node-llama-cpp 3.0.0-beta.35 → 3.0.0-beta.37

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (89) hide show
  1. package/bins/linux-arm64/_nlcBuildMetadata.json +1 -1
  2. package/bins/linux-arm64/libllama.so +0 -0
  3. package/bins/linux-arm64/llama-addon.node +0 -0
  4. package/bins/linux-armv7l/_nlcBuildMetadata.json +1 -1
  5. package/bins/linux-armv7l/libllama.so +0 -0
  6. package/bins/linux-armv7l/llama-addon.node +0 -0
  7. package/bins/linux-x64/_nlcBuildMetadata.json +1 -1
  8. package/bins/linux-x64/libllama.so +0 -0
  9. package/bins/linux-x64/llama-addon.node +0 -0
  10. package/bins/linux-x64-vulkan/_nlcBuildMetadata.json +1 -1
  11. package/bins/linux-x64-vulkan/libllama.so +0 -0
  12. package/bins/linux-x64-vulkan/llama-addon.node +0 -0
  13. package/bins/mac-arm64-metal/_nlcBuildMetadata.json +1 -1
  14. package/bins/mac-arm64-metal/ggml-common.h +7 -7
  15. package/bins/mac-arm64-metal/ggml-metal.metal +0 -1
  16. package/bins/mac-arm64-metal/libggml.dylib +0 -0
  17. package/bins/mac-arm64-metal/libllama.dylib +0 -0
  18. package/bins/mac-arm64-metal/llama-addon.node +0 -0
  19. package/bins/mac-x64/_nlcBuildMetadata.json +1 -1
  20. package/bins/mac-x64/libllama.dylib +0 -0
  21. package/bins/mac-x64/llama-addon.node +0 -0
  22. package/bins/win-arm64/_nlcBuildMetadata.json +1 -1
  23. package/bins/win-arm64/ggml.dll +0 -0
  24. package/bins/win-arm64/llama-addon.node +0 -0
  25. package/bins/win-arm64/llama.dll +0 -0
  26. package/bins/win-x64/_nlcBuildMetadata.json +1 -1
  27. package/bins/win-x64/ggml.dll +0 -0
  28. package/bins/win-x64/llama-addon.node +0 -0
  29. package/bins/win-x64/llama.dll +0 -0
  30. package/bins/win-x64-vulkan/_nlcBuildMetadata.json +1 -1
  31. package/bins/win-x64-vulkan/ggml.dll +0 -0
  32. package/bins/win-x64-vulkan/llama-addon.node +0 -0
  33. package/bins/win-x64-vulkan/llama.dll +0 -0
  34. package/dist/bindings/AddonTypes.d.ts +1 -0
  35. package/dist/bindings/Llama.d.ts +2 -2
  36. package/dist/bindings/Llama.js +4 -1
  37. package/dist/bindings/Llama.js.map +1 -1
  38. package/dist/bindings/getLlama.d.ts +5 -2
  39. package/dist/bindings/getLlama.js +2 -0
  40. package/dist/bindings/getLlama.js.map +1 -1
  41. package/dist/bindings/types.d.ts +1 -0
  42. package/dist/bindings/types.js.map +1 -1
  43. package/dist/bindings/utils/getGpuTypesToUseForOption.d.ts +2 -1
  44. package/dist/bindings/utils/getGpuTypesToUseForOption.js +13 -4
  45. package/dist/bindings/utils/getGpuTypesToUseForOption.js.map +1 -1
  46. package/dist/cli/commands/ChatCommand.d.ts +1 -0
  47. package/dist/cli/commands/ChatCommand.js +12 -5
  48. package/dist/cli/commands/ChatCommand.js.map +1 -1
  49. package/dist/cli/commands/CompleteCommand.d.ts +1 -0
  50. package/dist/cli/commands/CompleteCommand.js +10 -3
  51. package/dist/cli/commands/CompleteCommand.js.map +1 -1
  52. package/dist/cli/commands/InfillCommand.d.ts +1 -0
  53. package/dist/cli/commands/InfillCommand.js +10 -3
  54. package/dist/cli/commands/InfillCommand.js.map +1 -1
  55. package/dist/cli/commands/inspect/commands/InspectMeasureCommand.d.ts +1 -0
  56. package/dist/cli/commands/inspect/commands/InspectMeasureCommand.js +19 -6
  57. package/dist/cli/commands/inspect/commands/InspectMeasureCommand.js.map +1 -1
  58. package/dist/cli/utils/printCommonInfoLines.js +4 -0
  59. package/dist/cli/utils/printCommonInfoLines.js.map +1 -1
  60. package/dist/evaluator/LlamaContext/LlamaContext.d.ts +1 -0
  61. package/dist/evaluator/LlamaContext/LlamaContext.js +15 -4
  62. package/dist/evaluator/LlamaContext/LlamaContext.js.map +1 -1
  63. package/dist/evaluator/LlamaContext/types.d.ts +14 -0
  64. package/dist/evaluator/LlamaModel/LlamaModel.d.ts +22 -0
  65. package/dist/evaluator/LlamaModel/LlamaModel.js +49 -4
  66. package/dist/evaluator/LlamaModel/LlamaModel.js.map +1 -1
  67. package/dist/gguf/insights/GgufInsights.d.ts +3 -1
  68. package/dist/gguf/insights/GgufInsights.js +18 -2
  69. package/dist/gguf/insights/GgufInsights.js.map +1 -1
  70. package/dist/gguf/insights/GgufInsightsConfigurationResolver.d.ts +6 -3
  71. package/dist/gguf/insights/GgufInsightsConfigurationResolver.js +8 -5
  72. package/dist/gguf/insights/GgufInsightsConfigurationResolver.js.map +1 -1
  73. package/dist/gguf/insights/utils/resolveContextContextSizeOption.d.ts +2 -1
  74. package/dist/gguf/insights/utils/resolveContextContextSizeOption.js +3 -1
  75. package/dist/gguf/insights/utils/resolveContextContextSizeOption.js.map +1 -1
  76. package/dist/gguf/insights/utils/resolveModelGpuLayersOption.d.ts +2 -1
  77. package/dist/gguf/insights/utils/resolveModelGpuLayersOption.js +16 -10
  78. package/dist/gguf/insights/utils/resolveModelGpuLayersOption.js.map +1 -1
  79. package/dist/gguf/types/GgufMetadataTypes.d.ts +15 -10
  80. package/dist/gguf/types/GgufMetadataTypes.js +15 -2
  81. package/dist/gguf/types/GgufMetadataTypes.js.map +1 -1
  82. package/dist/index.d.ts +3 -3
  83. package/dist/index.js.map +1 -1
  84. package/llama/CMakeLists.txt +9 -1
  85. package/llama/addon.cpp +4 -0
  86. package/llama/binariesGithubRelease.json +1 -1
  87. package/llama/gitRelease.bundle +0 -0
  88. package/llama/llama.cpp.info.json +1 -1
  89. package/package.json +3 -3
@@ -24,7 +24,7 @@ export class GgufInsightsConfigurationResolver {
24
24
  *
25
25
  * `contextSize` defaults to `4096` (if the model train context size is lower than this, the model train context size is used instead).
26
26
  */
27
- async scoreModelConfigurationCompatibility({ contextSize = Math.min(4096, this._ggufInsights.trainContextSize ?? 4096), embeddingContext = false } = {}, { getVramState = (() => this._ggufInsights._llama._vramOrchestrator.getMemoryState()), getRamState = (async () => ({ total: os.totalmem(), free: os.freemem() })), llamaVramPaddingSize = this._ggufInsights._llama.vramPaddingSize, llamaGpu = this._ggufInsights._llama.gpu, llamaSupportsGpuOffloading = this._ggufInsights._llama.supportsGpuOffloading } = {}) {
27
+ async scoreModelConfigurationCompatibility({ contextSize = Math.min(4096, this._ggufInsights.trainContextSize ?? 4096), embeddingContext = false, flashAttention = false } = {}, { getVramState = (() => this._ggufInsights._llama._vramOrchestrator.getMemoryState()), getRamState = (async () => ({ total: os.totalmem(), free: os.freemem() })), llamaVramPaddingSize = this._ggufInsights._llama.vramPaddingSize, llamaGpu = this._ggufInsights._llama.gpu, llamaSupportsGpuOffloading = this._ggufInsights._llama.supportsGpuOffloading } = {}) {
28
28
  const defaultTrainContextSize = 4096;
29
29
  const [vramState, ramState] = await Promise.all([
30
30
  getVramState(),
@@ -53,7 +53,8 @@ export class GgufInsightsConfigurationResolver {
53
53
  const estimatedContextResourceUsage = this._ggufInsights.estimateContextResourceRequirements({
54
54
  contextSize: resolvedContextSize,
55
55
  isEmbeddingContext: embeddingContext,
56
- modelGpuLayers: resolvedGpuLayers
56
+ modelGpuLayers: resolvedGpuLayers,
57
+ flashAttention
57
58
  });
58
59
  const rankPoints = {
59
60
  gpuLayers: 60,
@@ -94,17 +95,18 @@ export class GgufInsightsConfigurationResolver {
94
95
  }
95
96
  };
96
97
  }
97
- async resolveModelGpuLayers(gpuLayers, { ignoreMemorySafetyChecks = false, getVramState = (() => this._ggufInsights._llama._vramOrchestrator.getMemoryState()), llamaVramPaddingSize = this._ggufInsights._llama.vramPaddingSize, llamaGpu = this._ggufInsights._llama.gpu, llamaSupportsGpuOffloading = this._ggufInsights._llama.supportsGpuOffloading } = {}) {
98
+ async resolveModelGpuLayers(gpuLayers, { ignoreMemorySafetyChecks = false, getVramState = (() => this._ggufInsights._llama._vramOrchestrator.getMemoryState()), llamaVramPaddingSize = this._ggufInsights._llama.vramPaddingSize, llamaGpu = this._ggufInsights._llama.gpu, llamaSupportsGpuOffloading = this._ggufInsights._llama.supportsGpuOffloading, defaultContextFlashAttention = false } = {}) {
