@genai-fi/nanogpt 0.20.6 → 0.21.0

package/dist/main.js

@@ -1,4 +1,4 @@
-import { a as e, b as t, i as n, n as r, r as i, s as a, t as o, y as s } from "./main-C0YUnw98.js";
+import { a as e, b as t, i as n, n as r, r as i, s as a, t as o, y as s } from "./main-DdsoTJAW.js";
 import c from "./tokeniser/CharTokeniser.js";
 import l from "./tokeniser/bpe.js";
 import { AdamWOptimizer as u } from "./training/AdamW.js";

package/dist/{matMul16-BNfZSnNM.js → matMul16-DGtlziMn.js}

@@ -1,12 +1,12 @@
-import { Fi as e, ii as t } from "./dist-BewPQWjc.js";
-import { t as n } from "./gelu-Bf1HW1RY.js";
+import { Fi as e, ii as t } from "./dist-Da20xy8E.js";
+import { t as n } from "./gelu-B6637MKa.js";
 import { isPackedTensor as r } from "./utilities/packed.js";
 import { transpose16 as i } from "./ops/transpose16.js";
 import { reshape16 as a } from "./ops/reshape16.js";
 import { mul16 as o } from "./ops/mul16.js";
 import "./ops/webgl/matMul16.js";
 import "./ops/cpu/matMul16.js";
-import { t as s } from "./pack16-Ck-spx_F.js";
+import { t as s } from "./pack16-BhuXNUS7.js";
 //#region lib/ops/grads/matMul16.ts
 var c = {
 	kernelName: "MatMul16",

package/dist/{matMulGelu-CPTntosE.js → matMulGelu-BAIgQaRx.js}

@@ -1,6 +1,6 @@
-import { Ci as e, Ii as t, In as n, Ps as r, di as i, gr as a, ii as o, oc as s } from "./dist-BewPQWjc.js";
-import { a as c } from "./gpgpu_math-DvLcCH6u.js";
-import { t as l } from "./Reshape-CABOPB9d.js";
+import { Ci as e, Ii as t, In as n, Ps as r, di as i, gr as a, ii as o, oc as s } from "./dist-Da20xy8E.js";
+import { a as c } from "./gpgpu_math-DBYEAAdI.js";
+import { t as l } from "./Reshape-dnm9bO3B.js";
 //#region node_modules/@tensorflow/tfjs-backend-webgl/dist/mulmat_packed_gpu.js
 var u = class {
 	constructor(e, t, n, r = !1, i = !1, a = !1, o = null, s = !1, l = !1) {

package/dist/models/NanoGPTV1.js

@@ -1,2 +1,2 @@
-import { y as e } from "../main-C0YUnw98.js";
+import { y as e } from "../main-DdsoTJAW.js";
 export { e as default };

package/dist/models/NanoGPTV2.js

@@ -1,2 +1,2 @@
-import { v as e } from "../main-C0YUnw98.js";
+import { v as e } from "../main-DdsoTJAW.js";
 export { e as default };

package/dist/models/factory.js

@@ -1,2 +1,2 @@
-import { _ as e } from "../main-C0YUnw98.js";
+import { _ as e } from "../main-DdsoTJAW.js";
 export { e as default };

package/dist/models/model.js

@@ -1,2 +1,2 @@
-import { b as e } from "../main-C0YUnw98.js";
+import { b as e } from "../main-DdsoTJAW.js";
 export { e as default };

package/dist/ops/adamAdjust.js

@@ -1,4 +1,4 @@
-import { ii as e } from "../dist-BewPQWjc.js";
+import { ii as e } from "../dist-Da20xy8E.js";
 import "./cpu/adamAdjust.js";
 import "./webgl/adamAdjust.js";
 //#region lib/ops/adamAdjust.ts

package/dist/ops/adamMoments.js

@@ -1,4 +1,4 @@
-import { ii as e } from "../dist-BewPQWjc.js";
+import { ii as e } from "../dist-Da20xy8E.js";
 import "./cpu/adamMoments.js";
 import "./webgl/adamMoments.js";
 //#region lib/ops/adamMoments.ts

