node-llama-cpp 3.15.0 → 3.16.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/dist/bindings/AddonTypes.d.ts +8 -1
- package/dist/bindings/Llama.js +5 -1
- package/dist/bindings/Llama.js.map +1 -1
- package/dist/bindings/getLlama.d.ts +1 -1
- package/dist/bindings/getLlama.js +1 -1
- package/dist/bindings/getLlama.js.map +1 -1
- package/dist/bindings/utils/testBindingBinary.js +2 -3
- package/dist/bindings/utils/testBindingBinary.js.map +1 -1
- package/dist/chatWrappers/generic/utils/extractFunctionCallSettingsFromJinjaTemplate.js +67 -8
- package/dist/chatWrappers/generic/utils/extractFunctionCallSettingsFromJinjaTemplate.js.map +1 -1
- package/dist/chatWrappers/generic/utils/extractSegmentSettingsFromTokenizerAndChatTemplate.js +2 -1
- package/dist/chatWrappers/generic/utils/extractSegmentSettingsFromTokenizerAndChatTemplate.js.map +1 -1
- package/dist/cli/commands/ChatCommand.d.ts +6 -0
- package/dist/cli/commands/ChatCommand.js +66 -3
- package/dist/cli/commands/ChatCommand.js.map +1 -1
- package/dist/cli/commands/CompleteCommand.d.ts +6 -0
- package/dist/cli/commands/CompleteCommand.js +66 -4
- package/dist/cli/commands/CompleteCommand.js.map +1 -1
- package/dist/cli/commands/InfillCommand.d.ts +6 -0
- package/dist/cli/commands/InfillCommand.js +66 -4
- package/dist/cli/commands/InfillCommand.js.map +1 -1
- package/dist/cli/utils/parseXtcArg.d.ts +5 -0
- package/dist/cli/utils/parseXtcArg.js +16 -0
- package/dist/cli/utils/parseXtcArg.js.map +1 -0
- package/dist/config.d.ts +1 -0
- package/dist/config.js +3 -0
- package/dist/config.js.map +1 -1
- package/dist/evaluator/LlamaChat/LlamaChat.d.ts +36 -1
- package/dist/evaluator/LlamaChat/LlamaChat.js +29 -10
- package/dist/evaluator/LlamaChat/LlamaChat.js.map +1 -1
- package/dist/evaluator/LlamaChatSession/LlamaChatSession.d.ts +83 -2
- package/dist/evaluator/LlamaChatSession/LlamaChatSession.js +11 -5
- package/dist/evaluator/LlamaChatSession/LlamaChatSession.js.map +1 -1
- package/dist/evaluator/LlamaChatSession/utils/LlamaChatSessionPromptCompletionEngine.d.ts +2 -0
- package/dist/evaluator/LlamaChatSession/utils/LlamaChatSessionPromptCompletionEngine.js.map +1 -1
- package/dist/evaluator/LlamaCompletion.d.ts +36 -3
- package/dist/evaluator/LlamaCompletion.js +7 -4
- package/dist/evaluator/LlamaCompletion.js.map +1 -1
- package/dist/evaluator/LlamaContext/LlamaContext.js +67 -46
- package/dist/evaluator/LlamaContext/LlamaContext.js.map +1 -1
- package/dist/evaluator/LlamaContext/LlamaSampler.js +8 -0
- package/dist/evaluator/LlamaContext/LlamaSampler.js.map +1 -1
- package/dist/evaluator/LlamaContext/tokenPredictors/DraftSequenceTokenPredictor.d.ts +1 -1
- package/dist/evaluator/LlamaContext/types.d.ts +115 -0
- package/dist/evaluator/LlamaModel/LlamaModel.d.ts +2 -2
- package/dist/evaluator/LlamaModel/LlamaModel.js +1 -14
- package/dist/evaluator/LlamaModel/LlamaModel.js.map +1 -1
- package/dist/gguf/insights/GgufInsights.js +4 -0
- package/dist/gguf/insights/GgufInsights.js.map +1 -1
- package/dist/gguf/types/GgufMetadataTypes.d.ts +5 -0
- package/dist/gguf/types/GgufMetadataTypes.js +5 -0
- package/dist/gguf/types/GgufMetadataTypes.js.map +1 -1
- package/dist/tsconfig.tsbuildinfo +1 -1
- package/dist/types.d.ts +51 -0
- package/dist/types.js.map +1 -1
- package/dist/utils/cmake.js +6 -3
- package/dist/utils/cmake.js.map +1 -1
- package/llama/CMakeLists.txt +1 -0
- package/llama/addon/AddonContext.cpp +19 -5
- package/llama/addon/AddonContext.h +1 -1
- package/llama/addon/AddonModel.cpp +7 -6
- package/llama/addon/AddonModelLora.cpp +1 -3
- package/llama/addon/AddonSampler.cpp +158 -0
- package/llama/addon/AddonSampler.h +13 -1
- package/llama/addon/globals/addonLog.cpp +5 -5
- package/llama/addon/globals/getGpuInfo.cpp +1 -1
- package/llama/binariesGithubRelease.json +1 -1
- package/llama/cmake/addVariantSuffix.cmake +1 -1
- package/llama/gitRelease.bundle +0 -0
- package/llama/gpuInfo/vulkan-gpu-info.cpp +12 -5
- package/llama/llama.cpp.info.json +1 -1
- package/package.json +63 -62
- package/templates/packed/electron-typescript-react.json +1 -1
- package/templates/packed/node-typescript.json +1 -1
|
@@ -1 +1 @@
|
|
|
1
|
-
{"root":["../src/index.ts","../src/ChatWrapper.ts","../src/apiDocsIndex.ts","../src/commands.ts","../src/config.ts","../src/consts.ts","../src/state.ts","../src/types.ts","../src/bindings/AddonTypes.ts","../src/bindings/Llama.ts","../src/bindings/consts.ts","../src/bindings/getLlama.ts","../src/bindings/types.ts","../src/bindings/utils/MemoryOrchestrator.ts","../src/bindings/utils/NoBinaryFoundError.ts","../src/bindings/utils/asyncEvery.ts","../src/bindings/utils/asyncSome.ts","../src/bindings/utils/binariesGithubRelease.ts","../src/bindings/utils/clearAllLocalBuilds.ts","../src/bindings/utils/cloneLlamaCppRepo.ts","../src/bindings/utils/compileLLamaCpp.ts","../src/bindings/utils/detectAvailableComputeLayers.ts","../src/bindings/utils/detectBuildTools.ts","../src/bindings/utils/detectGlibc.ts","../src/bindings/utils/getBestComputeLayersAvailable.ts","../src/bindings/utils/getBuildFolderNameForBuildOptions.ts","../src/bindings/utils/getCanUsePrebuiltBinaries.ts","../src/bindings/utils/getExampleUsageCodeOfGetLlama.ts","../src/bindings/utils/getGpuTypesToUseForOption.ts","../src/bindings/utils/getLinuxDistroInfo.ts","../src/bindings/utils/getLlamaGpuTypes.ts","../src/bindings/utils/getLlamaWithoutBackend.ts","../src/bindings/utils/getPlatform.ts","../src/bindings/utils/getPlatformInfo.ts","../src/bindings/utils/hasBuildingFromSourceDependenciesInstalled.ts","../src/bindings/utils/hasFileInPath.ts","../src/bindings/utils/lastBuildInfo.ts","../src/bindings/utils/logBinaryUsageExampleToConsole.ts","../src/bindings/utils/logDistroInstallInstruction.ts","../src/bindings/utils/resolveActualBindingBinaryPath.ts","../src/bindings/utils/resolveCustomCmakeOptions.ts","../src/bindings/utils/testBindingBinary.ts","../src/bindings/utils/testCmakeBinary.ts","../src/chatWrappers/AlpacaChatWrapper.ts","../src/chatWrappers/ChatMLChatWrapper.ts","../src/chatWrappers/DeepSeekChatWrapper.ts","../src/chatWrappers/EmptyChatWrapper.ts","../src/chatWrappers/FalconChatWrapper.ts","../src/chatWrappers/FunctionaryChatWrapper.ts","../src/chatWrappers/GemmaChatWrapper.ts","../src/chatWrappers/GeneralChatWrapper.ts","../src/chatWrappers/HarmonyChatWrapper.ts","../src/chatWrappers/Llama2ChatWrapper.ts","../src/chatWrappers/Llama3ChatWrapper.ts","../src/chatWrappers/Llama3_1ChatWrapper.ts","../src/chatWrappers/Llama3_2LightweightChatWrapper.ts","../src/chatWrappers/MistralChatWrapper.ts","../src/chatWrappers/QwenChatWrapper.ts","../src/chatWrappers/SeedChatWrapper.ts","../src/chatWrappers/generic/JinjaTemplateChatWrapper.ts","../src/chatWrappers/generic/TemplateChatWrapper.ts","../src/chatWrappers/generic/utils/UniqueIdGenerator.ts","../src/chatWrappers/generic/utils/chatHistoryFunctionCallMessageTemplate.ts","../src/chatWrappers/generic/utils/extractFunctionCallSettingsFromJinjaTemplate.ts","../src/chatWrappers/generic/utils/extractSegmentSettingsFromTokenizerAndChatTemplate.ts","../src/chatWrappers/generic/utils/getFirstValidResult.ts","../src/chatWrappers/generic/utils/squashChatHistoryItems.ts","../src/chatWrappers/generic/utils/templateSegmentOptionsToChatWrapperSettings.ts","../src/chatWrappers/utils/ChatModelFunctionsDocumentationGenerator.ts","../src/chatWrappers/utils/chunkChatItems.ts","../src/chatWrappers/utils/getModelLinageNames.ts","../src/chatWrappers/utils/isJinjaTemplateEquivalentToSpecializedChatWrapper.ts","../src/chatWrappers/utils/isLlama3_2LightweightModel.ts","../src/chatWrappers/utils/jsonDumps.ts","../src/chatWrappers/utils/resolveChatWrapper.ts","../src/cli/cli.ts","../src/cli/projectTemplates.ts","../src/cli/recommendedModels.ts","../src/cli/startCreateCli.ts","