@genai-fi/nanogpt 0.6.0 → 0.6.2

package/dist/ops/cpu/matMulMul.js

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

package/dist/ops/cpu/mulDropout.js

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

package/dist/ops/cpu/normRMS.js

@@ -1,4 +1,4 @@
-import { r as o, t as d } from "../../index-bMBtI-WR.js";
+import { r as o, t as d } from "../../index-BAzbokzv.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 { r as q } from "../../index-bMBtI-WR.js";
-import { r as o } from "../../reshape-T4yDEqoF.js";
-import { s as x } from "../../split-dcks18H1.js";
+import { r as q } from "../../index-BAzbokzv.js";
+import { r as o } from "../../reshape-krWGKraP.js";
+import { s as x } from "../../split-eiktj-6L.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 { r as S } from "../../index-bMBtI-WR.js";
-import { r as F } from "../../range-DMaG9A3G.js";
-import { g as I } from "../../gather-DZCMHZuN.js";
-import { s as E } from "../../stack-lpJ5kYvE.js";
-import { c as T } from "../../concat-DhZfF1GY.js";
+import { r as S } from "../../index-BAzbokzv.js";
+import { r as F } from "../../range-DYsrnfiy.js";
+import { g as I } from "../../gather-DjyCjmOD.js";
+import { s as E } from "../../stack-dfEEz2OY.js";
+import { c as T } from "../../concat-5aPGqw3Z.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

@@ -1,39 +1,8 @@
-import { o as l, n as g, j as h, E as k, a5 as w, r as $, s as d, b as m } from "../../index-bMBtI-WR.js";
-import { r as b } from "../../range-DMaG9A3G.js";
-import { s as E } from "../../stack-lpJ5kYvE.js";
-import { o as D } from "../../ones-g0K8jVwm.js";
-function N(a, r, t) {
-  const s = r.rank > 1 ? r.shape[r.rank - 1] : 1, e = r.rank > 1 ? r.rank - 1 : 1, o = `Must have updates.shape = indices.shape[:batchDim] + shape[sliceDim:], got updates.shape: ${t.shape}, indices.shape: ${r.shape}, shape: ${a}, sliceDim: ${s}, and batchDim: ${e}.`;
-  if (t.rank < e)
-    throw new Error(o + ` update.rank < ${e}. `);
-  if (a.length < s + (t.rank - e))
-    throw new Error(o + ` Output shape length < ${s + (t.rank - e)}`);
-  if (t.rank !== e + a.length - s)
-    throw new Error(o + ` update.rank != ${e + a.length - s}`);
-  for (let n = 0; n < e; ++n)
-    if (t.shape[n] !== r.shape[n])
-      throw new Error(o + ` updates.shape[${n}] (${t.shape[n]}) != indices.shape[${n}] (${r.shape[n]}).`);
-  for (let n = 0; n < t.rank - e; ++n)
-    if (t.shape[n + e] !== a[n + s])
-      throw new Error(o + ` updates.shape[${n + e}] (${t.shape[n + e]}) != shape[${n + e}] (${a[n + e]})`);
-}
-function S(a, r, t) {
-  if (r.rank < 1)
-    throw new Error(`tf.scatterND() expects the indices to be rank 1 or higher, but the rank was ${r.rank}.`);
-  if (a.rank < 1)
-    throw new Error(`tf.scatterND() expects the updates to be rank 1 or higher, but the rank was ${a.rank}.`);
-  if (r.dtype !== "int32")
-    throw new Error(`The dtype of 'indices' should be int32, but got dtype: ${r.dtype}`);
-  if (t.length < 1)
-    throw new Error(`Output rank must be greater or equal to 1, but got shape: ${t}`);
-  if (t.length === 0) {
-    if (r.size === 0)
-      throw new Error(`Indices specified for empty output. indices shape: ${r.shape}`);
-    if (a.size === 0)
-      throw new Error(`Updates specified for empty output. updates shape: ${a.shape}`);
-  }
-  N(t, r, a);
-}
+import { o as f, x as g, q as r, E as l, X as N, r as b, b as S, a as h } from "../../index-BAzbokzv.js";
+import { v as D } from "../../scatter_nd_util-93ln7Hut.js";
+import { r as k } from "../../range-DYsrnfiy.js";
+import { s as v } from "../../stack-dfEEz2OY.js";
+import { o as E } from "../../ones-BBlSRqn1.js";
 /**
  * @license
  * Copyright 2018 Google LLC. All Rights Reserved.
@@ -50,21 +19,21 @@ function S(a, r, t) {
  * limitations under the License.
  * =============================================================================
  */
-function y(a, r, t) {
-  g(t);
-  const s = h(a, "indices", "scatterND", "int32"), e = h(r, "updates", "scatterND");
-  S(e, s, t);
-  const o = { indices: s, updates: e }, n = { shape: t };
-  return k.runKernel(w, o, n);
+function I(a, e, s) {
+  g(s);
+  const n = r(a, "indices", "scatterND", "int32"), t = r(e, "updates", "scatterND");
+  D(t, n, s);
+  const c = { indices: n, updates: t }, o = { shape: s };
+  return l.runKernel(N, c, o);
 }
-const v = /* @__PURE__ */ l({ scatterND_: y });
-function I(a) {
-  const { logits: r, labels: t, dy: s } = a.inputs, e = t.shape[0], o = r.shape[1], n = b(0, e, 1, "int32"), i = E([n, t], 1), c = D([e]), p = v(i, c, [e, o]), f = d(r, p), u = s.reshape([e, 1]);
-  return m(f, u);
+const K = /* @__PURE__ */ f({ scatterND_: I });
+function L(a) {
+  const { logits: e, labels: s, dy: n } = a.inputs, t = s.shape[0], c = e.shape[1], o = k(0, t, 1, "int32"), i = v([o, s], 1), d = E([t]), u = K(i, d, [t, c]), p = S(e, u), m = n.reshape([t, 1]);
+  return h(p, m);
 }
 const T = {
   kernelName: "EfficientScatterSub",
   backendName: "cpu",
-  kernelFunc: I
+  kernelFunc: L
 };
-$(T);
+b(T);

