@genai-fi/nanogpt 0.6.0 → 0.6.2

package/dist/loader/load.d.ts

@@ -0,0 +1,13 @@
+import { default as NanoGPT } from '../NanoGPTModel';
+import { ITokeniser } from '../tokeniser/type';
+export declare const VERSION = 2;
+export interface Metadata {
+    version: string;
+    application: string;
+    name?: string;
+}
+export declare function loadModel(data: Blob | Buffer | string): Promise<{
+    model: NanoGPT;
+    tokeniser: ITokeniser;
+    name?: string;
+}>;

package/dist/loader/load.js

@@ -0,0 +1,27 @@
+import { j as o } from "../jszip.min-CjP2V1VV.js";
+import i from "./oldZipLoad.js";
+import s from "./newZipLoad.js";
+import n from "./loadHF.js";
+const u = 2;
+async function f(t) {
+  const e = await fetch(t);
+  if (!e.ok)
+    throw new Error(`Failed to fetch ${t}: ${e.statusText}`);
+  return e.arrayBuffer();
+}
+async function m(t) {
+  if (typeof t == "string")
+    if (t.startsWith("http://") || t.startsWith("https://")) {
+      const e = await f(t), r = await o.loadAsync(e);
+      return r.file("manifest.json") ? i(r) : s(r);
+    } else
+      return n(t);
+  else {
+    const e = await o.loadAsync(t);
+    return e.file("manifest.json") ? i(e) : s(e);
+  }
+}
+export {
+  u as VERSION,
+  m as loadModel
+};

package/dist/loader/loadHF.d.ts

@@ -0,0 +1,7 @@
+import { default as NanoGPT } from '../NanoGPTModel';
+import { ITokeniser } from '../main';
+export default function loadHuggingFace(name: string): Promise<{
+    model: NanoGPT;
+    tokeniser: ITokeniser;
+    name?: string;
+}>;

package/dist/loader/loadHF.js

@@ -0,0 +1,22 @@
+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/{utilities/load.d.ts → loader/loadTransformers.d.ts}

@@ -1,7 +1,5 @@
-import { default as zip } from 'jszip';
-import { default as NanoGPT } from '../NanoGPTModel';
 import { ITokeniser } from '../tokeniser/type';
-export declare const VERSION = 2;
+import { default as NanoGPT } from '../NanoGPTModel';
 export interface TransformersConfig {
     model_type: string;
     vocab_size: number;
@@ -15,16 +13,18 @@ export interface TransformersConfig {
     mlpFactor: number;
     useRope: boolean;
 }
-export interface Metadata {
-    version: string;
-    application: string;
+export interface TransformersTokeniser {
+    type: 'char' | 'bpe';
+    vocab: string[];
+    merges: [string, string][];
+}
+export interface TransformersMetadata {
     name?: string;
+    version: number;
+    application: string;
+    [key: string]: unknown;
 }
-export declare function loadOldModel(zipFile: zip): Promise<{
-    model: NanoGPT;
-    tokeniser: ITokeniser;
-}>;
-export declare function loadModel(data: Blob | Buffer | string): Promise<{
+export default function loadTransformers(config: TransformersConfig, tokeniser: TransformersTokeniser, metadata: TransformersMetadata, weightData: ArrayBuffer): Promise<{
     model: NanoGPT;
     tokeniser: ITokeniser;
     name?: string;

package/dist/loader/loadTransformers.js

@@ -0,0 +1,28 @@
+import b from "../NanoGPTModel.js";
+import c from "../tokeniser/CharTokeniser.js";
+import l from "../tokeniser/bpe.js";
+import { load_safetensors as u } from "../utilities/safetensors.js";
+import { U as y } from "../index-BAzbokzv.js";
+import { dummyPassAsync as h } from "../utilities/dummy.js";
+async function L(e, a, r, t) {
+  const n = {
+    vocabSize: e.vocab_size,
+    blockSize: e.block_size,
+    nLayer: e.num_hidden_layers,
+    nHead: e.num_attention_heads,
+    nEmbed: e.hidden_size,
+    dropout: e.dropout,
+    biasInLinear: e.biasInLinear,
+    biasInLayerNorm: e.biasInLayerNorm,
+    mlpFactor: e.mlpFactor,
+    useRope: e.useRope
+  }, m = (a.type ?? "char") === "char" ? new c(a.vocab) : new l(a.vocab, a.merges), i = await u(t), s = /* @__PURE__ */ new Map();
+  for (const [p, d] of Object.entries(i))
+    s.set(p, [d]);
+  y();
+  const o = new b(n);
+  return await h(o), o.loadWeights(s), { model: o, tokeniser: m, name: r.name };
+}
+export {
+  L as default
+};

