguf/types/GgufTensorInfoTypes.js";
|
|
39
|
-
export { Llama, getLlama, LlamaLogLevel, NoBinaryFoundError, LlamaModel, LlamaModelTokens, LlamaModelInfillTokens, LlamaGrammar, LlamaJsonSchemaGrammar, LlamaJsonSchemaValidationError, LlamaGrammarEvaluationState, LlamaContext, LlamaContextSequence, LlamaEmbeddingContext, LlamaEmbedding, LlamaChatSession, defineChatSessionFunction, LlamaChat, LlamaCompletion, TokenMeter, UnsupportedError, InsufficientMemoryError, DisposedError, ChatWrapper, EmptyChatWrapper,
|
|
42
|
+
export { Llama, getLlama, LlamaLogLevel, NoBinaryFoundError, LlamaModel, LlamaModelTokens, LlamaModelInfillTokens, LlamaGrammar, LlamaJsonSchemaGrammar, LlamaJsonSchemaValidationError, LlamaGrammarEvaluationState, LlamaContext, LlamaContextSequence, TokenBias, LlamaEmbeddingContext, LlamaEmbedding, LlamaChatSession, defineChatSessionFunction, LlamaChat, LlamaCompletion, TokenMeter, UnsupportedError, InsufficientMemoryError, DisposedError, ChatWrapper, EmptyChatWrapper, Llama3ChatWrapper, Llama2ChatWrapper, GeneralChatWrapper, ChatMLChatWrapper, FalconChatWrapper, AlpacaChatWrapper, FunctionaryChatWrapper, GemmaChatWrapper, TemplateChatWrapper, JinjaTemplateChatWrapper, resolveChatWrapper, resolvableChatWrapperTypeNames, specializedChatWrapperTypeNames, templateChatWrapperTypeNames, ChatModelFunctionsDocumentationGenerator, LlamaText, SpecialTokensText, SpecialToken, isLlamaText, tokenizeText, appendUserMessageToChatHistory, getModuleVersion, isChatModelResponseFunctionCall, LlamaVocabularyType, LlamaLogLevelGreaterThan, LlamaLogLevelGreaterThanOrEqual, readGgufFileInfo, GgufArchitectureType, GgufFileType, GgufMetadataTokenizerTokenType, GgufMetadataArchitecturePoolingType, GgmlType, isGgufMetadataOfArchitectureType };
|
|
40
43
|
//# sourceMappingURL=index.js.map
|
package/dist/index.js.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"index.js","sourceRoot":"","sources":["../src/index.ts"],"names":[],"mappings":"AAAA,OAAO,EAAC,aAAa,EAAC,MAAM,iBAAiB,CAAC;AAC9C,OAAO,EAAC,KAAK,EAAC,MAAM,qBAAqB,CAAC;AAC1C,OAAO,EAAC,QAAQ,EAAe,MAAM,wBAAwB,CAAC;AAC9D,OAAO,EAAC,kBAAkB,EAAC,MAAM,wCAAwC,CAAC;AAC1E,OAAO,EAAC,aAAa,EAAE,wBAAwB,EAAE,+BAA+B,EAAE,mBAAmB,EAAC,MAAM,qBAAqB,CAAC;AAClI,OAAO,EAAC,UAAU,EAAE,sBAAsB,EAA0B,gBAAgB,EAAC,MAAM,2BAA2B,CAAC;AACvH,OAAO,EAAC,YAAY,EAA2B,MAAM,6BAA6B,CAAC;AACnF,OAAO,EAAC,sBAAsB,EAAC,MAAM,uCAAuC,CAAC;AAC7E,OAAO,EAAC,8BAA