@genai-fi/nanogpt 0.19.0 → 0.20.0

package/dist/layers/PositionEmbedding.d.ts

@@ -1,8 +0,0 @@
-import { Tensor } from '@tensorflow/tfjs-core';
-import { default as BaseLayer } from './BaseLayer';
-import { GPTConfig, ModelForwardAttributes } from '../main';
-export default class PositionEmbedding extends BaseLayer {
-    private wpe?;
-    constructor(config: GPTConfig, name?: string, parent?: BaseLayer);
-    forward(attrs: ModelForwardAttributes, x: Tensor): Tensor;
-}

package/dist/layers/PositionEmbedding.js

@@ -1,31 +0,0 @@
-import { t as c, an as p, a as s } from "../index-CUXkjxiT.js";
-import f from "./BaseLayer.js";
-import { E as u } from "../random_width-DN5ZtQkM.js";
-import { r as b } from "../exports_initializers-CYzKLjN7.js";
-import { m as g } from "../mod-CSdCpRjf.js";
-import { r as l } from "../range-C-CjF-LI.js";
-function h(e) {
-  return new u(e);
-}
-class k extends f {
-  wpe;
-  // Position embeddings
-  constructor(n, t = "", o) {
-    super(n, o), this.wpe = h({
-      inputDim: this.config.blockSize,
-      outputDim: this.config.nEmbed,
-      name: t,
-      embeddingsInitializer: b({ mean: 0, stddev: 0.02 })
-    });
-  }
-  forward(n, t) {
-    const o = n.cache?.[0]?.length ?? 0;
-    return c(() => {
-      const [, i] = t.shape, r = this.config.blockSize, m = l(0, i, 1, "int32"), a = g(p(m, s(o, "int32")), s(r, "int32")), d = this.wpe.apply(a);
-      return t.add(d);
-    });
-  }
-}
-export {
-  k as default
-};

package/dist/layers/RMSNorm.d.ts

@@ -1,12 +0,0 @@
-import { Tensor } from '@tensorflow/tfjs-core';
-import { default as BaseLayer, ForwardAttributes } from './BaseLayer';
-import { GPTConfig } from '../main';
-export interface RMSNormConfig {
-    useGamma?: boolean;
-}
-export default class RMSNorm extends BaseLayer {
-    private GAMMA;
-    private rmsConfig;
-    constructor(config: GPTConfig, rmsConfig: RMSNormConfig, name?: string, parent?: BaseLayer);
-    forward(_: ForwardAttributes, x: Tensor): Tensor;
-}

package/dist/layers/RMSNorm.js

@@ -1,22 +0,0 @@
-import { t as i } from "../index-CUXkjxiT.js";
-import e from "./BaseLayer.js";
-import { normRMS as m } from "../ops/normRMS.js";
-import { v as a } from "../variable-Ck482e3n.js";
-import { o as M } from "../ones-BytntneX.js";
-class l extends e {
-  GAMMA;
-  rmsConfig;
-  constructor(r, t, s = "", o) {
-    super(r, o), this.GAMMA = s, this.rmsConfig = t, (this.rmsConfig.useGamma ?? !0) && this.addVariable(this.GAMMA, a(M([r.nEmbed]), !0, this.GAMMA, "float32"));
-  }
-  forward(r, t) {
-    return i(() => {
-      this.startMemory();
-      const s = m(t, this.rmsConfig.useGamma ?? !0 ? this.getVariable(this.GAMMA) : void 0);
-      return this.endMemory("RMSNorm"), s;
-    });
-  }
-}
-export {
-  l as default
-};

package/dist/layers/RoPECache.d.ts

@@ -1,18 +0,0 @@
-import { Tensor } from '@tensorflow/tfjs-core';
-import { GPTConfig } from '../models/config';
-export default class RoPECache {
-    private readonly config;
-    readonly rotaryDim: number;
-    private ropeBase;
-    private ropeInvFreq;
-    private ropeCos;
-    private ropeSin;
-    private ropeNegSin;
-    private ropeCacheLen;
-    constructor(config: GPTConfig);
-    ensureRopeCache(needed: number): void;
-    getCos(): Tensor | null;
-    getSin(): Tensor | null;
-    getNegSin(): Tensor | null;
-    dispose(): void;
-}

