@genai-fi/nanogpt 0.7.3 → 0.8.1

package/dist/models/NanoGPTV1.js

@@ -0,0 +1,71 @@
+import { defaultConfig as m } from "./config.js";
+import f from "../layers/TransformerBlock.js";
+import u from "../layers/TiedEmbedding.js";
+import g from "../layers/RoPECache.js";
+import b from "../layers/RMSNorm.js";
+import { t as l, k as p } from "../index-DdmHGZjq.js";
+import w from "./model.js";
+import k from "../layers/PositionEmbedding.js";
+class R extends w {
+  wte;
+  // Token embeddings
+  wpe;
+  // Position embeddings
+  // private drop: layers.Layer; // Dropout
+  blocks;
+  lnF;
+  // Final layer norm
+  ropeCache;
+  constructor(e = {}) {
+    super({ ...m, ...e }), this.wte = new u(this.config, "token_embedding", this), this.config.useRope === !1 ? this.wpe = new k(this.config, "positional_embedding", this) : this.ropeCache = new g(this.config), this.blocks = [];
+    for (let i = 0; i < this.config.nLayer; i++)
+      this.blocks.push(new f(i, this.config, this));
+    this.lnF = new b(this.config, "final_rms_norm", this);
+  }
+  getClassName() {
+    return "GenAI_NanoGPT_v1";
+  }
+  inputPhase(e, i) {
+    return l(() => {
+      const s = this.wte.embed(e);
+      if (this.config.useRope === !1) {
+        const o = this.wpe.call(i, s);
+        if (Array.isArray(o))
+          throw new Error("PositionEmbedding output should not be an array");
+        return o;
+      }
+      return s;
+    });
+  }
+  forward(e, i, s) {
+    return this.validateInput(i), e.ropeCache = this.ropeCache, e.outputEmbeddings && (e.embeddings = []), l(() => {
+      this.startMemory();
+      let o = this.inputPhase(i, e);
+      if (e.cache && e.cache.length !== this.blocks.length)
+        throw console.error("Cache", e.cache), new Error(
+          `Cache length ${e.cache.length} does not match number of blocks ${this.blocks.length}`
+        );
+      for (let t = 0; t < this.blocks.length; t++) {
+        const c = this.blocks[t], d = Math.random() * 1e9, r = {
+          ropeCache: e.ropeCache,
+          training: e.training,
+          seed: d,
+          attentionScores: e.attentionScores,
+          pastKV: e.cache ? e.cache[t] : void 0
+        }, a = e.checkpointing && e.training ? c.callCheckpoint(r, o) : c.call(r, o);
+        e.outputEmbeddings ? (p(o), e.embeddings.push(o)) : o.dispose(), o = a;
+      }
+      o = this.lnF.call(e, o);
+      const n = this.wte.project(o);
+      e.outputEmbeddings ? (p(o), e.embeddings.push(o)) : o.dispose();
+      let h;
+      return s && (h = this.calculateLoss(n, s)), this.endMemory("Forward"), h ? [n, h] : [n];
+    });
+  }
+  dispose() {
+    this.wte.dispose(), this.wpe && this.wpe.dispose(), this.blocks.forEach((e) => e.dispose()), this.lnF.dispose();
+  }
+}
+export {
+  R as default
+};

package/dist/{config.d.ts → models/config.d.ts}

@@ -1,4 +1,5 @@
 export interface GPTConfig {
+    modelType?: string;
     vocabSize: number;
     blockSize: number;
     nLayer: number;

package/dist/{config.js → models/config.js}

@@ -1,4 +1,5 @@
 const e = {
+  modelType: "GenAI_NanoGPT_v1",
   vocabSize: 2e3,
   blockSize: 128,
   // Maximum sequence length

package/dist/models/factory.d.ts

@@ -0,0 +1,3 @@
+import { GPTConfig } from './config';
+import { default as Model, ModelForwardAttributes } from './model';
+export default function createModelInstance(config: GPTConfig): Model<ModelForwardAttributes>;