../src/cli/commands/ChatCommand.ts","../src/cli/commands/CompleteCommand.ts","../src/cli/commands/DebugCommand.ts","../src/cli/commands/InfillCommand.ts","../src/cli/commands/InitCommand.ts","../src/cli/commands/OnPostInstallCommand.ts","../src/cli/commands/PullCommand.ts","../src/cli/commands/inspect/InspectCommand.ts","../src/cli/commands/inspect/commands/InspectEstimateCommand.ts","../src/cli/commands/inspect/commands/InspectGgufCommand.ts","../src/cli/commands/inspect/commands/InspectGpuCommand.ts","../src/cli/commands/inspect/commands/InspectMeasureCommand.ts","../src/cli/commands/source/SourceCommand.ts","../src/cli/commands/source/commands/BuildCommand.ts","../src/cli/commands/source/commands/ClearCommand.ts","../src/cli/commands/source/commands/DownloadCommand.ts","../src/cli/utils/ConsoleInteraction.ts","../src/cli/utils/ConsoleTable.ts","../src/cli/utils/basicChooseFromListConsoleInteraction.ts","../src/cli/utils/consolePromptQuestion.ts","../src/cli/utils/getReadablePath.ts","../src/cli/utils/interactivelyAskForModel.ts","../src/cli/utils/isRunningUnderRosetta.ts","../src/cli/utils/logUsedGpuTypeOption.ts","../src/cli/utils/printCommonInfoLines.ts","../src/cli/utils/printInfoLine.ts","../src/cli/utils/printModelDestination.ts","../src/cli/utils/projectTemplates.ts","../src/cli/utils/renderModelCompatibilityPercentageWithColors.ts","../src/cli/utils/resolveCommandGgufPath.ts","../src/cli/utils/resolveHeaderFlag.ts","../src/cli/utils/resolveModelRecommendationFileOptions.ts","../src/cli/utils/splitAnsiToLines.ts","../src/cli/utils/toBytes.ts","../src/cli/utils/withCliCommandDescriptionDocsUrl.ts","../src/evaluator/LlamaCompletion.ts","../src/evaluator/LlamaEmbedding.ts","../src/evaluator/LlamaEmbeddingContext.ts","../src/evaluator/LlamaGrammar.ts","../src/evaluator/LlamaGrammarEvaluationState.ts","../src/evaluator/LlamaJsonSchemaGrammar.ts","../src/evaluator/LlamaRankingContext.ts","../src/evaluator/TokenBias.ts","../src/evaluator/TokenMeter.ts","../src/evaluator/LlamaChat/LlamaChat.ts","../src/evaluator/LlamaChat/utils/FunctionCallNameGrammar.ts","../src/evaluator/LlamaChat/utils/FunctionCallParamsGrammar.ts","../src/evaluator/LlamaChat/utils/LlamaFunctionCallValidationError.ts","../src/evaluator/LlamaChat/utils/contextShiftStrategies/eraseFirstResponseAndKeepFirstSystemChatContextShiftStrategy.ts","../src/evaluator/LlamaChatSession/LlamaChatSession.ts","../src/evaluator/LlamaChatSession/utils/LlamaChatSessionPromptCompletionEngine.ts","../src/evaluator/LlamaChatSession/utils/defineChatSessionFunction.ts","../src/evaluator/LlamaContext/LlamaContext.ts","../src/evaluator/LlamaContext/LlamaSampler.ts","../src/evaluator/LlamaContext/TokenPredictor.ts","../src/evaluator/LlamaContext/types.ts","../src/evaluator/LlamaContext/tokenPredictors/DraftSequenceTokenPredictor.ts","../src/evaluator/LlamaContext/tokenPredictors/InputLookupTokenPredictor.ts","../src/evaluator/LlamaContext/utils/padSafeContextSize.ts","../src/evaluator/LlamaContext/utils/resolveBatchItemsPrioritizationStrategy.ts","../src/evaluator/LlamaContext/utils/batchItemsPrioritizationStrategies/firstInFirstOutStrategy.ts","../src/evaluator/LlamaContext/utils/batchItemsPrioritizationStrategies/maximumParallelismStrategy.ts","../src/evaluator/LlamaModel/LlamaModel.ts","../src/evaluator/LlamaModel/utils/TokenAttributes.ts","../src/evaluator/utils/chunkDocument.ts","../src/gguf/consts.ts","../src/gguf/readGgufFileInfo.ts","../src/gguf/errors/InvalidGgufMagicError.ts","../src/gguf/errors/UnsupportedGgufValueTypeError.ts","../src/gguf/fileReaders/GgufFileReader.ts","../src/gguf/fileReaders/GgufFsFileReader.ts","../src/gguf/fileReaders/GgufNetworkFetchFileReader.ts","../src/gguf/insights/GgufInsights.ts","../src/gguf/insights/GgufInsightsConfigurationResolver.ts","../src/gguf/insights/GgufInsightsTokens.ts","../src/gguf/insights/utils/getRamUsageFromUnifiedVram.ts","../src/gguf/insights/utils/resolveContextContextSizeOption.ts","../src/gguf/insights/utils/resolveModelGpuLayersOption.ts","../src/gguf/insights/utils/scoreLevels.ts","../src/gguf/parser/GgufV2Parser.ts","../src/gguf/parser/GgufV3Parser.ts","../src/gguf/parser/parseGguf.ts","../src/gguf/types/GgufFileInfoTypes.ts","../src/gguf/types/GgufMetadataTypes.ts","../src/gguf/types/GgufTensorInfoTypes.ts","../src/gguf/utils/GgufReadOffset.ts","../src/gguf/utils/convertMetadataKeyValueRecordToNestedObject.ts","../src/gguf/utils/getGgufFileTypeName.ts","../src/gguf/utils/getGgufMetadataArchitectureData.ts","../src/gguf/utils/getGgufMetadataKeyValue.ts","../src/gguf/utils/ggufQuantNames.ts","../src/gguf/utils/normalizeGgufDownloadUrl.ts","../src/gguf/utils/resolveBinarySplitGgufPartUrls.ts","../src/gguf/utils/resolveSplitGgufParts.ts","../src/utils/DisposeGuard.ts","../src/utils/InsufficientMemoryError.ts","../src/utils/LlamaText.ts","../src/utils/LruCache.ts","../src/utils/OpenAIFormat.ts","../src/utils/OverridesObject.ts","../src/utils/ReplHistory.ts","../src/utils/StopGenerationDetector.ts","../src/utils/ThreadsSplitter.ts","../src/utils/TokenStreamRegulator.ts","../src/utils/UnsupportedError.ts","../src/utils/appendUserMessageToChatHistory.ts","../src/utils/clearTempFolder.ts","../src/utils/cmake.ts","../src/utils/compareTokens.ts","../src/utils/createModelDownloader.ts","../src/utils/findBestOption.ts","../src/utils/findCharacterRemovalCountToFitChatHistoryInContext.ts","../src/utils/getBuildDefaults.ts","../src/utils/getChatWrapperSegmentDefinition.ts","../src/utils/getConsoleLogPrefix.ts","../src/utils/getGrammarsFolder.ts","../src/utils/getLlamaClasses.ts","../src/utils/getModuleVersion.ts","../src/utils/getQueuedTokensBeforeStopTrigger.ts","../src/utils/getReadableContextSize.ts","../src/utils/getTypeScriptTypeStringForGbnfJsonSchema.ts","../src/utils/gitReleaseBundles.ts","../src/utils/hashString.ts","../src/utils/includesText.ts","../src/utils/isLockfileActive.ts","../src/utils/isToken.ts","../src/utils/isUrl.ts","../src/utils/mergeUnionTypes.ts","../src/utils/modelDownloadEndpoints.ts","../src/utils/modelFileAccessTokens.ts","../src/utils/optionsMatrix.ts","../src/utils/parseModelFileName.ts","../src/utils/parseModelUri.ts","../src/utils/parseTextTemplate.ts","../src/utils/prettyPrintObject.ts","../src/utils/pushAll.ts","../src/utils/removeNullFields.ts","../src/utils/resolveGithubRelease.ts","../src/utils/resolveLastTokens.ts","../src/utils/resolveModelDestination.ts","../src/utils/resolveModelFile.ts","../src/utils/runtime.ts","../src/utils/safeEventCallback.ts","../src/utils/signalSleep.ts","../src/utils/spawnCommand.ts","../src/utils/tokenizeInput.ts","../src/utils/tokenizerUtils.ts","../src/utils/transformPromisable.ts","../src/utils/truncateTextAndRoundToWords.ts","../src/utils/utilTypes.ts","../src/utils/waitForLockfileRelease.ts","../src/utils/withLockfile.ts","../src/utils/withOra.ts","../src/utils/withProgressLog.ts","../src/utils/withStatusLogs.ts","../src/utils/wrapAbortSignal.ts","../src/utils/gbnfJson/GbnfGrammarGenerator.ts","../src/utils/gbnfJson/GbnfTerminal.ts","../src/utils/gbnfJson/getGbnfGrammarForGbnfJsonSchema.ts","../src/utils/gbnfJson/types.ts","../src/utils/gbnfJson/terminals/GbnfAnyJson.ts","../src/utils/gbnfJson/terminals/GbnfArray.ts","../src/utils/gbnfJson/terminals/GbnfBoolean.ts","../src/utils/gbnfJson/terminals/GbnfBooleanValue.ts","../src/utils/gbnfJson/terminals/GbnfCommaWhitespace.ts","../src/utils/gbnfJson/terminals/GbnfFormatString.ts","../src/utils/gbnfJson/terminals/GbnfGrammar.ts","../src/utils/gbnfJson/terminals/GbnfInsideStringChar.ts","../src/utils/gbnfJson/terminals/GbnfNull.ts","../src/utils/gbnfJson/terminals/GbnfNumber.ts","../src/utils/gbnfJson/terminals/GbnfNumberValue.ts","../src/utils/gbnfJson/terminals/GbnfObjectMap.ts","../src/utils/gbnfJson/terminals/GbnfOr.ts","../src/utils/gbnfJson/terminals/GbnfRef.ts","../src/utils/gbnfJson/terminals/GbnfRepetition.ts","../src/utils/gbnfJson/terminals/GbnfString.ts","../src/utils/gbnfJson/terminals/GbnfStringValue.ts","../src/utils/gbnfJson/terminals/GbnfVerbatimText.ts","../src/utils/gbnfJson/terminals/GbnfWhitespace.ts","../src/utils/gbnfJson/terminals/gbnfConsts.ts","../src/utils/gbnfJson/utils/GbnfJsonScopeState.ts","../src/utils/gbnfJson/utils/defsScope.ts","../src/utils/gbnfJson/utils/getGbnfJsonTerminalForGbnfJsonSchema.ts","../src/utils/gbnfJson/utils/getGbnfJsonTerminalForLiteral.ts","../src/utils/gbnfJson/utils/validateObjectAgainstGbnfSchema.ts"],"version":"5.9.2"}