package/dist/ops/add16.js

@@ -1,4 +1,4 @@
-import { ii as e, qr as t } from "../dist-BewPQWjc.js";
+import { ii as e, qr as t } from "../dist-Da20xy8E.js";
 import { isPackedTensor as n } from "../utilities/packed.js";
 import "./grads/add16.js";
 //#region lib/ops/add16.ts

package/dist/ops/appendCache.js

@@ -1,4 +1,4 @@
-import { _r as e, ii as t, kt as n } from "../dist-BewPQWjc.js";
+import { _r as e, ii as t, kt as n } from "../dist-Da20xy8E.js";
 import { isPackedTensor as r } from "../utilities/packed.js";
 import "./cpu/appendCache.js";
 import "./webgl/appendCache.js";

package/dist/ops/attentionMask.js

@@ -1,4 +1,4 @@
-import { ii as e } from "../dist-BewPQWjc.js";
+import { ii as e } from "../dist-Da20xy8E.js";
 import "./cpu/attentionMask.js";
 import "./webgl/attentionMask.js";
 import "./grads/attentionMask.js";

package/dist/ops/concat16.js

@@ -1,4 +1,4 @@
-import { _r as e, ii as t } from "../dist-BewPQWjc.js";
+import { _r as e, ii as t } from "../dist-Da20xy8E.js";
 import { isPackedTensor as n } from "../utilities/packed.js";
 //#region lib/ops/concat16.ts
 function r(r, i) {

package/dist/ops/cpu/adamAdjust.js

@@ -1,4 +1,4 @@
-import { Gr as e, Ii as t, Wr as n, gn as r, qr as i } from "../../dist-BewPQWjc.js";
+import { Gr as e, Ii as t, Wr as n, gn as r, qr as i } from "../../dist-Da20xy8E.js";
 //#region lib/ops/cpu/adamAdjust.ts
 function a(t) {
 	let { moments: a, value: o } = t.inputs, { beta1: s, beta2: c, epsilon: l, learningRate: u } = t.attrs, d = a.shape.length, f = Array(d).fill(0), p = a.shape.slice();

package/dist/ops/cpu/adamMoments.js

@@ -1,4 +1,4 @@
-import { Ii as e, V as t } from "../../dist-BewPQWjc.js";
+import { Ii as e, V as t } from "../../dist-Da20xy8E.js";
 //#region lib/ops/cpu/adamMoments.ts
 function n(e) {
 	let { moments: n, gradient: r } = e.inputs, { beta1: i, beta2: a } = e.attrs, o = n.shape.length, s = Array(o).fill(0), c = n.shape.slice();

package/dist/ops/cpu/appendCache.js

@@ -1,4 +1,4 @@
-import { Ii as e, _r as t } from "../../dist-BewPQWjc.js";
+import { Ii as e, _r as t } from "../../dist-Da20xy8E.js";
 //#region lib/ops/cpu/appendCache.ts
 function n(e) {
 	let { cache: n, item: r } = e.inputs, { maxSize: i, pastLen: a } = e.attrs, o = n.shape[0], s = n.shape[1], c = n.shape[3], l = r.shape[2];

package/dist/ops/cpu/attentionMask.js

@@ -1,4 +1,4 @@
-import { Ii as e, Ot as t, Pn as n, _n as r, ar as i, gr as a, kt as o, p as s } from "../../dist-BewPQWjc.js";
+import { Ii as e, Ot as t, Pn as n, _n as r, ar as i, gr as a, kt as o, p as s } from "../../dist-Da20xy8E.js";
 //#region lib/ops/cpu/attentionMask.ts
 function c(e) {
 	let { q: c, k: l } = e.inputs, { divisor: u } = e.attrs, d = c.shape[2], f = l.shape[2], p = n(s.bandPart(t([f, f]), -1, 0).cast("bool"), o([f, f]), i([f, f], -Infinity)), m = a(c, l, !1, !0).mul(r(u)), h = p.slice([0, 0], [d, f]).expandDims(0).expandDims(0);

package/dist/ops/cpu/fusedSoftmax.js

@@ -1,4 +1,4 @@
-import { Ii as e, Y as t } from "../../dist-BewPQWjc.js";
+import { Ii as e, Y as t } from "../../dist-Da20xy8E.js";
 //#region lib/ops/cpu/fusedSoftmax.ts
 function n(e) {
 	let { inputs: n, attrs: r } = e, { logits: i } = n, { dim: a, dropoutRate: o } = r;

