@genai-fi/nanogpt 0.17.5 → 0.18.1

package/dist/layers/TiedEmbedding.js

@@ -1,13 +1,13 @@
-import "../random_width-DN5ZtQkM.js";
-import "../index-CUXkjxiT.js";
-import { r as s } from "../exports_initializers-CYzKLjN7.js";
+import "../random_width-B_fVXhGx.js";
+import "../index-DSGwv2Yx.js";
+import { r as s } from "../exports_initializers-DIOZQt_L.js";
 import a from "./BaseLayer.js";
 import { dot16 as o } from "../ops/dot16.js";
 import { isPackedTensor as r } from "../utilities/packed.js";
-import { p as m } from "../pack16-bqltoUlR.js";
+import { p as m } from "../pack16-WlOSOuZA.js";
 import { transpose16 as d } from "../ops/transpose16.js";
-import { v as p } from "../variable-Ck482e3n.js";
-import { g as h } from "../gather-qIqEqaGn.js";
+import { v as p } from "../variable-SSATClyt.js";
+import { g as h } from "../gather-DEyjXNb1.js";
 class g extends a {
   vocabSize;
   embedDim;

package/dist/layers/TransformerBlock.js

@@ -2,7 +2,7 @@ import p from "./CausalSelfAttention.js";
 import h from "./MLP.js";
 import l from "./RMSNorm.js";
 import m from "./BaseLayer.js";
-import { k as n, t as u } from "../index-CUXkjxiT.js";
+import { k as n, t as u } from "../index-DSGwv2Yx.js";
 import { add16 as d } from "../ops/add16.js";
 class _ extends m {
   ln1;

package/dist/layers/WeightStore.js

@@ -1,6 +1,6 @@
-import { w as o } from "../index-CUXkjxiT.js";
+import { u as o } from "../index-DSGwv2Yx.js";
 import { p as h } from "../index-DmeWGGmS.js";
-import { v as b } from "../variable-Ck482e3n.js";
+import { v as b } from "../variable-SSATClyt.js";
 class d {
   _variables = /* @__PURE__ */ new Map();
   touchedVariables = /* @__PURE__ */ new Set();

package/dist/loader/load.d.ts

@@ -1,12 +1,6 @@
-import { ITokeniser } from '../tokeniser/type';
-import { default as Model, ModelForwardAttributes } from '../models/model';
-import { TransformersMetadata } from './types';
+import { LoadResult } from './types';
 export declare const VERSION = 2;
 export interface LoadModelOptions {
     sourceURL?: string;
 }
-export declare function loadModel(data: Blob | Buffer | string, options?: LoadModelOptions): Promise<{
-    model: Model<ModelForwardAttributes>;
-    tokeniser: ITokeniser;
-    metaData: TransformersMetadata;
-}>;
+export declare function loadModel(data: Blob | Buffer | string, options?: LoadModelOptions): Promise<LoadResult>;

package/dist/loader/loadTransformers.d.ts

@@ -1,10 +1,4 @@
 import { GPTConfig } from '../models/config';
-import { ITokeniser } from '../tokeniser/type';
-import { default as Model, ModelForwardAttributes } from '../models/model';
-import { TransformersConfig, TransformersMetadata, TransformersTokeniser } from './types';
+import { LoadResult, TransformersConfig, TransformersMetadata, TransformersTokeniser } from './types';
 export declare function mapTransformersConfigToGPTConfig(config: TransformersConfig): GPTConfig;
-export default function loadTransformers(config: TransformersConfig, tokeniser: TransformersTokeniser, metadata: TransformersMetadata, weightData: ArrayBuffer): Promise<{
-    model: Model<ModelForwardAttributes, GPTConfig>;
-    tokeniser: ITokeniser;
-    metaData: TransformersMetadata;
-}>;
+export default function loadTransformers(config: TransformersConfig, tokeniser: TransformersTokeniser, metadata: TransformersMetadata, weightData: ArrayBuffer): Promise<LoadResult>;