package/dist/models/factory.js

@@ -0,0 +1,14 @@
+import o from "./NanoGPTV1.js";
+function a(e) {
+  const n = e.modelType || "GenAI_NanoGPT_v1";
+  switch (n) {
+    case "GenAI_NanoGPT_1":
+    case "GenAI_NanoGPT_v1":
+      return new o(e);
+    default:
+      throw new Error(`Unsupported model type: ${n}`);
+  }
+}
+export {
+  a as default
+};

package/dist/models/model.d.ts

@@ -0,0 +1,26 @@
+import { Tensor } from '@tensorflow/tfjs-core';
+import { ForwardAttributes, default as BaseLayer } from '../layers/BaseLayer';
+import { AttentionScores, KVCache } from '../layers/CausalSelfAttention';
+export interface ModelForwardAttributes extends ForwardAttributes {
+    cache?: KVCache[];
+    attentionScores?: AttentionScores;
+    seed?: number;
+    outputEmbeddings?: boolean;
+    embeddings?: Tensor[];
+}
+interface TrainingState {
+    steps: number;
+    learningRate: number;
+    batchSize: number;
+    loss: number;
+}
+export default abstract class Model<T extends ModelForwardAttributes> extends BaseLayer<T> {
+    trainingState: TrainingState | null;
+    abstract getClassName(): string;
+    abstract forward(attrs: T, idx: Tensor, targets?: Tensor): Tensor[];
+    abstract dispose(): void;
+    getNumParams(): number;
+    protected validateInput(idx: Tensor): void;
+    protected calculateLoss(logits: Tensor, targets: Tensor): Tensor;
+}
+export {};

package/dist/models/model.js

@@ -0,0 +1,70 @@
+import i from "../layers/BaseLayer.js";
+import "../index-DdmHGZjq.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-DKGeiFuR.js";
+import "../register_all_kernels-Do9VvZmo.js";
+import "../index-Tf7vU29b.js";
+import "../dataset-DPPl-iLT.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/normRMS.js";
+import "../ops/webgl/normRMS.js";
+import "../ops/grads/normRMS.js";
+import "../jszip.min-CjP2V1VV.js";
+import "../index-Dwqa6Zy2.js";
+import "../Generator.js";
+import "../ops/cpu/adamAdjust.js";
+import "../ops/webgl/adamAdjust.js";
+import "../ops/cpu/adamMoments.js";
+import "../ops/webgl/adamMoments.js";
+import "../papaparse.min-C8l2Kvo1.js";
+import { estimateParameterCount as p } from "../utilities/parameters.js";
+import "../ops/cpu/scatterSub.js";
+import "../ops/webgl/scatterSub.js";
+import "../ops/cpu/gatherSub.js";
+import "../ops/webgl/gatherSub.js";
+import "../ops/cpu/gelu.js";
+import "../ops/webgl/gelu.js";
+import "../gelu-BkcmEEyD.js";
+import "../ops/webgl/log.js";
+import "../checks/normRMS.js";
+import "../checks/normRMSGrad.js";
+import { createSoftmaxCrossEntropyWithGrad as m } from "../training/sparseCrossEntropy.js";
+class x extends i {
+  trainingState = null;
+  getNumParams() {
+    return p(this.config);
+  }
+  validateInput(t) {
+    if (t.shape.length !== 2)
+      throw new Error(`Invalid input shape: expected [batch_size, sequence_length], got ${t.shape}`);
+    if (t.shape[1] > this.config.blockSize)
+      throw new Error(`Input sequence length ${t.shape[1]} isn't block size ${this.config.blockSize}`);
+    if (t.dtype !== "int32")
+      throw new Error(`Input tensor must be of type int32, got ${t.dtype}`);
+  }
+  calculateLoss(t, o) {
+    try {
+      return m()(t, o).mean();
+    } catch (r) {
+      throw console.error("Error computing loss:", r), new Error(`Loss computation failed: ${r}`);
+    }
+  }
+}
+export {
+  x as default
+};

package/dist/{mulmat_packed_gpu-VSekgsNv.js → mulmat_packed_gpu-q_Gmwyld.js}

@@ -1,4 +1,4 @@
-import { u as z } from "./gpgpu_math-DGNLNL4I.js";
+import { u as z } from "./gpgpu_math-D_ODOLix.js";
 /**
  * @license
  * Copyright 2018 Google LLC. All Rights Reserved.

