@genai-fi/nanogpt 0.8.0 → 0.8.2

package/dist/checks/qkv.js

@@ -0,0 +1,50 @@
+import { z as i, A as u, B as c, s as l, e as h } from "../index-DdmHGZjq.js";
+import { t as f } from "../tensor2d-CObBWBkW.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.
+ * =============================================================================
+ */
+function m(t, e, n) {
+  if (i(t), e != null && e.length !== 3)
+    throw new Error("tensor3d() requires shape to have three numbers");
+  const r = u(t, n);
+  if (r.length !== 3 && r.length !== 1)
+    throw new Error("tensor3d() requires values to be number[][][] or flat/TypedArray");
+  if (r.length === 1 && e == null)
+    throw new Error("tensor3d() requires shape to be provided when `values` are a flat array");
+  return c(t, e, r, n);
+}
+async function y(t) {
+  await l(t);
+  const e = m(
+    [
+      [
+        [0.1, 0.2],
+        [0.3, 0.4]
+      ]
+    ],
+    [1, 2, 2]
+  ), n = f(
+    [
+      [0.5, 0.6, 0.9, 1, 1.3, 1.4],
+      [0.7, 0.8, 1.1, 1.2, 1.5, 1.6]
+    ],
+    [2, 6]
+  ), r = h().runKernel("QKV", { x: e, kernel: n }, { heads: 1 }), o = await r[0].array(), a = await r[1].array(), s = await r[2].array();
+  return [o, a, s];
+}
+export {
+  y as execute
+};

package/dist/checks/rope.d.ts

@@ -0,0 +1 @@
+export declare function execute(backend: string): Promise<number | number[] | number[][] | number[][][] | number[][][][] | number[][][][][] | number[][][][][][] | Promise<number | number[] | number[][] | number[][][] | number[][][][] | number[][][][][] | number[][][][][][]>[]>;

package/dist/checks/rope.js

@@ -0,0 +1,38 @@
+import t from "../layers/RoPECache.js";
+import { s as c, e as i } from "../index-DdmHGZjq.js";
+import { t as p } from "../tensor4d-DLtk7Nxh.js";
+async function y(a) {
+  await c(a);
+  const o = p(
+    [
+      [
+        [
+          [0.1, 0.2],
+          [0.3, 0.4]
+        ]
+      ]
+    ],
+    [1, 1, 2, 2]
+  ), n = {
+    biasInLayerNorm: !1,
+    vocabSize: 20,
+    nEmbed: 16,
+    nHead: 2,
+    nLayer: 1,
+    biasInLinear: !1,
+    dropout: 0,
+    blockSize: 128,
+    mlpFactor: 4,
+    useRope: !0
+  }, e = new t(n);
+  e.ensureRopeCache(120);
+  const r = i().runKernel(
+    "Rope",
+    { x: o, sin: e.getSin(), cos: e.getCos() },
+    { pastLen: 20 }
+  );
+  return Array.isArray(r) ? r.map((s) => s.array()) : r.array();
+}
+export {
+  y as execute
+};

package/dist/checks/weights.d.ts

@@ -0,0 +1,16 @@
+import { default as BaseLayer } from '../layers/BaseLayer';
+import { Tensor } from '@tensorflow/tfjs-core';
+export interface TensorStatistics {
+    mean: number;
+    std: number;
+    min: number;
+    max: number;
+    sparsity: number;
+    isFinite: boolean;
+    hasNaN: boolean;
+    closeToZeroCount: number;
+}
+export declare function createTensorStatistics(weight: Tensor | number[]): Promise<TensorStatistics>;
+export declare function createWeightStatistics(layer: BaseLayer): Promise<{
+    [key: string]: TensorStatistics;
+}>;

package/dist/checks/weights.js

@@ -0,0 +1,29 @@
+async function d(s) {
+  const e = Array.isArray(s) ? s : await s.data(), a = e.length;
+  let n = 0, l = 0, i = e[0], r = e[0], u = 0, h = !0, f = !1, m = 0;
+  for (let c = 0; c < a; c++) {
+    const t = e[c];
+    n += t, l += t * t, t < i && (i = t), t > r && (r = t), t === 0 && u++, Math.abs(t) < 1e-8 && m++, Number.isNaN(t) && (f = !0);
+  }
+  const o = n / a, y = l / a - o * o, N = Math.sqrt(y), b = u / a;
+  return {
+    mean: o,
+    std: N,
+    min: i,
+    max: r,
+    sparsity: b,
+    isFinite: h,
+    hasNaN: f,
+    closeToZeroCount: m
+  };
+}
+async function S(s) {
+  const e = s.trainableVariables, a = {};
+  for (const n of e)
+    a[n.name] = await d(n);
+  return a;
+}
+export {
+  d as createTensorStatistics,
+  S as createWeightStatistics
+};

package/dist/{concat-BmDqqFsa.js → concat-pHiVqR3L.js}

@@ -1,4 +1,4 @@
-import { B as s, n as a, D as p, F as i, E as l, H as f } from "./index-CUQrfsw_.js";
+import { C as s, n as a, F as p, H as i, E as l, I as f } from "./index-DdmHGZjq.js";
 /**
  * @license
  * Copyright 2020 Google LLC. All Rights Reserved.