package/dist/layers/RoPECache.js

@@ -1,50 +0,0 @@
-import { i as a, a as n, p as c, t as p, k as r } from "../index-CUXkjxiT.js";
-import { r as l } from "../zeros_like-CWjDdwr-.js";
-import { c as f, s as m } from "../unsorted_segment_sum-DJvk5xnh.js";
-import { r as h } from "../range-C-CjF-LI.js";
-class x {
-  constructor(s) {
-    this.config = s;
-    const o = this.config.nEmbed / this.config.nHead;
-    if (this.rotaryDim = o, this.rotaryDim % 2 !== 0)
-      throw new Error("rotaryDim must be even");
-    this.ropeBase = 1e4;
-    const i = h(0, this.rotaryDim, 2, "float32"), e = a(i, n(this.rotaryDim, "float32")), t = c(n(this.ropeBase, "float32"), e);
-    this.ropeInvFreq = l(t), e.dispose(), t.dispose(), i.dispose(), (this.config.modelType === "GenAI_NanoGPT_v1" ? this.config.useRope ?? !0 : !0) === !1 ? (this.ropeCos = null, this.ropeSin = null, this.ropeNegSin = null, this.ropeCacheLen = 0) : p(() => {
-      this.ensureRopeCache(this.config.blockSize * 4);
-    });
-  }
-  rotaryDim;
-  ropeBase;
-  ropeInvFreq;
-  ropeCos = null;
-  // [cacheLen, rotaryDim/2]
-  ropeSin = null;
-  // [cacheLen, rotaryDim/2]
-  ropeNegSin = null;
-  // [cacheLen, rotaryDim/2]
-  ropeCacheLen = 0;
-  ensureRopeCache(s) {
-    p(() => {
-      if (s <= this.ropeCacheLen) return;
-      this.ropeCos && this.ropeCos.dispose(), this.ropeSin && this.ropeSin.dispose();
-      const o = Math.max(s, this.ropeCacheLen + this.config.blockSize * 4), e = h(0, o, 1, "float32").expandDims(1).mul(this.ropeInvFreq.expandDims(0));
-      this.ropeCos = r(f(e).expandDims(-1)), this.ropeSin = r(m(e).expandDims(-1)), this.ropeNegSin = r(this.ropeSin.neg()), this.ropeCacheLen = o;
-    });
-  }
-  getCos() {
-    return this.ropeCos;
-  }
-  getSin() {
-    return this.ropeSin;
-  }
-  getNegSin() {
-    return this.ropeNegSin;
-  }
-  dispose() {
-    this.ropeCos && this.ropeCos.dispose(), this.ropeSin && this.ropeSin.dispose(), this.ropeInvFreq.dispose();
-  }
-}
-export {
-  x as default
-};

package/dist/layers/TiedEmbedding.d.ts

@@ -1,13 +0,0 @@
-import { Tensor } from '@tensorflow/tfjs-core';
-import { default as BaseLayer, ForwardAttributes } from './BaseLayer';
-import { GPTConfig } from '../models/config';
-export default class TiedEmbeddingOutputLayer extends BaseLayer {
-    private vocabSize;
-    private embedDim;
-    private initializer;
-    private WEIGHTS;
-    constructor(config: GPTConfig, name: string, parent?: BaseLayer);
-    embed(inputs: Tensor): Tensor;
-    project(inputs: Tensor): Tensor;
-    forward(_: ForwardAttributes, x: Tensor): Tensor;
-}