98
99
  return resolveModelGpuLayersOption(gpuLayers, {
99
100
  ggufInsights: this._ggufInsights,
100
101
  ignoreMemorySafetyChecks,
101
102
  getVramState,
102
103
  llamaVramPaddingSize,
103
104
  llamaGpu,
104
- llamaSupportsGpuOffloading
105
+ llamaSupportsGpuOffloading,
106
+ defaultContextFlashAttention
105
107
  });
106
108
  }
107
- async resolveContextContextSize(contextSize, { modelGpuLayers, batchSize, modelTrainContextSize, getVramState = (() => this._ggufInsights._llama._vramOrchestrator.getMemoryState()), llamaGpu = this._ggufInsights._llama.gpu, ignoreMemorySafetyChecks = false, isEmbeddingContext = false, sequences = getDefaultContextSequences() }) {
109
+ async resolveContextContextSize(contextSize, { modelGpuLayers, batchSize, modelTrainContextSize, flashAttention = false, getVramState = (() => this._ggufInsights._llama._vramOrchestrator.getMemoryState()), llamaGpu = this._ggufInsights._llama.gpu, ignoreMemorySafetyChecks = false, isEmbeddingContext = false, sequences = getDefaultContextSequences() }) {
108
110
  return await resolveContextContextSizeOption({
109
111
  contextSize,
110
112
  batchSize,
@@ -112,6 +114,7 @@ export class GgufInsightsConfigurationResolver {
112
114
  modelFileInsights: this._ggufInsights,
113
115
  modelGpuLayers,
114
116
  modelTrainContextSize,
117
+ flashAttention,
115
118
  getVramState,
116
119
  llamaGpu,
117
120
  ignoreMemorySafetyChecks,
@@ -1 +1 @@
1
- {"version":3,"file":"GgufInsightsConfigurationResolver.js","sourceRoot":"","sources":["../../../src/gguf/insights/GgufInsightsConfigurationResolver.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,MAAM,IAAI,CAAC;AAIpB,OAAO,EAAC,0BAA0B,EAAC,MAAM,8CAA8C,CAAC;AACxF,OAAO,EAAC,2BAA2B,EAAC,MAAM,wCAAwC,CAAC;AACnF,OAAO,EAAC,+BAA+B,EAAC,MAAM,4CAA4C,CAAC;AAC3F,OAAO,EAAC,WAAW,EAAC,MAAM,wBAAwB,CAAC;AAInD,MAAM,OAAO,iCAAiC;IAC1C,gBAAgB,CAAkB,aAAa,CAAe;IAE9D,YAAoB,YAA0B;QAC1C,IAAI,CAAC,aAAa,GAAG,YAAY,CAAC;IACtC,CAAC;IAED,IAAW,YAAY;QACnB,OAAO,IAAI,CAAC,aAAa,CAAC;IAC9B,CAAC;IAED;;;;;;;;;;;;OAYG;IACI,KAAK,CAAC,oCAAoC,CAAC,EAC9C,WAAW,GAAG,IAAI,CAAC,GAAG,CAAC,IAAI,EAAE,IAAI,CAAC,aAAa,CAAC,gBAAgB,IAAI,IAAI,CAAC,EACzE,gBAAgB,GAAG,KAAK,KAIxB,EAAE,EAAE,EACJ,YAAY,GAAG,CAAC,GAAG,EAAE,CAAC,IAAI,CAAC,aAAa,CAAC,MAAM,CAAC,iBAAiB,CAAC,cAAc,EAAE,CAAC,EACnF,WAAW,GAAG,CAAC,KAAK,IAAI,EAAE,CAAC,CAAC,EAAC,KAAK,EAAE,EAAE,CAAC,QAAQ,EAAE,EAAE,IAAI,EAAE,EAAE,CAAC,OAAO,EAAE,EAAC,CAAC,CAAC,EACxE,oBAAoB,GAAG,IAAI,CAAC,aAAa,CAAC,MAAM,CAAC,eAAe,EAChE,QAAQ,GAAG,IAAI,CAAC,aAAa,CAAC,MAAM,CAAC,GAAG,EACxC,0BAA0B,GAAG,IAAI,CAAC,aAAa,CAAC,MAAM,CAAC,qBAAqB,KAO5E,EAAE;QAiCF,MAAM,uBAAuB,GAAG,IAAI,CAAC;QAErC,MAAM,CACF,SAAS,EACT,QAAQ,CACX,GAAG,MAAM,OAAO,CAAC,GAAG,CAAC;YAClB,YAAY,EAAE;YACd,WAAW,EAAE;SAChB,CAAC,CAAC;QACH,MAAM,iBAAiB,GAAG,MAAM,IAAI,CAAC,qBAAqB,CAAC,MAAM,EAAE;YAC/D,YAAY,EAAE,KAAK,IAAI,EAAE,CAAC,SAAS;YACnC,oBAAoB;YACpB,QAAQ;YACR,0BAA0B;SAC7B,CAAC,CAAC;QACH,MAAM,SAAS,GAAG,0BAA0B,IAAI,QAAQ,KAAK,KAAK,CAAC;QACnE,MAAM,2BAA2B,GAAG,IAAI,CAAC,aAAa,CAAC,iCAAiC,CAAC;YACrF,SAAS,EAAE,iBAAiB;SAC/B,CAAC,CAAC;QAEH,MAAM,mBAAmB,GAAG,MAAM,IAAI,CAAC,yBAAyB,CAAC,MAAM,EAAE;YACrE,YAAY,EAAE,KAAK,IAAI,EAAE,CAAC,CAAC;gBACvB,KAAK,EAAE,SAAS,CAAC,KAAK;gBACtB,IAAI,EAAE,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,SAAS,CAAC,IAAI,GAAG,2BAA2B,CAAC,OAAO,CAAC;aAC1E,CAAC;YACF,QAAQ;YACR,kBAAkB,EAAE,gBAAgB;YACpC,cAAc,EAAE,iBAAiB;YACjC,qBAAqB,EAAE,IAAI,CAAC,aAAa,CAAC,gBAAgB,IAAI,uBAAuB;SACxF,CAAC,CAAC;QACH,MAAM,6BAA6B,GAAG,IAAI,CAAC,aAAa,CAAC,mCAAmC,CAAC;YACzF,WAAW,EAAE,mBAAmB;YAChC,kBAAkB,EAAE,gBAAgB;YACpC,cAAc,EAAE,iBAAiB;SACpC,CAAC,CAAC;QAEH,MAAM,UAAU,GAAG;YACf,SAAS,EAAE,EAAE;YACb,qBAAqB,EAAE,EAAE;YACzB,WAAW,EAAE,EAAE;YACf,iBAAiB,EAAE,EAAE;YACrB,qBAAqB,EAAE,EAAE,EAAE,sDAAsD;YACjF,gBAAgB,EAAE,EAAE;SACd,CAAC;QAEX,MAAM,eAAe,GAAG,UAAU,CAAC,SAAS,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,iBAAiB,GAAG,IAAI,CAAC,aAAa,CAAC,WAAW,CAAC,CAAC;QAC/G,MAAM,2BAA2B,GAAG,UAAU,CAAC,qBAAqB,GAAG,CACnE,iBAAiB,KAAK,IAAI,CAAC,aAAa,CAAC,WAAW,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAC/D,CAAC;QACF,MAAM,iBAAiB,GAAG,UAAU,CAAC,WAAW,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,mBAAmB,GAAG,WAAW,CAAC,CAAC;QAClG,MAAM,uBAAuB,GAAG,UAAU,CAAC,iBAAiB,GAAG,CAC3D,2BAA2B,CAAC,MAAM,IAAI,QAAQ,CAAC,IAAI;YAC/C,CAAC,CAAC,CAAC;YACH,CAAC,CAAC,2BAA2B,CAAC,MAAM,IAAI,QAAQ,CAAC,KAAK;gBAClD,CAAC,CAAC,GAAG;gBACL,CAAC,CAAC,CACE,GAAG,GAAG,IAAI,CAAC,GAAG,CACV,GAAG,EACH,GAAG,GAAG,CACF,CAAC,2BAA2B,CAAC,MAAM,GAAG,QAAQ,CAAC,KAAK,CAAC,GAAG,QAAQ,CAAC,KAAK,CACzE,CACJ,CACJ,CACZ,CAAC;QACF,MAAM,sBAAsB,GAAG,EAAE,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,mBAAmB,GAAG,WAAW,CAAC,GAAG,WAAW,CAAC,CAAC;QAE9G,MAAM,kBAAkB,GAAG,SAAS;YAChC,CAAC,CAAC,CACE,CAAC,eAAe,GAAG,2BAA2B,GAAG,iBAAiB,GAAG,uBAAuB,CAAC;gBAC7F,CAAC,UAAU,CAAC,SAAS,GAAG,UAAU,CAAC,qBAAqB,GAAG,UAAU,CAAC,WAAW,GAAG,UAAU,CAAC,iBAAiB,CAAC,CACpH;YACD,CAAC,CAAC,CACE,CAAC,iBAAiB,GAAG,uBAAuB,GAAG,6BAA6B,CAAC,IAAI,CAAC,aAAa,CAAC,SAAS,CAAC,CAAC;gBAC3G,CAAC,UAAU,CAAC,WAAW,GAAG,UAAU,CAAC,iBAAiB,GAAG,UAAU,CAAC,qBAAqB,CAAC,CAAC,CAAC;QACpG,MAAM,UAAU,GAAG,sBAAsB,GAAG,UAAU,CAAC,gBAAgB,CAAC;QAExE,OAAO;YACH,kBAAkB;YAClB,UAAU;YACV,UAAU,EAAE,kBAAkB,GAAG,UAAU;YAE3C,cAAc,EAAE;gBACZ,SAAS,EAAE,iBAAiB;gBAC5B,WAAW,EAAE,mBAAmB;gBAEhC,aAAa,EAAE,2BAA2B,CAAC,MAAM;gBACjD,eAAe,EAAE,6BAA6B,CAAC,MAAM;gBACrD,aAAa,EAAE,2BAA2B,CAAC,MAAM,GAAG,6BAA6B,CAAC,MAAM;gBAExF,cAAc,EAAE,2BAA2B,CAAC,OAAO;gBACnD,gBAAgB,EAAE,6BAA6B,CAAC,OAAO;gBACvD,cAAc,EAAE,2BAA2B,CAAC,OAAO,GAAG,6BAA6B,CAAC,OAAO;aAC9F;SACJ,CAAC;IACN,CAAC;IAEM,KAAK,CAAC,qBAAqB,CAAC,SAAyC,EAAE,EAC1E,wBAAwB,GAAG,KAAK,EAChC,YAAY,GAAG,CAAC,GAAG,EAAE,CAAC,IAAI,CAAC,aAAa,CAAC,MAAM,CAAC,iBAAiB,CAAC,cAAc,EAAE,CAAC,EACnF,oBAAoB,GAAG,IAAI,CAAC,aAAa,CAAC,MAAM,CAAC,eAAe,EAAE,QAAQ,GAAG,IAAI,CAAC,aAAa,CAAC,MAAM,CAAC,GAAG,EAC1G,0BAA0B,GAAG,IAAI,CAAC,aAAa,CAAC,MAAM,CAAC,qBAAqB,KAI5E,EAAE;QACF,OAAO,2BAA2B,CAAC,SAAS,EAAE;YAC1C,YAAY,EAAE,IAAI,CAAC,aAAa;YAChC,wBAAwB;YACxB,YAAY;YACZ,oBAAoB;YACpB,QAAQ;YACR,0BAA0B;SAC7B,CAAC,CAAC;IACP,CAAC;IAEM,KAAK,CAAC,yBAAyB,CAAC,WAA+C,EAAE,EACpF,cAAc,EACd,SAAS,EACT,qBAAqB,EACrB,YAAY,GAAG,CAAC,GAAG,EAAE,CAAC,IAAI,CAAC,aAAa,CAAC,MAAM,CAAC,iBAAiB,CAAC,cAAc,EAAE,CAAC,EACnF,QAAQ,GAAG,IAAI,CAAC,aAAa,CAAC,MAAM,CAAC,GAAG,EACxC,wBAAwB,GAAG,KAAK,EAChC,kBAAkB,GAAG,KAAK,EAC1B,SAAS,GAAG,0BAA0B,EAAE,EAU3C;QACG,OAAO,MAAM,+BAA+B,CAAC;YACzC,WAAW;YACX,SAAS;YACT,SAAS;YACT,iBAAiB,EAAE,IAAI,CAAC,aAAa;YACrC,cAAc;YACd,qBAAqB;YACrB,YAAY;YACZ,QAAQ;YACR,wBAAwB;YACxB,kBAAkB;SACrB,CAAC,CAAC;IACP,CAAC;IAED,gBAAgB;IACT,MAAM,CAAC,OAAO,CAAC,YAA0B;QAC5C,OAAO,IAAI,iCAAiC,CAAC,YAAY,CAAC,CAAC;IAC/D,CAAC;CACJ;AAED,SAAS,6BAA6B,CAAC,SAAiB;IACpD,MAAM,IAAI,GAAG,IAAI,CAAC,GAAG,CAAC,IAAI,EAAE,CAAC,CAAC,CAAC;IAC/B,OAAO,EAAE,GAAG,WAAW,CAAC,SAAS,EAAE,CAAC;YAChC,KAAK,EAAE,IAAI;YACX,GAAG,EAAE,IAAI,GAAG,GAAG;YACf,MAAM,EAAE,EAAE;SACb,EAAE;YACC,KAAK,EAAE,IAAI,GAAG,GAAG;YACjB,GAAG,EAAE,IAAI,GAAG,CAAC;YACb,MAAM,EAAE,EAAE;SACb,EAAE;YACC,KAAK,EAAE,IAAI,GAAG,CAAC;YACf,MAAM,EAAE,CAAC;SACZ,CAAC,CAAC,CAAC;AACR,CAAC"}
1