package/dist/ops/cpu/gatherSub.js

@@ -1,4 +1,4 @@
-import { Bt as e, Ii as t, S as n, V as r, dt as i } from "../../dist-BewPQWjc.js";
+import { Bt as e, Ii as t, S as n, V as r, dt as i } from "../../dist-Da20xy8E.js";
 //#region lib/ops/cpu/gatherSub.ts
 function a(t) {
 	let { values: a, labels: o, logits: s } = t.inputs, c = o.shape[0];

package/dist/ops/cpu/gelu.js

@@ -1,4 +1,4 @@
-import { Ii as e, di as t } from "../../dist-BewPQWjc.js";
+import { Ii as e, di as t } from "../../dist-Da20xy8E.js";
 //#region lib/ops/cpu/gelu.ts
 var n = .7978845608028654, r = .044715;
 function i(e) {

package/dist/ops/cpu/matMul16.js

@@ -1,4 +1,4 @@
-import { Ii as e, gr as t } from "../../dist-BewPQWjc.js";
+import { Ii as e, gr as t } from "../../dist-Da20xy8E.js";
 import { isPackedTensor as n } from "../../utilities/packed.js";
 //#region lib/ops/cpu/matMul16.ts
 function r(e) {

package/dist/ops/cpu/matMulGelu.js

@@ -1,5 +1,5 @@
-import { Ii as e, di as t, gr as n } from "../../dist-BewPQWjc.js";
-import { n as r, t as i } from "../../gelu-Bf1HW1RY.js";
+import { Ii as e, di as t, gr as n } from "../../dist-Da20xy8E.js";
+import { n as r, t as i } from "../../gelu-B6637MKa.js";
 //#region lib/ops/cpu/matMulGelu.ts
 function a(e) {
 	let { inputs: i } = e, { x: a, kernel: o } = i, s = a, c = o;

package/dist/ops/cpu/matMulMul.js

@@ -1,4 +1,4 @@
-import { Ii as e, di as t } from "../../dist-BewPQWjc.js";
+import { Ii as e, di as t } from "../../dist-Da20xy8E.js";
 //#region lib/ops/cpu/matMulMul.ts
 function n(e) {
 	let { inputs: n, attrs: r } = e, { transposeA: i, transposeB: a } = r, { x: o, kernel: s, y: c } = n, l = o, u = s, d = c;

package/dist/ops/cpu/mulDropout.js

@@ -1,4 +1,4 @@
-import { Ii as e, Wr as t } from "../../dist-BewPQWjc.js";
+import { Ii as e, Wr as t } from "../../dist-Da20xy8E.js";
 //#region lib/ops/cpu/mulDropout.ts
 function n(e) {
 	let { inputs: n } = e, { a: r, b: i } = n;

package/dist/ops/cpu/normRMS.js

@@ -1,4 +1,4 @@
-import { Ii as e, di as t } from "../../dist-BewPQWjc.js";
+import { Ii as e, di as t } from "../../dist-Da20xy8E.js";
 //#region lib/ops/cpu/normRMS.ts
 function n(e) {
 	let { inputs: n } = e, { x: r, gamma: i } = n, a = r, o = i;

package/dist/ops/cpu/qkv.js

@@ -1,4 +1,4 @@
-import { G as e, Ii as t, br as n } from "../../dist-BewPQWjc.js";
+import { G as e, Ii as t, br as n } from "../../dist-Da20xy8E.js";
 //#region lib/ops/cpu/qkv.ts
 function r(t) {
 	let { x: r, kernel: i } = t.inputs, { heads: a, packed: o } = t.attrs;

package/dist/ops/cpu/rope.js

@@ -1,4 +1,4 @@
-import { Ii as e, V as t, _r as n, an as r, dt as i } from "../../dist-BewPQWjc.js";
+import { Ii as e, V as t, _r as n, an as r, dt as i } from "../../dist-Da20xy8E.js";
 //#region lib/ops/cpu/rope.ts
 function a(e, a, o, s, c) {
 	let l = s.shape[3], u = o;