package/dist/ops/fusedSoftmax.js

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

package/dist/ops/gelu.js

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

package/dist/ops/grads/attentionMask.js

@@ -1,4 +1,4 @@
-import { h as m, f as i } from "../../index-bMBtI-WR.js";
+import { c as m, s as i } from "../../index-BAzbokzv.js";
 import { matMulMul as u } from "../matMulMul.js";
 const p = {
   kernelName: "AttentionMask",

package/dist/ops/grads/fusedSoftmax.js

@@ -1,14 +1,14 @@
-import { h as f, b as i, s as l } from "../../index-bMBtI-WR.js";
+import { c as f, a as i, b as l } from "../../index-BAzbokzv.js";
 import { mulDrop as g } from "../mulDrop.js";
-import { s as T } from "../../sum-CutF5lj2.js";
+import { s as T } from "../../sum-BE_Irnim.js";
 const Y = {
   kernelName: "FusedSoftmax",
   outputsToSave: [!0],
   gradFunc: (o, n, a) => {
-    const [s] = n, { dim: p, dropoutRate: t, seed: e } = a, c = !0, r = t && e ? g(o, s, t, e) : i(o, s);
+    const [s] = n, { dim: c, dropoutRate: t, seed: e } = a, p = !0, r = t && e ? g(o, s, t, e) : i(o, s);
     return {
       logits: () => {
-        const m = T(r, [p], c), u = i(m, s);
+        const m = T(r, [c], p), u = i(m, s);
         m.dispose();
         const d = l(r, u);
         return u.dispose(), d;