+ {"version":3,"file":"GgufInsightsConfigurationResolver.js","sourceRoot":"","sources":["../../../src/gguf/insights/GgufInsightsConfigurationResolver.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,MAAM,IAAI,CAAC;AAIpB,OAAO,EAAC,0BAA0B,EAAC,MAAM,8CAA8C,CAAC;AACxF,OAAO,EAAC,2BAA2B,EAAC,MAAM,wCAAwC,CAAC;AACnF,OAAO,EAAC,+BAA+B,EAAC,MAAM,4CAA4C,CAAC;AAC3F,OAAO,EAAC,WAAW,EAAC,MAAM,wBAAwB,CAAC;AAInD,MAAM,OAAO,iCAAiC;IAC1C,gBAAgB,CAAkB,aAAa,CAAe;IAE9D,YAAoB,YAA0B;QAC1C,IAAI,CAAC,aAAa,GAAG,YAAY,CAAC;IACtC,CAAC;IAED,IAAW,YAAY;QACnB,OAAO,IAAI,CAAC,aAAa,CAAC;IAC9B,CAAC;IAED;;;;;;;;;;;;OAYG;IACI,KAAK,CAAC,oCAAoC,CAAC,EAC9C,WAAW,GAAG,IAAI,CAAC,GAAG,CAAC,IAAI,EAAE,IAAI,CAAC,aAAa,CAAC,gBAAgB,IAAI,IAAI,CAAC,EACzE,gBAAgB,GAAG,KAAK,EACxB,cAAc,GAAG,KAAK,KAKtB,EAAE,EAAE,EACJ,YAAY,GAAG,CAAC,GAAG,EAAE,CAAC,IAAI,CAAC,aAAa,CAAC,MAAM,CAAC,iBAAiB,CAAC,cAAc,EAAE,CAAC,EACnF,WAAW,GAAG,CAAC,KAAK,IAAI,EAAE,CAAC,CAAC,EAAC,KAAK,EAAE,EAAE,CAAC,QAAQ,EAAE,EAAE,IAAI,EAAE,EAAE,CAAC,OAAO,EAAE,EAAC,CAAC,CAAC,EACxE,oBAAoB,GAAG,IAAI,CAAC,aAAa,CAAC,MAAM,CAAC,eAAe,EAChE,QAAQ,GAAG,IAAI,CAAC,aAAa,CAAC,MAAM,CAAC,GAAG,EACxC,0BAA0B,GAAG,IAAI,CAAC,aAAa,CAAC,MAAM,CAAC,qBAAqB,KAO5E,EAAE;QAiCF,MAAM,uBAAuB,GAAG,IAAI,CAAC;QAErC,MAAM,CACF,SAAS,EACT,QAAQ,CACX,GAAG,MAAM,OAAO,CAAC,GAAG,CAAC;YAClB,YAAY,EAAE;YACd,WAAW,EAAE;SAChB,CAAC,CAAC;QACH,MAAM,iBAAiB,GAAG,MAAM,IAAI,CAAC,qBAAqB,CAAC,MAAM,EAAE;YAC/D,YAAY,EAAE,KAAK,IAAI,EAAE,CAAC,SAAS;YACnC,oBAAoB;YACpB,QAAQ;YACR,0BAA0B;SAC7B,CAAC,CAAC;QACH,MAAM,SAAS,GAAG,0BAA0B,IAAI,QAAQ,KAAK,KAAK,CAAC;QACnE,MAAM,2BAA2B,GAAG,IAAI,CAAC,aAAa,CAAC,iCAAiC,CAAC;YACrF,SAAS,EAAE,iBAAiB;SAC/B,CAAC,CAAC;QAEH,MAAM,mBAAmB,GAAG,MAAM,IAAI,CAAC,yBAAyB,CAAC,MAAM,EAAE;YACrE,YAAY,EAAE,KAAK,IAAI,EAAE,CAAC,CAAC;gBACvB,KAAK,EAAE,SAAS,CAAC,KAAK;gBACtB,IAAI,EAAE,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,SAAS,CAAC,IAAI,GAAG,2BAA2B,CAAC,OAAO,CAAC;aAC1E,CAAC;YACF,QAAQ;YACR,kBAAkB,EAAE,gBAAgB;YACpC,cAAc,EAAE,iBAAiB;YACjC,qBAAqB,EAAE,IAAI,CAAC,aAAa,CAAC,gBAAgB,IAAI,uBAAuB;SACxF,CAAC,CAAC;QACH,MAAM,6BAA6B,GAAG,IAAI,CAAC,aAAa,CAAC,mCAAmC,CAAC;YACzF,WAAW,EAAE,mBAAmB;YAChC,kBAAkB,EAAE,gBAAgB;YACpC,cAAc,EAAE,iBAAiB;YACjC,cAAc;SACjB,CAAC,CAAC;QAEH,MAAM,UAAU,GAAG;YACf,SAAS,EAAE,EAAE;YACb,qBAAqB,EAAE,EAAE;YACzB,WAAW,EAAE,EAAE;YACf,iBAAiB,EAAE,EAAE;YACrB,qBAAqB,EAAE,EAAE,EAAE,sDAAsD;YACjF,gBAAgB,EAAE,EAAE;SACd,CAAC;QAEX,MAAM,eAAe,GAAG,UAAU,CAAC,SAAS,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,iBAAiB,GAAG,IAAI,CAAC,aAAa,CAAC,WAAW,CAAC,CAAC;QAC/G,MAAM,2BAA2B,GAAG,UAAU,CAAC,qBAAqB,GAAG,CACnE,iBAAiB,KAAK,IAAI,CAAC,aAAa,CAAC,WAAW,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAC/D,CAAC;QACF,MAAM,iBAAiB,GAAG,UAAU,CAAC,WAAW,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,mBAAmB,GAAG,WAAW,CAAC,CAAC;QAClG,MAAM,uBAAuB,GAAG,UAAU,CAAC,iBAAiB,GAAG,CAC3D,2BAA2B,CAAC,MAAM,IAAI,QAAQ,CAAC,IAAI;YAC/C,CAAC,CAAC,CAAC;YACH,CAAC,CAAC,2BAA2B,CAAC,MAAM,IAAI,QAAQ,CAAC,KAAK;gBAClD,CAAC,CAAC,GAAG;gBACL,CAAC,CAAC,CACE,GAAG,GAAG,IAAI,CAAC,GAAG,CACV,GAAG,EACH,GAAG,GAAG,CACF,CAAC,2BAA2B,CAAC,MAAM,GAAG,QAAQ,CAAC,KAAK,CAAC,GAAG,QAAQ,CAAC,KAAK,CACzE,CACJ,CACJ,CACZ,CAAC;QACF,MAAM,sBAAsB,GAAG,EAAE,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,mBAAmB,GAAG,WAAW,CAAC,GAAG,WAAW,CAAC,CAAC;QAE9G,MAAM,kBAAkB,GAAG,SAAS;YAChC,CAAC,CAAC,CACE,CAAC,eAAe,GAAG,2BAA2B,GAAG,iBAAiB,GAAG,uBAAuB,CAAC;gBAC7F,CAAC,UAAU,CAAC,SAAS,GAAG,UAAU,CAAC,qBAAqB,GAAG,UAAU,CAAC,WAAW,GAAG,UAAU,CAAC,iBAAiB,CAAC,CACpH;YACD,CAAC,CAAC,CACE,CAAC,iBAAiB,GAAG,uBAAuB,GAAG,6BAA6B,CAAC,IAAI,CAAC,aAAa,CAAC,SAAS,CAAC,CAAC;gBAC3G,CAAC,UAAU,CAAC,WAAW,GAAG,UAAU,CAAC,iBAAiB,GAAG,UAAU,CAAC,qBAAqB,CAAC,CAAC,CAAC;QACpG,MAAM,UAAU,GAAG,sBAAsB,GAAG,UAAU,CAAC,gBAAgB,CAAC;QAExE,OAAO;YACH,kBAAkB;YAClB,UAAU;YACV,UAAU,EAAE,kBAAkB,GAAG,UAAU;YAE3C,cAAc,EAAE;gBACZ,SAAS,EAAE,iBAAiB;gBAC5B,WAAW,EAAE,mBAAmB;gBAEhC,aAAa,EAAE,2BAA2B,CAAC,MAAM;gBACjD,eAAe,EAAE,6BAA6B,CAAC,MAAM;gBACrD,aAAa,EAAE,2BAA2B,CAAC,MAAM,GAAG,6BAA6B,CAAC,MAAM;gBAExF,cAAc,EAAE,2BAA2B,CAAC,OAAO;gBACnD,gBAAgB,EAAE,6BAA6B,CAAC,OAAO;gBACvD,cAAc,EAAE,2BAA2B,CAAC,OAAO,GAAG,6BAA6B,CAAC,OAAO;aAC9F;SACJ,CAAC;IACN,CAAC;IAEM,KAAK,CAAC,qBAAqB,CAAC,SAAyC,EAAE,EAC1E,wBAAwB,GAAG,KAAK,EAChC,YAAY,GAAG,CAAC,GAAG,EAAE,CAAC,IAAI,CAAC,aAAa,CAAC,MAAM,CAAC,iBAAiB,CAAC,cAAc,EAAE,CAAC,EACnF,oBAAoB,GAAG,IAAI,CAAC,aAAa,CAAC,MAAM,CAAC,eAAe,EAAE,QAAQ,GAAG,IAAI,CAAC,aAAa,CAAC,MAAM,CAAC,GAAG,EAC1G,0BAA0B,GAAG,IAAI,CAAC,aAAa,CAAC,MAAM,CAAC,qBAAqB,EAC5E,4BAA4B,GAAG,KAAK,KAIpC,EAAE;QACF,OAAO,2BAA2B,CAAC,SAAS,EAAE;YAC1C,YAAY,EAAE,IAAI,CAAC,aAAa;YAChC,wBAAwB;YACxB,YAAY;YACZ,oBAAoB;YACpB,QAAQ;YACR,0BAA0B;YAC1B,4BAA4B;SAC/B,CAAC,CAAC;IACP,CAAC;IAEM,KAAK,CAAC,yBAAyB,CAAC,WAA+C,EAAE,EACpF,cAAc,EACd,SAAS,EACT,qBAAqB,EACrB,cAAc,GAAG,KAAK,EACtB,YAAY,GAAG,CAAC,GAAG,EAAE,CAAC,IAAI,CAAC,aAAa,CAAC,MAAM,CAAC,iBAAiB,CAAC,cAAc,EAAE,CAAC,EACnF,QAAQ,GAAG,IAAI,CAAC,aAAa,CAAC,MAAM,CAAC,GAAG,EACxC,wBAAwB,GAAG,KAAK,EAChC,kBAAkB,GAAG,KAAK,EAC1B,SAAS,GAAG,0BAA0B,EAAE,EAW3C;QACG,OAAO,MAAM,+BAA+B,CAAC;YACzC,WAAW;YACX,SAAS;YACT,SAAS;YACT,iBAAiB,EAAE,IAAI,CAAC,aAAa;YACrC,cAAc;YACd,qBAAqB;YACrB,cAAc;YACd,YAAY;YACZ,QAAQ;YACR,wBAAwB;YACxB,kBAAkB;SACrB,CAAC,CAAC;IACP,CAAC;IAED,gBAAgB;IACT,MAAM,CAAC,OAAO,CAAC,YAA0B;QAC5C,OAAO,IAAI,iCAAiC,CAAC,YAAY,CAAC,CAAC;IAC/D,CAAC;CACJ;AAED,SAAS,6BAA6B,CAAC,SAAiB;IACpD,MAAM,IAAI,GAAG,IAAI,CAAC,GAAG,CAAC,IAAI,EAAE,CAAC,CAAC,CAAC;IAC/B,OAAO,EAAE,GAAG,WAAW,CAAC,SAAS,EAAE,CAAC;YAChC,KAAK,EAAE,IAAI;YACX,GAAG,EAAE,IAAI,GAAG,GAAG;YACf,MAAM,EAAE,EAAE;SACb,EAAE;YACC,KAAK,EAAE,IAAI,GAAG,GAAG;YACjB,GAAG,EAAE,IAAI,GAAG,CAAC;YACb,MAAM,EAAE,EAAE;SACb,EAAE;YACC,KAAK,EAAE,IAAI,GAAG,CAAC;YACf,MAAM,EAAE,CAAC;SACZ,CAAC,CAAC,CAAC;AACR,CAAC"}
@@ -1,13 +1,14 @@
1
1
  import { LlamaContextOptions } from "../../../evaluator/LlamaContext/types.js";
2
2
  import { GgufInsights } from "../GgufInsights.js";
3
3
  import { BuildGpu } from "../../../bindings/types.js";
4
- export declare function resolveContextContextSizeOption({ contextSize, batchSize, sequences, modelFileInsights, modelGpuLayers, modelTrainContextSize, getVramState, llamaGpu, ignoreMemorySafetyChecks, isEmbeddingContext }: {
4
+ export declare function resolveContextContextSizeOption({ contextSize, batchSize, sequences, modelFileInsights, modelGpuLayers, modelTrainContextSize, flashAttention, getVramState, llamaGpu, ignoreMemorySafetyChecks, isEmbeddingContext }: {
5
5
  contextSize?: LlamaContextOptions["contextSize"];
6
6
  batchSize?: LlamaContextOptions["batchSize"];
7
7
  sequences: number;
8
8
  modelFileInsights: GgufInsights;
9
9
  modelGpuLayers: number;
10
10
  modelTrainContextSize: number;
11
+ flashAttention: boolean;
11
12
  getVramState(): Promise<{
12
13
  total: number;
13
14
  free: number;
@@ -1,6 +1,6 @@
1
1
  import { minAllowedContextSizeInCalculations } from "../../../config.js";
2
2
  import { getDefaultContextBatchSize, getDefaultModelContextSize } from "../../../evaluator/LlamaContext/LlamaContext.js";
3
- export async function resolveContextContextSizeOption({ contextSize, batchSize, sequences, modelFileInsights, modelGpuLayers, modelTrainContextSize, getVramState, llamaGpu, ignoreMemorySafetyChecks = false, isEmbeddingContext = false }) {
3
+ export async function resolveContextContextSizeOption({ contextSize, batchSize, sequences, modelFileInsights, modelGpuLayers, modelTrainContextSize, flashAttention, getVramState, llamaGpu, ignoreMemorySafetyChecks = false, isEmbeddingContext = false }) {
4
4
  if (contextSize == null)
5
5
  contextSize = "auto";
6
6
  if (typeof contextSize === "number") {
@@ -13,6 +13,7 @@ export async function resolveContextContextSizeOption({ contextSize, batchSize,
13
13
  batchSize: batchSize ?? getDefaultContextBatchSize({ contextSize: resolvedContextSize, sequences }),
14
14
  modelGpuLayers: modelGpuLayers,
15
15
  sequences,
16
+ flashAttention,
16
17
  isEmbeddingContext
17
18
  }).gpuVram;
18
19
  if (contextVram > vramState.free)
@@ -40,6 +41,7 @@ export async function resolveContextContextSizeOption({ contextSize, batchSize,
40
41
  batchSize: batchSize ?? getDefaultContextBatchSize({ contextSize: testContextSize, sequences }),
41
42
  modelGpuLayers: modelGpuLayers,
42
43
  sequences,
44
+ flashAttention,
43
45
  isEmbeddingContext
44
46
  }).gpuVram;
45
47
  if (contextVram <= freeVram) {
@@ -1 +1 @@
1