package/dist/ops/cpu/scatterSub.js

@@ -1,4 +1,4 @@
-import { Bt as e, C as t, Ii as n, Ot as r, V as i, Wr as a, dt as o } from "../../dist-BewPQWjc.js";
+import { Bt as e, C as t, Ii as n, Ot as r, V as i, Wr as a, dt as o } from "../../dist-Da20xy8E.js";
 //#region lib/ops/cpu/scatterSub.ts
 function s(n) {
 	let { logits: s, labels: c, dy: l } = n.inputs, u = c.shape[0], d = s.shape[1];

package/dist/ops/dot16.js

@@ -1,8 +1,8 @@
-import { o as e } from "../tfjs_backend-BjuQ5FqB.js";
+import { o as e } from "../tfjs_backend-CydPRQTc.js";
 import { isPackedTensor as t } from "../utilities/packed.js";
 import { transpose16 as n } from "./transpose16.js";
 import { reshape16 as r } from "./reshape16.js";
-import { t as i } from "../matMul16-BNfZSnNM.js";
+import { t as i } from "../matMul16-DGtlziMn.js";
 //#region lib/ops/dot16.ts
 function a(a, o, s = !1, c = !1) {
 	if (!t(a) && !t(o)) return e(a, o);

package/dist/ops/dropout.js

@@ -1,4 +1,4 @@
-import { Gr as e, Ps as t, ft as n, mi as r, on as i, qr as a, wi as o, x as s } from "../dist-BewPQWjc.js";
+import { Gr as e, Ps as t, ft as n, mi as r, on as i, qr as a, wi as o, x as s } from "../dist-Da20xy8E.js";
 import { mul16 as c } from "./mul16.js";
 //#region lib/ops/dropout.ts
 function l(l, u, d, f) {

package/dist/ops/dropout16.js

@@ -1,4 +1,4 @@
-import { Fi as e, ii as t } from "../dist-BewPQWjc.js";
+import { Fi as e, ii as t } from "../dist-Da20xy8E.js";
 import { isPackedTensor as n } from "../utilities/packed.js";
 import "./webgl/dropout16.js";
 //#region lib/ops/grads/dropout16.ts

package/dist/ops/gatherSub.js

@@ -1,4 +1,4 @@
-import { ii as e } from "../dist-BewPQWjc.js";
+import { ii as e } from "../dist-Da20xy8E.js";
 import "./cpu/gatherSub.js";
 import "./webgl/gatherSub.js";
 //#region lib/ops/gatherSub.ts

package/dist/ops/gelu.js

@@ -1,2 +1,2 @@
-import { n as e, t } from "../gelu-Bf1HW1RY.js";
+import { n as e, t } from "../gelu-B6637MKa.js";
 export { t as dGelu, e as gelu };

package/dist/ops/globalNorm.js

@@ -1,4 +1,4 @@
-import { di as e, ii as t, kt as n } from "../dist-BewPQWjc.js";
+import { di as e, ii as t, kt as n } from "../dist-Da20xy8E.js";
 //#region lib/ops/globalNorm.ts
 function r(r, i, a = 1) {
 	return e(() => {

package/dist/ops/grads/add16.js

@@ -1,4 +1,4 @@
-import { Fi as e, In as t, Rn as n } from "../../dist-BewPQWjc.js";
+import { Fi as e, In as t, Rn as n } from "../../dist-Da20xy8E.js";
 import { reshape16 as r } from "../reshape16.js";
 import { sum16 as i } from "../sum16.js";
 //#region lib/ops/grads/add16.ts

package/dist/ops/grads/attentionMask.js

@@ -1,6 +1,6 @@
-import { Fi as e } from "../../dist-BewPQWjc.js";
+import { Fi as e } from "../../dist-Da20xy8E.js";
 import { transpose16 as t } from "../transpose16.js";
-import { a as n } from "../../matMul16-BNfZSnNM.js";
+import { a as n } from "../../matMul16-DGtlziMn.js";
 //#region lib/ops/grads/attentionMask.ts
 e({
 	kernelName: "AttentionMask",

package/dist/ops/grads/gelu.js

@@ -1,2 +1,2 @@
-import { r as e } from "../../gelu-Bf1HW1RY.js";
+import { r as e } from "../../gelu-B6637MKa.js";
 export { e as geluGradConfig };