package/dist/loader/newZipLoad.d.ts

@@ -0,0 +1,8 @@
+import { ITokeniser } from '../main';
+import { default as NanoGPT } from '../NanoGPTModel';
+import { default as zip } from 'jszip';
+export default function loadZipFile(zipFile: zip): Promise<{
+    model: NanoGPT;
+    tokeniser: ITokeniser;
+    name?: string;
+}>;

package/dist/loader/newZipLoad.js

@@ -0,0 +1,21 @@
+import c from "./loadTransformers.js";
+async function g(r) {
+  const e = await r.file("config.json")?.async("string");
+  if (!e)
+    throw new Error("Config file not found in the zip archive");
+  const t = JSON.parse(e), n = await r.file("tokeniser.json")?.async("string");
+  if (!n)
+    throw new Error("Tokeniser file not found in the zip archive");
+  const i = JSON.parse(n), s = await r.file("model.safetensors").async("arraybuffer"), o = await r.file("meta.json")?.async("string");
+  let a = { version: 0, application: "" };
+  if (o)
+    try {
+      a = JSON.parse(o);
+    } catch (f) {
+      console.error("Error parsing meta file:", f);
+    }
+  return c(t, i, a, s);
+}
+export {
+  g as default
+};

package/dist/loader/oldZipLoad.d.ts

@@ -0,0 +1,7 @@
+import { default as zip } from 'jszip';
+import { ITokeniser } from '../main';
+import { default as NanoGPT } from '../NanoGPTModel';
+export default function loadOldModel(zipFile: zip): Promise<{
+    model: NanoGPT;
+    tokeniser: ITokeniser;
+}>;

package/dist/loader/oldZipLoad.js

@@ -0,0 +1,76 @@
+import d from "../NanoGPTModel.js";
+import "../jszip.min-CjP2V1VV.js";
+import h from "../tokeniser/CharTokeniser.js";
+import { U as k } from "../index-BAzbokzv.js";
+import b from "../tokeniser/bpe.js";
+import { dummyPassAsync as u } from "../utilities/dummy.js";
+import "../Generator.js";
+import "../index-Dwqa6Zy2.js";
+import "../dataset-pgqp-YfL.js";
+import "../index-Tf7vU29b.js";
+import "../papaparse.min-C8l2Kvo1.js";
+import "../ops/cpu/scatterSub.js";
+import "../ops/webgl/scatterSub.js";
+import "../ops/cpu/gatherSub.js";
+import "../ops/webgl/gatherSub.js";
+import "../ops/cpu/attentionMask.js";
+import "../ops/webgl/attentionMask.js";
+import "../ops/grads/attentionMask.js";
+import "../ops/cpu/qkv.js";
+import "../ops/webgl/qkv.js";
+import "../ops/grads/qkv.js";
+import "../random_width-DWzaOgrn.js";
+import "../register_all_kernels-BfFCQAqs.js";
+import "../ops/cpu/rope.js";
+import "../ops/webgl/rope.js";
+import "../ops/grads/rope.js";
+import "../ops/cpu/appendCache.js";
+import "../ops/webgl/appendCache.js";
+import "../ops/cpu/fusedSoftmax.js";
+import "../ops/webgl/fusedSoftmax.js";
+import "../ops/grads/fusedSoftmax.js";
+import "../ops/cpu/matMulGelu.js";
+import "../ops/webgl/matMulGelu.js";
+import "../ops/grads/matMulGelu.js";
+import "../ops/cpu/gelu.js";
+import "../ops/webgl/gelu.js";
+import "../ops/grads/gelu.js";
+import "../ops/cpu/normRMS.js";
+import "../ops/webgl/normRMS.js";
+import "../ops/grads/normRMS.js";
+import "../ops/webgl/log.js";
+import { importWeights as O } from "../utilities/weights.js";
+async function ft(o) {
+  const n = /* @__PURE__ */ new Map(), s = await o.file("manifest.json")?.async("string");
+  if (!s)
+    throw new Error("Manifest file not found in the zip archive");
+  const m = JSON.parse(s);
+  for (const [t, r] of Object.entries(m.weightSpec))
+    n.set(t, { spec: r, data: new Float32Array() });
+  const p = await o.file("tokeniser.json")?.async("string");
+  if (!p)
+    throw new Error("Tokeniser file not found in the zip archive");
+  const i = JSON.parse(p), l = (i.type ?? "char") === "char" ? new h(i.vocab) : new b(i.vocab, i.merges), c = /* @__PURE__ */ new Map();
+  for (const t of Object.keys(o.files))
+    if (t.endsWith(".bin")) {
+      const r = t.replace(".bin", ""), w = await o.file(t).async("arraybuffer"), g = new Float32Array(w), a = n.get(r) || { spec: [], data: new Float32Array() };
+      a.data = g, n.set(r, a);
+      const y = await O(a);
+      c.set(r, y);
+    }
+  k();
+  const e = new d(m.config);
+  await u(e), e.loadWeights(c);
+  const f = await o.file("log.json")?.async("string");
+  if (f)
+    try {
+      const t = JSON.parse(f);
+      e.log = t;
+    } catch (t) {
+      throw console.error("Error parsing training log:", t), new Error(`Failed to parse training log: ${t}`);
+    }
+  return { model: e, tokeniser: l };
+}
+export {
+  ft as default
+};