- {"version":3,"file":"resolveContextContextSizeOption.js","sourceRoot":"","sources":["../../../../src/gguf/insights/utils/resolveContextContextSizeOption.ts"],"names":[],"mappings":"AAGA,OAAO,EAAC,mCAAmC,EAAC,MAAM,oBAAoB,CAAC;AACvE,OAAO,EAAC,0BAA0B,EAAE,0BAA0B,EAAC,MAAM,iDAAiD,CAAC;AAEvH,MAAM,CAAC,KAAK,UAAU,+BAA+B,CAAC,EAClD,WAAW,EAAE,SAAS,EAAE,SAAS,EAAE,iBAAiB,EAAE,cAAc,EAAE,qBAAqB,EAAE,YAAY,EAAE,QAAQ,EACnH,wBAAwB,GAAG,KAAK,EAAE,kBAAkB,GAAG,KAAK,EAY/D;IACG,IAAI,WAAW,IAAI,IAAI;QACnB,WAAW,GAAG,MAAM,CAAC;IAEzB,IAAI,OAAO,WAAW,KAAK,QAAQ,EAAE,CAAC;QAClC,MAAM,mBAAmB,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,IAAI,CAAC,KAAK,CAAC,WAAW,CAAC,CAAC,CAAC;QAEjE,IAAI,wBAAwB;YACxB,OAAO,mBAAmB,CAAC;QAE/B,MAAM,SAAS,GAAG,MAAM,YAAY,EAAE,CAAC;QACvC,MAAM,WAAW,GAAG,iBAAiB,CAAC,mCAAmC,CAAC;YACtE,WAAW,EAAE,mBAAmB;YAChC,SAAS,EAAE,SAAS,IAAI,0BAA0B,CAAC,EAAC,WAAW,EAAE,mBAAmB,EAAE,SAAS,EAAC,CAAC;YACjG,cAAc,EAAE,cAAc;YAC9B,SAAS;YACT,kBAAkB;SACrB,CAAC,CAAC,OAAO,CAAC;QAEX,IAAI,WAAW,GAAG,SAAS,CAAC,IAAI;YAC5B,MAAM,IAAI,KAAK,CAAC,uBAAuB,mBAAmB,GAAG,SAAS,GAAG,CAAC,CAAC,CAAC,CAAC,SAAS,SAAS,YAAY,CAAC,CAAC,CAAC,EAAE,sCAAsC,CAAC,CAAC;QAE5J,OAAO,mBAAmB,CAAC;IAC/B,CAAC;SAAM,IAAI,WAAW,KAAK,MAAM,IAAI,OAAO,WAAW,KAAK,QAAQ,EAAE,CAAC;QACnE,IAAI,QAAQ,KAAK,KAAK;YAClB,OAAO,qBAAqB,CAAC;QAEjC,MAAM,SAAS,GAAG,MAAM,YAAY,EAAE,CAAC;QAEvC,IAAI,SAAS,CAAC,KAAK,KAAK,CAAC;YACrB,OAAO,qBAAqB,CAAC;QAEjC,MAAM,QAAQ,GAAG,SAAS,CAAC,IAAI,CAAC;QAEhC,MAAM,cAAc,GAAG,WAAW,KAAK,MAAM;YACzC,CAAC,CAAC,0BAA0B,CAAC,EAAC,gBAAgB,EAAE,qBAAqB,EAAC,CAAC;YACvE,CAAC,CAAC,IAAI,CAAC,GAAG,CACN,WAAW,CAAC,GAAG,IAAI,0BAA0B,CAAC,EAAC,gBAAgB,EAAE,qBAAqB,EAAC,CAAC,EACxF,0BAA0B,CAAC,EAAC,gBAAgB,EAAE,qBAAqB,EAAC,CAAC,CACxE,CAAC;QAEN,MAAM,cAAc,GAAG,WAAW,KAAK,MAAM;YACzC,CAAC,CAAC,mCAAmC;YACrC,CAAC,CAAC,IAAI,CAAC,GAAG,CACN,WAAW,CAAC,GAAG,IAAI,mCAAmC,EACtD,mCAAmC,CACtC,CAAC;QAEN,IAAI,4BAA4B,GAAkB,IAAI,CAAC;QACvD,IAAI,IAAI,GAAG,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,IAAI,CAAC,KAAK,CAAC,CAAC,cAAc,GAAG,cAAc,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC;QAC3E,KAAK,IAAI,eAAe,GAAG,cAAc,EAAE,eAAe,IAAI,cAAc,IAAI,eAAe,IAAI,cAAc,GAAG,CAAC;YACjH,MAAM,WAAW,GAAG,iBAAiB,CAAC,mCAAmC,CAAC;gBACtE,WAAW,EAAE,eAAe;gBAC5B,SAAS,EAAE,SAAS,IAAI,0BAA0B,CAAC,EAAC,WAAW,EAAE,eAAe,EAAE,SAAS,EAAC,CAAC;gBAC7F,cAAc,EAAE,cAAc;gBAC9B,SAAS;gBACT,kBAAkB;aACrB,CAAC,CAAC,OAAO,CAAC;YAEX,IAAI,WAAW,IAAI,QAAQ,EAAE,CAAC;gBAC1B,IAAI,4BAA4B,IAAI,IAAI,IAAI,eAAe,GAAG,4BAA4B,EAAE,CAAC;oBACzF,4BAA4B,GAAG,eAAe,CAAC;oBAE/C,IAAI,IAAI,KAAK,CAAC,CAAC;wBACX,MAAM;yBACL,IAAI,IAAI,GAAG,CAAC;wBACb,IAAI,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,IAAI,CAAC,KAAK,CAAC,CAAC,IAAI,GAAG,CAAC,CAAC,CAAC,CAAC;gBAClD,CAAC;YACL,CAAC;iBAAM,IAAI,IAAI,GAAG,CAAC;gBACf,IAAI,GAAG,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,IAAI,CAAC,KAAK,CAAC,IAAI,GAAG,CAAC,CAAC,CAAC,CAAC;YAE9C,IAAI,eAAe,IAAI,cAAc,IAAI,IAAI,KAAK,CAAC,CAAC;gBAChD,MAAM;YAEV,eAAe,IAAI,IAAI,CAAC;YACxB,IAAI,eAAe,GAAG,cAAc,EAAE,CAAC;gBACnC,eAAe,GAAG,cAAc,CAAC;gBACjC,IAAI,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,GAAG,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC;YACvD,CAAC;iBAAM,IAAI,eAAe,GAAG,cAAc,EAAE,CAAC;gBAC1C,eAAe,GAAG,cAAc,CAAC;gBACjC,IAAI,GAAG,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,GAAG,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC;YACxD,CAAC;QACL,CAAC;QAED,IAAI,4BAA4B,IAAI,IAAI;YACpC,OAAO,4BAA4B,CAAC;QAExC,IAAI,wBAAwB;YACxB,OAAO,cAAc,CAAC;QAE1B,MAAM,IAAI,KAAK,CAAC,8DAA8D,cAAc,GAAG,SAAS,GAAG,CAAC,CAAC,CAAC,CAAC,SAAS,SAAS,YAAY,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC;IAC1J,CAAC;IAED,MAAM,IAAI,KAAK,CAAC,0BAA0B,WAAW,GAAG,CAAC,CAAC;AAC9D,CAAC"}
1
+ {"version":3,"file":"resolveContextContextSizeOption.js","sourceRoot":"","sources":["../../../../src/gguf/insights/utils/resolveContextContextSizeOption.ts"],"names":[],"mappings":"AAGA,OAAO,EAAC,mCAAmC,EAAC,MAAM,oBAAoB,CAAC;AACvE,OAAO,EAAC,0BAA0B,EAAE,0BAA0B,EAAC,MAAM,iDAAiD,CAAC;AAEvH,MAAM,CAAC,KAAK,UAAU,+BAA+B,CAAC,EAClD,WAAW,EAAE,SAAS,EAAE,SAAS,EAAE,iBAAiB,EAAE,cAAc,EAAE,qBAAqB,EAAE,cAAc,EAAE,YAAY,EAAE,QAAQ,EACnI,wBAAwB,GAAG,KAAK,EAAE,kBAAkB,GAAG,KAAK,EAa/D;IACG,IAAI,WAAW,IAAI,IAAI;QACnB,WAAW,GAAG,MAAM,CAAC;IAEzB,IAAI,OAAO,WAAW,KAAK,QAAQ,EAAE,CAAC;QAClC,MAAM,mBAAmB,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,IAAI,CAAC,KAAK,CAAC,WAAW,CAAC,CAAC,CAAC;QAEjE,IAAI,wBAAwB;YACxB,OAAO,mBAAmB,CAAC;QAE/B,MAAM,SAAS,GAAG,MAAM,YAAY,EAAE,CAAC;QACvC,MAAM,WAAW,GAAG,iBAAiB,CAAC,mCAAmC,CAAC;YACtE,WAAW,EAAE,mBAAmB;YAChC,SAAS,EAAE,SAAS,IAAI,0BAA0B,CAAC,EAAC,WAAW,EAAE,mBAAmB,EAAE,SAAS,EAAC,CAAC;YACjG,cAAc,EAAE,cAAc;YAC9B,SAAS;YACT,cAAc;YACd,kBAAkB;SACrB,CAAC,CAAC,OAAO,CAAC;QAEX,IAAI,WAAW,GAAG,SAAS,CAAC,IAAI;YAC5B,MAAM,IAAI,KAAK,CAAC,uBAAuB,mBAAmB,GAAG,SAAS,GAAG,CAAC,CAAC,CAAC,CAAC,SAAS,SAAS,YAAY,CAAC,CAAC,CAAC,EAAE,sCAAsC,CAAC,CAAC;QAE5J,OAAO,mBAAmB,CAAC;IAC/B,CAAC;SAAM,IAAI,WAAW,KAAK,MAAM,IAAI,OAAO,WAAW,KAAK,QAAQ,EAAE,CAAC;QACnE,IAAI,QAAQ,KAAK,KAAK;YAClB,OAAO,qBAAqB,CAAC;QAEjC,MAAM,SAAS,GAAG,MAAM,YAAY,EAAE,CAAC;QAEvC,IAAI,SAAS,CAAC,KAAK,KAAK,CAAC;YACrB,OAAO,qBAAqB,CAAC;QAEjC,MAAM,QAAQ,GAAG,SAAS,CAAC,IAAI,CAAC;QAEhC,MAAM,cAAc,GAAG,WAAW,KAAK,MAAM;YACzC,CAAC,CAAC,0BAA0B,CAAC,EAAC,gBAAgB,EAAE,qBAAqB,EAAC,CAAC;YACvE,CAAC,CAAC,IAAI,CAAC,GAAG,CACN,WAAW,CAAC,GAAG,IAAI,0BAA0B,CAAC,EAAC,gBAAgB,EAAE,qBAAqB,EAAC,CAAC,EACxF,0BAA0B,CAAC,EAAC,gBAAgB,EAAE,qBAAqB,EAAC,CAAC,CACxE,CAAC;QAEN,MAAM,cAAc,GAAG,WAAW,KAAK,MAAM;YACzC,CAAC,CAAC,mCAAmC;YACrC,CAAC,CAAC,IAAI,CAAC,GAAG,CACN,WAAW,CAAC,GAAG,IAAI,mCAAmC,EACtD,mCAAmC,CACtC,CAAC;QAEN,IAAI,4BAA4B,GAAkB,IAAI,CAAC;QACvD,IAAI,IAAI,GAAG,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,IAAI,CAAC,KAAK,CAAC,CAAC,cAAc,GAAG,cAAc,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC;QAC3E,KAAK,IAAI,eAAe,GAAG,cAAc,EAAE,eAAe,IAAI,cAAc,IAAI,eAAe,IAAI,cAAc,GAAG,CAAC;YACjH,MAAM,WAAW,GAAG,iBAAiB,CAAC,mCAAmC,CAAC;gBACtE,WAAW,EAAE,eAAe;gBAC5B,SAAS,EAAE,SAAS,IAAI,0BAA0B,CAAC,EAAC,WAAW,EAAE,eAAe,EAAE,SAAS,EAAC,CAAC;gBAC7F,cAAc,EAAE,cAAc;gBAC9B,SAAS;gBACT,cAAc;gBACd,kBAAkB;aACrB,CAAC,CAAC,OAAO,CAAC;YAEX,IAAI,WAAW,IAAI,QAAQ,EAAE,CAAC;gBAC1B,IAAI,4BAA4B,IAAI,IAAI,IAAI,eAAe,GAAG,4BAA4B,EAAE,CAAC;oBACzF,4BAA4B,GAAG,eAAe,CAAC;oBAE/C,IAAI,IAAI,KAAK,CAAC,CAAC;wBACX,MAAM;yBACL,IAAI,IAAI,GAAG,CAAC;wBACb,IAAI,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,IAAI,CAAC,KAAK,CAAC,CAAC,IAAI,GAAG,CAAC,CAAC,CAAC,CAAC;gBAClD,CAAC;YACL,CAAC;iBAAM,IAAI,IAAI,GAAG,CAAC;gBACf,IAAI,GAAG,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,IAAI,CAAC,KAAK,CAAC,IAAI,GAAG,CAAC,CAAC,CAAC,CAAC;YAE9C,IAAI,eAAe,IAAI,cAAc,IAAI,IAAI,KAAK,CAAC,CAAC;gBAChD,MAAM;YAEV,eAAe,IAAI,IAAI,CAAC;YACxB,IAAI,eAAe,GAAG,cAAc,EAAE,CAAC;gBACnC,eAAe,GAAG,cAAc,CAAC;gBACjC,IAAI,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,GAAG,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC;YACvD,CAAC;iBAAM,IAAI,eAAe,GAAG,cAAc,EAAE,CAAC;gBAC1C,eAAe,GAAG,cAAc,CAAC;gBACjC,IAAI,GAAG,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,GAAG,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC;YACxD,CAAC;QACL,CAAC;QAED,IAAI,4BAA4B,IAAI,IAAI;YACpC,OAAO,4BAA4B,CAAC;QAExC,IAAI,wBAAwB;YACxB,OAAO,cAAc,CAAC;QAE1B,MAAM,IAAI,KAAK,CAAC,8DAA8D,cAAc,GAAG,SAAS,GAAG,CAAC,CAAC,CAAC,CAAC,SAAS,SAAS,YAAY,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC;IAC1J,CAAC;IAED,MAAM,IAAI,KAAK,CAAC,0BAA0B,WAAW,GAAG,CAAC,CAAC;AAC9D,CAAC"}
@@ -1,7 +1,7 @@
1
1
  import { LlamaModelOptions } from "../../../evaluator/LlamaModel/LlamaModel.js";
2
2
  import { BuildGpu } from "../../../bindings/types.js";
3
3
  import type { GgufInsights } from "../GgufInsights.js";
4
- export declare function resolveModelGpuLayersOption(gpuLayers: LlamaModelOptions["gpuLayers"], { ggufInsights, ignoreMemorySafetyChecks, getVramState, llamaVramPaddingSize, llamaGpu, llamaSupportsGpuOffloading }: {
4
+ export declare function resolveModelGpuLayersOption(gpuLayers: LlamaModelOptions["gpuLayers"], { ggufInsights, ignoreMemorySafetyChecks, getVramState, llamaVramPaddingSize, llamaGpu, llamaSupportsGpuOffloading, defaultContextFlashAttention }: {
5
5
  ggufInsights: GgufInsights;
6
6
  ignoreMemorySafetyChecks?: boolean;
7
7
  getVramState(): Promise<{
@@ -11,4 +11,5 @@ export declare function resolveModelGpuLayersOption(gpuLayers: LlamaModelOptions
11
11
  llamaVramPaddingSize: number;
12
12
  llamaGpu: BuildGpu;
13
13
  llamaSupportsGpuOffloading: boolean;
14
+ defaultContextFlashAttention: boolean;
14
15
  }): Promise<number>;
@@ -4,7 +4,7 @@ import { getDefaultContextBatchSize, getDefaultModelContextSize } from "../../..