|
|
1
|
+
{"root":["../src/index.ts","../src/ChatWrapper.ts","../src/apiDocsIndex.ts","../src/commands.ts","../src/config.ts","../src/consts.ts","../src/state.ts","../src/types.ts","../src/bindings/AddonTypes.ts","../src/bindings/Llama.ts","../src/bindings/consts.ts","../src/bindings/getLlama.ts","../src/bindings/types.ts","../src/bindings/utils/MemoryOrchestrator.ts","../src/bindings/utils/NoBinaryFoundError.ts","../src/bindings/utils/asyncEvery.ts","../src/bindings/utils/asyncSome.ts","../src/bindings/utils/binariesGithubRelease.ts","../src/bindings/utils/clearAllLocalBuilds.ts","../src/bindings/utils/cloneLlamaCppRepo.ts","../src/bindings/utils/compileLLamaCpp.ts","../src/bindings/utils/detectAvailableComputeLayers.ts","../src/bindings/utils/detectBuildTools.ts","../src/bindings/utils/detectGlibc.ts","../src/bindings/utils/getBestComputeLayersAvailable.ts","../src/bindings/utils/getBuildFolderNameForBuildOptions.ts","../src/bindings/utils/getCanUsePrebuiltBinaries.ts","../src/bindings/utils/getExampleUsageCodeOfGetLlama.ts","../src/bindings/utils/getGpuTypesToUseForOption.ts","../src/bindings/utils/getLinuxDistroInfo.ts","../src/bindings/utils/getLlamaGpuTypes.ts","../src/bindings/utils/getLlamaWithoutBackend.ts","../src/bindings/utils/getPlatform.ts","../src/bindings/utils/getPlatformInfo.ts","../src/bindings/utils/hasBuildingFromSourceDependenciesInstalled.ts","../src/bindings/utils/hasFileInPath.ts","../src/bindings/utils/lastBuildInfo.ts","../src/bindings/utils/logBinaryUsageExampleToConsole.ts","../src/bindings/utils/logDistroInstallInstruction.ts","../src/bindings/utils/resolveActualBindingBinaryPath.ts","../src/bindings/utils/resolveCustomCmakeOptions.ts","../src/bindings/utils/testBindingBinary.ts","../src/bindings/utils/testCmakeBinary.ts","../src/chatWrappers/AlpacaChatWrapper.ts","../src/chatWrappers/ChatMLChatWrapper.ts","../src/chatWrappers/DeepSeekChatWrapper.ts","../src/chatWrappers/EmptyChatWrapper.ts","../src/chatWrappers/FalconChatWrapper.ts","../src/chatWrappers/FunctionaryChatWrapper.ts","../src/chatWrappers/GemmaChatWrapper.ts","../src/chatWrappers/GeneralChatWrapper.ts","../src/chatWrappers/HarmonyChatWrapper.ts","../src/chatWrappers/Llama2ChatWrapper.ts","../src/chatWrappers/Llama3ChatWrapper.ts","../src/chatWrappers/Llama3_1ChatWrapper.ts","../src/chatWrappers/Llama3_2LightweightChatWrapper.ts","../src/chatWrappers/MistralChatWrapper.ts","../src/chatWrappers/QwenChatWrapper.ts","../src/chatWrappers/SeedChatWrapper.ts","../src/chatWrappers/generic/JinjaTemplateChatWrapper.ts","../src/chatWrappers/generic/TemplateChatWrapper.ts","../src/chatWrappers/generic/utils/UniqueIdGenerator.ts","../src/chatWrappers/generic/utils/chatHistoryFunctionCallMessageTemplate.ts","../src/chatWrappers/generic/utils/extractFunctionCallSettingsFromJinjaTemplate.ts","../src/chatWrappers/generic/utils/extractSegmentSettingsFromTokenizerAndChatTemplate.ts","../src/chatWrappers/generic/utils/getFirstValidResult.ts","../src/chatWrappers/generic/utils/squashChatHistoryItems.ts","../src/chatWrappers/generic/utils/templateSegmentOptionsToChatWrapperSettings.ts","../src/chatWrappers/utils/ChatModelFunctionsDocumentationGenerator.ts","../src/chatWrappers/utils/chunkChatItems.ts","../src/chatWrappers/utils/getModelLinageNames.ts","../src/chatWrappers/utils/isJinjaTemplateEquivalentToSpecializedChatWrapper.ts","../src/chatWrappers/utils/isLlama3_2LightweightModel.ts","../src/chatWrappers/utils/jsonDumps.ts","../src/chatWrappers/utils/resolveChatWrapper.ts","../src/cli/cli.ts","../src/cli/projectTemplates.ts","../src/cli/recommendedModels.ts","../src/cli/startCreateCli.ts","../src/cli/commands/ChatCommand.ts","../src/cli/commands/CompleteCommand.ts","../src/cli/commands/DebugCommand.ts","../src/cli/commands/InfillCommand.ts","../src/cli/commands/InitCommand.ts","../src/cli/commands/OnPostInstallCommand.ts","../src/cli/commands/PullCommand.ts","../src/cli/commands/inspect/InspectCommand.ts","../src/cli/commands/inspect/commands/InspectEstimateCommand.ts","../src/cli/commands/inspect/commands/InspectGgufCommand.ts","../src/cli/commands/inspect/commands/InspectGpuCommand.ts","../src/cli/commands/inspect/commands/InspectMeasureCommand.ts","../src/cli/commands/source/SourceCommand.ts","../src/cli/commands/source/commands/BuildCommand.ts","../src/cli/commands/source/commands/ClearCommand.ts","../src/cli/commands/source/commands/DownloadCommand.ts","../src/cli/utils/ConsoleInteraction.ts","../src/cli/utils/ConsoleTable.ts","../src/cli/utils/basicChooseFromListConsoleInteraction.ts","../src/cli/utils/consolePromptQuestion.ts","../src/cli/utils/getReadablePath.ts","../src/cli/utils/interactivelyAskForModel.ts","../src/cli/utils/isRunningUnderRosetta.ts","../src/cli/utils/logUsedGpuTypeOption.ts","../src/cli/utils/parseXtcArg.ts","../src/cli/utils/printCommonInfoLines.ts","../src/cli/utils/printInfoLine.ts","../src/cli/utils/printModelDestination.ts","../src/cli/utils/projectTemplates.ts","../src/cli/utils/renderModelCompatibilityPercentageWithColors.ts","../src/cli/utils/resolveCommandGgufPath.ts","../src/cli/utils/resolveHeaderFlag.ts","../src/cli/utils/resolveModelRecommendationFileOptions.ts","../src/cli/utils/splitAnsiToLines.ts","../src/cli/utils/toBytes.ts","../src/cli/utils/withCliCommandDescriptionDocsUrl.ts","../src/evaluator/LlamaCompletion.ts","../src/evaluator/LlamaEmbedding.ts","../src/evaluator/LlamaEmbeddingContext.ts","../src/evaluator/LlamaGrammar.ts","../src/evaluator/LlamaGrammarEvaluationState.ts","../src/evaluator/LlamaJsonSchemaGrammar.ts","../src/evaluator/LlamaRankingContext.ts","../src/evaluator/TokenBias.ts","../src/evaluator/TokenMeter.ts","../src/evaluator/LlamaChat/LlamaChat.ts","../src/evaluator/LlamaChat/utils/FunctionCallNameGrammar.ts","../src/evaluator/LlamaChat/utils/FunctionCallParamsGrammar.ts","../src/evaluator/LlamaChat/utils/LlamaFunctionCallValidationError.ts","../src/evaluator/LlamaChat/utils/contextShiftStrategies/eraseFirstResponseAndKeepFirstSystemChatContextShiftStrategy.ts","../src/evaluator/LlamaChatSession/LlamaChatSession.ts","../src/evaluator/LlamaChatSession/utils/LlamaChatSessionPromptCompletionEngine.ts","../src/evaluator/LlamaChatSession/utils/defineChatSessionFunction.ts","../src/evaluator/LlamaContext/LlamaContext.ts","../src/evaluator/LlamaContext/LlamaSampler.ts","../src/evaluator/LlamaContext/TokenPredictor.ts","../src/evaluator/LlamaContext/types.ts","../src/evaluator/LlamaContext/tokenPredictors/DraftSequenceTokenPredictor.ts","../src/evaluator/LlamaContext/tokenPredictors/InputLookupTokenPredictor.ts","../src/evaluator/LlamaContext/utils/padSafeContextSize.ts","../src/evaluator/LlamaContext/utils/resolveBatchItemsPrioritizationStrategy.ts","../src/evaluator/LlamaContext/utils/batchItemsPrioritizationStrategies/firstInFirstOutStrategy.ts","../src/evaluator/LlamaContext/utils/batchItemsPrioritizationStrategies/maximumParallelismStrategy.ts","../src/evaluator/LlamaModel/LlamaModel.ts","../src/evaluator/LlamaModel/utils/TokenAttributes.ts","../src/evaluator/utils/chunkDocument.ts","../src/gguf/consts.ts","../src/gguf/readGgufFileInfo.ts","../src/