package/dist/loader/loadTransformers.js

@@ -1,12 +1,12 @@
 import p from "../tokeniser/CharTokeniser.js";
 import _ from "../tokeniser/bpe.js";
 import { load_safetensors as b } from "../utilities/safetensors.js";
-import { aa as c } from "../index-CUXkjxiT.js";
+import { aa as c } from "../index-DSGwv2Yx.js";
 import { dummyPassAsync as u } from "../utilities/dummy.js";
 import y from "../models/factory.js";
 function h(e) {
-  let o;
-  return e.model_type === "GenAI_NanoGPT_v1" ? o = {
+  let a;
+  return e.model_type === "GenAI_NanoGPT_v1" ? a = {
     modelType: "GenAI_NanoGPT_v1",
     vocabSize: e.vocab_size,
     blockSize: e.block_size,
@@ -15,7 +15,7 @@ function h(e) {
     nEmbed: e.hidden_size,
     mlpFactor: e.mlpFactor,
     useRope: e.useRope
-  } : o = {
+  } : a = {
     modelType: "GenAI_NanoGPT_v2",
     vocabSize: e.vocab_size,
     blockSize: e.block_size,
@@ -26,17 +26,19 @@ function h(e) {
     loraConfig: e.loraConfig ? new Map(Object.entries(e.loraConfig)) : void 0,
     loraName: e.loraName,
     windowSize: e.windowSize
-  }, o;
+  }, a;
 }
-async function S(e, o, r, s) {
-  const n = h(e), m = (o.type ?? "char") === "char" ? new p(o.vocab) : new _(o.vocab, o.merges), i = await b(s), t = /* @__PURE__ */ new Map();
-  for (const [l, d] of Object.entries(i))
-    t.set(l, [d]);
+async function G(e, a, r, n) {
+  const m = h(e), o = (a.type ?? "char") === "char" ? new p(a.vocab) : new _(a.vocab, a.merges);
+  a.datasetID && (o.datasetID = a.datasetID), a.id ? o.id = a.id : o.generateID();
+  const d = await b(n), s = /* @__PURE__ */ new Map();
+  for (const [i, l] of Object.entries(d))
+    s.set(i, [l]);
   c();
-  const a = y(n);
-  return a.metaData = r, await u(a), a.weightStore.loadWeights(t, !!r.url), { model: a, tokeniser: m, metaData: r };
+  const t = y(m);
+  return t.metaData = r, await u(t), t.weightStore.loadWeights(s, !!r.url), { model: t, tokeniser: o, metaData: r };
 }
 export {
-  S as default,
+  G as default,
   h as mapTransformersConfigToGPTConfig
 };

package/dist/loader/newZipLoad.d.ts

@@ -1,9 +1,3 @@
-import { ITokeniser } from '../main';
 import { default as zip } from 'jszip';
-import { default as Model, ModelForwardAttributes } from '../models/model';
-import { TransformersMetadata } from './types';
-export default function loadZipFile(zipFile: zip, metaData: TransformersMetadata): Promise<{
-    model: Model<ModelForwardAttributes>;
-    tokeniser: ITokeniser;
-    metaData: TransformersMetadata;
-}>;
+import { LoadResult, TransformersMetadata } from './types';
+export default function loadZipFile(zipFile: zip, metaData: TransformersMetadata): Promise<LoadResult>;

package/dist/loader/newZipLoad.js

@@ -1,16 +1,31 @@
-import s from "./loadTransformers.js";
-async function c(n, i) {
-  const o = await n.file("config.json")?.async("string");
-  if (!o)
+import h from "./loadTransformers.js";
+import { AdamWOptimizer as m } from "../training/AdamW.js";
+async function u(i, c) {
+  const t = await i.file("config.json")?.async("string");
+  if (!t)
     throw new Error("Config file not found in the zip archive");
-  const t = JSON.parse(o), e = await n.file("tokeniser.json")?.async("string");
-  if (!e)
+  const g = JSON.parse(t), n = await i.file("tokeniser.json")?.async("string");
+  if (!n)
     throw new Error("Tokeniser file not found in the zip archive");
-  const a = JSON.parse(e), r = await n.file("model.safetensors")?.async("arraybuffer");
-  if (!r)
+  const w = JSON.parse(n), e = await i.file("model.safetensors")?.async("arraybuffer");
+  if (!e)
     throw new Error("Model weights not found in the zip archive");
-  return s(t, a, i, r);
+  const r = await i.file("optimizer_config.json")?.async("string");
+  let o;
+  if (r) {
+    const l = JSON.parse(r), f = await i.file("optimizer.safetensors")?.async("arraybuffer");
+    if (!f)
+      throw new Error("Optimizer weights not found in the zip archive");
+    o = new m(l), await o.loadMoments(f);
+  }
+  const a = await i.file("training_log.json")?.async("string");
+  let s;
+  return a && (s = JSON.parse(a)), {
+    ...await h(g, w, c, e),
+    optimizer: o,
+    log: s
+  };
 }
 export {
-  c as default
+  u as default
 };