package/dist/{ones-Dj0SDhHf.js → ones-BAqVh-eA.js}

@@ -1,5 +1,5 @@
-import { F as n, H as t, j as m, E as i } from "./index-BoWRt-10.js";
-import { z as c, c as f } from "./zeros--BdLQ3oG.js";
+import { M as n, N as t, j as m, E as i } from "./index-DdmHGZjq.js";
+import { z as c, c as f } from "./zeros-Dnwix0p4.js";
 /**
  * @license
  * Copyright 2018 Google LLC. All Rights Reserved.

package/dist/ops/adamAdjust.js

@@ -1,4 +1,4 @@
-import { e as i } from "../index-BoWRt-10.js";
+import { e as i } from "../index-DdmHGZjq.js";
 import "./cpu/adamAdjust.js";
 import "./webgl/adamAdjust.js";
 function p(r, t, e, n, m, o) {

package/dist/ops/adamMoments.js

@@ -1,4 +1,4 @@
-import { e as o } from "../index-BoWRt-10.js";
+import { e as o } from "../index-DdmHGZjq.js";
 import "./cpu/adamMoments.js";
 import "./webgl/adamMoments.js";
 function p(e, n, r, m) {

package/dist/ops/appendCache.js

@@ -1,8 +1,8 @@
-import { e as a } from "../index-BoWRt-10.js";
+import { e as a } from "../index-DdmHGZjq.js";
 import "./cpu/appendCache.js";
 import "./webgl/appendCache.js";
-import { c as s } from "../concat-CsxrgovM.js";
-import { z as c } from "../zeros--BdLQ3oG.js";
+import { c as s } from "../concat-pHiVqR3L.js";
+import { z as c } from "../zeros-Dnwix0p4.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-BoWRt-10.js";
+import { e as o } from "../index-DdmHGZjq.js";
 import "./cpu/attentionMask.js";
 import "./webgl/attentionMask.js";
 import "./grads/attentionMask.js";

package/dist/ops/cpu/adamAdjust.js

@@ -1,18 +1,18 @@
-import { f as k, n as t, o as i, a as q, q as z } from "../../index-BoWRt-10.js";
-function A(a) {
-  const { moments: n, value: r } = a.inputs, { beta1: l, beta2: u, epsilon: m, learningRate: d } = a.attrs, e = n.shape.length, c = new Array(e).fill(0), s = n.shape.slice();
-  s[e - 1] = 1;
+import { f as k, o as t, q as i, a as q, w } from "../../index-DdmHGZjq.js";
+function z(a) {
+  const { moments: s, value: r } = a.inputs, { beta1: l, beta2: u, epsilon: m, learningRate: d } = a.attrs, e = s.shape.length, c = new Array(e).fill(0), n = s.shape.slice();
+  n[e - 1] = 1;
   const o = c.slice();
   o[e - 1] = 1;
-  const b = s.slice(), p = n.slice(c, s).squeeze([e - 1]), M = n.slice(o, b).squeeze([e - 1]), f = t(p, l), g = t(M, u);
+  const b = n.slice(), p = s.slice(c, n).squeeze([e - 1]), M = s.slice(o, b).squeeze([e - 1]), f = t(p, l), g = t(M, u);
   return i(
-    q(t(f, i(z(g), m ?? 1e-8)), -d),
+    q(t(f, i(w(g), m ?? 1e-8)), -d),
     r
   );
 }
-const C = {
+const A = {
   kernelName: "AdamAdjust",
   backendName: "cpu",
-  kernelFunc: A
+  kernelFunc: z
 };
-k(C);
+k(A);

package/dist/ops/cpu/adamMoments.js

@@ -1,5 +1,5 @@
-import { f as p } from "../../index-BoWRt-10.js";
-import { s as f } from "../../stack-B17UN2nn.js";
+import { f as p } from "../../index-DdmHGZjq.js";
+import { s as f } from "../../stack-BNwLzE43.js";
 function b(t) {
   const { moments: n, gradient: c } = t.inputs, { beta1: o, beta2: m } = t.attrs, e = n.shape.length, a = new Array(e).fill(0), s = n.shape.slice();
   s[e - 1] = 1;