package/dist/layers/TiedEmbedding.js

@@ -1,36 +0,0 @@
-import "../random_width-DN5ZtQkM.js";
-import "../index-CUXkjxiT.js";
-import { r as s } from "../exports_initializers-CYzKLjN7.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 { transpose16 as d } from "../ops/transpose16.js";
-import { v as p } from "../variable-Ck482e3n.js";
-import { g as h } from "../gather-qIqEqaGn.js";
-class g extends a {
-  vocabSize;
-  embedDim;
-  initializer;
-  WEIGHTS;
-  constructor(t, i, e) {
-    super(t, e), this.WEIGHTS = i, this.vocabSize = t.vocabSize, this.embedDim = t.nEmbed, this.initializer = s({
-      mean: 0,
-      stddev: 0.02
-    }), this.addVariable(this.WEIGHTS, p(this.initializer.apply([this.vocabSize, this.embedDim]), !0, i));
-  }
-  embed(t) {
-    return h(this.getVariable(this.WEIGHTS), t, 0);
-  }
-  project(t) {
-    const i = r(t) ? m(this.getVariable(this.WEIGHTS), void 0, 32) : this.getVariable(this.WEIGHTS), e = d(i);
-    return r(t) && i.dispose(), o(t, e);
-  }
-  // Dummy, should not be used.
-  forward(t, i) {
-    return this.project(i);
-  }
-}
-export {
-  g as default
-};

package/dist/layers/TransformerBlock.d.ts

@@ -1,27 +0,0 @@
-import { AttentionScores, CausalSelfAttentionConfig, KVCache } from './CausalSelfAttention';
-import { MLPConfig } from './MLP';
-import { RMSNormConfig } from './RMSNorm';
-import { default as BaseLayer, ForwardAttributes } from './BaseLayer';
-import { Tensor } from '@tensorflow/tfjs-core';
-import { GPTConfig } from '../models/config';
-interface BlockAttributes extends ForwardAttributes {
-    pastKV?: KVCache;
-    seed?: number;
-    attentionScores?: AttentionScores;
-    ropePositionOffset?: number;
-}
-export type TransformerBlockConfig = MLPConfig & RMSNormConfig & CausalSelfAttentionConfig;
-export default class Block extends BaseLayer<BlockAttributes> {
-    private ln1;
-    private attn;
-    private ln2;
-    private mlp;
-    private index;
-    skipped: boolean;
-    private blockConfig;
-    constructor(index: number, config: GPTConfig, blockConfig: TransformerBlockConfig, parent?: BaseLayer);
-    private getMLPOutput;
-    forward(attrs: BlockAttributes, x: Tensor): Tensor;
-    dispose(): void;
-}
-export {};

package/dist/layers/TransformerBlock.js

@@ -1,40 +0,0 @@
-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 { add16 as d } from "../ops/add16.js";
-class _ extends m {
-  ln1;
-  attn;
-  ln2;
-  mlp;
-  index;
-  skipped = !1;
-  blockConfig;
-  constructor(i, s, e, t) {
-    super(s, t), this.index = i, this.blockConfig = e, this.ln1 = new l(s, this.blockConfig, `block_${this.index}_rms1`, this), this.attn = new p(this.index, s, this.blockConfig, this), this.ln2 = new l(s, this.blockConfig, `block_${this.index}_rms2`, this), this.mlp = new h(this.index, s, this.blockConfig, this);
-  }
-  getMLPOutput(i, s) {
-    const e = this.ln2.call({ training: s.training }, i), t = this.mlp.call(s, e);
-    s.outputEmbeddings ? (n(e), s.embeddings.push({ name: `block_ln2_${this.index}`, tensor: e })) : e.dispose();
-    const o = d(i, t);
-    return i.dispose(), s.outputEmbeddings ? (n(t), s.embeddings.push({ name: `block_mlp_out_${this.index}`, tensor: t })) : t.dispose(), o;
-  }
-  forward(i, s) {
-    return u(() => {
-      if (this.skipped)
-        return s;
-      const e = this.ln1.call(i, s), t = this.attn.call(i, e);
-      i.outputEmbeddings ? (n(e), i.embeddings.push({ name: `block_ln1_${this.index}`, tensor: e })) : e.dispose();
-      const o = d(s, t);
-      return i.outputEmbeddings ? (n(t), i.embeddings.push({ name: `block_attn_out_${this.index}`, tensor: t })) : t.dispose(), this.getMLPOutput(o, i);
-    });
-  }
-  dispose() {
-    this.ln1.dispose(), this.attn.dispose(), this.ln2.dispose(), this.mlp.dispose();
-  }
-}
-export {
-  _ as default
-};