package/dist/loader/oldZipLoad.js

@@ -1,22 +1,22 @@
-import { aa as y } from "../index-CUXkjxiT.js";
-import "../random_width-DN5ZtQkM.js";
-import "../zeros_like-CWjDdwr-.js";
+import { aa as y } from "../index-DSGwv2Yx.js";
+import "../random_width-B_fVXhGx.js";
+import "../zeros_like-rOHr54NY.js";
 import "../Generator.js";
 import "../index-Cp39cXWe.js";
-import "../dataset-CGGp1z9P.js";
+import "../dataset-DlqAN81i.js";
 import "../ops/cpu/attentionMask.js";
 import "../ops/webgl/attentionMask.js";
 import "../ops/grads/attentionMask.js";
 import "../ops/cpu/rope.js";
 import "../ops/webgl/rope.js";
-import "../rope-0j_f1TPm.js";
+import "../rope-CC5RjmKU.js";
 import "../ops/cpu/appendCache.js";
 import "../ops/webgl/appendCache.js";
 import "../ops/grads/softmax16.js";
-import "../matMul16-BcVC_E62.js";
+import "../matMul16-BIT70Vya.js";
 import "../ops/webgl/matMul16.js";
 import "../ops/cpu/matMul16.js";
-import "../pack16-bqltoUlR.js";
+import "../pack16-WlOSOuZA.js";
 import "../ops/transpose16.js";
 import "../ops/reshape16.js";
 import "../ops/cpu/qkv.js";
@@ -33,22 +33,22 @@ import g from "../tokeniser/CharTokeniser.js";
 import k from "../tokeniser/bpe.js";
 import { dummyPassAsync as u } from "../utilities/dummy.js";
 import b from "../models/factory.js";
+import "../ops/cpu/adamAdjust.js";
+import "../ops/webgl/adamAdjust.js";
+import "../ops/cpu/adamMoments.js";
+import "../ops/webgl/adamMoments.js";
 import "../index-DvYrXKkX.js";
 import "../ops/cpu/gatherSub.js";
 import "../ops/webgl/gatherSub.js";
 import "../ops/cpu/scatterSub.js";
 import "../ops/webgl/scatterSub.js";
-import "../ops/cpu/adamAdjust.js";
-import "../ops/webgl/adamAdjust.js";
-import "../ops/cpu/adamMoments.js";
-import "../ops/webgl/adamMoments.js";
 import "../papaparse.min-C0cScC2i.js";
 import "../ops/cpu/matMulGelu.js";
-import "../matMulGelu-JNLZqKQp.js";
+import "../matMulGelu-CsZnh18H.js";
 import "../ops/grads/matMulGelu.js";
 import "../ops/cpu/gelu.js";
 import "../ops/webgl/gelu.js";
-import "../gelu-B220X1Go.js";
+import "../gelu-DpTCC3eB.js";
 import "../ops/webgl/log.js";
 import "../checks/normRMS.js";
 import "../checks/normRMSGrad.js";