package/dist/ops/cpu/appendCache.js

@@ -1,5 +1,5 @@
-import { f as d } from "../../index-BoWRt-10.js";
-import { c as h } from "../../concat-CsxrgovM.js";
+import { f as d } from "../../index-DdmHGZjq.js";
+import { c as h } from "../../concat-pHiVqR3L.js";
 function u(p) {
   const { cache: n, item: s } = p.inputs, { maxSize: i, pastLen: c } = p.attrs, t = n.shape[0], o = n.shape[1], a = n.shape[3], e = s.shape[2];
   if (c + e <= i) {

package/dist/ops/cpu/attentionMask.js

@@ -1,8 +1,8 @@
-import { f as a, h as p, b as u } from "../../index-BoWRt-10.js";
-import { l as N, w as b } from "../../ops-BFGCx8Ri.js";
-import { o as A } from "../../ones-Dj0SDhHf.js";
-import { z as I } from "../../zeros--BdLQ3oG.js";
-import { m as g } from "../../mat_mul-8m8pfdcx.js";
+import { f as a, h as p, b as u } from "../../index-DdmHGZjq.js";
+import { l as N, w as b } from "../../ops-542ai2vG.js";
+import { o as A } from "../../ones-BAqVh-eA.js";
+import { z as I } from "../../zeros-Dnwix0p4.js";
+import { m as g } from "../../mat_mul-Dpy2mMRu.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(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);

package/dist/ops/cpu/fusedSoftmax.js

@@ -1,5 +1,5 @@
-import { f as e } from "../../index-BoWRt-10.js";
-import { s as f } from "../../softmax-Bv_6lyMX.js";
+import { f as e } from "../../index-DdmHGZjq.js";
+import { s as f } from "../../softmax-BfsyI4As.js";
 function n(t) {
   const { inputs: s, attrs: a } = t, { logits: o } = s, { dim: i, dropoutRate: r } = a;
   if (!o)

package/dist/ops/cpu/gatherSub.js

@@ -1,6 +1,6 @@
-import { B as u, C as c, E as g, a5 as h, f as m, c as p } from "../../index-BoWRt-10.js";
-import { r as f } from "../../range-CRuAh-gd.js";
-import { s as l } from "../../stack-B17UN2nn.js";
+import { C as u, D as c, E as g, a8 as h, f as m, c as p } from "../../index-DdmHGZjq.js";
+import { r as f } from "../../range-BcUvLuf5.js";
+import { s as l } from "../../stack-BNwLzE43.js";
 /**
  * @license
  * Copyright 2018 Google LLC. All Rights Reserved.
@@ -26,9 +26,9 @@ function d(e) {
   const { values: t, labels: n, logits: r } = e.inputs, s = n.shape[0], a = f(0, s, 1, "int32"), i = l([a, n], 1), o = b(r, i);
   return p(t, o);
 }
-const k = {
+const D = {
   kernelName: "EfficientGatherSub",
   backendName: "cpu",
   kernelFunc: d
 };
-m(k);
+m(D);

package/dist/ops/cpu/gelu.js

@@ -1,4 +1,4 @@
-import { f as t, t as d } from "../../index-BoWRt-10.js";
+import { f as t, t as d } from "../../index-DdmHGZjq.js";
 const o = 0.7978845608028654, c = 0.044715;
 function m(r) {
   const { inputs: u } = r, { x: n } = u, e = n;

package/dist/ops/cpu/matMulGelu.js

@@ -1,5 +1,5 @@
-import { f as e, t as m } from "../../index-BoWRt-10.js";
-import { g as d, d as M } from "../../gelu-C-dPj6Ku.js";
+import { f as e, t as m } from "../../index-DdmHGZjq.js";
+import { g as d, d as M } from "../../gelu-BkcmEEyD.js";
 function c(t) {
   const { inputs: u } = t, { x: n, kernel: l } = u, a = n, r = l;
   return m(() => {

package/dist/ops/cpu/matMulMul.js

@@ -1,4 +1,4 @@
-import { f as e, t as i } from "../../index-BoWRt-10.js";
+import { f as e, t as i } from "../../index-DdmHGZjq.js";
 function n(t) {
   const { inputs: r, attrs: l } = t, { transposeA: o, transposeB: s } = l, { x: u, kernel: a, y: c } = r, m = u, k = a, M = c;
   return i(() => m.matMul(k, o, s).mul(M));