package/dist/layers/WeightStore.d.ts

@@ -1,20 +0,0 @@
-import { Tensor, Variable } from '@tensorflow/tfjs-core';
-export default class WeightStore {
-    private _variables;
-    private touchedVariables;
-    onWeightRead?: (name: string, variable: Variable) => Tensor;
-    saveWeights(map: Map<string, Tensor[]>): void;
-    loadWeights(weights: Map<string, Tensor[]>, reference: boolean, trainable?: boolean): void;
-    addVariable(name: string, variable?: Variable): void;
-    deleteVariable(name: string): void;
-    get variables(): Variable[];
-    get variableNames(): string[];
-    get trainableVariables(): Variable[];
-    getRawVariable(name: string): Variable;
-    getVariable(name: string): Tensor;
-    setTrainable(names: string[]): void;
-    hasVariable(name: string): boolean;
-    setVariable(name: string, variable: Variable): void;
-    touchVariables(names: string[]): void;
-    dispose(): void;
-}

package/dist/layers/WeightStore.js

@@ -1,76 +0,0 @@
-import { w as o } from "../index-CUXkjxiT.js";
-import { p as h } from "../index-DmeWGGmS.js";
-import { v as b } from "../variable-Ck482e3n.js";
-class d {
-  _variables = /* @__PURE__ */ new Map();
-  touchedVariables = /* @__PURE__ */ new Set();
-  // Hooks
-  onWeightRead;
-  saveWeights(a) {
-    this._variables.forEach((e, i) => {
-      e && this.touchedVariables.has(i) && a.set(i, [o(e)]);
-    });
-  }
-  loadWeights(a, e, i = !0) {
-    a.forEach((s, r) => {
-      const t = s[0], l = this._variables.get(r);
-      l ? l.assign(t) : this._variables.set(r, b(t, i, r)), e ? this.touchedVariables.delete(r) : this.touchedVariables.add(r);
-    });
-  }
-  addVariable(a, e) {
-    this._variables.set(a, e || null);
-  }
-  deleteVariable(a) {
-    const e = this._variables.get(a);
-    e && e.dispose(), this._variables.delete(a), this.touchedVariables.delete(a);
-  }
-  get variables() {
-    return Array.from(this._variables.values()).filter((e) => e !== null);
-  }
-  get variableNames() {
-    return Array.from(this._variables.keys());
-  }
-  get trainableVariables() {
-    return Array.from(this._variables.values()).filter(
-      (e) => e !== null && e.trainable
-    );
-  }
-  getRawVariable(a) {
-    const e = this._variables.get(a);
-    if (!e)
-      throw new Error(`Variable ${a} not found`);
-    return e;
-  }
-  getVariable(a) {
-    const e = this._variables.get(a);
-    if (!e)
-      throw new Error(`Variable ${a} not found`);
-    return this.onWeightRead ? this.onWeightRead(a, e) : e;
-  }
-  setTrainable(a) {
-    const e = h(a);
-    this._variables.forEach((i, s) => {
-      i && (i.trainable = e(s));
-    });
-  }
-  hasVariable(a) {
-    return !!this._variables.get(a);
-  }
-  setVariable(a, e) {
-    if (!this._variables.has(a))
-      throw new Error(`Variable ${a} not found`);
-    this._variables.set(a, e);
-  }
-  touchVariables(a) {
-    for (const e of a)
-      this._variables.get(e) && this.touchedVariables.add(e);
-  }
-  dispose() {
-    this._variables.forEach((a) => {
-      a?.dispose();
-    }), this._variables.clear();
-  }
-}
-export {
-  d as default
-};