package/dist/ops/grads/gelu.js

@@ -1,4 +1,4 @@
-import { h as t, e as n } from "../../index-bMBtI-WR.js";
+import { c as t, e as n } from "../../index-BAzbokzv.js";
 import "../cpu/gelu.js";
 import "../webgl/gelu.js";
 const o = {

package/dist/ops/grads/matMulGelu.js

@@ -1,4 +1,4 @@
-import { h as a, e as o } from "../../index-bMBtI-WR.js";
+import { c as a, e as o } from "../../index-BAzbokzv.js";
 function s(e, n, r) {
   return o().runKernel("MatMulGeluGrad", { dy: e, x: n, kernel: r });
 }

package/dist/ops/grads/normRMS.js

@@ -1,4 +1,4 @@
-import { h as t, e as g } from "../../index-bMBtI-WR.js";
+import { c as t, e as g } from "../../index-BAzbokzv.js";
 function s(r, a, n) {
   return g().runKernel("RMSNormGrad", { dy: r, x: a, gamma: n });
 }

package/dist/ops/grads/qkv.js

@@ -1,4 +1,4 @@
-import { h as Q } from "../../index-bMBtI-WR.js";
+import { c as Q } from "../../index-BAzbokzv.js";
 const V = {
   kernelName: "QKV",
   inputsToSave: ["x", "kernel"],

package/dist/ops/grads/rope.js

@@ -1,4 +1,4 @@
-import { h as a, e as i } from "../../index-bMBtI-WR.js";
+import { c as a, e as i } from "../../index-BAzbokzv.js";
 function p(n, e, s, o) {
   return i().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-bMBtI-WR.js";
+import { e as u } from "../index-BAzbokzv.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-bMBtI-WR.js";
+import { e as u } from "../index-BAzbokzv.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-bMBtI-WR.js";
+import { e as t } from "../index-BAzbokzv.js";
 import "./cpu/mulDropout.js";
 import "./webgl/mulDropout.js";
 function m(r, o, e, n) {

package/dist/ops/node/sparseCrossEntropy.js

@@ -1,4 +1,4 @@
-import { r as o } from "../../index-bMBtI-WR.js";
+import { r as o } from "../../index-BAzbokzv.js";
 function r(e) {
   const { logits: t, labels: n } = e.inputs;
   return e.backend.executeMultipleOutputs("SparseSoftmaxCrossEntropyWithLogits", [], [t, n], 2);

package/dist/ops/normRMS.js

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

package/dist/ops/rope.js

@@ -1,10 +1,14 @@
-import { engine as n } from "@tensorflow/tfjs";
+import { e as p } from "../index-BAzbokzv.js";
+import "../random_width-DWzaOgrn.js";
+import "../register_all_kernels-BfFCQAqs.js";
+import "../index-Tf7vU29b.js";
+import "../dataset-pgqp-YfL.js";
 import "./cpu/rope.js";
 import "./webgl/rope.js";
 import "./grads/rope.js";
-function s(o, e, r) {
-  return e.ensureRopeCache(o.shape[1] + r), n().runKernel("Rope", { x: o, sin: e.getSin(), cos: e.getCos() }, { pastLen: r });
+function C(r, o, e) {
+  return o.ensureRopeCache(r.shape[1] + e), p().runKernel("Rope", { x: r, sin: o.getSin(), cos: o.getCos() }, { pastLen: e });
 }
 export {
-  s as rope
+  C as rope
 };

package/dist/ops/scatterSub.js

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

package/dist/ops/webgl/appendCache.js

@@ -1,4 +1,4 @@
-import { r as p } from "../../index-bMBtI-WR.js";
+import { r as p } from "../../index-BAzbokzv.js";
 class m {
   variableNames = ["cache", "item"];
   outputShape;

package/dist/ops/webgl/attentionMask.js

@@ -1,4 +1,4 @@
-import { r as m } from "../../index-bMBtI-WR.js";
+import { r as m } from "../../index-BAzbokzv.js";
 class h {
   variableNames = ["q", "k"];
   outputShape;