package/dist/loader/save.d.ts

@@ -1,9 +1,16 @@
 import { ITokeniser } from '../tokeniser/type';
 import { default as Model, ModelForwardAttributes } from '../models/model';
-import { GPTConfig } from '../main';
+import { AdamWOptimizer } from '../training/AdamW';
+import { TrainingLogEntry } from '../training/types';
+import { GPTConfig } from '../models/config';
 export interface SaveOptions {
     name?: string;
     metadata?: Record<string, unknown>;
     files?: Record<string, unknown>;
+    includeOptimizer?: boolean;
 }
-export declare function saveModel(model: Model<ModelForwardAttributes, GPTConfig>, tokeniser: ITokeniser, options?: SaveOptions): Promise<Blob>;
+export interface ExtraSaveItems {
+    optimizer?: AdamWOptimizer;
+    trainingLog?: TrainingLogEntry[];
+}
+export declare function saveModel(model: Model<ModelForwardAttributes, GPTConfig>, tokeniser: ITokeniser, options?: SaveOptions, extraItems?: ExtraSaveItems): Promise<Blob>;

package/dist/loader/save.js

@@ -1,73 +1,83 @@
-import { z as p } from "../jszip.min-BZhlzntC.js";
-import l from "../tokeniser/CharTokeniser.js";
+import { z as y } from "../jszip.min-BZhlzntC.js";
+import b from "../tokeniser/CharTokeniser.js";
 import { save_safetensors as _ } from "../utilities/safetensors.js";
-import { VERSION as b } from "./load.js";
-async function z(e, i, a) {
-  const r = /* @__PURE__ */ new Map();
-  e.weightStore.saveWeights(r);
-  const n = new p(), s = {};
-  r.forEach((o, f) => {
-    o.length === 1 && (s[f] = o[0]);
+import { VERSION as m } from "./load.js";
+async function d(i, a, o, t) {
+  const g = /* @__PURE__ */ new Map();
+  i.weightStore.saveWeights(g);
+  const n = new y();
+  if (t?.optimizer) {
+    const e = await t.optimizer.saveMoments();
+    n.file("optimizer.safetensors", e, { binary: !0 }), n.file("optimizer_config.json", JSON.stringify(t.optimizer.serializeConfig()), {
+      binary: !1
+    });
+  }
+  t?.trainingLog && n.file("training_log.json", JSON.stringify(t.trainingLog, void 0, 4), {
+    binary: !1
+  });
+  const s = {};
+  g.forEach((e, f) => {
+    e.length === 1 && (s[f] = e[0]);
   });
   const c = await _(s);
   n.file("model.safetensors", c, { binary: !0 });
-  const g = e.config.modelType;
-  let t;
-  if (g === "GenAI_NanoGPT_v1" ? t = {
+  const p = i.config.modelType;
+  let r;
+  p === "GenAI_NanoGPT_v1" ? r = {
     model_type: "GenAI_NanoGPT_v1",
-    vocab_size: i.getVocab().length,
-    hidden_size: e.config.nEmbed,
-    num_hidden_layers: e.config.nLayer,
-    num_attention_heads: e.config.nHead,
-    block_size: e.config.blockSize,
-    mlpFactor: e.config.mlpFactor,
-    useRope: e.config.useRope
-  } : t = {
+    vocab_size: a.getVocab().length,
+    hidden_size: i.config.nEmbed,
+    num_hidden_layers: i.config.nLayer,
+    num_attention_heads: i.config.nHead,
+    block_size: i.config.blockSize,
+    mlpFactor: i.config.mlpFactor,
+    useRope: i.config.useRope
+  } : r = {
     model_type: "GenAI_NanoGPT_v2",
-    vocab_size: i.getVocab().length,
-    hidden_size: e.config.nEmbed,
-    num_hidden_layers: e.config.nLayer,
-    num_attention_heads: e.config.nHead,
-    block_size: e.config.blockSize,
-    mlpFactor: e.config.mlpFactor,
-    loraConfig: e.config.loraConfig ? Object.fromEntries(e.config.loraConfig) : void 0,
-    loraName: e.config.loraName,
-    windowSize: e.config.windowSize
-  }, n.file("config.json", JSON.stringify(t, void 0, 4), {
+    vocab_size: a.getVocab().length,
+    hidden_size: i.config.nEmbed,
+    num_hidden_layers: i.config.nLayer,
+    num_attention_heads: i.config.nHead,
+    block_size: i.config.blockSize,
+    mlpFactor: i.config.mlpFactor,
+    loraConfig: i.config.loraConfig ? Object.fromEntries(i.config.loraConfig) : void 0,
+    loraName: i.config.loraName,
+    windowSize: i.config.windowSize
+  }, n.file("config.json", JSON.stringify(r, void 0, 4), {
+    binary: !1
+  });
+  const l = {
+    version: m,
+    application: "@genai-fi/nanogpt",
+    meta: o?.metadata,
+    name: o?.name,
+    training: i.metaData?.training || void 0,
+    reference: i.metaData?.url || void 0,
+    phase: i.metaData?.phase || void 0,
+    pretrainingData: i.metaData?.pretrainingData || void 0,
+    pretrainingSettings: i.metaData?.pretrainingSettings || void 0,
+    generationSettings: i.metaData?.generationSettings || void 0,
+    actionLog: i.metaData?.actionLog || void 0
+  };
+  if (n.file("meta.json", JSON.stringify(l, void 0, 4), {
     binary: !1
   }), n.file(
-    "meta.json",
-    JSON.stringify(
-      {
-        version: b,
-        application: "@genai-fi/nanogpt",
-        meta: a?.metadata,
-        name: a?.name,
-        training: e.trainingState || void 0,
-        reference: e.metaData?.url || void 0,
-        phase: e.metaData?.phase || void 0
-      },
-      void 0,
-      4
-    ),
-    {
-      binary: !1
-    }
-  ), n.file(
     "tokeniser.json",
     JSON.stringify({
-      type: i instanceof l ? "char" : "bpe",
-      vocab: i.getVocab(),
-      merges: await i.getMerges()
+      type: a instanceof b ? "char" : "bpe",
+      vocab: a.getVocab(),
+      merges: a.getMerges(),
+      datasetID: a.datasetID,
+      id: a.id
     }),
     {
       binary: !1
     }
-  ), a?.files)
-    for (const [o, f] of Object.entries(a.files))
-      n.file(o, JSON.stringify(f), { binary: !1 });
+  ), o?.files)
+    for (const [e, f] of Object.entries(o.files))
+      n.file(e, JSON.stringify(f), { binary: !1 });
   return n.generateAsync({ type: "blob" });
 }
 export {
-  z as saveModel
+  d as saveModel
 };