package/dist/loader/load.d.ts

@@ -1,6 +0,0 @@
-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<LoadResult>;

package/dist/loader/load.js

@@ -1,68 +0,0 @@
-import { z as i } from "../jszip.min-BZhlzntC.js";
-import u from "./oldZipLoad.js";
-import d from "./newZipLoad.js";
-import w from "./loadHF.js";
-import { loadZipMeta as s } from "./loadZipMeta.js";
-import { load_safetensors as g } from "../utilities/safetensors.js";
-const j = 2;
-async function h(e) {
-  const t = await fetch(e);
-  if (!t.ok)
-    throw new Error(`Failed to fetch ${e}: ${t.statusText}`);
-  return t.arrayBuffer();
-}
-async function p(e, t, o) {
-  const a = e.file("model.safetensors");
-  if (!a)
-    return;
-  const r = await a.async("arraybuffer"), c = await g(r), n = /* @__PURE__ */ new Map();
-  for (const [l, m] of Object.entries(c))
-    n.set(l, [m]);
-  t.weightStore.loadWeights(n, o);
-}
-async function y(e, t) {
-  const o = e.file("config.json");
-  if (!o)
-    return;
-  const a = await o.async("string"), r = JSON.parse(a);
-  if (r.loraName) {
-    if (t.hasLoRA())
-      throw new Error("Model already has LoRA attached");
-    t.attachLoRA(r.loraName);
-  }
-}
-async function f(e, t) {
-  if (t.reference) {
-    const o = await L(t.reference);
-    return await p(e, o.model, !!t.url), await y(e, o.model), o.model.config.loraName && o.model.attachLoRA(o.model.config.loraName), {
-      ...o,
-      metaData: {
-        ...o.metaData,
-        ...t
-      }
-    };
-  } else {
-    if (e.file("manifest.json"))
-      return u(e, t);
-    {
-      const o = await d(e, t);
-      return o.model.config.loraName && o.model.attachLoRA(o.model.config.loraName), o;
-    }
-  }
-}
-async function L(e, t) {
-  if (typeof e == "string")
-    if (e.startsWith("http://") || e.startsWith("https://")) {
-      const o = await h(e), a = await i.loadAsync(o), r = await s(a);
-      return r.url = e, f(a, r);
-    } else
-      return w(e);
-  else {
-    const o = await i.loadAsync(e), a = await s(o);
-    return a.url = t?.sourceURL || void 0, f(o, a);
-  }
-}
-export {
-  j as VERSION,
-  L as loadModel
-};

package/dist/loader/loadHF.d.ts

@@ -1,8 +0,0 @@
-import { ITokeniser } from '../main';
-import { default as Model, ModelForwardAttributes } from '../models/model';
-import { TransformersMetadata } from './types';
-export default function loadHuggingFace(name: string): Promise<{
-    model: Model<ModelForwardAttributes>;
-    tokeniser: ITokeniser;
-    metaData: TransformersMetadata;
-}>;

package/dist/loader/loadHF.js

@@ -1,22 +0,0 @@
-import w from "./loadTransformers.js";
-async function u(t) {
-  const r = `https://huggingface.co/${t}/resolve/main/config.json`, a = `https://huggingface.co/${t}/resolve/main/tokeniser.json`, i = `https://huggingface.co/${t}/resolve/main/meta.json`, c = `https://huggingface.co/${t}/resolve/main/model.safetensors`, [o, e, s, n] = await Promise.all([
-    fetch(r),
-    fetch(a),
-    fetch(i),
-    fetch(c)
-  ]);
-  if (!o.ok)
-    throw new Error(`Failed to fetch config from ${r}: ${o.statusText}`);
-  if (!e.ok)
-    throw new Error(`Failed to fetch tokeniser from ${a}: ${e.statusText}`);
-  if (!s.ok)
-    throw new Error(`Failed to fetch meta from ${i}: ${s.statusText}`);
-  if (!n.ok)
-    throw new Error(`Failed to fetch weights from ${c}: ${n.statusText}`);
-  const f = await o.json(), g = await e.json(), h = await s.json(), l = await n.arrayBuffer();
-  return w(f, g, h, l);
-}
-export {
-  u as default
-};