gguf/errors/InvalidGgufMagicError.ts","../src/gguf/errors/UnsupportedGgufValueTypeError.ts","../src/gguf/fileReaders/GgufFileReader.ts","../src/gguf/fileReaders/GgufFsFileReader.ts","../src/gguf/fileReaders/GgufNetworkFetchFileReader.ts","../src/gguf/insights/GgufInsights.ts","../src/gguf/insights/GgufInsightsConfigurationResolver.ts","../src/gguf/insights/GgufInsightsTokens.ts","../src/gguf/insights/utils/getRamUsageFromUnifiedVram.ts","../src/gguf/insights/utils/resolveContextContextSizeOption.ts","../src/gguf/insights/utils/resolveModelGpuLayersOption.ts","../src/gguf/insights/utils/scoreLevels.ts","../src/gguf/parser/GgufV2Parser.ts","../src/gguf/parser/GgufV3Parser.ts","../src/gguf/parser/parseGguf.ts","../src/gguf/types/GgufFileInfoTypes.ts","../src/gguf/types/GgufMetadataTypes.ts","../src/gguf/types/GgufTensorInfoTypes.ts","../src/gguf/utils/GgufReadOffset.ts","../src/gguf/utils/convertMetadataKeyValueRecordToNestedObject.ts","../src/gguf/utils/getGgufFileTypeName.ts","../src/gguf/utils/getGgufMetadataArchitectureData.ts","../src/gguf/utils/getGgufMetadataKeyValue.ts","../src/gguf/utils/ggufQuantNames.ts","../src/gguf/utils/normalizeGgufDownloadUrl.ts","../src/gguf/utils/resolveBinarySplitGgufPartUrls.ts","../src/gguf/utils/resolveSplitGgufParts.ts","../src/utils/DisposeGuard.ts","../src/utils/InsufficientMemoryError.ts","../src/utils/LlamaText.ts","../src/utils/LruCache.ts","../src/utils/OpenAIFormat.ts","../src/utils/OverridesObject.ts","../src/utils/ReplHistory.ts","../src/utils/StopGenerationDetector.ts","../src/utils/ThreadsSplitter.ts","../src/utils/TokenStreamRegulator.ts","../src/utils/UnsupportedError.ts","../src/utils/appendUserMessageToChatHistory.ts","../src/utils/clearTempFolder.ts","../src/utils/cmake.ts","../src/utils/compareTokens.ts","../src/utils/createModelDownloader.ts","../src/utils/findBestOption.ts","../src/utils/findCharacterRemovalCountToFitChatHistoryInContext.ts","../src/utils/getBuildDefaults.ts","../src/utils/getChatWrapperSegmentDefinition.ts","../src/utils/getConsoleLogPrefix.ts","../src/utils/getGrammarsFolder.ts","../src/utils/getLlamaClasses.ts","../src/utils/getModuleVersion.ts","../src/utils/getQueuedTokensBeforeStopTrigger.ts","../src/utils/getReadableContextSize.ts","../src/utils/getTypeScriptTypeStringForGbnfJsonSchema.ts","../src/utils/gitReleaseBundles.ts","../src/utils/hashString.ts","../src/utils/includesText.ts","../src/utils/isLockfileActive.ts","../src/utils/isToken.ts","../src/utils/isUrl.ts","../src/utils/mergeUnionTypes.ts","../src/utils/modelDownloadEndpoints.ts","../src/utils/modelFileAccessTokens.ts","../src/utils/optionsMatrix.ts","../src/utils/parseModelFileName.ts","../src/utils/parseModelUri.ts","../src/utils/parseTextTemplate.ts","../src/utils/prettyPrintObject.ts","../src/utils/pushAll.ts","../src/utils/removeNullFields.ts","../src/utils/resolveGithubRelease.ts","../src/utils/resolveLastTokens.ts","../src/utils/resolveModelDestination.ts","../src/utils/resolveModelFile.ts","../src/utils/runtime.ts","../src/utils/safeEventCallback.ts","../src/utils/signalSleep.ts","../src/utils/spawnCommand.ts","../src/utils/tokenizeInput.ts","../src/utils/tokenizerUtils.ts","../src/utils/transformPromisable.ts","../src/utils/truncateTextAndRoundToWords.ts","../src/utils/utilTypes.ts","../src/utils/waitForLockfileRelease.ts","../src/utils/withLockfile.ts","../src/utils/withOra.ts","../src/utils/withProgressLog.ts","../src/utils/withStatusLogs.ts","../src/utils/wrapAbortSignal.ts","../src/utils/gbnfJson/GbnfGrammarGenerator.ts","../src/utils/gbnfJson/GbnfTerminal.ts","../src/utils/gbnfJson/getGbnfGrammarForGbnfJsonSchema.ts","../src/utils/gbnfJson/types.ts","../src/utils/gbnfJson/terminals/GbnfAnyJson.ts","../src/utils/gbnfJson/terminals/GbnfArray.ts","../src/utils/gbnfJson/terminals/GbnfBoolean.ts","../src/utils/gbnfJson/terminals/GbnfBooleanValue.ts","../src/utils/gbnfJson/terminals/GbnfCommaWhitespace.ts","../src/utils/gbnfJson/terminals/GbnfFormatString.ts","../src/utils/gbnfJson/terminals/GbnfGrammar.ts","../src/utils/gbnfJson/terminals/GbnfInsideStringChar.ts","../src/utils/gbnfJson/terminals/GbnfNull.ts","../src/utils/gbnfJson/terminals/GbnfNumber.ts","../src/utils/gbnfJson/terminals/GbnfNumberValue.ts","../src/utils/gbnfJson/terminals/GbnfObjectMap.ts","../src/utils/gbnfJson/terminals/GbnfOr.ts","../src/utils/gbnfJson/terminals/GbnfRef.ts","../src/utils/gbnfJson/terminals/GbnfRepetition.ts","../src/utils/gbnfJson/terminals/GbnfString.ts","../src/utils/gbnfJson/terminals/GbnfStringValue.ts","../src/utils/gbnfJson/terminals/GbnfVerbatimText.ts","../src/utils/gbnfJson/terminals/GbnfWhitespace.ts","../src/utils/gbnfJson/terminals/gbnfConsts.ts","../src/utils/gbnfJson/utils/GbnfJsonScopeState.ts","../src/utils/gbnfJson/utils/defsScope.ts","../src/utils/gbnfJson/utils/getGbnfJsonTerminalForGbnfJsonSchema.ts","../src/utils/gbnfJson/utils/getGbnfJsonTerminalForLiteral.ts","../src/utils/gbnfJson/utils/validateObjectAgainstGbnfSchema.ts"],"version":"5.9.3"}
|
package/dist/types.d.ts
CHANGED
|
@@ -72,6 +72,11 @@ export type ChatWrapperSettings = {
|
|
|
72
72
|
readonly parallelism?: {
|
|
73
73
|
readonly call: {
|
|
74
74
|
readonly sectionPrefix: string | LlamaText;
|
|
75
|
+
/**
|
|
76
|
+
* Alternate section prefixes that can be used to detect a function call section,
|
|
77
|
+
* but won't be used to construct the context when building it from scratch.
|
|
78
|
+
*/
|
|
79
|
+
readonly sectionPrefixAlternateMatches?: Array<string | LlamaText>;
|
|
75
80
|
readonly betweenCalls?: string | LlamaText;
|
|
76
81
|
readonly sectionSuffix?: string | LlamaText;
|
|
77
82
|
};
|
|
@@ -354,3 +359,49 @@ export type LLamaContextualRepeatPenalty = {
|
|
|
354
359
|
*/
|
|
355
360
|
presencePenalty?: number;
|
|
356
361
|
};
|
|
362
|
+
export type LLamaContextualDryRepeatPenalty = {
|
|
363
|
+
/**
|
|
364
|
+
* A number between `0` and `1` representing the strength of the DRY (Don't Repeat Yourself) effect.
|
|
365
|
+
*
|
|
366
|
+
* Setting this to `0` will disable the DRY penalty completely.
|
|
367
|
+
*
|
|
368
|
+
* The recommended value is `0.8`.
|
|
369
|
+
*/
|
|
370
|
+
strength: number;
|
|
371
|
+
/**
|
|
372
|
+
* The base value for the exponential penality calculation.
|
|
373
|
+
*
|
|
374
|
+
* A higher value will lead to more aggressive penalization of repetitions.
|
|
375
|
+
*
|
|
376
|
+
* Defaults to `1.75`.
|
|
377
|
+
*/
|
|
378
|
+
base?: number;
|
|
379
|
+
/**
|
|
380
|
+
* The maximum sequence length (in tokens) that will be allowed to be repeated without being penalized.
|
|
381
|
+
*
|
|
382
|
+
* Repetitions shorter than or equal to this length will not be penalized,
|
|
383
|
+
* allowing for natural repetitions of short phrases and common words.
|
|
384
|
+
*
|
|
385
|
+
* Defaults to `2`.
|
|
386
|
+
*/
|
|
387
|
+
allowedLength?: number;
|
|
388
|
+
/**
|
|
389
|
+
* Number of recent tokens generated by the model to consider for sequence repetition matching.
|
|
390
|
+
*
|
|
391
|
+
* When set to `null`, the entire context sequence history will be considered for repetition matching.
|
|
392
|
+
* Setting to `0` will disable DRY (Don't Repeat Yourself) penalty.
|
|
393
|
+
*
|
|
394
|
+
* Defaults to `null`.
|
|
395
|
+
*/
|
|
396
|
+
lastTokens?: number | null;
|
|
397
|
+
/**
|
|
398
|
+
* Text sequences that will be considered as breakers for the repeated sequences.
|
|
399
|
+
* These will never be penalized for being repeated, and are used to mark the boundaries of the repeated sequences.
|
|
400
|
+
*
|
|
401
|
+
* For example, setting this to `["\n", "*"]` will allow the model to make as many lists as it wants,
|
|
402
|
+
* without being penalized for repeating the list item marker (like `*`).
|
|
403
|
+
*
|
|
404
|
+
* Defaults to `["\n", ":", '"', "*"]`.