4
4
  import { minAllowedContextSizeInCalculations } from "../../../config.js";
5
5
  import { scoreLevels } from "./scoreLevels.js";
6
6
  const fitContextExtraMemoryPaddingPercentage = 0.5;
7
- export async function resolveModelGpuLayersOption(gpuLayers, { ggufInsights, ignoreMemorySafetyChecks = false, getVramState, llamaVramPaddingSize, llamaGpu, llamaSupportsGpuOffloading }) {
7
+ export async function resolveModelGpuLayersOption(gpuLayers, { ggufInsights, ignoreMemorySafetyChecks = false, getVramState, llamaVramPaddingSize, llamaGpu, llamaSupportsGpuOffloading, defaultContextFlashAttention }) {
8
8
  if (gpuLayers == null)
9
9
  gpuLayers = "auto";
10
10
  if (!llamaSupportsGpuOffloading)
@@ -19,7 +19,8 @@ export async function resolveModelGpuLayersOption(gpuLayers, { ggufInsights, ign
19
19
  const maxLayersRequirements = getVramRequiredForGpuLayers({
20
20
  gpuLayers: resolvedGpuLayers,
21
21
  ggufInsights,
22
- currentVram: vramState.free
22
+ currentVram: vramState.free,
23
+ defaultContextFlashAttention
23
24
  });
24
25
  if (maxLayersRequirements == null)
25
26
  throw new InsufficientMemoryError("Not enough VRAM to fit the model with the specified settings");
@@ -48,7 +49,8 @@ export async function resolveModelGpuLayersOption(gpuLayers, { ggufInsights, ign
48
49
  : undefined,
49
50
  maxGpuLayers: typeof gpuLayers === "object"
50
51
  ? gpuLayers.max
51
- : undefined
52
+ : undefined,
53
+ defaultContextFlashAttention
52
54
  });
53
55
  const hasGpuLayersRequirements = typeof gpuLayers === "object" &&
54
56
  (gpuLayers.min != null || gpuLayers.max != null || gpuLayers.fitContext?.contextSize != null);
@@ -58,7 +60,7 @@ export async function resolveModelGpuLayersOption(gpuLayers, { ggufInsights, ign
58
60
  }
59
61
  throw new Error(`Invalid gpuLayers value: ${gpuLayers}`);
60
62
  }
61
- function getBestGpuLayersForFreeVram({ ggufInsights, freeVram, fitContext, minGpuLayers, maxGpuLayers }) {
63
+ function getBestGpuLayersForFreeVram({ ggufInsights, freeVram, fitContext, minGpuLayers, maxGpuLayers, defaultContextFlashAttention }) {
62
64
  return findBestOption({
63
65
  *generator() {
64
66
  const minLayers = Math.floor(Math.max(0, minGpuLayers ?? 0));
@@ -74,7 +76,8 @@ function getBestGpuLayersForFreeVram({ ggufInsights, freeVram, fitContext, minGp
74
76
  gpuLayers: option.gpuLayers,
75
77
  ggufInsights,
76
78
  currentVram: freeVram,
77
- fitContext
79
+ fitContext,
80
+ defaultContextFlashAttention
78
81
  });
79
82
  if (layersRequirements == null)
80
83
  return null;
@@ -121,7 +124,7 @@ function scoreGpuLayersAndContextCombination({ gpuLayers, contextSize }, { total
121
124
  }
122
125
  return scoreGpuLayers() + scoreContextSize();
123
126
  }
124
- function getVramRequiredForGpuLayers({ gpuLayers, ggufInsights, currentVram, fitContext }) {
127
+ function getVramRequiredForGpuLayers({ gpuLayers, ggufInsights, currentVram, fitContext, defaultContextFlashAttention = false }) {
125
128
  const modelVram = ggufInsights.estimateModelResourceRequirements({ gpuLayers }).gpuVram;
126
129
  if (modelVram > currentVram)
127
130
  return null;
@@ -131,7 +134,8 @@ function getVramRequiredForGpuLayers({ gpuLayers, ggufInsights, currentVram, fit
131
134
  batchSize: getDefaultContextBatchSize({ contextSize: fitContext.contextSize, sequences: 1 }),
132
135
  modelGpuLayers: gpuLayers,
133
136
  sequences: 1,
134
- isEmbeddingContext: fitContext.embeddingContext ?? false
137
+ isEmbeddingContext: fitContext.embeddingContext ?? false,
138
+ flashAttention: defaultContextFlashAttention
135
139
  }).gpuVram;
136
140
  const totalVram = modelVram + contextVram;
137
141
  if (totalVram > currentVram)
@@ -146,7 +150,8 @@ function getVramRequiredForGpuLayers({ gpuLayers, ggufInsights, currentVram, fit
146
150
  gpuLayers,
147
151
  ggufInsights,
148
152
  vram: currentVram - modelVram,
149
- isEmbeddingContext: fitContext?.embeddingContext ?? false
153
+ isEmbeddingContext: fitContext?.embeddingContext ?? false,
154
+ flashAttention: defaultContextFlashAttention
150
155
  });
151
156
  if (maxContext == null || modelVram + maxContext.vram > currentVram)
152
157
  return null;
@@ -156,7 +161,7 @@ function getVramRequiredForGpuLayers({ gpuLayers, ggufInsights, currentVram, fit
156
161
  totalVram: modelVram + maxContext.vram
157
162
  };
158
163
  }
159
- function findMaxPossibleContextSizeForVram({ gpuLayers, ggufInsights, vram, isEmbeddingContext }) {
164
+ function findMaxPossibleContextSizeForVram({ gpuLayers, ggufInsights, vram, isEmbeddingContext, flashAttention }) {
160
165
  const maxContextSize = getDefaultModelContextSize({ trainContextSize: ggufInsights.trainContextSize });
161
166
  for (let contextSize = maxContextSize; contextSize >= minAllowedContextSizeInCalculations; contextSize--) {
162
167
  const contextVram = ggufInsights.estimateContextResourceRequirements({
@@ -164,7 +169,8 @@ function findMaxPossibleContextSizeForVram({ gpuLayers, ggufInsights, vram, isEm
164
169
  batchSize: getDefaultContextBatchSize({ contextSize, sequences: 1 }),
165
170
  modelGpuLayers: gpuLayers,
166
171
  sequences: 1,
167
- isEmbeddingContext
172
+ isEmbeddingContext,
173
+ flashAttention
168
174
  }).gpuVram;
169
175
  if (contextVram <= vram)
170
176
  return {
@@ -1 +1 @@
1
- {"version":3,"file":"resolveModelGpuLayersOption.js","sourceRoot":"","sources":["../../../../src/gguf/insights/utils/resolveModelGpuLayersOption.ts"],"names":[],"mappings":"AAEA,OAAO,EAAC,uBAAuB,EAAC,MAAM,2CAA2C,CAAC;AAClF,OAAO,EAAC,cAAc,EAAC,MAAM,kCAAkC,CAAC;AAChE,OAAO,EAAC,0BAA0B,EAAE,0BAA0B,EAAC,MAAM,iDAAiD,CAAC;AACvH,OAAO,EAAC,mCAAmC,EAAC,MAAM,oBAAoB,CAAC;AACvE,OAAO,EAAC,WAAW,EAAC,MAAM,kBAAkB,CAAC;AAG7C,MAAM,sCAAsC,GAAG,GAAG,CAAC;AAEnD,MAAM,CAAC,KAAK,UAAU,2BAA2B,CAAC,SAAyC,EAAE,EACzF,YAAY,EAAE,wBAAwB,GAAG,KAAK,EAAE,YAAY,EAAE,oBAAoB,EAClF,QAAQ,EAAE,0BAA0B,EAKvC;IACG,IAAI,SAAS,IAAI,IAAI;QACjB,SAAS,GAAG,MAAM,CAAC;IAEvB,IAAI,CAAC,0BAA0B;QAC3B,OAAO,CAAC,CAAC;IAEb,IAAI,SAAS,KAAK,KAAK,IAAI,OAAO,SAAS,KAAK,QAAQ,EAAE,CAAC;QACvD,MAAM,iBAAiB,GAAG,OAAO,SAAS,KAAK,QAAQ;YACnD,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,IAAI,CAAC,GAAG,CAAC,YAAY,CAAC,WAAW,EAAE,SAAS,CAAC,CAAC;YAC5D,CAAC,CAAC,YAAY,CAAC,WAAW,CAAC;QAE/B,IAAI,wBAAwB;YACxB,OAAO,iBAAiB,CAAC;QAE7B,MAAM,SAAS,GAAG,MAAM,YAAY,EAAE,CAAC;QACvC,MAAM,qBAAqB,GAAG,2BAA2B,CAAC;YACtD,SAAS,EAAE,iBAAiB;YAC5B,YAAY;YACZ,WAAW,EAAE,SAAS,CAAC,IAAI;SAC9B,CAAC,CAAC;QAEH,IAAI,qBAAqB,IAAI,IAAI;YAC7B,MAAM,IAAI,uBAAuB,CAAC,8DAA8D,CAAC,CAAC;QAEtG,OAAO,iBAAiB,CAAC;IAC7B,CAAC;SAAM,IAAI,SAAS,KAAK,MAAM,IAAI,OAAO,SAAS,KAAK,QAAQ,EAAE,CAAC;QAC/D,IAAI,QAAQ,KAAK,KAAK;YAClB,OAAO,CAAC,CAAC;QAEb,MAAM,SAAS,GAAG,MAAM,YAAY,EAAE,CAAC;QACvC,IAAI,SAAS,CAAC,KAAK,KAAK,CAAC;YACrB,OAAO,CAAC,CAAC;QAEb,IAAI,QAAQ,GAAG,SAAS,CAAC,IAAI,CAAC;QAC9B,IAAI,OAAO,SAAS,KAAK,QAAQ,IAAI,SAAS,CAAC,UAAU,EAAE,WAAW,IAAI,IAAI,EAAE,CAAC;YAC7E,QAAQ,IAAI,oBAAoB,GAAG,sCAAsC,CAAC;YAE1E,IAAI,QAAQ,GAAG,CAAC;gBACZ,QAAQ,GAAG,CAAC,CAAC;QACrB,CAAC;QAED,MAAM,mBAAmB,GAAG,2BAA2B,CAAC;YACpD,YAAY;YACZ,QAAQ;YACR,UAAU,EAAE,OAAO,SAAS,KAAK,QAAQ;gBACrC,CAAC,CAAC,SAAS,CAAC,UAAU;gBACtB,CAAC,CAAC,SAAS;YACf,YAAY,EAAE,OAAO,SAAS,KAAK,QAAQ;gBACvC,CAAC,CAAC,SAAS,CAAC,GAAG;gBACf,CAAC,CAAC,SAAS;YACf,YAAY,EAAE,OAAO,SAAS,KAAK,QAAQ;gBACvC,CAAC,CAAC,SAAS,CAAC,GAAG;gBACf,CAAC,CAAC,SAAS;SAClB,CAAC,CAAC;QAEH,MAAM,wBAAwB,GAAG,OAAO,SAAS,KAAK,QAAQ;YAC1D,CAAC,SAAS,CAAC,GAAG,IAAI,IAAI,IAAI,SAAS,CAAC,GAAG,IAAI,IAAI,IAAI,SAAS,CAAC,UAAU,EAAE,WAAW,IAAI,IAAI,CAAC,CAAC;QAElG,IAAI,CAAC,wBAAwB,IAAI,mBAAmB,IAAI,IAAI,IAAI,wBAAwB;YACpF,MAAM,IAAI,uBAAuB,CAAC,8DAA8D,CAAC,CAAC;QAEtG,OAAO,mBAAmB,IAAI,CAAC,CAAC;IACpC,CAAC;IAED,MAAM,IAAI,KAAK,CAAC,4BAA4B,SAAS,EAAE,CAAC,CAAC;AAC7D,CAAC;AAED,SAAS,2BAA2B,CAAC,EACjC,YAAY,EACZ,QAAQ,EACR,UAAU,EACV,YAAY,EACZ,YAAY,EAOf;IACG,OAAO,cAAc,CAAC;QAClB,CAAC,SAAS;YACN,MAAM,SAAS,GAAG,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,YAAY,IAAI,CAAC,CAAC,CAAC,CAAC;YAC7D,MAAM,SAAS,GAAG,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,GAAG,CAAC,YAAY,CAAC,WAAW,EAAE,YAAY,IAAI,YAAY,CAAC,WAAW,CAAC,CAAC,CAAC;YAE3G,KAAK,IAAI,MAAM,GAAG,SAAS,EAAE,MAAM,IAAI,SAAS,EAAE,MAAM,EAAE,EAAE,CAAC;gBACzD,MAAM;oBACF,SAAS,EAAE,MAAM;iBACpB,CAAC;YACN,CAAC;QACL,CAAC;QACD,KAAK,CAAC,MAAM;YACR,MAAM,kBAAkB,GAAG,2BAA2B,CAAC;gBACnD,SAAS,EAAE,MAAM,CAAC,SAAS;gBAC3B,YAAY;gBACZ,WAAW,EAAE,QAAQ;gBACrB,UAAU;aACb,CAAC,CAAC;YAEH,IAAI,kBAAkB,IAAI,IAAI;gBAC1B,OAAO,IAAI,CAAC;YAEhB,OAAO,mCAAmC,CAAC,EAAC,SAAS,EAAE,MAAM,CAAC,SAAS,EAAE,WAAW,EAAE,kBAAkB,CAAC,WAAW,EAAC,EAAE;gBACnH,cAAc,EAAE,YAAY,CAAC,WAAW;gBACxC,gBAAgB,EAAE,0BAA0B,CAAC,EAAC,gBAAgB,EAAE,YAAY,CAAC,gBAAgB,EAAC,CAAC;aAClG,CAAC,CAAC;QACP,CAAC;KACJ,CAAC,EAAE,SAAS,IAAI,IAAI,CAAC;AAC1B,CAAC;AAED,SAAS,mCAAmC,CAAC,EAAC,SAAS,EAAE,WAAW,EAA2C,EAAE,EAC7G,cAAc,EAAE,gBAAgB,EAGnC;IACG,SAAS,cAAc;QACnB,OAAO,WAAW,CAAC,SAAS,EAAE,CAAC;gBAC3B,KAAK,EAAE,CAAC;gBACR,MAAM,EAAE,CAAC;aACZ,EAAE;gBACC,KAAK,EAAE,CAAC;gBACR,MAAM,EAAE,EAAE;aACb,EAAE;gBACC,KAAK,EAAE,cAAc;gBACrB,MAAM,EAAE,EAAE;gBACV,GAAG,EAAE,cAAc;aACtB,CAAC,CAAC,CAAC;IACR,CAAC;IAED,SAAS,gBAAgB;QACrB,MAAM,mBAAmB,GAAG,SAAS,GAAG,cAAc,CAAC;QAEvD,OAAO,WAAW,CAAC,WAAW,EAAE,CAAC;gBAC7B,KAAK,EAAE,CAAC;gBACR,MAAM,EAAE,CAAC;aACZ,EAAE;gBACC,KAAK,EAAE,IAAI;gBACX,MAAM,EAAE,CAAC;aACZ,EAAE;gBACC,KAAK,EAAE,IAAI;gBACX,MAAM,EAAE,mBAAmB,GAAG,GAAG,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;aAC5C,EAAE;gBACC,KAAK,EAAE,IAAI;gBACX,MAAM,EAAE,mBAAmB,GAAG,GAAG,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE;aAC7C,EAAE;gBACC,KAAK,EAAE,IAAI;gBACX,MAAM,EAAE,mBAAmB,GAAG,GAAG,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;gBACzC,GAAG,EAAE,IAAI,CAAC,GAAG,CAAC,gBAAgB,EAAE,KAAK,CAAC;aACzC,CAAC,CAAC,CAAC;IACR,CAAC;IAED,OAAO,cAAc,EAAE,GAAG,gBAAgB,EAAE,CAAC;AACjD,CAAC;AAED,SAAS,2BAA2B,CAAC,EACjC,SAAS,EAAE,YAAY,EAAE,WAAW,EAAE,UAAU,EAGnD;IACG,MAAM,SAAS,GAAG,YAAY,CAAC,iCAAiC,CAAC,EAAC,SAAS,EAAC,CAAC,CAAC,OAAO,CAAC;IAEtF,IAAI,SAAS,GAAG,WAAW;QACvB,OAAO,IAAI,CAAC;IAEhB,IAAI,UAAU,IAAI,IAAI,IAAI,UAAU,CAAC,WAAW,IAAI,IAAI,EAAE,CAAC;QACvD,MAAM,WAAW,GAAG,YAAY,CAAC,mCAAmC,CAAC;YACjE,WAAW,EAAE,UAAU,CAAC,WAAW;YACnC,SAAS,EAAE,0BAA0B,CAAC,EAAC,WAAW,EAAE,UAAU,CAAC,WAAW,EAAE,SAAS,EAAE,CAAC,EAAC,CAAC;YAC1F,cAAc,EAAE,SAAS;YACzB,SAAS,EAAE,CAAC;YACZ,kBAAkB,EAAE,UAAU,CAAC,gBAAgB,IAAI,KAAK;SAC3D,CAAC,CAAC,OAAO,CAAC;QAEX,MAAM,SAAS,GAAG,SAAS,GAAG,WAAW,CAAC;QAC1C,IAAI,SAAS,GAAG,WAAW;YACvB,OAAO,IAAI,CAAC;QAEhB,OAAO;YACH,WAAW,EAAE,UAAU,CAAC,WAAW;YACnC,WAAW;YACX,SAAS;SACZ,CAAC;IACN,CAAC;IAED,MAAM,UAAU,GAAG,iCAAiC,CAAC;QACjD,SAAS;QACT,YAAY;QACZ,IAAI,EAAE,WAAW,GAAG,SAAS;QAC7B,kBAAkB,EAAE,UAAU,EAAE,gBAAgB,IAAI,KAAK;KAC5D,CAAC,CAAC;IAEH,IAAI,UAAU,IAAI,IAAI,IAAI,SAAS,GAAG,UAAU,CAAC,IAAI,GAAG,WAAW;QAC/D,OAAO,IAAI,CAAC;IAEhB,OAAO;QACH,WAAW,EAAE,UAAU,CAAC,WAAW;QACnC,WAAW,EAAE,UAAU,CAAC,IAAI;QAC5B,SAAS,EAAE,SAAS,GAAG,UAAU,CAAC,IAAI;KACzC,CAAC;AACN,CAAC;AAED,SAAS,iCAAiC,CAAC,EAAC,SAAS,EAAE,YAAY,EAAE,IAAI,EAAE,kBAAkB,EAE5F;IACG,MAAM,cAAc,GAAG,0BAA0B,CAAC,EAAC,gBAAgB,EAAE,YAAY,CAAC,gBAAgB,EAAC,CAAC,CAAC;IAErG,KAAK,IAAI,WAAW,GAAG,cAAc,EAAE,WAAW,IAAI,mCAAmC,EAAE,WAAW,EAAE,EAAE,CAAC;QACvG,MAAM,WAAW,GAAG,YAAY,CAAC,mCAAmC,CAAC;YACjE,WAAW;YACX,SAAS,EAAE,0BAA0B,CAAC,EAAC,WAAW,EAAE,SAAS,EAAE,CAAC,EAAC,CAAC;YAClE,cAAc,EAAE,SAAS;YACzB,SAAS,EAAE,CAAC;YACZ,kBAAkB;SACrB,CAAC,CAAC,OAAO,CAAC;QAEX,IAAI,WAAW,IAAI,IAAI;YACnB,OAAO;gBACH,WAAW;gBACX,IAAI,EAAE,WAAW;aACpB,CAAC;IACV,CAAC;IAED,OAAO,IAAI,CAAC;AAChB,CAAC"}
1
+ {"version":3,"file":"resolveModelGpuLayersOption.js","sourceRoot":"","sources":["../../../../src/gguf/insights/utils/resolveModelGpuLayersOption.ts"],"names":[],"mappings":"AAEA,OAAO,EAAC,uBAAuB,EAAC,MAAM,2CAA2C,CAAC;AAClF,OAAO,EAAC,cAAc,EAAC,MAAM,kCAAkC,CAAC;AAChE,OAAO,EAAC,0BAA0B,EAAE,0BAA0B,EAAC,MAAM,iDAAiD,CAAC;AACvH,OAAO,EAAC,mCAAmC,EAAC,MAAM,oBAAoB,CAAC;AACvE,OAAO,EAAC,WAAW,EAAC,MAAM,kBAAkB,CAAC;AAG7C,MAAM,sCAAsC,GAAG,GAAG,CAAC;AAEnD,MAAM,CAAC,KAAK,UAAU,2BAA2B,CAAC,SAAyC,EAAE,EACzF,YAAY,EAAE,wBAAwB,GAAG,KAAK,EAAE,YAAY,EAAE,oBAAoB,EAClF,QAAQ,EAAE,0BAA0B,EAAE,4BAA4B,EAKrE;IACG,IAAI,SAAS,IAAI,IAAI;QACjB,SAAS,GAAG,MAAM,CAAC;IAEvB,IAAI,CAAC,0BAA0B;QAC3B,OAAO,CAAC,CAAC;IAEb,IAAI,SAAS,KAAK,KAAK,IAAI,OAAO,SAAS,KAAK,QAAQ,EAAE,CAAC;QACvD,MAAM,iBAAiB,GAAG,OAAO,SAAS,KAAK,QAAQ;YACnD,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,IAAI,CAAC,GAAG,CAAC,YAAY,CAAC,WAAW,EAAE,SAAS,CAAC,CAAC;YAC5D,CAAC,CAAC,YAAY,CAAC,WAAW,CAAC;QAE/B,IAAI,wBAAwB;YACxB,OAAO,iBAAiB,CAAC;QAE7B,MAAM,SAAS,GAAG,MAAM,YAAY,EAAE,CAAC;QACvC,MAAM,qBAAqB,GAAG,2BAA2B,CAAC;YACtD,SAAS,EAAE,iBAAiB;YAC5B,YAAY;YACZ,WAAW,EAAE,SAAS,CAAC,IAAI;YAC3B,4BAA4B;SAC/B,CAAC,CAAC;QAEH,IAAI,qBAAqB,IAAI,IAAI;YAC7B,MAAM,IAAI,uBAAuB,CAAC,8DAA8D,CAAC,CAAC;QAEtG,OAAO,iBAAiB,CAAC;IAC7B,CAAC;SAAM,IAAI,SAAS,KAAK,MAAM,IAAI,OAAO,SAAS,KAAK,QAAQ,EAAE,CAAC;QAC/D,IAAI,QAAQ,KAAK,KAAK;YAClB,OAAO,CAAC,CAAC;QAEb,MAAM,SAAS,GAAG,MAAM,YAAY,EAAE,CAAC;QACvC,IAAI,SAAS,CAAC,KAAK,KAAK,CAAC;YACrB,OAAO,CAAC,CAAC;QAEb,IAAI,QAAQ,GAAG,SAAS,CAAC,IAAI,CAAC;QAC9B,IAAI,OAAO,SAAS,KAAK,QAAQ,IAAI,SAAS,CAAC,UAAU,EAAE,WAAW,IAAI,IAAI,EAAE,CAAC;YAC7E,QAAQ,IAAI,oBAAoB,GAAG,sCAAsC,CAAC;YAE1E,IAAI,QAAQ,GAAG,CAAC;gBACZ,QAAQ,GAAG,CAAC,CAAC;QACrB,CAAC;QAED,MAAM,mBAAmB,GAAG,2BAA2B,CAAC;YACpD,YAAY;YACZ,QAAQ;YACR,UAAU,EAAE,OAAO,SAAS,KAAK,QAAQ;gBACrC,CAAC,CAAC,SAAS,CAAC,UAAU;gBACtB,CAAC,CAAC,SAAS;YACf,YAAY,EAAE,OAAO,SAAS,KAAK,QAAQ;gBACvC,CAAC,CAAC,SAAS,CAAC,GAAG;gBACf,CAAC,CAAC,SAAS;YACf,YAAY,EAAE,OAAO,SAAS,KAAK,QAAQ;gBACvC,CAAC,CAAC,SAAS,CAAC,GAAG;gBACf,CAAC,CAAC,SAAS;YACf,4BAA4B;SAC/B,CAAC,CAAC;QAEH,MAAM,wBAAwB,GAAG,OAAO,SAAS,KAAK,QAAQ;YAC1D,CAAC,SAAS,CAAC,GAAG,IAAI,IAAI,IAAI,SAAS,CAAC,GAAG,IAAI,IAAI,IAAI,SAAS,CAAC,UAAU,EAAE,WAAW,IAAI,IAAI,CAAC,CAAC;QAElG,IAAI,CAAC,wBAAwB,IAAI,mBAAmB,IAAI,IAAI,IAAI,wBAAwB;YACpF,MAAM,IAAI,uBAAuB,CAAC,8DAA8D,CAAC,CAAC;QAEtG,OAAO,mBAAmB,IAAI,CAAC,CAAC;IACpC,CAAC;IAED,MAAM,IAAI,KAAK,CAAC,4BAA4B,SAAS,EAAE,CAAC,CAAC;AAC7D,CAAC;AAED,SAAS,2BAA2B,CAAC,EACjC,YAAY,EACZ,QAAQ,EACR,UAAU,EACV,YAAY,EACZ,YAAY,EACZ,4BAA4B,EAQ/B;IACG,OAAO,cAAc,CAAC;QAClB,CAAC,SAAS;YACN,MAAM,SAAS,GAAG,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,YAAY,IAAI,CAAC,CAAC,CAAC,CAAC;YAC7D,MAAM,SAAS,GAAG,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,GAAG,CAAC,YAAY,CAAC,WAAW,EAAE,YAAY,IAAI,YAAY,CAAC,WAAW,CAAC,CAAC,CAAC;YAE3G,KAAK,IAAI,MAAM,GAAG,SAAS,EAAE,MAAM,IAAI,SAAS,EAAE,MAAM,EAAE,EAAE,CAAC;gBACzD,MAAM;oBACF,SAAS,EAAE,MAAM;iBACpB,CAAC;YACN,CAAC;QACL,CAAC;QACD,KAAK,CAAC,MAAM;YACR,MAAM,kBAAkB,GAAG,2BAA2B,CAAC;gBACnD,SAAS,EAAE,MAAM,CAAC,SAAS;gBAC3B,YAAY;gBACZ,WAAW,EAAE,QAAQ;gBACrB,UAAU;gBACV,4BAA4B;aAC/B,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,EAAE,4BAA4B,GAAG,KAAK,EAIzF;IACG,MAAM,SAAS,GAAG,YAAY,CAAC,iCAAiC,CAAC,EAAC,SAAS,EAAC,CAAC,CAAC,OAAO,CAAC;IAEtF,IAAI,SAAS,GAAG,WAAW;QACvB,OAAO,IAAI,CAAC;IAEhB,IAAI,UAAU,IAAI,IAAI,IAAI,UAAU,CAAC,WAAW,IAAI,IAAI,EAAE,CAAC;QACvD,MAAM,WAAW,GAAG,YAAY,CAAC,mCAAmC,CAAC;YACjE,WAAW,EAAE,UAAU,CAAC,WAAW;YACnC,SAAS,EAAE,0BAA0B,CAAC,EAAC,WAAW,EAAE,UAAU,CAAC,WAAW,EAAE,SAAS,EAAE,CAAC,EAAC,CAAC;YAC1F,cAAc,EAAE,SAAS;YACzB,SAAS,EAAE,CAAC;YACZ,kBAAkB,EAAE,UAAU,CAAC,gBAAgB,IAAI,KAAK;YACxD,cAAc,EAAE,4BAA4B;SAC/C,CAAC,CAAC,OAAO,CAAC;QAEX,MAAM,SAAS,GAAG,SAAS,GAAG,WAAW,CAAC;QAC1C,IAAI,SAAS,GAAG,WAAW;YACvB,OAAO,IAAI,CAAC;QAEhB,OAAO;YACH,WAAW,EAAE,UAAU,CAAC,WAAW;YACnC,WAAW;YACX,SAAS;SACZ,CAAC;IACN,CAAC;IAED,MAAM,UAAU,GAAG,iCAAiC,CAAC;QACjD,SAAS;QACT,YAAY;QACZ,IAAI,EAAE,WAAW,GAAG,SAAS;QAC7B,kBAAkB,EAAE,UAAU,EAAE,gBAAgB,IAAI,KAAK;QACzD,cAAc,EAAE,4BAA4B;KAC/C,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,EAAE,cAAc,EAE5G;IACG,MAAM,cAAc,GAAG,0BAA0B,CAAC,EAAC,gBAAgB,EAAE,YAAY,CAAC,gBAAgB,EAAC,CAAC,CAAC;IAErG,KAAK,IAAI,WAAW,GAAG,cAAc,EAAE,WAAW,IAAI,mCAAmC,EAAE,WAAW,EAAE,EAAE,CAAC;QACvG,MAAM,WAAW,GAAG,YAAY,CAAC,mCAAmC,CAAC;YACjE,WAAW;YACX,SAAS,EAAE,0BAA0B,CAAC,EAAC,WAAW,EAAE,SAAS,EAAE,CAAC,EAAC,CAAC;YAClE,cAAc,EAAE,SAAS;YACzB,SAAS,EAAE,CAAC;YACZ,kBAAkB;YAClB,cAAc;SACjB,CAAC,CAAC,OAAO,CAAC;QAEX,IAAI,WAAW,IAAI,IAAI;YACnB,OAAO;gBACH,WAAW;gBACX,IAAI,EAAE,WAAW;aACpB,CAAC;IACV,CAAC;IAED,OAAO,IAAI,CAAC;AAChB,CAAC"}
@@ -1,31 +1,44 @@
1
1
  export declare const enum GgufArchitectureType {
2
2
  llama = "llama",
3
3
  falcon = "falcon",
4
+ grok = "grok",
4
5
  gpt2 = "gpt2",
5
6
  gptj = "gptj",
6
7
  gptneox = "gptneox",
7
8
  mpt = "mpt",
8
9
  baichuan = "baichuan",
9
10
  starcoder = "starcoder",
10
- persimmon = "persimmon",
11
11
  refact = "refact",
12
12
  bert = "bert",
13
13
  nomicBert = "nomic-bert",
14
+ jinaBertV2 = "jina-bert-v2",
14
15
  bloom = "bloom",
15
16
  stablelm = "stablelm",
16
17
  qwen = "qwen",
17
18
  qwen2 = "qwen2",
19
+ qwen2moe = "qwen2moe",
18
20
  phi2 = "phi2",
21
+ phi3 = "phi3",
19
22
  plamo = "plamo",
20
23
  codeshell = "codeshell",
21
24
  orion = "orion",
22
25
  internlm2 = "internlm2",
23
26
  minicpm = "minicpm",
24
27
  gemma = "gemma",
28
+ gemma2 = "gemma2",
25
29
  starcoder2 = "starcoder2",
26
30
  mamba = "mamba",
31
+ xverse = "xverse",
27
32
  commandR = "command-r",
28
- rwkv = "rwkv"
33
+ dbrx = "dbrx",
34
+ olmo = "olmo",
35
+ openelm = "openelm",
36
+ arctic = "arctic",
37
+ deepseek2 = "deepseek2",
38
+ bitnet = "bitnet",
39
+ t5 = "t5",
40
+ jais = "jais",
41
+ unknown = "(unknown)"
29
42
  }
30
43
  export type GgufMetadata<A extends GgufArchitectureType = GgufArchitectureType> = {
31
44
  readonly general: GgufMetadataGeneral<A>;
@@ -44,7 +57,6 @@ export type GgufMetadataLlmToType = {
44
57
  [GgufArchitectureType.bloom]: GgufMetadataBloom;
45
58
  [GgufArchitectureType.falcon]: GgufMetadataFalcon;
46
59
  [GgufArchitectureType.mamba]: GgufMetadataMamba;
47
- [GgufArchitectureType.rwkv]: GgufMetadataRWKV;
48
60
  };
49
61