package/dist/ops/grads/matMul16.js

@@ -1,2 +1,2 @@
-import { o as e } from "../../matMul16-BNfZSnNM.js";
+import { o as e } from "../../matMul16-DGtlziMn.js";
 export { e as matMul16GradConfig };

package/dist/ops/grads/matMulGelu.js

@@ -1,4 +1,4 @@
-import { Fi as e, ii as t } from "../../dist-BewPQWjc.js";
+import { Fi as e, ii as t } from "../../dist-Da20xy8E.js";
 //#region lib/ops/grads/matMulGelu.ts
 function n(e, n, r) {
 	return t().runKernel("MatMulGeluGrad", {

package/dist/ops/grads/normRMS.js

@@ -1,4 +1,4 @@
-import { Fi as e, ii as t } from "../../dist-BewPQWjc.js";
+import { Fi as e, ii as t } from "../../dist-Da20xy8E.js";
 //#region lib/ops/grads/normRMS.ts
 function n(e, n, r) {
 	return t().runKernel("RMSNormGrad", r ? {

package/dist/ops/grads/pack16.js

@@ -1,2 +1,2 @@
-import { n as e } from "../../pack16-Ck-spx_F.js";
+import { n as e } from "../../pack16-BhuXNUS7.js";
 export { e as packGradConfig };

package/dist/ops/grads/qkv.js

@@ -1,6 +1,6 @@
-import { Fi as e, H as t } from "../../dist-BewPQWjc.js";
+import { Fi as e, H as t } from "../../dist-Da20xy8E.js";
 import { sum16 as n } from "../sum16.js";
-import { o as r } from "../../matMul16-BNfZSnNM.js";
+import { o as r } from "../../matMul16-DGtlziMn.js";
 import { concat16 as i } from "../concat16.js";
 //#region lib/ops/grads/qkv.ts
 var a = {

package/dist/ops/grads/rope.js

@@ -1,2 +1,2 @@
-import { n as e } from "../../rope-CbeGlsV8.js";
+import { n as e } from "../../rope-C8EAiSGY.js";
 export { e as ropeGradConfig };

package/dist/ops/grads/softmax16.js

@@ -1,4 +1,4 @@
-import { Fi as e, ii as t } from "../../dist-BewPQWjc.js";
+import { Fi as e, ii as t } from "../../dist-Da20xy8E.js";
 import { isPackedTensor as n } from "../../utilities/packed.js";
 //#region lib/ops/grads/softmax16.ts
 function r(e, n) {

package/dist/ops/grads/unpack16.js

@@ -1,2 +1,2 @@
-import { i as e } from "../../pack16-Ck-spx_F.js";
+import { i as e } from "../../pack16-BhuXNUS7.js";
 export { e as unpackGradConfig };

package/dist/ops/matMul16.js

@@ -1,2 +1,2 @@
-import { a as e, i as t, n, r, t as i } from "../matMul16-BNfZSnNM.js";
+import { a as e, i as t, n, r, t as i } from "../matMul16-DGtlziMn.js";
 export { i as matMul16, n as matMul16Gelu, r as matMul16ScaleA, t as matMul16ScaleB, e as matMul16Scaled };

package/dist/ops/matMulGelu.js

@@ -1,5 +1,5 @@
-import { ii as e } from "../dist-BewPQWjc.js";
-import "../matMulGelu-CPTntosE.js";
+import { ii as e } from "../dist-Da20xy8E.js";
+import "../matMulGelu-BAIgQaRx.js";
 import "./cpu/matMulGelu.js";
 import "./grads/matMulGelu.js";
 //#region lib/ops/matMulGelu.ts

package/dist/ops/matMulMul.js

@@ -1,4 +1,4 @@
-import { ii as e } from "../dist-BewPQWjc.js";
+import { ii as e } from "../dist-Da20xy8E.js";
 import "./cpu/matMulMul.js";
 import "./webgl/matMulMul.js";
 //#region lib/ops/matMulMul.ts