|
|
405
|
+
*/
|
|
406
|
+
sequenceBreakers?: string[];
|
|
407
|
+
};
|
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":"AA2VA,MAAM,CAAC,MAAM,eAAe,GAAG,CAAC,SAAS,EAAE,SAAS,CAAoD,CAAC;AACzG,KAAM,IAAsF,CAAC;AA8B7F,MAAM,UAAU,+BAA+B,CAAC,IAAuD;IACnG,IAAI,IAAI,IAAI,IAAI,IAAI,OAAO,IAAI,KAAK,QAAQ;QACxC,OAAO,KAAK,CAAC;IAEjB,OAAO,IAAI,CAAC,IAAI,KAAK,cAAc,CAAC;AACxC,CAAC;AAED,MAAM,UAAU,0BAA0B,CAAC,IAAuD;IAC9F,IAAI,IAAI,IAAI,IAAI,IAAI,OAAO,IAAI,KAAK,QAAQ;QACxC,OAAO,KAAK,CAAC;IAEjB,OAAO,IAAI,CAAC,IAAI,KAAK,SAAS,CAAC;AACnC,CAAC"}
|
package/dist/utils/cmake.js
CHANGED
|
@@ -35,9 +35,12 @@ export async function getCmakePath() {
|
|
|
35
35
|
}
|
|
36
36
|
catch (err) { }
|
|
37
37
|
try {
|
|
38
|
-
let resolvedPath = await which("cmake", {
|
|
39
|
-
path: path.join(llamaDirectory, "xpack", "xpacks", ".bin")
|
|
40
|
-
|
|
38
|
+
let resolvedPath = (await which("cmake", {
|
|
39
|
+
path: path.join(llamaDirectory, "xpack", "xpacks", ".bin"),
|
|
40
|
+
nothrow: true
|
|
41
|
+
})) || (await which("cmake", {
|
|
42
|
+
path: path.join(llamaDirectory, "xpack", "xpacks", "@xpack-dev-tools", "cmake", ".content", "bin")
|
|
43
|
+
}));
|
|
41
44
|
if (resolvedPath.toLowerCase().endsWith(".cmd"))
|
|
42
45
|
resolvedPath = (await getBinFromWindowCmd(resolvedPath, "cmake.exe")) ?? "";
|
|
43
46
|
else if (resolvedPath.toLowerCase().endsWith(".ps1")) {
|
package/dist/utils/cmake.js.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"cmake.js","sourceRoot":"","sources":["../../src/utils/cmake.ts"],"names":[],"mappings":"AAAA,OAAO,IAAI,MAAM,MAAM,CAAC;AACxB,OAAO,EAAE,MAAM,UAAU,CAAC;AAC1B,OAAO,KAAK,MAAM,OAAO,CAAC;AAC1B,OAAO,KAAK,MAAM,OAAO,CAAC;AAC1B,OAAO,EAAC,MAAM,EAAC,MAAM,SAAS,CAAC;AAC/B,OAAO,EACH,2BAA2B,EAAE,2BAA2B,EAAE,cAAc,EAAE,yBAAyB,EAAE,yBAAyB,EAC9H,cAAc,EAAE,UAAU,EAC7B,MAAM,cAAc,CAAC;AACtB,OAAO,EAAC,2BAA2B,EAAC,MAAM,kDAAkD,CAAC;AAC7F,OAAO,EAAC,WAAW,EAAC,MAAM,kCAAkC,CAAC;AAC7D,OAAO,EAAC,kCAAkC,EAAC,MAAM,uCAAuC,CAAC;AACzF,OAAO,EAAC,YAAY,EAAC,MAAM,mBAAmB,CAAC;AAC/C,OAAO,cAAc,MAAM,qBAAqB,CAAC;AACjD,OAAO,EAAC,YAAY,EAAC,MAAM,mBAAmB,CAAC;AAG/C,MAAM,CAAC,KAAK,UAAU,eAAe;IACjC,IAAI,CAAC;QACD,MAAM,YAAY,GAAG,MAAM,KAAK,CAAC,OAAO,CAAC,CAAC;QAC1C,OAAO,YAAY,KAAK,EAAE,CAAC;IAC/B,CAAC;IAAC,OAAO,GAAG,EAAE,CAAC;QACX,OAAO,KAAK,CAAC;IACjB,CAAC;AACL,CAAC;AAED,MAAM,CAAC,KAAK,UAAU,YAAY;IAC9B,IAAI,CAAC;QACD,MAAM,YAAY,GAAG,MAAM,KAAK,CAAC,OAAO,EAAE;YACtC,OAAO,EAAE,IAAI;SAChB,CAAC,CAAC;QAEH,IAAI,YAAY,KAAK,EAAE,IAAI,YAAY,IAAI,IAAI;YAC3C,OAAO,YAAY,CAAC;IAC5B,CAAC;IAAC,OAAO,GAAG,EAAE,CAAC,CAAA,CAAC;IAEhB,IAAI,CAAC;QACD,MAAM,aAAa,GAAG,MAAM,iBAAiB,EAAE,CAAC;QAChD,IAAI,aAAa,IAAI,IAAI;YACrB,OAAO,aAAa,CAAC;IAC7B,CAAC;IAAC,OAAO,GAAG,EAAE,CAAC,CAAA,CAAC;IAEhB,IAAI,CAAC;QACD,IAAI,YAAY,GAAG,MAAM,KAAK,CAAC,OAAO,EAAE;
|
|
1
|
+
{"version":3,"file":"cmake.js","sourceRoot":"","sources":["../../src/utils/cmake.ts"],"names":[],"mappings":"AAAA,OAAO,IAAI,MAAM,MAAM,CAAC;AACxB,OAAO,EAAE,MAAM,UAAU,CAAC;AAC1B,OAAO,KAAK,MAAM,OAAO,CAAC;AAC1B,OAAO,KAAK,MAAM,OAAO,CAAC;AAC1B,OAAO,EAAC,MAAM,EAAC,MAAM,SAAS,CAAC;AAC/B,OAAO,EACH,2BAA2B,EAAE,2BAA2B,EAAE,cAAc,EAAE,yBAAyB,EAAE,yBAAyB,EAC9H,cAAc,EAAE,UAAU,EAC7B,MAAM,cAAc,CAAC;AACtB,OAAO,EAAC,2BAA2B,EAAC,MAAM,kDAAkD,CAAC;AAC7F,OAAO,EAAC,WAAW,EAAC,MAAM,kCAAkC,CAAC;AAC7D,OAAO,EAAC,kCAAkC,EAAC,MAAM,uCAAuC,CAAC;AACzF,OAAO,EAAC,YAAY,EAAC,MAAM,mBAAmB,CAAC;AAC/C,OAAO,cAAc,MAAM,qBAAqB,CAAC;AACjD,OAAO,EAAC,YAAY,EAAC,MAAM,mBAAmB,CAAC;AAG/C,MAAM,CAAC,KAAK,UAAU,eAAe;IACjC,IAAI,CAAC;QACD,MAAM,YAAY,GAAG,MAAM,KAAK,CAAC,OAAO,CAAC,CAAC;QAC1C,OAAO,YAAY,KAAK,EAAE,CAAC;IAC/B,CAAC;IAAC,OAAO,GAAG,EAAE,CAAC;QACX,OAAO,KAAK,CAAC;IACjB,CAAC;AACL,CAAC;AAED,MAAM,CAAC,KAAK,UAAU,YAAY;IAC9B,IAAI,CAAC;QACD,MAAM,YAAY,GAAG,MAAM,KAAK,CAAC,OAAO,EAAE;YACtC,OAAO,EAAE,IAAI;SAChB,CAAC,CAAC;QAEH,IAAI,YAAY,KAAK,EAAE,IAAI,YAAY,IAAI,IAAI;YAC3C,OAAO,YAAY,CAAC;IAC5B,CAAC;IAAC,OAAO,GAAG,EAAE,CAAC,CAAA,CAAC;IAEhB,IAAI,CAAC;QACD,MAAM,aAAa,GAAG,MAAM,iBAAiB,EAAE,CAAC;QAChD,IAAI,aAAa,IAAI,IAAI;YACrB,OAAO,aAAa,CAAC;IAC7B,CAAC;IAAC,OAAO,GAAG,EAAE,CAAC,CAAA,CAAC;IAEhB,IAAI,CAAC;QACD,IAAI,YAAY,GAAG,CACf,MAAM,KAAK,CAAC,OAAO,EAAE;YACjB,IAAI,EAAE,IAAI,CAAC,IAAI,CAAC,cAAc,EAAE,OAAO,EAAE,QAAQ,EAAE,MAAM,CAAC;YAC1D,OAAO,EAAE,IAAI;SAChB,CAAC,CACL,IAAI,CACD,MAAM,KAAK,CAAC,OAAO,EAAE;YACjB,IAAI,EAAE,IAAI,CAAC,IAAI,CAAC,cAAc,EAAE,OAAO,EAAE,QAAQ,EAAE,kBAAkB,EAAE,OAAO,EAAE,UAAU,EAAE,KAAK,CAAC;SACrG,CAAC,CACL,CAAC;QAEF,IAAI,YAAY,CAAC,WAAW,EAAE,CAAC,QAAQ,CAAC,MAAM,CAAC;YAC3C,YAAY,GAAG,CAAC,MAAM,mBAAmB,CAAC,YAAY,EAAE,WAAW,CAAC,CAAC,IAAI,EAAE,CAAC;aAC3E,IAAI,YAAY,CAAC,WAAW,EAAE,CAAC,QAAQ,CAAC,MAAM,CAAC,EAAE,CAAC;YACnD,MAAM,WAAW,GAAG,YAAY,CAAC,KAAK,CAAC,CAAC,EAAE,CAAC,MAAM,CAAC,MAAM,CAAC,GAAG,MAAM,CAAC;YAEnE,IAAI,MAAM,EAAE,CAAC,UAAU,CAAC,WAAW,CAAC;gBAChC,YAAY,GAAG,CAAC,MAAM,mBAAmB,CAAC,WAAW,EAAE,WAAW,CAAC,CAAC,IAAI,EAAE,CAAC;QACnF,CAAC;QAED,IAAI,YAAY,KAAK,EAAE;YACnB,OAAO,YAAY,CAAC;IAC5B,CAAC;IAAC,OAAO,GAAG,EAAE,CAAC,CAAA,CAAC;IAEhB,MAAM,IAAI,KAAK,CAAC,iBAAiB,CAAC,CAAC;AACvC,CAAC;AAED,MAAM,CAAC,KAAK,UAAU,qBAAqB,CAAC,qBAA8B,KAAK;IAC3E,IAAI,CAAC;QACD,MAAM,YAAY,EAAE,CAAC;QACrB,OAAO;IACX,CAAC;IAAC,OAAO,GAAG,EAAE,CAAC,CAAA,CAAC;IAEhB,IAAI,CAAC,kBAAkB;QACnB,MAAM,aAAa,CAAC,EAAC,YAAY,EAAE,kBAAkB,EAAC,CAAC,CAAC;SACvD,CAAC;QACF,IAAI,CAAC;YACD,MAAM,cAAc,CAAC;gBACjB,OAAO,EAAE,KAAK,CAAC,IAAI,CAAC,mBAAmB,CAAC;gBACxC,OAAO,EAAE,KAAK,CAAC,IAAI,CAAC,kBAAkB,CAAC;gBACvC,IAAI,EAAE,KAAK,CAAC,IAAI,CAAC,0BAA0B,CAAC;aAC/C,EAAE,KAAK,IAAI,EAAE;gBACV,MAAM,aAAa,CAAC,EAAC,YAAY,EAAE,kBAAkB,EAAC,CAAC,CAAC;YAC5D,CAAC,CAAC,CAAC;QACP,CAAC;QAAC,OAAO,GAAG,EAAE,CAAC;YACX,MAAM,2BAA2B,CAAC,sBAAsB,EAAE;gBACtD,aAAa,EAAE,EAAC,GAAG,EAAE,CAAC,OAAO,CAAC,EAAE,GAAG,EAAE,CAAC,OAAO,CAAC,EAAC;gBAC/C,aAAa,EAAE,EAAC,IAAI,EAAE,CAAC,OAAO,CAAC,EAAC;aACnC,CAAC,CAAC;YACH,MAAM,GAAG,CAAC;QACd,CAAC;IACL,CAAC;AACL,CAAC;AAED,MAAM,CAAC,KAAK,UAAU,eAAe;IACjC,MAAM,EAAE,CAAC,MAAM,CAAC,yBAAyB,CAAC,CAAC;IAC3C,MAAM,EAAE,CAAC,MAAM,CAAC,yBAAyB,CAAC,CAAC;IAC3C,MAAM,EAAE,CAAC,MAAM,CAAC,IAAI,CAAC,IAAI,CAAC,cAAc,EAAE,QAAQ,CAAC,CAAC,CAAC;AACzD,CAAC;AAED;;;;GAIG;AACH,MAAM,CAAC,KAAK,UAAU,mBAAmB;IACrC,IAAI,CAAC;QACD,MAAM,MAAM,CAAC,yBAAyB,EAAE,KAAK,CAAC,CAAC;QAC/C,MAAM,MAAM,CAAC,yBAAyB,EAAE,KAAK,CAAC,CAAC;QAC/C,MAAM,MAAM,CAAC,IAAI,CAAC,IAAI,CAAC,cAAc,EAAE,QAAQ,CAAC,EAAE,KAAK,CAAC,CAAC;IAC7D,CAAC;IAAC,OAAO,GAAG,EAAE,CAAC,CAAA,CAAC;AACpB,CAAC;AAED,KAAK,UAAU,iBAAiB;IAC5B,MAAM,QAAQ,GAAG,WAAW,EAAE,CAAC;IAE/B,IAAI,QAAQ,KAAK,KAAK,EAAE,CAAC;QACrB,MAAM,EAAC,cAAc,EAAC,GAAG,MAAM,kCAAkC,EAAE,CAAC;QAEpE,MAAM,mBAAmB,GAAG,cAAc,CAAC,GAAG,CAAC,CAAC,WAAW,EAAE,EAAE,CAAC,CAC5D,IAAI,CAAC,IAAI,CAAC,WAAW,EAAE,SAAS,EAAE,KAAK,EAAE,kBAAkB,EAAE,WAAW,EAAE,OAAO,EAAE,OAAO,EAAE,KAAK,EAAE,