package/dist/{log_sum_exp-BHdkCb4s.js → log_sum_exp-YEo2h3gb.js}

@@ -1,8 +1,8 @@
-import { o as r, j as p, E as u, a6 as E, a7 as h, a1 as S, s as $, a8 as d } from "./index-bMBtI-WR.js";
-import { e as K } from "./axis_util-D17qZRQm.js";
-import { m as T } from "./max-DechV4Bc.js";
-import { r as m } from "./reshape-T4yDEqoF.js";
-import { s as _ } from "./sum-CutF5lj2.js";
+import { o as r, q as p, E as u, a8 as E, a9 as h, p as S, b as $, a7 as d } from "./index-BAzbokzv.js";
+import { e as b } from "./axis_util-Bu4h7XWV.js";
+import { m as K } from "./max-DtlIuVeW.js";
+import { r as m } from "./reshape-krWGKraP.js";
+import { s as T } from "./sum-BE_Irnim.js";
 /**
  * @license
  * Copyright 2018 Google LLC. All Rights Reserved.
@@ -19,11 +19,11 @@ import { s as _ } from "./sum-CutF5lj2.js";
  * limitations under the License.
  * =============================================================================
  */
-function b(s) {
+function _(s) {
   const o = { x: p(s, "x", "exp") };
   return u.runKernel(E, o);
 }
-const N = /* @__PURE__ */ r({ exp_: b });
+const N = /* @__PURE__ */ r({ exp_: _ });
 /**
  * @license
  * Copyright 2018 Google LLC. All Rights Reserved.
@@ -40,11 +40,11 @@ const N = /* @__PURE__ */ r({ exp_: b });
  * limitations under the License.
  * =============================================================================
  */
-function j(s) {
+function q(s) {
   const o = { x: p(s, "x", "log", "float32") };
   return u.runKernel(h, o);
 }
-const v = /* @__PURE__ */ r({ log_: j });
+const v = /* @__PURE__ */ r({ log_: q });
 /**
  * @license
  * Copyright 2020 Google LLC. All Rights Reserved.
@@ -61,15 +61,15 @@ const v = /* @__PURE__ */ r({ log_: j });
  * limitations under the License.
  * =============================================================================
  */
-function w(s, a = null, o = !1) {
-  const n = p(s, "x", "logSumExp"), t = S(a, n.shape), x = T(
-    n,
+function w(s, n = null, o = !1) {
+  const a = p(s, "x", "logSumExp"), t = S(n, a.shape), x = K(
+    a,
     t,
     !0
     /* keepDims */
-  ), i = $(n, x), l = N(i), f = _(l, t), c = v(f), e = d(m(x, c.shape), c);
+  ), i = $(a, x), l = N(i), f = T(l, t), c = v(f), e = d(m(x, c.shape), c);
   if (o) {
-    const g = K(e.shape, t);
+    const g = b(e.shape, t);
     return m(e, g);
   }
   return e;