package/dist/ops/mul16.js

@@ -1,4 +1,4 @@
-import { Fi as e, In as t, Rn as n, Wr as r, ii as i } from "../dist-BewPQWjc.js";
+import { Fi as e, In as t, Rn as n, Wr as r, ii as i } from "../dist-Da20xy8E.js";
 import { isPackedTensor as a } from "../utilities/packed.js";
 import { reshape16 as o } from "./reshape16.js";
 import { sum16 as s } from "./sum16.js";

package/dist/ops/mulDrop.js

@@ -1,4 +1,4 @@
-import { ii as e } from "../dist-BewPQWjc.js";
+import { ii as e } from "../dist-Da20xy8E.js";
 import "./cpu/mulDropout.js";
 import "./webgl/mulDropout.js";
 //#region lib/ops/mulDrop.ts

package/dist/ops/normRMS.js

@@ -1,4 +1,4 @@
-import { ii as e } from "../dist-BewPQWjc.js";
+import { ii as e } from "../dist-Da20xy8E.js";
 import "./cpu/normRMS.js";
 import "./webgl/normRMS.js";
 import "./grads/normRMS.js";

package/dist/ops/pack16.js

@@ -1,2 +1,2 @@
-import { t as e } from "../pack16-Ck-spx_F.js";
+import { t as e } from "../pack16-BhuXNUS7.js";
 export { e as pack16 };

package/dist/ops/qkv.js

@@ -1,4 +1,4 @@
-import { ii as e } from "../dist-BewPQWjc.js";
+import { ii as e } from "../dist-Da20xy8E.js";
 import "./cpu/qkv.js";
 import "./webgl/qkv.js";
 import "./grads/qkv.js";

package/dist/ops/reshape16.js

@@ -1,4 +1,4 @@
-import { Fi as e, Ii as t, br as n, ii as r } from "../dist-BewPQWjc.js";
+import { Fi as e, Ii as t, br as n, ii as r } from "../dist-Da20xy8E.js";
 //#region lib/ops/reshape16.ts
 e({
 	kernelName: "Reshape16",

package/dist/ops/rope.js

@@ -1,2 +1,2 @@
-import { t as e } from "../rope-CbeGlsV8.js";
+import { t as e } from "../rope-C8EAiSGY.js";
 export { e as rope };

package/dist/ops/scatterSub.js

@@ -1,4 +1,4 @@
-import { ii as e } from "../dist-BewPQWjc.js";
+import { ii as e } from "../dist-Da20xy8E.js";
 import "./cpu/scatterSub.js";
 import "./webgl/scatterSub.js";
 //#region lib/ops/scatterSub.ts

package/dist/ops/slice16.js

@@ -1,4 +1,4 @@
-import { ii as e, mr as t } from "../dist-BewPQWjc.js";
+import { ii as e, mr as t } from "../dist-Da20xy8E.js";
 import { isPackedTensor as n } from "../utilities/packed.js";
 //#region lib/ops/slice16.ts
 function r(r, i, a) {

package/dist/ops/softmax16.js

@@ -1,4 +1,4 @@
-import { ii as e } from "../dist-BewPQWjc.js";
+import { ii as e } from "../dist-Da20xy8E.js";
 import { isPackedTensor as t } from "../utilities/packed.js";
 import "./grads/softmax16.js";
 //#region lib/ops/softmax16.ts

package/dist/ops/sub16.js

@@ -1,4 +1,4 @@
-import { Bt as e, ii as t } from "../dist-BewPQWjc.js";
+import { Bt as e, ii as t } from "../dist-Da20xy8E.js";
 import { isPackedTensor as n } from "../utilities/packed.js";
 //#region lib/ops/sub16.ts
 function r(r, i) {

package/dist/ops/sum16.js

@@ -1,4 +1,4 @@
-import { ii as e, mn as t } from "../dist-BewPQWjc.js";
+import { ii as e, mn as t } from "../dist-Da20xy8E.js";
 import { isPackedTensor as n } from "../utilities/packed.js";
 //#region lib/ops/sum16.ts
 function r(r, i, a = !1) {

package/dist/ops/transpose16.js

@@ -1,4 +1,4 @@
-import { Fi as e, Ii as t, On as n, ii as r, w as i } from "../dist-BewPQWjc.js";
+import { Fi as e, Ii as t, On as n, ii as r, w as i } from "../dist-Da20xy8E.js";
 import { forceFloat as a, forcePacked as o } from "./grads/utils.js";
 import { isPackedTensor as s } from "../utilities/packed.js";
 //#region lib/ops/transpose16.ts