WAAW,CAAC,CAClH,CAAC,CAAC;QAEH,MAAM,UAAU,GAAG,CAAC,MAAM,OAAO,CAAC,GAAG,CACjC,mBAAmB,CAAC,GAAG,CAAC,KAAK,EAAE,SAAS,EAAE,EAAE;YACxC,IAAI,MAAM,EAAE,CAAC,UAAU,CAAC,SAAS,CAAC;gBAC9B,OAAO,SAAS,CAAC;YAErB,OAAO,IAAI,CAAC;QAChB,CAAC,CAAC,CACL,CAAC;aACG,MAAM,CAAC,CAAC,SAAS,EAAE,EAAE,CAAC,SAAS,IAAI,IAAI,CAAC,CAAC;QAE9C,OAAO,UAAU,CAAC,CAAC,CAAC,CAAC;IACzB,CAAC;IAED,OAAO,SAAS,CAAC;AACrB,CAAC;AAED,KAAK,UAAU,aAAa,CAAC,EAAC,YAAY,GAAG,IAAI,KAA8B,EAAE;IAC7E,MAAM,YAAY,CAAC;QACf,YAAY,EAAE,IAAI,CAAC,IAAI,CAAC,cAAc,EAAE,cAAc,CAAC;KAC1D,EAAE,KAAK,IAAI,EAAE;QACV,MAAM,MAAM,GAAsB;YAC9B,GAAG,OAAO,CAAC,GAAG;YACd,mBAAmB,EAAE,2BAA2B;YAChD,mBAAmB,EAAE,2BAA2B;SACnD,CAAC;QAEF,MAAM,YAAY,CAAC,KAAK,EAAE,CAAC,MAAM,EAAE,OAAO,EAAE,IAAI,EAAE,OAAO,UAAU,EAAE,EAAE,SAAS,EAAE,+BAA+B,EAAE,WAAW,CAAC,EAAE,cAAc,EAAE,MAAM,EAAE,YAAY,CAAC,CAAC;QAEvK,MAAM,EAAE,CAAC,MAAM,CAAC,yBAAyB,CAAC,CAAC;QAC3C,MAAM,mBAAmB,EAAE,CAAC;IAChC,CAAC,CAAC,CAAC;AACP,CAAC;AAED,KAAK,UAAU,mBAAmB,CAAC,WAAmB,EAAE,OAAe;IACnE,MAAM,WAAW,GAAW,MAAM,EAAE,CAAC,QAAQ,CAAC,WAAW,EAAE,MAAM,CAAC,CAAC;IACnE,MAAM,oBAAoB,GAAG,WAAW,CAAC,WAAW,EAAE,CAAC;IAEvD,IAAI,CAAC,oBAAoB,CAAC,QAAQ,CAAC,OAAO,CAAC;QACvC,OAAO,IAAI,CAAC;IAEhB,MAAM,kBAAkB,GAAG,oBAAoB,CAAC,WAAW,CAAC,OAAO,CAAC,CAAC;IACrE,MAAM,qBAAqB,GAAG,WAAW,CAAC,kBAAkB,GAAG,OAAO,CAAC,MAAM,CAAC,CAAC;IAE/E,IAAI,qBAAqB,KAAK,GAAG,IAAI,qBAAqB,KAAK,GAAG;QAC9D,OAAO,IAAI,CAAC;IAEhB,MAAM,oBAAoB,GAAG,WAAW,CAAC,WAAW,CAAC,qBAAqB,EAAE,kBAAkB,CAAC,CAAC;IAEhG,MAAM,OAAO,GAAG,WAAW,CAAC,KAAK,CAAC,oBAAoB,GAAG,CAAC,EAAE,kBAAkB,GAAG,OAAO,CAAC,MAAM,CAAC,CAAC;IAEjG,IAAI,CAAC,MAAM,EAAE,CAAC,UAAU,CAAC,OAAO,CAAC;QAC7B,OAAO,IAAI,CAAC;IAEhB,OAAO,OAAO,CAAC;AACnB,CAAC"}
|
package/llama/CMakeLists.txt
CHANGED
|
@@ -942,11 +942,25 @@ Napi::Value AddonContext::EnsureDraftContextIsCompatibleForSpeculative(const Nap
|
|
|
942
942
|
return info.Env().Undefined();
|
|
943
943
|
}
|
|
944
944
|
|
|
945
|
-
Napi::Value AddonContext::
|
|
946
|
-
|
|
947
|
-
|
|
945
|
+
Napi::Value AddonContext::SetLoras(const Napi::CallbackInfo& info) {
|
|
946
|
+
Napi::Array loraArray = info[0].As<Napi::Array>();
|
|
947
|
+
Napi::Array scaleArray = info[1].As<Napi::Array>();
|
|
948
948
|
|
|
949
|
-
|
|
949
|
+
std::vector<llama_adapter_lora *> loras;
|
|
950
|
+
std::vector<float> scales;
|
|
951
|
+
|
|
952
|
+
loras.reserve(loraArray.Length());
|
|
953
|
+
scales.reserve(scaleArray.Length());
|
|
954
|
+
|
|
955
|
+
for (size_t i = 0; i < loraArray.Length() && i < scaleArray.Length(); i++) {
|
|
956
|
+
AddonModelLora* lora = Napi::ObjectWrap<AddonModelLora>::Unwrap(loraArray.Get(i).As<Napi::Object>());
|
|
957
|
+
float scale = scaleArray.Get(i).As<Napi::Number>().FloatValue();
|
|
958
|
+
|
|
959
|
+
loras.push_back(lora->lora_adapter);
|
|
960
|
+
scales.push_back(scale);
|
|
961
|
+
}
|
|
962
|
+
|
|
963
|
+
llama_set_adapters_lora(ctx, loras.data(), loras.size(), scales.data());
|
|
950
964
|
|
|
951
965
|
return info.Env().Undefined();
|
|
952
966
|
}
|
|
@@ -977,7 +991,7 @@ void AddonContext::init(Napi::Object exports) {
|
|
|
977
991
|
InstanceMethod("ensureDraftContextIsCompatibleForSpeculative", &AddonContext::EnsureDraftContextIsCompatibleForSpeculative),
|
|
978
992
|
InstanceMethod("saveSequenceStateToFile", &AddonContext::SaveSequenceStateToFile),
|
|
979
993
|
InstanceMethod("loadSequenceStateFromFile", &AddonContext::LoadSequenceStateFromFile),
|
|
980
|
-
InstanceMethod("
|
|
994
|
+
InstanceMethod("setLoras", &AddonContext::SetLoras),
|
|
981
995
|
InstanceMethod("dispose", &AddonContext::Dispose),
|
|
982
996
|
}
|
|
983
997
|
)
|
|
@@ -52,7 +52,7 @@ class AddonContext : public Napi::ObjectWrap<AddonContext> {
|
|
|
52
52
|
Napi::Value PrintTimings(const Napi::CallbackInfo& info);
|
|
53
53
|
Napi::Value EnsureDraftContextIsCompatibleForSpeculative(const Napi::CallbackInfo& info);
|
|
54
54
|
|
|
55
|
-
Napi::Value
|
|
55
|
+
Napi::Value SetLoras(const Napi::CallbackInfo& info);
|
|
56
56
|
|
|
57
57
|
static void init(Napi::Object exports);
|
|
58
58
|
};
|
|
@@ -362,6 +362,13 @@ void AddonModel::dispose() {
|
|
|
362
362
|
}
|
|
363
363
|
|
|
364
364
|
disposed = true;
|
|
365
|
+
|
|
366
|
+
if (data != nullptr) {
|
|
367
|
+
auto currentData = data;
|
|
368
|
+
data = nullptr;
|
|
369
|
+
delete currentData;
|
|
370
|
+
}
|
|
371
|
+
|
|
365
372
|
if (modelLoaded) {
|
|
366
373
|
modelLoaded = false;
|
|
367
374
|
llama_model_free(model);
|
|
@@ -370,12 +377,6 @@ void AddonModel::dispose() {
|
|
|
370
377
|
loadedModelSize = 0;
|
|
371
378
|
}
|
|
372
379
|
|
|
373
|
-
if (data != nullptr) {
|
|
374
|
-
auto currentData = data;
|
|
375
|
-
data = nullptr;
|
|
376
|
-
delete currentData;
|
|
377
|
-
}
|
|
378
|
-
|
|
379
380
|
if (hasAddonExportsRef) {
|
|
380
381
|
addonExportsRef.Unref();
|
|
381
382
|
hasAddonExportsRef = false;
|
|
@@ -53,10 +53,8 @@ AddonModelLora::~AddonModelLora() {
|
|
|
53
53
|
|
|
54
54
|
void AddonModelLora::dispose(bool skipErase) {
|
|
55
55
|
if (lora_adapter != nullptr) {
|
|
56
|
-
auto loraAdapterToDispose = lora_adapter;
|
|
57
56
|
lora_adapter = nullptr;
|
|
58
|
-
|
|
59
|
-
|
|
57
|
+
|
|
60
58
|
if (!skipErase && model->data != nullptr) {
|
|
61
59
|
model->data->removeLora(this);
|
|
62
60
|
}
|
|
@@ -28,6 +28,11 @@ void AddonSampler::dispose() {
|
|
|
28
28
|
model->Unref();
|
|
29
29
|
freeChain();
|
|
30
30
|
|
|
31
|
+
if (xtcSampler != nullptr) {
|
|
32
|
+
llama_sampler_free(xtcSampler);
|
|
33
|
+
xtcSampler = nullptr;
|
|
34
|
+
}
|
|
35
|
+
|
|
31
36
|
if (temperatureSampler != nullptr) {
|
|
32
37
|
llama_sampler_free(temperatureSampler);
|
|
33
38
|
temperatureSampler = nullptr;
|
|
@@ -63,6 +68,11 @@ void AddonSampler::dispose() {
|
|
|
63
68
|
repeatPenaltySampler = nullptr;
|
|
64
69
|
}
|
|
65
70
|
|
|
71
|
+
if (dryRepeatPenaltySampler != nullptr) {
|
|
72
|
+
llama_sampler_free(dryRepeatPenaltySampler);
|
|
73
|
+
dryRepeatPenaltySampler = nullptr;
|
|
74
|
+
}
|
|
75
|
+
|
|
66
76
|
if (tokenBiasSampler != nullptr) {
|
|
67
77
|
llama_sampler_free(tokenBiasSampler);
|
|
68
78
|
tokenBiasSampler = nullptr;
|
|
@@ -108,11 +118,19 @@ void AddonSampler::rebuildChainIfNeeded() {
|
|
|
108
118
|
llama_sampler_chain_add(chain, repeatPenaltySampler);
|
|
109
119
|
}
|
|
110
120
|
|
|
121
|
+
if (dryRepeatPenaltySampler != nullptr) {
|
|
122
|
+
llama_sampler_chain_add(chain, dryRepeatPenaltySampler);
|
|
123
|
+
}
|
|
124
|
+
|
|
111
125
|
if (grammarEvaluationState != nullptr) {
|
|
112
126
|
llama_sampler_chain_add(chain, grammarEvaluationState->sampler);
|
|
113
127
|
}
|
|
114
128
|
|
|
115
129
|