package/dist/main.js

@@ -1,11 +1,11 @@
-import { default as F } from "./NanoGPTModel.js";
-import { default as N } from "./TeachableLLM.js";
-import { default as j } from "./tokeniser/CharTokeniser.js";
-import { default as z } from "./tokeniser/bpe.js";
-import { default as H } from "./utilities/waitForModel.js";
-import { default as J } from "./data/textLoader.js";
-import { estimateMemoryUsage as O, estimateParameterCount as Q, estimateResources as S, estimateTrainingMemoryUsage as V, validateConfig as W } from "./utilities/parameters.js";
-import "./index-bMBtI-WR.js";
+import { default as R } from "./NanoGPTModel.js";
+import { default as q } from "./TeachableLLM.js";
+import { default as A } from "./tokeniser/CharTokeniser.js";
+import { default as I } from "./tokeniser/bpe.js";
+import { default as K } from "./utilities/waitForModel.js";
+import { default as Q } from "./data/textLoader.js";
+import { estimateMemoryUsage as V, estimateParameterCount as W, estimateResources as X, estimateTrainingMemoryUsage as Y, validateConfig as Z } from "./utilities/parameters.js";
+import "./index-BAzbokzv.js";
 import "./ops/cpu/scatterSub.js";
 import "./ops/webgl/scatterSub.js";
 import "./ops/cpu/gatherSub.js";
@@ -16,7 +16,10 @@ import "./ops/grads/attentionMask.js";
 import "./ops/cpu/qkv.js";
 import "./ops/webgl/qkv.js";
 import "./ops/grads/qkv.js";
-import "@tensorflow/tfjs";
+import "./random_width-DWzaOgrn.js";
+import "./register_all_kernels-BfFCQAqs.js";
+import "./index-Tf7vU29b.js";
+import "./dataset-pgqp-YfL.js";
 import "./ops/cpu/rope.js";
 import "./ops/webgl/rope.js";
 import "./ops/grads/rope.js";
@@ -36,15 +39,15 @@ import "./ops/webgl/normRMS.js";
 import "./ops/grads/normRMS.js";
 import "./ops/webgl/log.js";
 export {
-  z as BPETokeniser,
-  j as CharTokeniser,
-  F as NanoGPT,
-  N as TeachableLLM,
-  O as estimateMemoryUsage,
-  Q as estimateParameterCount,
-  S as estimateResources,
-  V as estimateTrainingMemoryUsage,
-  J as loadTextData,
-  W as validateConfig,
-  H as waitForModel
+  I as BPETokeniser,
+  A as CharTokeniser,
+  R as NanoGPT,
+  q as TeachableLLM,
+  V as estimateMemoryUsage,
+  W as estimateParameterCount,
+  X as estimateResources,
+  Y as estimateTrainingMemoryUsage,
+  Q as loadTextData,
+  Z as validateConfig,
+  K as waitForModel
 };

package/dist/{mat_mul-BsrLfy81.js → mat_mul-7121rsJk.js}

@@ -1,4 +1,4 @@
-import { o as m, j as s, u as c, E as M, B as p } from "./index-bMBtI-WR.js";
+import { o as m, q as s, B as c, E as M, C as p } from "./index-BAzbokzv.js";
 /**
  * @license
  * Copyright 2020 Google LLC. All Rights Reserved.

package/dist/{max-DechV4Bc.js → max-DtlIuVeW.js}

@@ -1,4 +1,4 @@
-import { o as r, j as e, E as x, M as c } from "./index-bMBtI-WR.js";
+import { o as r, q as e, E as x, M as c } from "./index-BAzbokzv.js";
 /**
  * @license
  * Copyright 2020 Google LLC. All Rights Reserved.