8B,EAAC,MAAM,2DAA2D,CAAC;AACzG,OAAO,EAAC,2BAA2B,EAAqC,MAAM,4CAA4C,CAAC;AAC3H,OAAO,EAAC,YAAY,EAAE,oBAAoB,EAAC,MAAM,0CAA0C,CAAC;AAC5F,OAAO,EACH,qBAAqB,EAAqC,cAAc,EAC3E,MAAM,sCAAsC,CAAC;AAM9C,OAAO,EACH,gBAAgB,EAEnB,MAAM,kDAAkD,CAAC;AAC1D,OAAO,EAAC,yBAAyB,EAAC,MAAM,iEAAiE,CAAC;AAC1G,OAAO,EACH,SAAS,EAEZ,MAAM,oCAAoC,CAAC;AAC5C,OAAO,EACH,eAAe,EAClB,MAAM,gCAAgC,CAAC;AACxC,OAAO,EAAC,UAAU,EAAuB,MAAM,2BAA2B,CAAC;AAC3E,OAAO,EAAC,gBAAgB,EAAC,MAAM,6BAA6B,CAAC;AAC7D,OAAO,EAAC,uBAAuB,EAAC,MAAM,oCAAoC,CAAC;AAC3E,OAAO,EAAC,WAAW,
|
|
1
|
+
{"version":3,"file":"index.js","sourceRoot":"","sources":["../src/index.ts"],"names":[],"mappings":"AAAA,OAAO,EAAC,aAAa,EAAC,MAAM,iBAAiB,CAAC;AAC9C,OAAO,EAAC,KAAK,EAAC,MAAM,qBAAqB,CAAC;AAC1C,OAAO,EAAC,QAAQ,EAAe,MAAM,wBAAwB,CAAC;AAC9D,OAAO,EAAC,kBAAkB,EAAC,MAAM,wCAAwC,CAAC;AAC1E,OAAO,EAAC,aAAa,EAAE,wBAAwB,EAAE,+BAA+B,EAAE,mBAAmB,EAAC,MAAM,qBAAqB,CAAC;AAClI,OAAO,EAAC,UAAU,EAAE,sBAAsB,EAA0B,gBAAgB,EAAC,MAAM,2BAA2B,CAAC;AACvH,OAAO,EAAC,YAAY,EAA2B,MAAM,6BAA6B,CAAC;AACnF,OAAO,EAAC,sBAAsB,EAAC,MAAM,uCAAuC,CAAC;AAC7E,OAAO,EAAC,8BAA8B,EAAC,MAAM,2DAA2D,CAAC;AACzG,OAAO,EAAC,2BAA2B,EAAqC,MAAM,4CAA4C,CAAC;AAC3H,OAAO,EAAC,YAAY,EAAE,oBAAoB,EAAC,MAAM,0CAA0C,CAAC;AAC5F,OAAO,EACH,qBAAqB,EAAqC,cAAc,EAC3E,MAAM,sCAAsC,CAAC;AAM9C,OAAO,EAAC,SAAS,EAAC,MAAM,0BAA0B,CAAC;AACnD,OAAO,EACH,gBAAgB,EAEnB,MAAM,kDAAkD,CAAC;AAC1D,OAAO,EAAC,yBAAyB,EAAC,MAAM,iEAAiE,CAAC;AAC1G,OAAO,EACH,SAAS,EAEZ,MAAM,oCAAoC,CAAC;AAC5C,OAAO,EACH,eAAe,EAClB,MAAM,gCAAgC,CAAC;AACxC,OAAO,EAAC,UAAU,EAAuB,MAAM,2BAA2B,CAAC;AAC3E,OAAO,EAAC,gBAAgB,EAAC,MAAM,6BAA6B,CAAC;AAC7D,OAAO,EAAC,uBAAuB,EAAC,MAAM,oCAAoC,CAAC;AAC3E,OAAO,EAAC,WAAW,EAAC,MAAM,kBAAkB,CAAC;AAC7C,OAAO,EAAC,gBAAgB,EAAC,MAAM,oCAAoC,CAAC;AACpE,OAAO,EAAC,iBAAiB,EAAC,MAAM,qCAAqC,CAAC;AACtE,OAAO,EAAC,iBAAiB,EAAC,MAAM,qCAAqC,CAAC;AACtE,OAAO,EAAC,kBAAkB,EAAC,MAAM,sCAAsC,CAAC;AACxE,OAAO,EAAC,iBAAiB,EAAC,MAAM,qCAAqC,CAAC;AACtE,OAAO,EAAC,iBAAiB,EAAC,MAAM,qCAAqC,CAAC;AACtE,OAAO,EAAC,iBAAiB,EAAC,MAAM,qCAAqC,CAAC;AACtE,OAAO,EAAC,sBAAsB,EAAC,MAAM,0CAA0C,CAAC;AAChF,OAAO,EAAC,gBAAgB,EAAC,MAAM,oCAAoC,CAAC;AACpE,OAAO,EAAC,mBAAmB,EAAkC,MAAM,+CAA+C,CAAC;AACnH,OAAO,EAAC,wBAAwB,EAAuC,MAAM,oDAAoD,CAAC;AAClI,OAAO,EACH,8BAA8B,EAAsC,+BAA+B,EAC9D,4BAA4B,EAAoC,kBAAkB,EAE1H,MAAM,4CAA4C,CAAC;AACpD,OAAO,EAAC,wCAAwC,EAAC,MAAM,kEAAkE,CAAC;AAC1H,OAAO,EACH,SAAS,EAAE,iBAAiB,EAAE,YAAY,EAAE,WAAW,EAAE,YAAY,EAExE,MAAM,sBAAsB,CAAC;AAC9B,OAAO,EAAC,8BAA8B,EAAC,MAAM,2CAA2C,CAAC;AACzF,OAAO,EAAC,gBAAgB,EAAC,MAAM,6BAA6B,CAAC;AAC7D,OAAO,EAAC,gBAAgB,EAAC,MAAM,4BAA4B,CAAC;AAE5D,OAAO,EAGS,+BAA+B,EAC9C,MAAM,YAAY,CAAC;AAMpB,OAAO,EAC4C,oBAAoB,EAAE,YAAY,EAAE,8BAA8B,EACjH,mCAAmC,EAE6D,gCAAgC,EACnI,MAAM,mCAAmC,CAAC;AAC3C,OAAO,EAAC,QAAQ,EAAsB,MAAM,qCAAqC,CAAC;AAGlF,OAAO,EACH,KAAK,EACL,QAAQ,EAER,aAAa,EACb,kBAAkB,EAClB,UAAU,EACV,gBAAgB,EAChB,sBAAsB,EAEtB,YAAY,EAEZ,sBAAsB,EACtB,8BAA8B,EAC9B,2BAA2B,EAE3B,YAAY,EACZ,oBAAoB,EAWpB,SAAS,EACT,qBAAqB,EAErB,cAAc,EAEd,gBAAgB,EAChB,yBAAyB,EAKzB,SAAS,EAOT,eAAe,EAIf,UAAU,EAEV,gBAAgB,EAChB,uBAAuB,EACvB,aAAa,EACb,WAAW,EAEX,gBAAgB,EAChB,iBAAiB,EACjB,iBAAiB,EACjB,kBAAkB,EAClB,iBAAiB,EACjB,iBAAiB,EACjB,iBAAiB,EACjB,sBAAsB,EACtB,gBAAgB,EAChB,mBAAmB,EAEnB,wBAAwB,EAExB,kBAAkB,EAElB,8BAA8B,EAE9B,+BAA+B,EAE/B,4BAA4B,EAE5B,wCAAwC,EACxC,SAAS,EACT,iBAAiB,EACjB,YAAY,EACZ,WAAW,EACX,YAAY,EAKZ,8BAA8B,EAC9B,gBAAgB,EAUhB,+BAA+B,EAU/B,mBAAmB,EACnB,wBAAwB,EACxB,+BAA+B,EAC/B,gBAAgB,EAKhB,oBAAoB,EACpB,YAAY,EACZ,8BAA8B,EAC9B,mCAAmC,EAanC,QAAQ,EACR,gCAAgC,EACnC,CAAC"}
|
package/dist/types.d.ts
CHANGED
|
@@ -3,10 +3,30 @@ import { BuiltinSpecialTokenValue } from "./utils/LlamaText.js";
|
|
|
3
3
|
export type Token = number & {
|
|
4
4
|
__token: never;
|
|
5
5
|
};
|
|
6
|
+
export type Detokenizer = {
|
|
7
|
+
detokenize(tokens: readonly Token[], specialTokens?: boolean): string;
|
|
8
|
+
}["detokenize"];
|
|
6
9
|
export type Tokenizer = {
|
|
7
10
|
tokenize(text: string, specialTokens?: boolean, options?: "trimLeadingSpace"): Token[];
|
|
8
11
|
tokenize(text: BuiltinSpecialTokenValue, specialTokens: "builtin"): Token[];
|
|
9
|
-
}["tokenize"]
|
|
12
|
+
}["tokenize"] & {
|
|
13
|
+
readonly detokenize: Detokenizer;
|
|
14
|
+
isSpecialToken(token: Token): boolean;
|
|
15
|
+
};
|
|
16
|
+
export type ChatWrapperSettings = {
|
|
17
|
+
readonly functions: {
|
|
18
|
+
readonly call: {
|
|
19
|
+
readonly optionalPrefixSpace: boolean;
|
|
20
|
+
readonly prefix: string;
|
|
21
|
+
readonly paramsPrefix: string;
|
|
22
|
+
readonly suffix: string;
|
|
23
|
+
};
|
|
24
|
+
readonly result: {
|
|
25
|
+
readonly prefix: string;
|
|
26
|
+
readonly suffix: string;
|
|
27
|
+
};
|
|
28
|
+
};
|
|
29
|
+
};
|
|
10
30
|
export type ChatHistoryItem = ChatSystemMessage | ChatUserMessage | ChatModelResponse;
|
|
11
31
|
export type ChatSystemMessage = {
|
|
12
32
|
type: "system";
|