if (greedySampler != nullptr) {
|
|
130
|
+
if (xtcSampler != nullptr) {
|
|
131
|
+
llama_sampler_chain_add(chain, xtcSampler);
|
|
132
|
+
}
|
|
133
|
+
|
|
116
134
|
llama_sampler_chain_add(chain, greedySampler);
|
|
117
135
|
} else {
|
|
118
136
|
if (topKSampler != nullptr) {
|
|
@@ -127,6 +145,10 @@ void AddonSampler::rebuildChainIfNeeded() {
|
|
|
127
145
|
llama_sampler_chain_add(chain, minPSampler);
|
|
128
146
|
}
|
|
129
147
|
|
|
148
|
+
if (xtcSampler != nullptr) {
|
|
149
|
+
llama_sampler_chain_add(chain, xtcSampler);
|
|
150
|
+
}
|
|
151
|
+
|
|
130
152
|
if (temperatureSampler != nullptr) {
|
|
131
153
|
llama_sampler_chain_add(chain, temperatureSampler);
|
|
132
154
|
}
|
|
@@ -143,6 +165,10 @@ void AddonSampler::acceptToken(llama_token token) {
|
|
|
143
165
|
repeatPenalty_lastTokens.push_back(token);
|
|
144
166
|
}
|
|
145
167
|
|
|
168
|
+
if (dryRepeatPenaltySampler != nullptr) {
|
|
169
|
+
llama_sampler_accept(dryRepeatPenaltySampler, token);
|
|
170
|
+
}
|
|
171
|
+
|
|
146
172
|
if (grammarEvaluationState != nullptr && grammarEvaluationState->sampler != nullptr && !llama_vocab_is_eog(model->vocab, token)) {
|
|
147
173
|
llama_sampler_accept(grammarEvaluationState->sampler, token);
|
|
148
174
|
}
|
|
@@ -282,6 +308,33 @@ Napi::Value AddonSampler::ApplyConfig(const Napi::CallbackInfo& info) {
|
|
|
282
308
|
seedSampler = llama_sampler_init_dist(time(NULL));
|
|
283
309
|
}
|
|
284
310
|
|
|
311
|
+
if (config.Has("xtcProbability") && config.Has("xtcThreshold")) {
|
|
312
|
+
auto xtcProbability = config.Get("xtcProbability").As<Napi::Number>().FloatValue();
|
|
313
|
+
auto xtcThreshold = config.Get("xtcThreshold").As<Napi::Number>().FloatValue();
|
|
314
|
+
|
|
315
|
+
if (xtcProbability != xtcSampler_probability || xtcThreshold != xtcSampler_threshold || xtcSampler == nullptr) {
|
|
316
|
+
xtcSampler_probability = xtcProbability;
|
|
317
|
+
xtcSampler_threshold = xtcThreshold;
|
|
318
|
+
freeChain();
|
|
319
|
+
|
|
320
|
+
if (xtcSampler != nullptr) {
|
|
321
|
+
llama_sampler_free(xtcSampler);
|
|
322
|
+
xtcSampler = nullptr;
|
|
323
|
+
}
|
|
324
|
+
|
|
325
|
+
xtcSampler = llama_sampler_init_xtc(
|
|
326
|
+
xtcSampler_probability,
|
|
327
|
+
xtcSampler_threshold,
|
|
328
|
+
0,
|
|
329
|
+
seedSampler == nullptr ? LLAMA_DEFAULT_SEED : seedSampler_seed
|
|
330
|
+
);
|
|
331
|
+
}
|
|
332
|
+
} else if (xtcSampler != nullptr) {
|
|
333
|
+
freeChain();
|
|
334
|
+
llama_sampler_free(xtcSampler);
|
|
335
|
+
xtcSampler = nullptr;
|
|
336
|
+
}
|
|
337
|
+
|
|
285
338
|
if (config.Has("repeatPenaltyTokens")) {
|
|
286
339
|
Napi::Uint32Array repeat_penalty_tokens_uint32_array = config.Get("repeatPenaltyTokens").As<Napi::Uint32Array>();
|
|
287
340
|
auto repeatPenalty = config.Has("repeatPenalty")
|
|
@@ -374,6 +427,111 @@ Napi::Value AddonSampler::ApplyConfig(const Napi::CallbackInfo& info) {
|
|
|
374
427
|
repeatPenaltySampler = nullptr;
|
|
375
428
|
}
|
|
376
429
|
|
|
430
|
+
if (config.Has("dryRepeatPenaltyStrength")) {
|
|
431
|
+
float strength = config.Get("dryRepeatPenaltyStrength").As<Napi::Number>().FloatValue();
|
|
432
|
+
float base = config.Has("dryRepeatPenaltyBase")
|
|
433
|
+
? config.Get("dryRepeatPenaltyBase").As<Napi::Number>().FloatValue()
|
|
434
|
+
: 0;
|
|
435
|
+
int32_t allowedLength = config.Has("dryRepeatPenaltyAllowedLength")
|
|
436
|
+
? config.Get("dryRepeatPenaltyAllowedLength").As<Napi::Number>().Int32Value()
|
|
437
|
+
: 2;
|
|
438
|
+
int32_t lastTokens = config.Has("dryRepeatPenaltyLastTokens")
|
|
439
|
+
? config.Get("dryRepeatPenaltyLastTokens").As<Napi::Number>().Int32Value()
|
|
440
|
+
: -1;
|
|
441
|
+
|
|
442
|
+
bool sequenceBreaksIsTheSame = (
|
|
443
|
+
config.Has("dryRepeatPenaltySequenceBreakers") &&
|
|
444
|
+
config.Get("dryRepeatPenaltySequenceBreakers").IsBoolean() &&
|
|
445
|
+
config.Get("dryRepeatPenaltySequenceBreakers").As<Napi::Boolean>().Value() == false
|
|
446
|
+
) ? true : false;
|
|
447
|
+
|
|
448
|
+
std::vector<std::string> sequenceBreakers;
|
|
449
|
+
if (config.Has("dryRepeatPenaltySequenceBreakers") && config.Get("dryRepeatPenaltySequenceBreakers").IsArray()) {
|
|
450
|
+
Napi::Array sequenceBreakersArray = config.Get("dryRepeatPenaltySequenceBreakers").As<Napi::Array>();
|
|
451
|
+
sequenceBreaksIsTheSame = dryRepeatPenalty_sequenceBreakers.size() == sequenceBreakersArray.Length();
|
|
452
|
+
|
|
453
|
+
sequenceBreakers.reserve(sequenceBreakersArray.Length());
|
|
454
|
+
for (size_t i = 0; i < sequenceBreakersArray.Length(); i++) {
|
|
455
|
+
std::string breaker = sequenceBreakersArray.Get(i).As<Napi::String>().Utf8Value();
|
|
456
|
+
|
|
457
|
+
if (sequenceBreaksIsTheSame && dryRepeatPenalty_sequenceBreakers[i] != breaker) {
|
|
458
|
+
sequenceBreaksIsTheSame = false;
|
|
459
|
+
}
|
|
460
|
+
|
|
461
|
+
sequenceBreakers.push_back(std::move(breaker));
|
|
462
|
+
}
|
|
463
|
+
}
|
|
464
|
+
|
|
465
|
+
auto enabled = base != 0 && lastTokens != 0;
|
|
466
|
+
bool shouldCreateSampler = false;
|
|
467
|
+
|
|
468
|
+
if (!enabled) {
|
|
469
|
+
if (dryRepeatPenaltySampler != nullptr) {
|
|
470
|
+
freeChain();
|
|
471
|
+
llama_sampler_free(dryRepeatPenaltySampler);
|
|
472
|
+
dryRepeatPenaltySampler = nullptr;
|
|
473
|
+
}
|
|
474
|
+
} else if (dryRepeatPenaltySampler == nullptr) {
|
|
475
|
+
freeChain();
|
|
476
|
+
shouldCreateSampler = true;
|
|
477
|
+
} else {
|
|
478
|
+
bool existingSamplerMatchesConfig = true;
|
|
479
|
+
existingSamplerMatchesConfig &= dryRepeatPenalty_strength == strength;
|
|
480
|
+
existingSamplerMatchesConfig &= dryRepeatPenalty_base == base;
|
|
481
|
+
existingSamplerMatchesConfig &= dryRepeatPenalty_allowedLength == allowedLength;
|
|
482
|
+
existingSamplerMatchesConfig &= dryRepeatPenalty_lastTokens == lastTokens;
|
|
483
|
+
existingSamplerMatchesConfig &= sequenceBreaksIsTheSame;
|
|
484
|
+
|
|
485
|
+
if (!existingSamplerMatchesConfig) {
|
|
486
|
+
freeChain();
|
|
487
|
+
llama_sampler_free(dryRepeatPenaltySampler);
|
|
488
|
+
dryRepeatPenaltySampler = nullptr;
|
|
489
|
+
|
|
490
|
+
shouldCreateSampler = true;
|
|
491
|
+
}
|
|
492
|
+
}
|
|
493
|
+
|
|
494
|
+
if (shouldCreateSampler) {
|
|
495
|
+
std::vector<const char *> cSequenceBreakers;
|
|
496
|
+
|
|
497
|
+
if (sequenceBreaksIsTheSame) {
|
|
498
|
+
cSequenceBreakers.reserve(dryRepeatPenalty_sequenceBreakers.size());
|
|
499
|
+
for (const auto & str : dryRepeatPenalty_sequenceBreakers) {
|
|
500
|
+
cSequenceBreakers.push_back(str.c_str());
|