package/dist/mulmat_packed_gpu-D4nKF7Je.js

@@ -0,0 +1,71 @@
+import { u as z } from "./gpgpu_math-CNslybmD.js";
+/**
+ * @license
+ * Copyright 2018 Google LLC. All Rights Reserved.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ * =============================================================================
+ */
+class g {
+  constructor(e, s, v, a = !1, r = !1, c = !1, t = null, o = !1, l = !1) {
+    this.variableNames = ["matrixA", "matrixB"], this.packedInputs = !0, this.packedOutput = !0, this.outputShape = v, this.enableShapeUniforms = z(this.outputShape.length);
+    const m = a ? e[1] : e[2], u = Math.ceil(m / 2), b = a ? "i * 2, rc.y" : "rc.y, i * 2", x = r ? "rc.z, i * 2" : "i * 2, rc.z", n = a ? ["a.xxyy", "a.zzww"] : ["a.xxzz", "a.yyww"], p = r ? ["b.xzxz", "b.ywyw"] : ["b.xyxy", "b.zwzw"];
+    let i = "", h = "";
+    t && (o ? i = `vec4 activation(vec4 a) {
+          vec4 b = getPreluActivationWeightsAtOutCoords();
+          ${t}
+        }` : l ? i = `vec4 activation(vec4 a) {
+          vec4 b = getLeakyreluAlphaAtOutCoords();
+          ${t}
+        }` : i = `vec4 activation(vec4 x) {
+          ${t}
+        }`, h = "result = activation(result);");
+    const $ = c ? "result += getBiasAtOutCoords();" : "";
+    c && this.variableNames.push("bias"), o && this.variableNames.push("preluActivationWeights"), l && this.variableNames.push("leakyreluAlpha");
+    let d = "rc.x", f = "rc.x";
+    e[0] < s[0] ? d = `imod(rc.x, ${e[0]})` : s[0] < e[0] && (f = `imod(rc.x, ${s[0]})`), this.userCode = `
+      ${i}
+      // Don't use uniform for sharedDimensionPacked for performance.
+      const float sharedDimension = ${u}.0;
+
+      vec4 dot2x2ARowBCol(ivec3 rc) {
+        vec4 result = vec4(0);
+        int batchA = ${d};
+        int batchB = ${f};
+        for (int i = 0; i < ${u}; i++) {
+          vec4 a = getMatrixA(batchA, ${b});
+          vec4 b = getMatrixB(batchB, ${x});
+
+          // These swizzled products need to be separately added.
+          // See: https://github.com/tensorflow/tfjs/issues/1735
+          result += (${n[0]} * ${p[0]});
+          result += (${n[1]} * ${p[1]});
+        }
+        return result;
+      }
+
+      void main() {
+        ivec3 rc = getOutputCoords();
+        vec4 result = dot2x2ARowBCol(rc);
+
+        ${$}
+
+        ${h}
+
+        setOutput(result);
+      }
+    `;
+  }
+}
+export {
+  g as M
+};

package/dist/{norm-B9hWHZH1.js → norm-CzltS9Fz.js}

@@ -1,8 +1,8 @@
-import { o as l, j as c, E as y, a0 as E, a1 as w, a2 as o, a3 as u, W as v, f as I, a4 as A } from "./index-bMBtI-WR.js";
-import { e as $ } from "./axis_util-D17qZRQm.js";
-import { m as f } from "./max-DechV4Bc.js";
-import { r as h } from "./reshape-T4yDEqoF.js";
-import { s as t } from "./sum-CutF5lj2.js";
+import { o as l, q as c, E as y, a2 as E, p as w, a3 as o, a4 as u, j as v, s as I, a5 as A } from "./index-BAzbokzv.js";
+import { e as $ } from "./axis_util-Bu4h7XWV.js";
+import { m } from "./max-DtlIuVeW.js";
+import { r as h } from "./reshape-krWGKraP.js";
+import { s as t } from "./sum-BE_Irnim.js";
 /**
  * @license
  * Copyright 2020 Google Inc. All Rights Reserved.
@@ -40,26 +40,26 @@ const s = /* @__PURE__ */ l({ min_: k });
  * limitations under the License.
  * =============================================================================
  */
-function T(n, e = "euclidean", r = null, m = !1) {
+function q(n, e = "euclidean", r = null, f = !1) {
   n = c(n, "x", "norm");
-  const a = d(n, e, r);
+  const a = p(n, e, r);
   let i = a.shape;
-  if (m) {
-    const p = w(r, n.shape);
-    i = $(a.shape, p);
+  if (f) {
+    const d = w(r, n.shape);
+    i = $(a.shape, d);
   }
   return h(a, i);
 }
-function d(n, e, r = null) {
+function p(n, e, r = null) {
   if (n.rank === 0)
     return o(n);
   if (n.rank !== 1 && r === null)
-    return d(h(n, [-1]), e, r);
+    return p(h(n, [-1]), e, r);
   if (n.rank === 1 || typeof r == "number" || Array.isArray(r) && r.length === 1) {
     if (e === 1)
       return t(o(n), r);
     if (e === 1 / 0)
-      return f(o(n), r);
+      return m(o(n), r);
     if (e === -1 / 0)
       return s(o(n), r);
     if (e === "euclidean" || e === 2)
@@ -68,9 +68,9 @@ function d(n, e, r = null) {
   }
   if (Array.isArray(r) && r.length === 2) {
     if (e === 1)
-      return f(t(o(n), r[0]), r[1] - 1);
+      return m(t(o(n), r[0]), r[1] - 1);
     if (e === 1 / 0)
-      return f(t(o(n), r[1]), r[0]);
+      return m(t(o(n), r[1]), r[0]);
     if (e === -1 / 0)
       return s(t(o(n), r[1]), r[0]);
     if (e === "fro" || e === "euclidean")
@@ -79,7 +79,7 @@ function d(n, e, r = null) {
   }
   throw new Error(`Error in norm: invalid axis: ${r}`);
 }
-const N = /* @__PURE__ */ l({ norm_: T });
+const N = /* @__PURE__ */ l({ norm_: q });
 export {
   s as m,
   N as n