package/dist/ops/cpu/mulDropout.js

@@ -1,4 +1,4 @@
-import { f as e, a as t } from "../../index-BoWRt-10.js";
+import { f as e, a as t } from "../../index-DdmHGZjq.js";
 function n(o) {
   const { inputs: r } = o, { a: l, b: u } = r;
   return console.warn("Using fallback mulDrop implementation without dropout."), t(l, u);

package/dist/ops/cpu/normRMS.js

@@ -1,4 +1,4 @@
-import { f as o, t as d } from "../../index-BoWRt-10.js";
+import { f as o, t as d } from "../../index-DdmHGZjq.js";
 function i(t) {
   const { inputs: e } = t, { x: n, gamma: s } = e, r = n, a = s;
   return d(() => {

package/dist/ops/cpu/qkv.js

@@ -1,6 +1,6 @@
-import { f as q } from "../../index-BoWRt-10.js";
-import { r as o } from "../../reshape-CdBq1WJ6.js";
-import { s as x } from "../../split-B-dikLRw.js";
+import { f as q } from "../../index-DdmHGZjq.js";
+import { r as o } from "../../reshape-WeJkT3ja.js";
+import { s as x } from "../../split-BPxr8_8m.js";
 function v(p) {
   const { x: c, kernel: K } = p.inputs, { heads: n } = p.attrs, [s, e, t] = c.shape, a = o(c, [s * e, t]), i = a.dot(K);
   a.dispose();

package/dist/ops/cpu/rope.js

@@ -1,8 +1,8 @@
-import { f as S } from "../../index-BoWRt-10.js";
-import { r as F } from "../../range-CRuAh-gd.js";
-import { g as I } from "../../gather-CMMy2KEG.js";
-import { s as E } from "../../stack-B17UN2nn.js";
-import { c as T } from "../../concat-CsxrgovM.js";
+import { f as S } from "../../index-DdmHGZjq.js";
+import { r as F } from "../../range-BcUvLuf5.js";
+import { g as I } from "../../gather-CPg6ZlQA.js";
+import { s as E } from "../../stack-BNwLzE43.js";
+import { c as T } from "../../concat-pHiVqR3L.js";
 function U(t, c, p, o, r) {
   const n = o.shape[3], s = p;
   if (s > n) return o;

package/dist/ops/cpu/scatterSub.js

@@ -1,8 +1,8 @@
-import { B as f, F as g, C as r, E as l, a4 as N, f as b, c as S, a as h } from "../../index-BoWRt-10.js";
-import { v as D } from "../../scatter_nd_util-DUstGbU1.js";
-import { r as k } from "../../range-CRuAh-gd.js";
-import { s as v } from "../../stack-B17UN2nn.js";
-import { o as E } from "../../ones-Dj0SDhHf.js";
+import { C as f, M as g, D as r, E as l, a7 as N, f as b, c as S, a as D } from "../../index-DdmHGZjq.js";
+import { v as h } from "../../scatter_nd_util-B7yDhiQr.js";
+import { r as k } from "../../range-BcUvLuf5.js";
+import { s as v } from "../../stack-BNwLzE43.js";
+import { o as E } from "../../ones-BAqVh-eA.js";
 /**
  * @license
  * Copyright 2018 Google LLC. All Rights Reserved.
@@ -22,14 +22,14 @@ import { o as E } from "../../ones-Dj0SDhHf.js";
 function I(a, e, s) {
   g(s);
   const n = r(a, "indices", "scatterND", "int32"), t = r(e, "updates", "scatterND");
-  D(t, n, s);
+  h(t, n, s);
   const c = { indices: n, updates: t }, o = { shape: s };
   return l.runKernel(N, c, o);
 }
 const C = /* @__PURE__ */ f({ scatterND_: I });
 function K(a) {
   const { logits: e, labels: s, dy: n } = a.inputs, t = s.shape[0], c = e.shape[1], o = k(0, t, 1, "int32"), i = v([o, s], 1), d = E([t]), u = C(i, d, [t, c]), p = S(e, u), m = n.reshape([t, 1]);
-  return h(p, m);
+  return D(p, m);
 }
 const L = {
   kernelName: "EfficientScatterSub",