package/dist/loader/types.d.ts

@@ -1,5 +1,9 @@
+import { GenerateOptions } from '../inference/types';
 import { LoRAConfig } from '../models/config';
-import { TrainingState } from '../training/types';
+import { default as Model, ModelForwardAttributes, TrainingState } from '../models/model';
+import { ITokeniser } from '../tokeniser/type';
+import { AdamWOptimizer } from '../training/AdamW';
+import { TrainingLogEntry, TrainingOptions } from '../training/types';
 export interface TransformersConfigBase {
     model_type: 'GenAI_NanoGPT_v1' | 'GenAI_NanoGPT_v2';
     vocab_size: number;
@@ -24,15 +28,40 @@ export interface TransformersTokeniser {
     type: 'char' | 'bpe';
     vocab: string[];
     merges: [string, string][];
+    datasetID?: string;
+    id?: string;
 }
 export type ModelPhase = 'untrained' | 'pretrained' | 'finetuned';
+export interface DatasetMetadata {
+    id: string;
+    name: string;
+}
+export interface ActionLogEntry {
+    action: 'pretrain' | 'generate' | 'finetune';
+    timestamp: number;
+    duration: number;
+    tokensProcessed: number;
+    options: TrainingOptions | GenerateOptions;
+}
 export interface TransformersMetadata {
     name?: string;
     version: number;
     application: string;
     training?: TrainingState;
     reference?: string;
+    id?: string;
     url?: string;
     phase?: ModelPhase;
+    pretrainingData?: DatasetMetadata[];
+    pretrainingSettings?: TrainingOptions;
+    generationSettings?: GenerateOptions;
+    actionLog?: ActionLogEntry[];
     [key: string]: unknown;
 }
+export interface LoadResult {
+    model: Model<ModelForwardAttributes>;
+    tokeniser: ITokeniser;
+    metaData: TransformersMetadata;
+    optimizer?: AdamWOptimizer;
+    log?: TrainingLogEntry[];
+}