  export declare enum GgufFileType {
50
62
  ALL_F32 = 0,
@@ -325,11 +337,4 @@ export type GgufMetadataMamba = {
325
337
  readonly layer_norm_rms_epsilon: number;
326
338
  };
327
339
  };
328
- export type GgufMetadataRWKV = {
329
- readonly architecture_version: 4 | number;
330
- readonly context_length: number;
331
- readonly block_count: number;
332
- readonly embedding_length: number;
333
- readonly feed_forward_length: number;
334
- };
335
340
  export declare function isGgufMetadataOfArchitectureType<A extends GgufArchitectureType>(metadata: GgufMetadata, type: A): metadata is GgufMetadata<A>;
@@ -2,31 +2,44 @@ export var GgufArchitectureType;
2
2
  (function (GgufArchitectureType) {
3
3
  GgufArchitectureType["llama"] = "llama";
4
4
  GgufArchitectureType["falcon"] = "falcon";
5
+ GgufArchitectureType["grok"] = "grok";
5
6
  GgufArchitectureType["gpt2"] = "gpt2";
6
7
  GgufArchitectureType["gptj"] = "gptj";
7
8
  GgufArchitectureType["gptneox"] = "gptneox";
8
9
  GgufArchitectureType["mpt"] = "mpt";
9
10
  GgufArchitectureType["baichuan"] = "baichuan";
10
11
  GgufArchitectureType["starcoder"] = "starcoder";
11
- GgufArchitectureType["persimmon"] = "persimmon";
12
12
  GgufArchitectureType["refact"] = "refact";
13
13
  GgufArchitectureType["bert"] = "bert";
14
14
  GgufArchitectureType["nomicBert"] = "nomic-bert";
15
+ GgufArchitectureType["jinaBertV2"] = "jina-bert-v2";
15
16
  GgufArchitectureType["bloom"] = "bloom";
16
17
  GgufArchitectureType["stablelm"] = "stablelm";
17
18
  GgufArchitectureType["qwen"] = "qwen";
18
19
  GgufArchitectureType["qwen2"] = "qwen2";
20
+ GgufArchitectureType["qwen2moe"] = "qwen2moe";
19
21
  GgufArchitectureType["phi2"] = "phi2";
22
+ GgufArchitectureType["phi3"] = "phi3";
20
23
  GgufArchitectureType["plamo"] = "plamo";
21
24
  GgufArchitectureType["codeshell"] = "codeshell";
22
25
  GgufArchitectureType["orion"] = "orion";
23
26
  GgufArchitectureType["internlm2"] = "internlm2";
24
27
  GgufArchitectureType["minicpm"] = "minicpm";
25
28
  GgufArchitectureType["gemma"] = "gemma";
29
+ GgufArchitectureType["gemma2"] = "gemma2";
26
30
  GgufArchitectureType["starcoder2"] = "starcoder2";
27
31
  GgufArchitectureType["mamba"] = "mamba";
32
+ GgufArchitectureType["xverse"] = "xverse";
28
33
  GgufArchitectureType["commandR"] = "command-r";
29
- GgufArchitectureType["rwkv"] = "rwkv";
34
+ GgufArchitectureType["dbrx"] = "dbrx";
35
+ GgufArchitectureType["olmo"] = "olmo";
36
+ GgufArchitectureType["openelm"] = "openelm";
37
+ GgufArchitectureType["arctic"] = "arctic";
38
+ GgufArchitectureType["deepseek2"] = "deepseek2";
39
+ GgufArchitectureType["bitnet"] = "bitnet";
40
+ GgufArchitectureType["t5"] = "t5";
41
+ GgufArchitectureType["jais"] = "jais";
42
+ GgufArchitectureType["unknown"] = "(unknown)";
30
43
  })(GgufArchitectureType || (GgufArchitectureType = {}));
31
44
  // source: `enum llama_ftype` in `llama.h` in the `llama.cpp` source code
32
45
  export var GgufFileType;
@@ -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;AAiCD,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"}
1
+ {"version":3,"file":"GgufMetadataTypes.js","sourceRoot":"","sources":["../../../src/gguf/types/GgufMetadataTypes.ts"],"names":[],"mappings":"AAAA,MAAM,CAAN,IAAkB,oBAyCjB;AAzCD,WAAkB,oBAAoB;IAClC,uCAAe,CAAA;IACf,yCAAiB,CAAA;IACjB,qCAAa,CAAA;IACb,qCAAa,CAAA;IACb,qCAAa,CAAA;IACb,2CAAmB,CAAA;IACnB,mCAAW,CAAA;IACX,6CAAqB,CAAA;IACrB,+CAAuB,CAAA;IACvB,yCAAiB,CAAA;IACjB,qCAAa,CAAA;IACb,gDAAwB,CAAA;IACxB,mDAA2B,CAAA;IAC3B,uCAAe,CAAA;IACf,6CAAqB,CAAA;IACrB,qCAAa,CAAA;IACb,uCAAe,CAAA;IACf,6CAAqB,CAAA;IACrB,qCAAa,CAAA;IACb,qCAAa,CAAA;IACb,uCAAe,CAAA;IACf,+CAAuB,CAAA;IACvB,uCAAe,CAAA;IACf,+CAAuB,CAAA;IACvB,2CAAmB,CAAA;IACnB,uCAAe,CAAA;IACf,yCAAiB,CAAA;IACjB,iDAAyB,CAAA;IACzB,uCAAe,CAAA;IACf,yCAAiB,CAAA;IACjB,8CAAsB,CAAA;IACtB,qCAAa,CAAA;IACb,qCAAa,CAAA;IACb,2CAAmB,CAAA;IACnB,yCAAiB,CAAA;IACjB,+CAAuB,CAAA;IACvB,yCAAiB,CAAA;IACjB,iCAAS,CAAA;IACT,qCAAa,CAAA;IACb,6CAAqB,CAAA;AACzB,CAAC,EAzCiB,oBAAoB,KAApB,oBAAoB,QAyCrC;AA8BD,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;AAiCD,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;AA8MD,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
@@ -2,7 +2,7 @@ import { DisposedError } from "lifecycle-utils";
2
2
  import { Llama } from "./bindings/Llama.js";
3
3
  import { getLlama, type LlamaOptions, type LastBuildOptions } from "./bindings/getLlama.js";
4
4
  import { NoBinaryFoundError } from "./bindings/utils/NoBinaryFoundError.js";
5
- import { LlamaLogLevel, LlamaLogLevelGreaterThan, LlamaLogLevelGreaterThanOrEqual, LlamaVocabularyType } from "./bindings/types.js";
5
+ import { type LlamaGpuType, LlamaLogLevel, LlamaLogLevelGreaterThan, LlamaLogLevelGreaterThanOrEqual, LlamaVocabularyType } from "./bindings/types.js";
6
6
  import { LlamaModel, LlamaModelInfillTokens, type LlamaModelOptions, LlamaModelTokens } from "./evaluator/LlamaModel/LlamaModel.js";
7
7
  import { TokenAttributes } from "./evaluator/LlamaModel/utils/TokenAttributes.js";
8
8
  import { LlamaGrammar, type LlamaGrammarOptions } from "./evaluator/LlamaGrammar.js";
@@ -46,6 +46,6 @@ import { createModelDownloader, ModelDownloader, type ModelDownloaderOptions } f
46
46
  import { type ChatHistoryItem, type ChatModelFunctionCall, type ChatModelFunctions, type ChatModelResponse, type ChatSessionModelFunction, type ChatSessionModelFunctions, type ChatSystemMessage, type ChatUserMessage, type Token, type Tokenizer, type Detokenizer, isChatModelResponseFunctionCall, type LLamaContextualRepeatPenalty, type ChatWrapperSettings, type ChatWrapperGenerateContextStateOptions, type ChatWrapperGeneratedContextState } from "./types.js";
47
47
  import { type GbnfJsonArraySchema, type GbnfJsonBasicSchema, type GbnfJsonConstSchema, type GbnfJsonEnumSchema, type GbnfJsonObjectSchema, type GbnfJsonOneOfSchema, type GbnfJsonSchema, type GbnfJsonSchemaImmutableType, type GbnfJsonSchemaToType } from "./utils/gbnfJson/types.js";
48
48
  import { type GgufFileInfo } from "./gguf/types/GgufFileInfoTypes.js";
49
- 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";
49
+ 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, isGgufMetadataOfArchitectureType } from "./gguf/types/GgufMetadataTypes.js";
50
50
  import { GgmlType, type GgufTensorInfo } from "./gguf/types/GgufTensorInfoTypes.js";
51
- export { Llama, getLlama, type LlamaOptions, type LastBuildOptions, LlamaLogLevel, NoBinaryFoundError, LlamaModel, LlamaModelTokens, LlamaModelInfillTokens, TokenAttributes, 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, type LlamaEmbedding, LlamaChatSession, defineChatSessionFunction, type LlamaChatSessionOptions, type LlamaChatSessionContextShiftOptions, type LLamaChatPromptOptions, type LLamaChatCompletePromptOptions, type LlamaChatSessionRepeatPenalty, type LLamaChatPreloadPromptOptions, LlamaChat, type LlamaChatOptions, type LLamaChatGenerateResponseOptions, type LLamaChatLoadAndCompleteUserMessageOptions, type LLamaChatContextShiftOptions, type LLamaContextualRepeatPenalty, type LlamaChatResponse, type LlamaChatResponseFunctionCall, type LlamaChatLoadAndCompleteUserResponse, LlamaChatSessionPromptCompletionEngine, type LLamaChatPromptCompletionEngineOptions, LlamaCompletion, type LlamaCompletionOptions, type LlamaCompletionGenerationOptions, type LlamaInfillGenerationOptions, type LlamaCompletionResponse, TokenMeter, type TokenMeterState, UnsupportedError, InsufficientMemoryError, DisposedError, ChatWrapper, type ChatWrapperSettings, type ChatWrapperGenerateContextStateOptions, type ChatWrapperGeneratedContextState, EmptyChatWrapper, Llama3ChatWrapper, Llama2ChatWrapper, GeneralChatWrapper, ChatMLChatWrapper, FalconChatWrapper, AlpacaChatWrapper, FunctionaryChatWrapper, GemmaChatWrapper, TemplateChatWrapper, type TemplateChatWrapperOptions, JinjaTemplateChatWrapper, type JinjaTemplateChatWrapperOptions, type JinjaTemplateChatWrapperOptionsConvertMessageFormat, type ChatHistoryFunctionCallMessageTemplate, resolveChatWrapper, type ResolveChatWrapperOptions, resolvableChatWrapperTypeNames, type ResolvableChatWrapperTypeName, specializedChatWrapperTypeNames, type SpecializedChatWrapperTypeName, templateChatWrapperTypeNames, type TemplateChatWrapperTypeName, ChatModelFunctionsDocumentationGenerator, LlamaText, SpecialTokensText, SpecialToken, isLlamaText, tokenizeText, type LlamaTextValue, type LlamaTextInputValue, type LlamaTextJSON, type LlamaTextJSONValue, type LlamaTextSpecialTokensTextJSON, type LlamaTextSpecialTokenJSON, type BuiltinSpecialTokenValue, appendUserMessageToChatHistory, getModuleVersion, type ChatHistoryItem, type ChatModelFunctionCall, type ChatModelFunctions, type ChatModelResponse, type ChatSessionModelFunction, type ChatSessionModelFunctions, type ChatSystemMessage, type ChatUserMessage, type Token, type Tokenizer, type Detokenizer, 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, GgufInsights, type GgufInsightsResourceRequirements, GgufInsightsConfigurationResolver, createModelDownloader, ModelDownloader, type ModelDownloaderOptions };