package/dist/loader/loadTransformers.d.ts

@@ -1,4 +0,0 @@
-import { GPTConfig } from '../models/config';
-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<LoadResult>;

package/dist/loader/loadTransformers.js

@@ -1,44 +0,0 @@
-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 { dummyPassAsync as u } from "../utilities/dummy.js";
-import y from "../models/factory.js";
-function h(e) {
-  let a;
-  return e.model_type === "GenAI_NanoGPT_v1" ? a = {
-    modelType: "GenAI_NanoGPT_v1",
-    vocabSize: e.vocab_size,
-    blockSize: e.block_size,
-    nLayer: e.num_hidden_layers,
-    nHead: e.num_attention_heads,
-    nEmbed: e.hidden_size,
-    mlpFactor: e.mlpFactor,
-    useRope: e.useRope
-  } : a = {
-    modelType: "GenAI_NanoGPT_v2",
-    vocabSize: e.vocab_size,
-    blockSize: e.block_size,
-    nLayer: e.num_hidden_layers,
-    nHead: e.num_attention_heads,
-    nEmbed: e.hidden_size,
-    mlpFactor: e.mlpFactor,
-    loraConfig: e.loraConfig ? new Map(Object.entries(e.loraConfig)) : void 0,
-    loraName: e.loraName,
-    windowSize: e.windowSize
-  }, a;
-}
-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 t = y(m);
-  return t.metaData = r, await u(t), t.weightStore.loadWeights(s, !!r.url), { model: t, tokeniser: o, metaData: r };
-}
-export {
-  G as default,
-  h as mapTransformersConfigToGPTConfig
-};

package/dist/loader/loadZipMeta.d.ts

@@ -1,3 +0,0 @@
-import { default as zip } from 'jszip';
-import { TransformersMetadata } from './types';
-export declare function loadZipMeta(zipFile: zip): Promise<TransformersMetadata>;

package/dist/loader/loadZipMeta.js

@@ -1,16 +0,0 @@
-async function r(t) {
-  const e = await t.file("meta.json")?.async("string");
-  let a = { version: 0, application: "" };
-  if (e)
-    try {
-      a = JSON.parse(e);
-    } catch (n) {
-      throw console.error(n), new Error("Failed to parse meta.json in the zip archive");
-    }
-  else
-    console.warn("meta.json not found in the zip archive, using default metadata");
-  return a;
-}
-export {
-  r as loadZipMeta
-};

package/dist/loader/newZipLoad.d.ts

@@ -1,3 +0,0 @@
-import { default as zip } from 'jszip';
-import { LoadResult, TransformersMetadata } from './types';
-export default function loadZipFile(zipFile: zip, metaData: TransformersMetadata): Promise<LoadResult>;

package/dist/loader/newZipLoad.js

@@ -1,31 +0,0 @@
-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 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 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");
-  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 {
-  u as default
-};

package/dist/loader/oldZipLoad.d.ts

@@ -1,9 +0,0 @@
-import { default as zip } from 'jszip';
-import { ITokeniser } from '../main';
-import { default as Model, ModelForwardAttributes } from '../models/model';
-import { TransformersMetadata } from './types';
-export default function loadOldModel(zipFile: zip, metaData: TransformersMetadata): Promise<{
-    model: Model<ModelForwardAttributes>;
-    tokeniser: ITokeniser;
-    metaData: TransformersMetadata;
-}>;