package/dist/main.d.ts

@@ -12,7 +12,9 @@ export { default as TeachableLLM } from './TeachableLLM';
 export { default as CharTokeniser } from './tokeniser/CharTokeniser';
 export { default as BPETokeniser } from './tokeniser/bpe';
 export { default as waitForModel } from './utilities/waitForModel';
+export { default as generateDatasetID } from './utilities/datasetID';
 export { default as loadTextData } from './data/textLoader';
+export type { DatasetMetadata } from './loader/types';
 export { default as Generator, type IGenerator } from './Generator';
 export { default as Evaluator } from './training/Evaluator';
 export { default as Trainer } from './Trainer';

package/dist/main.js

@@ -1,21 +1,22 @@
-import "./index-CUXkjxiT.js";
-import "./random_width-DN5ZtQkM.js";
-import "./zeros_like-CWjDdwr-.js";
+import "./index-DSGwv2Yx.js";
+import "./random_width-B_fVXhGx.js";
+import "./zeros_like-rOHr54NY.js";
 import { default as io } from "./Generator.js";
 import "./index-Cp39cXWe.js";
-import "./dataset-CGGp1z9P.js";
+import "./dataset-DlqAN81i.js";
 import { default as fo } from "./models/NanoGPTV1.js";
 import { default as lo } from "./TeachableLLM.js";
 import { default as uo } from "./tokeniser/CharTokeniser.js";
 import { default as ko } from "./tokeniser/bpe.js";
 import { default as go } from "./utilities/waitForModel.js";
-import { default as Co } from "./data/textLoader.js";
-import { default as Eo } from "./training/Evaluator.js";
-import { default as Bo } from "./Trainer.js";
-import { default as vo } from "./models/model.js";
-import { estimateMemoryUsage as Lo, estimateParameterCount as So, estimateResources as Ao, estimateTrainingMemoryUsage as Fo, validateConfig as Go } from "./utilities/parameters.js";
-import { default as Uo } from "./utilities/topP.js";
-import { Task as zo, tokensFromTasks as Do } from "./training/tasks/Task.js";
+import { default as Co } from "./utilities/datasetID.js";
+import { default as Eo } from "./data/textLoader.js";
+import { default as Bo } from "./training/Evaluator.js";
+import { default as vo } from "./Trainer.js";
+import { default as Do } from "./models/model.js";
+import { estimateMemoryUsage as So, estimateParameterCount as Ao, estimateResources as Fo, estimateTrainingMemoryUsage as Go, validateConfig as Ro } from "./utilities/parameters.js";
+import { default as wo } from "./utilities/topP.js";
+import { Task as Io, tokensFromTasks as No } from "./training/tasks/Task.js";
 import o from "./training/tasks/PretrainingTask.js";
 import r from "./training/tasks/StartSentenceTask.js";
 import t from "./training/tasks/ConversationTask.js";
@@ -31,15 +32,15 @@ import "./ops/webgl/qkv.js";
 import "./ops/grads/qkv.js";
 import "./ops/cpu/rope.js";
 import "./ops/webgl/rope.js";
-import "./rope-0j_f1TPm.js";
+import "./rope-CC5RjmKU.js";
 import "./ops/cpu/appendCache.js";
 import "./ops/webgl/appendCache.js";
 import "./ops/cpu/matMulGelu.js";
-import "./matMulGelu-JNLZqKQp.js";
+import "./matMulGelu-CsZnh18H.js";
 import "./ops/grads/matMulGelu.js";
 import "./ops/cpu/gelu.js";
 import "./ops/webgl/gelu.js";
-import "./gelu-B220X1Go.js";
+import "./gelu-DpTCC3eB.js";
 import "./ops/cpu/normRMS.js";
 import "./ops/webgl/normRMS.js";
 import "./ops/grads/normRMS.js";
@@ -48,21 +49,21 @@ import "./ops/cpu/adamMoments.js";