package/dist/{ones-g0K8jVwm.js → ones-BBlSRqn1.js}

@@ -1,5 +1,5 @@
-import { n, p as t, q as m, E as i } from "./index-bMBtI-WR.js";
-import { z as c, c as f } from "./zeros-CAbHfODe.js";
+import { x as n, y as t, g as m, E as i } from "./index-BAzbokzv.js";
+import { z as c, c as f } from "./zeros-YJDE7oRb.js";
 /**
  * @license
  * Copyright 2018 Google LLC. All Rights Reserved.

package/dist/ops/appendCache.js

@@ -1,8 +1,8 @@
-import { e as a } from "../index-bMBtI-WR.js";
+import { e as a } from "../index-BAzbokzv.js";
 import "./cpu/appendCache.js";
 import "./webgl/appendCache.js";
-import { c as s } from "../concat-DhZfF1GY.js";
-import { z as c } from "../zeros-CAbHfODe.js";
+import { c as s } from "../concat-5aPGqw3Z.js";
+import { z as c } from "../zeros-YJDE7oRb.js";
 function i(r, p, n, o) {
   if (!o) {
     const e = r.shape[2];

package/dist/ops/attentionMask.js

@@ -1,4 +1,4 @@
-import { e as o } from "../index-bMBtI-WR.js";
+import { e as o } from "../index-BAzbokzv.js";
 import "./cpu/attentionMask.js";
 import "./webgl/attentionMask.js";
 import "./grads/attentionMask.js";

package/dist/ops/cpu/appendCache.js

@@ -1,5 +1,5 @@
-import { r as d } from "../../index-bMBtI-WR.js";
-import { c as h } from "../../concat-DhZfF1GY.js";
+import { r as d } from "../../index-BAzbokzv.js";
+import { c as h } from "../../concat-5aPGqw3Z.js";
 function u(p) {
   const { cache: n, item: s } = p.inputs, { maxSize: r, pastLen: c } = p.attrs, t = n.shape[0], o = n.shape[1], a = n.shape[3], e = s.shape[2];
   if (c + e <= r) {