package/dist/ops/fusedSoftmax.js

@@ -1,4 +1,4 @@
-import { e as t } from "../index-BoWRt-10.js";
+import { e as t } from "../index-DdmHGZjq.js";
 import "./cpu/fusedSoftmax.js";
 import "./webgl/fusedSoftmax.js";
 import "./grads/fusedSoftmax.js";

package/dist/ops/gatherSub.js

@@ -1,4 +1,4 @@
-import { e as n } from "../index-BoWRt-10.js";
+import { e as n } from "../index-DdmHGZjq.js";
 import "./cpu/gatherSub.js";
 import "./webgl/gatherSub.js";
 function f(r, e, t) {

package/dist/ops/gelu.js

@@ -1,7 +1,7 @@
-import "../index-BoWRt-10.js";
+import "../index-DdmHGZjq.js";
 import "./cpu/gelu.js";
 import "./webgl/gelu.js";
-import { d as e, g as i } from "../gelu-C-dPj6Ku.js";
+import { d as e, g as i } from "../gelu-BkcmEEyD.js";
 export {
   e as dGelu,
   i as gelu

package/dist/ops/grads/attentionMask.js

@@ -1,4 +1,4 @@
-import { i as m, b as i } from "../../index-BoWRt-10.js";
+import { i as m, b as i } from "../../index-DdmHGZjq.js";
 import { matMulMul as u } from "../matMulMul.js";
 const p = {
   kernelName: "AttentionMask",

package/dist/ops/grads/fusedSoftmax.js

@@ -1,6 +1,6 @@
-import { i as f, a as i, c as l } from "../../index-BoWRt-10.js";
+import { i as f, a as i, c as l } from "../../index-DdmHGZjq.js";
 import { mulDrop as g } from "../mulDrop.js";
-import { s as T } from "../../sum-66ew2byf.js";
+import { s as T } from "../../sum-ByFINZgi.js";
 const Y = {
   kernelName: "FusedSoftmax",
   outputsToSave: [!0],

package/dist/ops/grads/gelu.js

@@ -1,5 +1,5 @@
-import "../../index-BoWRt-10.js";
-import { a as e } from "../../gelu-C-dPj6Ku.js";
+import "../../index-DdmHGZjq.js";
+import { a as e } from "../../gelu-BkcmEEyD.js";
 export {
   e as geluGradConfig
 };

package/dist/ops/grads/matMulGelu.js

@@ -1,4 +1,4 @@
-import { i as a, e as o } from "../../index-BoWRt-10.js";
+import { i as a, e as o } from "../../index-DdmHGZjq.js";
 function i(e, n, r) {
   return o().runKernel("MatMulGeluGrad", { dy: e, x: n, kernel: r });
 }

package/dist/ops/grads/normRMS.js

@@ -1,4 +1,4 @@
-import { i as t, e as g } from "../../index-BoWRt-10.js";
+import { i as t, e as g } from "../../index-DdmHGZjq.js";
 function i(r, a, n) {
   return g().runKernel("RMSNormGrad", { dy: r, x: a, gamma: n });
 }

package/dist/ops/grads/qkv.js

@@ -1,4 +1,4 @@
-import { i as Q } from "../../index-BoWRt-10.js";
+import { i as Q } from "../../index-DdmHGZjq.js";
 const V = {
   kernelName: "QKV",
   inputsToSave: ["x", "kernel"],

package/dist/ops/grads/rope.js

@@ -1,4 +1,4 @@
-import { i, e as a } from "../../index-BoWRt-10.js";
+import { i, e as a } from "../../index-DdmHGZjq.js";
 function p(n, e, s, o) {
   return a().runKernel("Rope", { x: n, sin: e, cos: s }, { pastLen: o });
 }

package/dist/ops/matMulGelu.js

@@ -1,4 +1,4 @@
-import { e as u } from "../index-BoWRt-10.js";
+import { e as u } from "../index-DdmHGZjq.js";
 import "./cpu/matMulGelu.js";
 import "./webgl/matMulGelu.js";
 import "./grads/matMulGelu.js";