 import "./ops/webgl/adamMoments.js";
 import "./ops/cpu/adamAdjust.js";
 import "./ops/webgl/adamAdjust.js";
-import { u as e, p as m } from "./pack16-bqltoUlR.js";
+import { u as e, p as m } from "./pack16-WlOSOuZA.js";
 import "./ops/grads/softmax16.js";
-import "./matMul16-BcVC_E62.js";
+import "./matMul16-BIT70Vya.js";
 import "./ops/webgl/matMul16.js";
 import "./ops/cpu/matMul16.js";
 import "./ops/transpose16.js";
-import { selectBackend as Oo } from "./backend.js";
-import { default as jo } from "./utilities/performance.js";
-import p from "./layers/CausalSelfAttention.js";
-import a from "./layers/MLP.js";
+import { selectBackend as Wo } from "./backend.js";
+import { default as qo } from "./utilities/performance.js";
+import a from "./layers/CausalSelfAttention.js";
+import p from "./layers/MLP.js";
 import i from "./layers/TransformerBlock.js";
 import s from "./layers/RoPECache.js";
-import { AdamWOptimizer as Ho } from "./training/AdamW.js";
-import { default as Jo } from "./checks/index.js";
-import { sentenceEmbeddings as Qo, sentenceEmbeddingsTensor as Vo } from "./utilities/sentences.js";
+import { AdamWOptimizer as Jo } from "./training/AdamW.js";
+import { default as Qo } from "./checks/index.js";
+import { sentenceEmbeddings as Xo, sentenceEmbeddingsTensor as Yo } from "./utilities/sentences.js";
 const to = {
   PretrainingTask: o,
   StartSentenceTask: r,
@@ -71,37 +72,38 @@ const to = {
   pack16: m,
   unpack16: e
 }, mo = {
-  CausalSelfAttention: p,
-  MLP: a,
+  CausalSelfAttention: a,
+  MLP: p,
   TransformerBlock: i,
   RoPECache: s
 };
 export {
-  Ho as AdamWOptimizer,
+  Jo as AdamWOptimizer,
   ko as BPETokeniser,
   uo as CharTokeniser,
-  Eo as Evaluator,
+  Bo as Evaluator,
   io as Generator,
-  vo as Model,
+  Do as Model,
   fo as NanoGPT,
-  zo as Task,
+  Io as Task,
   lo as TeachableLLM,
-  Bo as Trainer,
-  Jo as checks,
-  Lo as estimateMemoryUsage,
-  So as estimateParameterCount,
-  Ao as estimateResources,
-  Fo as estimateTrainingMemoryUsage,
+  vo as Trainer,
+  Qo as checks,
+  So as estimateMemoryUsage,
+  Ao as estimateParameterCount,
+  Fo as estimateResources,
+  Go as estimateTrainingMemoryUsage,
+  Co as generateDatasetID,
   mo as layers,
-  Co as loadTextData,
+  Eo as loadTextData,
   eo as ops,
-  jo as performanceTest,
-  Oo as selectBackend,
-  Qo as sentenceEmbeddings,
-  Vo as sentenceEmbeddingsTensor,
+  qo as performanceTest,
+  Wo as selectBackend,
+  Xo as sentenceEmbeddings,
+  Yo as sentenceEmbeddingsTensor,
   to as tasks,
-  Do as tokensFromTasks,
-  Uo as topP,
-  Go as validateConfig,
+  No as tokensFromTasks,
+  wo as topP,
+  Ro as validateConfig,
   go as waitForModel
 };

package/dist/{matMul16-BcVC_E62.js → matMul16-BIT70Vya.js}

@@ -1,9 +1,9 @@
-import { e as y, h } from "./index-CUXkjxiT.js";
+import { c as y, e as h } from "./index-DSGwv2Yx.js";
 import "./ops/webgl/matMul16.js";
 import "./ops/cpu/matMul16.js";
 import { isPackedTensor as v } from "./utilities/packed.js";
-import { p as g } from "./pack16-bqltoUlR.js";
-import { d as k } from "./gelu-B220X1Go.js";
+import { p as g } from "./pack16-WlOSOuZA.js";
+import { d as k } from "./gelu-DpTCC3eB.js";
 import { transpose16 as S } from "./ops/transpose16.js";
 import { reshape16 as w } from "./ops/reshape16.js";
 import { mul16 as D } from "./ops/mul16.js";