package/dist/ops/cpu/attentionMask.js

@@ -1,10 +1,10 @@
-import { r as a, g as p, f as u } from "../../index-bMBtI-WR.js";
-import { l as N, w as b } from "../../ops-Mv7Ta72x.js";
-import { o as g } from "../../ones-g0K8jVwm.js";
-import { z as A } from "../../zeros-CAbHfODe.js";
-import { m as I } from "../../mat_mul-BsrLfy81.js";
+import { r as a, f as p, s as u } from "../../index-BAzbokzv.js";
+import { l as N, w as b } from "../../ops-C0sQEcPw.js";
+import { o as A } from "../../ones-BBlSRqn1.js";
+import { z as I } from "../../zeros-YJDE7oRb.js";
+import { m as g } from "../../mat_mul-7121rsJk.js";
 function o(n) {
-  const { q: s, k: e } = n.inputs, { divisor: r } = n.attrs, c = s.shape[2], t = e.shape[2], m = N.bandPart(g([t, t]), -1, 0).cast("bool"), l = A([t, t]), i = p([t, t], Number.NEGATIVE_INFINITY), f = b(m, l, i), k = I(s, e, !1, !0).mul(u(r)), d = f.slice([0, 0], [c, t]).expandDims(0).expandDims(0);
+  const { q: s, k: e } = n.inputs, { divisor: r } = n.attrs, c = s.shape[2], t = e.shape[2], m = N.bandPart(A([t, t]), -1, 0).cast("bool"), l = I([t, t]), i = p([t, t], Number.NEGATIVE_INFINITY), f = b(m, l, i), k = g(s, e, !1, !0).mul(u(r)), d = f.slice([0, 0], [c, t]).expandDims(0).expandDims(0);
   return k.add(d);
 }
 const w = {

package/dist/ops/cpu/fusedSoftmax.js

@@ -1,5 +1,5 @@
-import { r as n } from "../../index-bMBtI-WR.js";
-import { s as f } from "../../softmax-B2_IKPDR.js";
+import { r as n } from "../../index-BAzbokzv.js";
+import { s as f } from "../../softmax-fsdtf6JC.js";
 function r(t) {
   const { inputs: s, attrs: i } = t, { logits: o } = s, { dim: a, dropoutRate: e } = i;
   if (!o)

package/dist/ops/cpu/gatherSub.js

@@ -1,6 +1,6 @@
-import { o as u, j as c, E as g, O as h, r as m, s as p } from "../../index-bMBtI-WR.js";
-import { r as l } from "../../range-DMaG9A3G.js";
-import { s as N } from "../../stack-lpJ5kYvE.js";
+import { o as u, q as c, E as g, Y as h, r as m, b as p } from "../../index-BAzbokzv.js";
+import { r as l } from "../../range-DYsrnfiy.js";
+import { s as N } from "../../stack-dfEEz2OY.js";
 /**
  * @license
  * Copyright 2018 Google LLC. All Rights Reserved.
@@ -17,14 +17,14 @@ import { s as N } from "../../stack-lpJ5kYvE.js";
  * limitations under the License.
  * =============================================================================
  */
-function f(e, s) {
-  const n = c(s, "indices", "gatherND", "int32"), t = { params: c(e, "x", "gatherND", "string_or_numeric"), indices: n };
-  return g.runKernel(h, t);
+function b(e, t) {
+  const n = c(t, "indices", "gatherND", "int32"), s = { params: c(e, "x", "gatherND", "string_or_numeric"), indices: n };
+  return g.runKernel(h, s);
 }
-const b = /* @__PURE__ */ u({ gatherND_: f });
+const f = /* @__PURE__ */ u({ gatherND_: b });
 function d(e) {
-  const { values: s, labels: n, logits: r } = e.inputs, t = n.shape[0], o = l(0, t, 1, "int32"), a = N([o, n], 1), i = b(r, a);
-  return p(s, i);
+  const { values: t, labels: n, logits: r } = e.inputs, s = n.shape[0], o = l(0, s, 1, "int32"), a = N([o, n], 1), i = f(r, a);
+  return p(t, i);
 }
 const k = {
   kernelName: "EfficientGatherSub",

package/dist/ops/cpu/gelu.js

@@ -1,4 +1,4 @@
-import { r as t, t as d } from "../../index-bMBtI-WR.js";
+import { r as t, t as d } from "../../index-BAzbokzv.js";
 const o = 0.7978845608028654, c = 0.044715;
 function m(u) {
   const { inputs: l } = u, { x: e } = l, n = e;

package/dist/ops/cpu/matMulGelu.js

@@ -1,4 +1,4 @@
-import { r as a, t as i } from "../../index-bMBtI-WR.js";
+import { r as a, t as i } from "../../index-BAzbokzv.js";
 const c = 0.7978845608028654, m = 0.044715;
 function M(o) {
   const { inputs: s } = o, { x: t, kernel: l } = s, e = t, u = l;