package/dist/ops/unpack16.js

@@ -1,2 +1,2 @@
-import { r as e } from "../pack16-Ck-spx_F.js";
+import { r as e } from "../pack16-BhuXNUS7.js";
 export { e as unpack16 };

package/dist/ops/webgl/adamAdjust.js

@@ -1,5 +1,5 @@
-import { Ii as e } from "../../dist-BewPQWjc.js";
-import { t } from "../../Reshape-CABOPB9d.js";
+import { Ii as e } from "../../dist-Da20xy8E.js";
+import { t } from "../../Reshape-dnm9bO3B.js";
 //#region lib/ops/webgl/adamAdjust.ts
 var n = class {
 	variableNames = ["moments", "value"];

package/dist/ops/webgl/adamMoments.js

@@ -1,4 +1,4 @@
-import { Ii as e } from "../../dist-BewPQWjc.js";
+import { Ii as e } from "../../dist-Da20xy8E.js";
 //#region lib/ops/webgl/adamMoments.ts
 var t = class {
 	variableNames = ["moments", "gradient"];
@@ -19,8 +19,8 @@ var t = class {
             float m = getMomentsAtOutCoords();
             ${n} coords = getOutputCoords();
 
-            // Add gradient clipping here
-            float g = clamp(getGradient(${i}), -1.0, 1.0);
+            // TODO: Add scaling
+            float g = getGradient(${i});
             int which = ${r};
 
             float beta = which == 0 ? beta1 : beta2;

package/dist/ops/webgl/appendCache.js

@@ -1,4 +1,4 @@
-import { Ii as e } from "../../dist-BewPQWjc.js";
+import { Ii as e } from "../../dist-Da20xy8E.js";
 //#region lib/ops/webgl/appendCache.ts
 var t = class {
 	variableNames = ["cache", "item"];

package/dist/ops/webgl/attentionMask.js

@@ -1,4 +1,4 @@
-import { Ii as e } from "../../dist-BewPQWjc.js";
+import { Ii as e } from "../../dist-Da20xy8E.js";
 //#region lib/ops/webgl/attentionMask.ts
 var t = class {
 	variableNames = ["q", "k"];

package/dist/ops/webgl/dropout16.js

@@ -1,4 +1,4 @@
-import { Ii as e } from "../../dist-BewPQWjc.js";
+import { Ii as e } from "../../dist-Da20xy8E.js";
 //#region lib/ops/webgl/dropout16.ts
 function t(e) {
 	let { x: t } = e.inputs;

package/dist/ops/webgl/fusedSoftmax.js

@@ -1,6 +1,6 @@
-import { Ii as e, Tn as t, nc as n } from "../../dist-BewPQWjc.js";
-import { t as r } from "../../Reshape-CABOPB9d.js";
-import { a as i, r as a, t as o } from "../../RealDiv-DBu0FQqT.js";
+import { Ii as e, Tn as t, nc as n } from "../../dist-Da20xy8E.js";
+import { t as r } from "../../Reshape-dnm9bO3B.js";
+import { a as i, r as a, t as o } from "../../RealDiv-CNsvC4AU.js";
 //#region lib/ops/webgl/fusedSoftmax.ts
 var s = class {
 	variableNames = ["logits", "maxLogits"];

package/dist/ops/webgl/gatherSub.js

@@ -1,4 +1,4 @@
-import { Ii as e } from "../../dist-BewPQWjc.js";
+import { Ii as e } from "../../dist-Da20xy8E.js";
 //#region lib/ops/webgl/gatherSub.ts
 var t = class {
 	variableNames = [

package/dist/ops/webgl/gelu.js

@@ -1,5 +1,5 @@
-import { Ii as e } from "../../dist-BewPQWjc.js";
-import { _ as t, i as n } from "../../kernel_funcs_utils-HiXOOx3f.js";
+import { Ii as e } from "../../dist-Da20xy8E.js";
+import { _ as t, i as n } from "../../kernel_funcs_utils-D-mATnGR.js";
 //#region lib/ops/webgl/gelu.ts
 var r = .7978845608028654, i = .044715, a = n({ opSnippet: t + `
     float x3 = x * x * x;