|
501
|
+
}
|
|
502
|
+
} else {
|
|
503
|
+
cSequenceBreakers.reserve(sequenceBreakers.size());
|
|
504
|
+
for (const auto & str : sequenceBreakers) {
|
|
505
|
+
cSequenceBreakers.push_back(str.c_str());
|
|
506
|
+
}
|
|
507
|
+
}
|
|
508
|
+
|
|
509
|
+
dryRepeatPenaltySampler = llama_sampler_init_dry(
|
|
510
|
+
model->vocab,
|
|
511
|
+
llama_model_n_ctx_train(model->model),
|
|
512
|
+
strength,
|
|
513
|
+
base,
|
|
514
|
+
allowedLength,
|
|
515
|
+
lastTokens,
|
|
516
|
+
cSequenceBreakers.data(),
|
|
517
|
+
cSequenceBreakers.size()
|
|
518
|
+
);
|
|
519
|
+
|
|
520
|
+
dryRepeatPenalty_strength = strength;
|
|
521
|
+
dryRepeatPenalty_base = base;
|
|
522
|
+
dryRepeatPenalty_allowedLength = allowedLength;
|
|
523
|
+
dryRepeatPenalty_lastTokens = lastTokens;
|
|
524
|
+
|
|
525
|
+
if (!sequenceBreaksIsTheSame) {
|
|
526
|
+
dryRepeatPenalty_sequenceBreakers.swap(sequenceBreakers);
|
|
527
|
+
}
|
|
528
|
+
}
|
|
529
|
+
} else if (dryRepeatPenaltySampler != nullptr) {
|
|
530
|
+
freeChain();
|
|
531
|
+
llama_sampler_free(dryRepeatPenaltySampler);
|
|
532
|
+
dryRepeatPenaltySampler = nullptr;
|
|
533
|
+
}
|
|
534
|
+
|
|
377
535
|
if (config.Has("tokenBiasKeys") && config.Has("tokenBiasValues")) {
|
|
378
536
|
Napi::Uint32Array tokenBiasKeys = config.Get("tokenBiasKeys").As<Napi::Uint32Array>();
|
|
379
537
|
Napi::Float32Array tokenBiasValues = config.Get("tokenBiasValues").As<Napi::Float32Array>();
|
|
@@ -25,7 +25,11 @@ class AddonSampler : public Napi::ObjectWrap<AddonSampler> {
|
|
|
25
25
|
|
|
26
26
|
llama_sampler * topPSampler = nullptr;
|
|
27
27
|
float topPSampler_topP = 0.0f; // Top p sampling >=1.0 = disabled
|
|
28
|
-
|
|
28
|
+
|
|
29
|
+
llama_sampler * xtcSampler = nullptr;
|
|
30
|
+
float xtcSampler_probability = 0;
|
|
31
|
+
float xtcSampler_threshold = 0;
|
|
32
|
+
|
|
29
33
|
llama_sampler * seedSampler = nullptr;
|
|
30
34
|
uint32_t seedSampler_seed = 0;
|
|
31
35
|
|
|
@@ -36,6 +40,14 @@ class AddonSampler : public Napi::ObjectWrap<AddonSampler> {
|
|
|
36
40
|
float repeatPenalty_presencePenalty = 0.00f; // 0.0 = disabled
|
|
37
41
|
float repeatPenalty_frequencyPenalty = 0.00f; // 0.0 = disabled
|
|
38
42
|
|
|
43
|
+
llama_sampler * dryRepeatPenaltySampler = nullptr;
|
|
44
|
+
float dryRepeatPenalty_strength = 0.00f; // 0.00f = disabled
|
|
45
|
+
float dryRepeatPenalty_base = 0.00f;
|
|
46
|
+
int32_t dryRepeatPenalty_allowedLength = 1;
|
|
47
|
+
int32_t dryRepeatPenalty_lastTokens = 1;
|
|
48
|
+
bool dryRepeatPenalty_lastTokensAll = true;
|
|
49
|
+
std::vector<std::string> dryRepeatPenalty_sequenceBreakers;
|
|
50
|
+
|
|
39
51
|
llama_sampler * tokenBiasSampler = nullptr;
|
|
40
52
|
std::vector<llama_logit_bias> tokenBiasSampler_biases;
|
|
41
53
|
|
|
@@ -94,12 +94,12 @@ void addonLlamaCppLogCallback(ggml_log_level level, const char* text, void* user
|
|
|
94
94
|
}
|
|
95
95
|
|
|
96
96
|
Napi::Value setLogger(const Napi::CallbackInfo& info) {
|
|
97
|
-
if (
|
|
98
|
-
|
|
99
|
-
|
|
100
|
-
|
|
101
|
-
}
|
|
97
|
+
if (addonJsLoggerCallbackSet) {
|
|
98
|
+
addonJsLoggerCallbackSet = false;
|
|
99
|
+
addonThreadSafeLoggerCallback.Release();
|
|
100
|
+
}
|
|
102
101
|
|
|
102
|
+
if (info.Length() < 1 || !info[0].IsFunction()) {
|
|
103
103
|
return info.Env().Undefined();
|
|
104
104
|
}
|
|
105
105
|
|
|
@@ -99,7 +99,7 @@ std::pair<ggml_backend_dev_t, std::string> getGpuDevice() {
|
|
|
99
99
|
ggml_backend_dev_t device = ggml_backend_dev_get(i);
|
|
100
100
|
const auto deviceName = std::string(ggml_backend_dev_name(device));
|
|
101
101
|
|
|
102
|
-
if (deviceName == "Metal") {
|
|
102
|
+
if (std::string(deviceName).find("MTL") == 0 || deviceName == "Metal") {
|
|
103
103
|
return std::pair<ggml_backend_dev_t, std::string>(device, "metal");
|
|
104
104
|
} else if (std::string(deviceName).find("Vulkan") == 0) {
|
|
105
105
|
return std::pair<ggml_backend_dev_t, std::string>(device, "vulkan");
|
package/llama/gitRelease.bundle
CHANGED
|
Binary file
|
|
@@ -1,4 +1,5 @@
|
|
|
1
1
|
#include <stddef.h>
|
|
2
|
+
#include <cstdint>
|
|
2
3
|
#include <map>
|
|
3
4
|
#include <vector>
|
|
4
5
|
|
|
@@ -8,6 +9,7 @@ constexpr std::uint32_t VK_VENDOR_ID_AMD = 0x1002;
|
|
|
8
9
|
constexpr std::uint32_t VK_VENDOR_ID_APPLE = 0x106b;
|
|
9
10
|
constexpr std::uint32_t VK_VENDOR_ID_INTEL = 0x8086;
|
|
10
11
|
constexpr std::uint32_t VK_VENDOR_ID_NVIDIA = 0x10de;
|
|
12
|
+
constexpr std::uint32_t VK_VENDOR_ID_QUALCOMM = 0x5143;
|
|
11
13
|
|
|
12
14
|
typedef void (*gpuInfoVulkanWarningLogCallback_t)(const char* message);
|
|
13
15
|
|
|
@@ -35,7 +37,7 @@ static std::vector<vk::PhysicalDevice> dedupedDevices() {
|
|
|
35
37
|
auto oldDevice = std::find_if(
|
|
36
38
|
dedupedDevices.begin(),
|
|
37
39
|
dedupedDevices.end(),
|
|
38
|
-
[&newId](const vk::PhysicalDevice& oldDevice) {
|
|
40
|
+
[&newId, &newDriver](const vk::PhysicalDevice& oldDevice) {
|
|
39
41
|
vk::PhysicalDeviceProperties2 oldProps;
|
|
40
42
|
vk::PhysicalDeviceDriverProperties oldDriver;
|
|
41
43
|
vk::PhysicalDeviceIDProperties oldId;
|
|
@@ -43,13 +45,14 @@ static std::vector<vk::PhysicalDevice> dedupedDevices() {
|
|
|
43
45
|
oldDriver.pNext = &oldId;
|
|
44
46
|
oldDevice.getProperties2(&oldProps);
|
|
45
47
|
|
|
46
|
-
bool
|
|
47
|
-
|
|
48
|
+
bool sameUuid = std::equal(std::begin(oldId.deviceUUID), std::end(oldId.deviceUUID), std::begin(newId.deviceUUID));
|
|
49
|
+
sameUuid = sameUuid || (
|
|
48
50
|
oldId.deviceLUIDValid && newId.deviceLUIDValid &&
|
|
49
51
|
std::equal(std::begin(oldId.deviceLUID), std::end(oldId.deviceLUID), std::begin(newId.deviceLUID))
|
|
50
52
|
);
|
|
53
|
+
bool bothMoltenVk = (newDriver.driverID == vk::DriverId::eMoltenvk && oldDriver.driverID == vk::DriverId::eMoltenvk);
|
|
51
54
|
|
|
52
|
-
return
|
|
55
|
+
return sameUuid && !bothMoltenVk;
|
|
53
56
|
}
|
|
54
57
|
);
|
|
55
58
|
|
|
@@ -83,8 +86,12 @@ static std::vector<vk::PhysicalDevice> dedupedDevices() {
|
|
|
83
86
|
driverPriorities[vk::DriverId::eMesaNvk] = 2;
|
|
84
87
|
#endif
|
|
85
88
|
break;
|
|
89
|
+
case VK_VENDOR_ID_QUALCOMM:
|
|
90
|
+
driverPriorities[vk::DriverId::eQualcommProprietary] = 1;
|
|
91
|
+
driverPriorities[vk::DriverId::eMesaTurnip] = 2;
|
|
92
|
+
break;
|
|
86
93
|
}
|
|
87
|
-
driverPriorities[vk::DriverId::eMesaDozen] =
|
|
94
|
+
driverPriorities[vk::DriverId::eMesaDozen] = 100;
|
|
88
95
|
|
|
89
96
|
if (driverPriorities.count(oldDriver.driverID)) {
|
|
90
97
|
oldPriority = driverPriorities[oldDriver.driverID];
|