package/dist/ops/matMulMul.js

@@ -1,4 +1,4 @@
-import { e as u } from "../index-BoWRt-10.js";
+import { e as u } from "../index-DdmHGZjq.js";
 import "./cpu/matMulMul.js";
 import "./webgl/matMulMul.js";
 function m(e, r, t, l = !1, n = !1) {

package/dist/ops/mulDrop.js

@@ -1,4 +1,4 @@
-import { e as t } from "../index-BoWRt-10.js";
+import { e as t } from "../index-DdmHGZjq.js";
 import "./cpu/mulDropout.js";
 import "./webgl/mulDropout.js";
 function m(r, o, e, n) {

package/dist/ops/normRMS.js

@@ -1,4 +1,4 @@
-import { e as n } from "../index-BoWRt-10.js";
+import { e as n } from "../index-DdmHGZjq.js";
 import "./cpu/normRMS.js";
 import "./webgl/normRMS.js";
 import "./grads/normRMS.js";

package/dist/ops/qkv.js

@@ -1,4 +1,4 @@
-import { e as o } from "../index-BoWRt-10.js";
+import { e as o } from "../index-DdmHGZjq.js";
 import "./cpu/qkv.js";
 import "./webgl/qkv.js";
 import "./grads/qkv.js";

package/dist/ops/rope.js

@@ -1,8 +1,8 @@
-import { e as p } from "../index-BoWRt-10.js";
-import "../random_width-sZORGo5k.js";
-import "../register_all_kernels-BwDSRN-f.js";
+import { e as p } from "../index-DdmHGZjq.js";
+import "../random_width-DKGeiFuR.js";
+import "../register_all_kernels-Do9VvZmo.js";
 import "../index-Tf7vU29b.js";
-import "../dataset-CtdBYwjo.js";
+import "../dataset-DPPl-iLT.js";
 import "./cpu/rope.js";
 import "./webgl/rope.js";
 import "./grads/rope.js";

package/dist/ops/scatterSub.js

@@ -1,4 +1,4 @@
-import { e as i } from "../index-BoWRt-10.js";
+import { e as i } from "../index-DdmHGZjq.js";
 import "./cpu/scatterSub.js";
 import "./webgl/scatterSub.js";
 function c(t, r, e) {

package/dist/ops/webgl/adamAdjust.js

@@ -1,5 +1,5 @@
-import { r as n } from "../../Reshape-DH5srBP0.js";
-import { f } from "../../index-BoWRt-10.js";
+import { r as n } from "../../Reshape-Bh_jzKzV.js";
+import { f } from "../../index-DdmHGZjq.js";
 class v {
   variableNames = ["moments", "value"];
   outputShape;

package/dist/ops/webgl/adamMoments.js

@@ -1,4 +1,4 @@
-import { f as m } from "../../index-BoWRt-10.js";
+import { f as m } from "../../index-DdmHGZjq.js";
 class i {
   variableNames = ["moments", "gradient"];
   outputShape;

package/dist/ops/webgl/appendCache.js

@@ -1,4 +1,4 @@
-import { f as p } from "../../index-BoWRt-10.js";
+import { f as p } from "../../index-DdmHGZjq.js";
 class m {
   variableNames = ["cache", "item"];
   outputShape;

package/dist/ops/webgl/attentionMask.js

@@ -1,4 +1,4 @@
-import { f as m } from "../../index-BoWRt-10.js";
+import { f as m } from "../../index-DdmHGZjq.js";
 class h {
   variableNames = ["q", "k"];
   outputShape;

package/dist/ops/webgl/fusedSoftmax.js

@@ -1,7 +1,7 @@
-import { m as b, s as I, r as k } from "../../RealDiv-Dy0p8Bvo.js";
-import { r as v } from "../../Reshape-DH5srBP0.js";
-import { f as w, p as P } from "../../index-BoWRt-10.js";
-import { e as S } from "../../axis_util-BzbKo31C.js";
+import { m as b, s as I, r as k } from "../../RealDiv-D_q39E3A.js";
+import { r as v } from "../../Reshape-Bh_jzKzV.js";
+import { f as w, p as P } from "../../index-DdmHGZjq.js";
+import { e as S } from "../../axis_util-Did9235A.js";
 class T {
   variableNames = ["logits", "maxLogits"];
   outputShape;