package/dist/types.js.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"types.js","sourceRoot":"","sources":["../src/types.ts"],"names":[],"mappings":"
|
|
1
|
+
{"version":3,"file":"types.js","sourceRoot":"","sources":["../src/types.ts"],"names":[],"mappings":"AA0EA,MAAM,UAAU,+BAA+B,CAAC,IAA2C;IACvF,IAAI,OAAO,IAAI,KAAK,QAAQ;QACxB,OAAO,KAAK,CAAC;IAEjB,OAAO,IAAI,CAAC,IAAI,KAAK,cAAc,CAAC;AACxC,CAAC"}
|
|
@@ -35,6 +35,7 @@ export declare class SpecialTokensText {
|
|
|
35
35
|
constructor(value: string);
|
|
36
36
|
toString(): string;
|
|
37
37
|
tokenize(tokenizer: Tokenizer, trimLeadingSpace?: boolean): Token[];
|
|
38
|
+
tokenizeSpecialTokensOnly(tokenizer: Tokenizer): (string | Token)[];
|
|
38
39
|
toJSON(): LlamaTextSpecialTokensTextJSON;
|
|
39
40
|
static fromJSON(json: LlamaTextSpecialTokensTextJSON): SpecialTokensText;
|
|
40
41
|
static isSpecialTokensTextJSON(value: LlamaTextJSONValue): value is LlamaTextSpecialTokensTextJSON;
|
|
@@ -43,7 +44,7 @@ export declare class SpecialTokensText {
|
|
|
43
44
|
*/
|
|
44
45
|
static wrapIf(shouldWrap: boolean, value: string): SpecialTokensText | string;
|
|
45
46
|
}
|
|
46
|
-
export type BuiltinSpecialTokenValue = "BOS" | "EOS" | "NL";
|
|
47
|
+
export type BuiltinSpecialTokenValue = "BOS" | "EOS" | "NL" | "EOT";
|
|
47
48
|
export declare class SpecialToken {
|
|
48
49
|
readonly value: BuiltinSpecialTokenValue;
|
|
49
50
|
constructor(value: BuiltinSpecialTokenValue);
|
package/dist/utils/LlamaText.js
CHANGED
|
@@ -37,6 +37,26 @@ export class SpecialTokensText {
|
|
|
37
37
|
tokenize(tokenizer, trimLeadingSpace = false) {
|
|
38
38
|
return tokenizer(this.value, true, trimLeadingSpace ? "trimLeadingSpace" : undefined);
|
|
39
39
|
}
|
|
40
|
+
tokenizeSpecialTokensOnly(tokenizer) {
|
|
41
|
+
const tokens = this.tokenize(tokenizer, true);
|
|
42
|
+
const res = [];
|
|
43
|
+
let currentText = "";
|
|
44
|
+
for (const token of tokens) {
|
|
45
|
+
if (tokenizer.isSpecialToken(token)) {
|
|
46
|
+
if (currentText !== "") {
|
|
47
|
+
res.push(currentText);
|
|
48
|
+
currentText = "";
|
|
49
|
+
}
|
|
50
|
+
res.push(token);
|
|
51
|
+
}
|
|
52
|
+
else {
|
|
53
|
+
currentText += tokenizer.detokenize([token], false);
|
|
54
|
+
}
|
|
55
|
+
}
|
|
56
|
+
if (currentText !== "")
|
|
57
|
+
res.push(currentText);
|
|
58
|
+
return res;
|
|
59
|
+
}
|
|
40
60
|
toJSON() {
|
|
41
61
|
return {
|
|
42
62
|
type: "specialTokensText",
|