51
+ export { Llama, getLlama, type LlamaOptions, type LastBuildOptions, type LlamaGpuType, LlamaLogLevel, NoBinaryFoundError, LlamaModel, LlamaModelTokens, LlamaModelInfillTokens, TokenAttributes, 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, type LlamaEmbedding, LlamaChatSession, defineChatSessionFunction, type LlamaChatSessionOptions, type LlamaChatSessionContextShiftOptions, type LLamaChatPromptOptions, type LLamaChatCompletePromptOptions, type LlamaChatSessionRepeatPenalty, type LLamaChatPreloadPromptOptions, LlamaChat, type LlamaChatOptions, type LLamaChatGenerateResponseOptions, type LLamaChatLoadAndCompleteUserMessageOptions, type LLamaChatContextShiftOptions, type LLamaContextualRepeatPenalty, type LlamaChatResponse, type LlamaChatResponseFunctionCall, type LlamaChatLoadAndCompleteUserResponse, LlamaChatSessionPromptCompletionEngine, type LLamaChatPromptCompletionEngineOptions, LlamaCompletion, type LlamaCompletionOptions, type LlamaCompletionGenerationOptions, type LlamaInfillGenerationOptions, type LlamaCompletionResponse, TokenMeter, type TokenMeterState, UnsupportedError, InsufficientMemoryError, DisposedError, ChatWrapper, type ChatWrapperSettings, type ChatWrapperGenerateContextStateOptions, type ChatWrapperGeneratedContextState, EmptyChatWrapper, Llama3ChatWrapper, Llama2ChatWrapper, GeneralChatWrapper, ChatMLChatWrapper, FalconChatWrapper, AlpacaChatWrapper, FunctionaryChatWrapper, GemmaChatWrapper, TemplateChatWrapper, type TemplateChatWrapperOptions, JinjaTemplateChatWrapper, type JinjaTemplateChatWrapperOptions, type JinjaTemplateChatWrapperOptionsConvertMessageFormat, type ChatHistoryFunctionCallMessageTemplate, resolveChatWrapper, type ResolveChatWrapperOptions, resolvableChatWrapperTypeNames, type ResolvableChatWrapperTypeName, specializedChatWrapperTypeNames, type SpecializedChatWrapperTypeName, templateChatWrapperTypeNames, type TemplateChatWrapperTypeName, ChatModelFunctionsDocumentationGenerator, LlamaText, SpecialTokensText, SpecialToken, isLlamaText, tokenizeText, type LlamaTextValue, type LlamaTextInputValue, type LlamaTextJSON, type LlamaTextJSONValue, type LlamaTextSpecialTokensTextJSON, type LlamaTextSpecialTokenJSON, type BuiltinSpecialTokenValue, appendUserMessageToChatHistory, getModuleVersion, type ChatHistoryItem, type ChatModelFunctionCall, type ChatModelFunctions, type ChatModelResponse, type ChatSessionModelFunction, type ChatSessionModelFunctions, type ChatSystemMessage, type ChatUserMessage, type Token, type Tokenizer, type Detokenizer, 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, GgmlType, isGgufMetadataOfArchitectureType, GgufInsights, type GgufInsightsResourceRequirements, GgufInsightsConfigurationResolver, createModelDownloader, ModelDownloader, type ModelDownloaderOptions };
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,EAA2C,MAAM,wBAAwB,CAAC;AAC1F,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,sCAAsC,CAAC;AAClI,OAAO,EAAC,eAAe,EAAC,MAAM,iDAAiD,CAAC;AAChF,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,EAAC,qBAAqB,EAAyD,MAAM,sCAAsC,CAAC;AAMnI,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,sCAAsC,EACzC,MAAM,8EAA8E,CAAC;AACtF,OAAO,EACH,eAAe,EAElB,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,EACH,wBAAwB,EAC3B,MAAM,oDAAoD,CAAC;AAE5D,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,EAGxE,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;AAC5D,OAAO,EAAC,YAAY,EAAwC,MAAM,iCAAiC,CAAC;AACpG,OAAO,EAAC,iCAAiC,EAAC,MAAM,sDAAsD,CAAC;AACvG,OAAO,EAAC,qBAAqB,EAAE,eAAe,EAA8B,MAAM,kCAAkC,CAAC;AAErH,OAAO,EAG2C,+BAA+B,EAEhF,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,EAGR,aAAa,EACb,kBAAkB,EAClB,UAAU,EACV,gBAAgB,EAChB,sBAAsB,EACtB,eAAe,EAEf,YAAY,EAEZ,sBAAsB,EACtB,8BAA8B,EAC9B,2BAA2B,EAE3B,YAAY,EACZ,oBAAoB,EAWpB,SAAS,EACT,qBAAqB,EAGrB,gBAAgB,EAChB,yBAAyB,EAOzB,SAAS,EAST,sCAAsC,EAEtC,eAAe,EAKf,UAAU,EAEV,gBAAgB,EAChB,uBAAuB,EACvB,aAAa,EACb,WAAW,EAIX,gBAAgB,EAChB,iBAAiB,EACjB,iBAAiB,EACjB,kBAAkB,EAClB,iBAAiB,EACjB,iBAAiB,EACjB,iBAAiB,EACjB,sBAAsB,EACtB,gBAAgB,EAChB,mBAAmB,EAEnB,wBAAwB,EAIxB,kBAAkB,EAElB,8BAA8B,EAE9B,+BAA+B,EAE/B,4BAA4B,EAE5B,wCAAwC,EACxC,SAAS,EACT,iBAAiB,EACjB,YAAY,EACZ,WAAW,EACX,YAAY,EAQZ,8BAA8B,EAC9B,gBAAgB,EAYhB,+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,EAChC,YAAY,EAEZ,iCAAiC,EACjC,qBAAqB,EACrB,eAAe,EAElB,CAAC"}
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,EAA2C,MAAM,wBAAwB,CAAC;AAC1F,OAAO,EAAC,kBAAkB,EAAC,MAAM,wCAAwC,CAAC;AAC1E,OAAO,EACgB,aAAa,EAAE,wBAAwB,EAAE,+BAA+B,EAAE,mBAAmB,EACnH,MAAM,qBAAqB,CAAC;AAC7B,OAAO,EAAC,UAAU,EAAE,sBAAsB,EAA0B,gBAAgB,EAAC,MAAM,sCAAsC,CAAC;AAClI,OAAO,EAAC,eAAe,EAAC,MAAM,iDAAiD,CAAC;AAChF,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,EAAC,qBAAqB,EAAyD,MAAM,sCAAsC,CAAC;AAMnI,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,sCAAsC,EACzC,MAAM,8EAA8E,CAAC;AACtF,OAAO,EACH,eAAe,EAElB,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,EACH,wBAAwB,EAC3B,MAAM,oDAAoD,CAAC;AAE5D,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,EAGxE,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;AAC5D,OAAO,EAAC,YAAY,EAAwC,MAAM,iCAAiC,CAAC;AACpG,OAAO,EAAC,iCAAiC,EAAC,MAAM,sDAAsD,CAAC;AACvG,OAAO,EAAC,qBAAqB,EAAE,eAAe,EAA8B,MAAM,kCAAkC,CAAC;AAErH,OAAO,EAG2C,+BAA+B,EAEhF,MAAM,YAAY,CAAC;AAMpB,OAAO,EAC4C,oBAAoB,EAAE,YAAY,EAAE,8BAA8B,EACjH,mCAAmC,EAEsC,gCAAgC,EAC5G,MAAM,mCAAmC,CAAC;AAC3C,OAAO,EAAC,QAAQ,EAAsB,MAAM,qCAAqC,CAAC;AAGlF,OAAO,EACH,KAAK,EACL,QAAQ,EAIR,aAAa,EACb,kBAAkB,EAClB,UAAU,EACV,gBAAgB,EAChB,sBAAsB,EACtB,eAAe,EAEf,YAAY,EAEZ,sBAAsB,EACtB,8BAA8B,EAC9B,2BAA2B,EAE3B,YAAY,EACZ,oBAAoB,EAWpB,SAAS,EACT,qBAAqB,EAGrB,gBAAgB,EAChB,yBAAyB,EAOzB,SAAS,EAST,sCAAsC,EAEtC,eAAe,EAKf,UAAU,EAEV,gBAAgB,EAChB,uBAAuB,EACvB,aAAa,EACb,WAAW,EAIX,gBAAgB,EAChB,iBAAiB,EACjB,iBAAiB,EACjB,kBAAkB,EAClB,iBAAiB,EACjB,iBAAiB,EACjB,iBAAiB,EACjB,sBAAsB,EACtB,gBAAgB,EAChB,mBAAmB,EAEnB,wBAAwB,EAIxB,kBAAkB,EAElB,8BAA8B,EAE9B,+BAA+B,EAE/B,4BAA4B,EAE5B,wCAAwC,EACxC,SAAS,EACT,iBAAiB,EACjB,YAAY,EACZ,WAAW,EACX,YAAY,EAQZ,8BAA8B,EAC9B,gBAAgB,EAYhB,+BAA+B,EAU/B,mBAAmB,EACnB,wBAAwB,EACxB,+BAA+B,EAC/B,gBAAgB,EAKhB,oBAAoB,EACpB,YAAY,EACZ,8BAA8B,EAC9B,mCAAmC,EAYnC,QAAQ,EACR,gCAAgC,EAChC,YAAY,EAEZ,iCAAiC,EACjC,qBAAqB,EACrB,eAAe,EAElB,CAAC"}
@@ -131,7 +131,15 @@ endif()
131
131
 
132
132
  file(GLOB SOURCE_FILES "addon.cpp" ${GPU_INFO_SOURCES})
133
133
 
134
- set(CMAKE_BUILD_RPATH_USE_ORIGIN ON)
134
+ if(APPLE)
135
+ set(CMAKE_SKIP_BUILD_RPATH FALSE)
136
+ set(CMAKE_INSTALL_RPATH_USE_LINK_PATH FALSE)
137
+ set(CMAKE_BUILD_RPATH "@loader_path")
138
+ set(CMAKE_INSTALL_RPATH "@loader_path")
139
+ set(CMAKE_BUILD_WITH_INSTALL_RPATH TRUE)
140
+ else()
141
+ set(CMAKE_BUILD_RPATH_USE_ORIGIN ON)
142
+ endif()
135
143
 
136
144
  add_library(${PROJECT_NAME} SHARED ${SOURCE_FILES} ${CMAKE_JS_SRC} ${GPU_INFO_HEADERS})
137
145
  set_target_properties(${PROJECT_NAME} PROPERTIES PREFIX "" SUFFIX ".node")
package/llama/addon.cpp CHANGED
@@ -987,6 +987,10 @@ class AddonContext : public Napi::ObjectWrap<AddonContext> {
987
987
  context_params.embeddings = options.Get("embeddings").As<Napi::Boolean>().Value();
988
988
  }
989
989
 
990
+ if (options.Has("flashAttention")) {
991
+ context_params.flash_attn = options.Get("flashAttention").As<Napi::Boolean>().Value();
992
+ }
993
+
990
994
  if (options.Has("threads")) {
991
995
  const auto n_threads = options.Get("threads").As<Napi::Number>().Uint32Value();
992
996
  const auto resolved_n_threads = n_threads == 0 ? std::thread::hardware_concurrency() : n_threads;
@@ -1,3 +1,3 @@
1
1
  {
2
- "release": "b3266"
2
+ "release": "b3322"
3
3
  }
Binary file
@@ -1,4 +1,4 @@
1
1
  {
2
- "tag": "b3266",
2
+ "tag": "b3322",
3
3
  "llamaCppGithubRepo": "ggerganov/llama.cpp"
4
4
  }
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "node-llama-cpp",
3
- "version": "3.0.0-beta.35",
3
+ "version": "3.0.0-beta.37",
4
4
  "description": "Run AI models locally on your machine with node.js bindings for llama.cpp. Force a JSON schema on the model output on the generation level",
5
5
  "main": "./dist/index.js",
6
6
  "type": "module",
@@ -193,7 +193,7 @@
193
193
  }
194
194
  },
195
195
  "optionalDependencies": {
196
- "@node-llama-cpp/linux-x64-cuda": "3.0.0-beta.35",
197
- "@node-llama-cpp/win-x64-cuda": "3.0.0-beta.35"
196
+ "@node-llama-cpp/linux-x64-cuda": "3.0.0-beta.37",
197
+ "@node-llama-cpp/win-x64-cuda": "3.0.0-beta.37"
198
198
  }
199
199
  }