package/dist/ops/webgl/log.js

@@ -1,6 +1,6 @@
-import { Ii as e } from "../../dist-BewPQWjc.js";
-import { y as t } from "../../shared-DOgWaqvL.js";
-import { i as n, t as r } from "../../kernel_funcs_utils-HiXOOx3f.js";
+import { Ii as e } from "../../dist-Da20xy8E.js";
+import { y as t } from "../../shared-CfTzpULd.js";
+import { i as n, t as r } from "../../kernel_funcs_utils-D-mATnGR.js";
 //#region lib/ops/webgl/log.ts
 e({
 	kernelName: "Log",

package/dist/ops/webgl/matMul16.js

@@ -1,4 +1,4 @@
-import { Ii as e, Wr as t, _n as n, br as r, gr as i, hn as a, lt as o, w as s } from "../../dist-BewPQWjc.js";
+import { Ii as e, Wr as t, _n as n, br as r, gr as i, hn as a, lt as o, w as s } from "../../dist-Da20xy8E.js";
 import { isPackedTensor as c } from "../../utilities/packed.js";
 import { matMulMul as l } from "../matMulMul.js";
 import { matMulGelu as u } from "../matMulGelu.js";
@@ -7,12 +7,16 @@ function d(e) {
 	let { A: d, B: f } = e.inputs, { transposeA: p, transposeB: m, scale: h, activation: g, scaleA: _, scaleB: v, forceOutputShape: y, perm: b } = e.attrs, x = !c(d), S = !c(f);
 	if (x && S) {
 		let e = _ === void 0 ? d : t(d, n(_)), c = v === void 0 ? f : t(f, n(v)), x;
-		if (h !== void 0) x = l(e, c, n(h), p, m);
-		else if (g === "gelu") x = u(e, c);
+		if (g === "gelu") x = u(e, c);
 		else if (g === "relu2") {
 			let t = i(e, c, p, m), n = o(t);
 			t.dispose(), x = a(n), n.dispose();
-		} else x = g === "relu" ? o(i(e, c, p, m)) : i(e, c, p, m);
+		} else if (g === "relu") {
+			if (x = o(i(e, c, p, m)), h !== void 0) {
+				let e = t(x, n(h));
+				x.dispose(), x = e;
+			}
+		} else x = h === void 0 ? i(e, c, p, m) : l(e, c, n(h), p, m);
 		if (b) if (y) {
 			let e = r(x, y);
 			x.dispose();

package/dist/ops/webgl/matMulGelu.js

@@ -1,2 +1,2 @@
-import { n as e, r as t, t as n } from "../../matMulGelu-CPTntosE.js";
+import { n as e, r as t, t as n } from "../../matMulGelu-BAIgQaRx.js";
 export { n as MATMUL_SHARED_DIM_THRESHOLD, e as batchMatMulGeluImpl, t as batchMatMulKernel };

package/dist/ops/webgl/matMulMul.js

@@ -1,5 +1,5 @@
-import { Ii as e } from "../../dist-BewPQWjc.js";
-import { n as t } from "../../matMulGelu-CPTntosE.js";
+import { Ii as e } from "../../dist-Da20xy8E.js";
+import { n as t } from "../../matMulGelu-BAIgQaRx.js";
 //#region lib/ops/webgl/matMulMul.ts
 var n = "\n    return a * b;\n";
 function r(e) {

package/dist/ops/webgl/mulDropout.js

@@ -1,4 +1,4 @@
-import { Ii as e } from "../../dist-BewPQWjc.js";
+import { Ii as e } from "../../dist-Da20xy8E.js";
 //#region lib/ops/webgl/mulDropout.ts
 var t = class {
 	variableNames = ["a", "b"];

package/dist/ops/webgl/normRMS.js

@@ -1,4 +1,4 @@
-import { Ii as e, ii as t, mn as n } from "../../dist-BewPQWjc.js";
+import { Ii as e, ii as t, mn as n } from "../../dist-Da20xy8E.js";
 //#region lib/ops/webgl/normRMS.ts
 var r = class {
 	variableNames = ["x", "meanSquare"];