package/dist/{dataset-CJmEGu6D.js → dataset-DPPl-iLT.js}

@@ -1,7 +1,7 @@
-import { ai as S, T as h, af as k, d as v, aj as o, ak as p, al as g, n as N, t as y } from "./index-CUQrfsw_.js";
+import { ak as S, T as h, ag as k, d as v, al as o, am as p, an as g, n as N, t as y } from "./index-DdmHGZjq.js";
 import { s as R } from "./index-C4L8Cm77.js";
-import { s as $ } from "./stack-DkdFLq37.js";
-import { t as B } from "./tensor-BAQdLqoU.js";
+import { s as $ } from "./stack-BNwLzE43.js";
+import { t as B } from "./tensor-DbqgIV9B.js";
 /**
  * @license
  * Copyright 2018 Google LLC. All Rights Reserved.
@@ -381,7 +381,7 @@ class i {
    *   throw an `Error`.
    */
   handleErrors(t) {
-    return new j(this, t);
+    return new Z(this, t);
   }
   // TODO(soergel): Implement reduce() etc.
   /**
@@ -708,7 +708,7 @@ class K extends i {
     return { value: r, done: !1 };
   }
 }
-class j extends i {
+class Z extends i {
   constructor(t, e) {
     super(), this.upstream = t, this.handler = e, this.count = 0, this.lastRead = Promise.resolve({ value: null, done: !1 });
   }
@@ -745,7 +745,7 @@ class w extends i {
     return { value: r, done: !1 };
   }
 }
-class Z extends i {
+class j extends i {
   constructor() {
     super(), this.outputQueue = new f(), this.lastRead = Promise.resolve({ value: null, done: !1 });
   }
@@ -759,7 +759,7 @@ class Z extends i {
     return { value: this.outputQueue.shift(), done: !1 };
   }
 }
-class tt extends Z {
+class tt extends j {
   constructor(t, e) {
     super(), this.upstream = t, this.transform = e;
   }

package/dist/{dropout-sx0sjVAT.js → dropout-CcKSfOYE.js}

@@ -1,5 +1,5 @@
-import { B as l, C as h, E as m, am as p, L as c, an as d, ae as g, n as u, T as V, o as v, q as N, a as w } from "./index-CUQrfsw_.js";
-import { s as f } from "./index-C4L8Cm77.js";
+import { C as l, D as h, E as m, ao as p, M as c, ap as f, af as g, n as u, T as V, o as v, q as N, a as w } from "./index-DdmHGZjq.js";
+import { s as d } from "./index-C4L8Cm77.js";
 /**
  * @license
  * Copyright 2018 Google LLC. All Rights Reserved.
@@ -41,7 +41,7 @@ class T {
   constructor(t, e, s, n, o) {
     this.mean = t, this.stdDev = e, this.dtype = s, this.nextVal = NaN, this.truncated = n, this.truncated && (this.upper = this.mean + this.stdDev * 2, this.lower = this.mean - this.stdDev * 2);
     const a = o || Math.random();
-    this.random = f.alea(a.toString());
+    this.random = d.alea(a.toString());
   }
   /** Returns next sample from a Gaussian distribution. */
   nextValue() {
@@ -69,11 +69,11 @@ class T {
     return t <= this.upper && t >= this.lower;
   }
 }
-class $ {
+class M {
   constructor(t = 0, e = 1, s, n) {
     if (this.canReturnFloat = () => this.dtype == null || this.dtype === "float32", this.min = t, this.range = e - t, this.dtype = s, n == null && (n = Math.random()), typeof n == "number" && (n = n.toString()), !this.canReturnFloat() && this.range <= 1)
       throw new Error(`The difference between ${t} - ${e} <= 1 and dtype is not float`);
-    this.random = f.alea(n);
+    this.random = d.alea(n);
   }
   convertValue(t) {
     return this.canReturnFloat() ? t : Math.round(t);
@@ -98,15 +98,15 @@ class $ {
  * limitations under the License.
  * =============================================================================
  */
-function y(r, t = 0, e = 1, s, n) {
+function $(r, t = 0, e = 1, s, n) {
   if (c(r), s != null && s === "bool")
     throw new Error(`Unsupported data type ${s}`);
-  const o = new T(t, e, s, !1, n), a = d(r, s);
+  const o = new T(t, e, s, !1, n), a = f(r, s);
   for (let i = 0; i < a.values.length; i++)
     a.values[i] = o.nextValue();
   return a.toTensor();
 }
-const _ = /* @__PURE__ */ l({ randomNormal_: y });
+const _ = /* @__PURE__ */ l({ randomNormal_: $ });
 /**
  * @license
  * Copyright 2020 Google LLC. All Rights Reserved.
@@ -123,14 +123,14 @@ const _ = /* @__PURE__ */ l({ randomNormal_: y });
  * limitations under the License.
  * =============================================================================
  */
-function M(r, t = 0, e = 1, s = "float32", n) {
+function y(r, t = 0, e = 1, s = "float32", n) {
   c(r);
-  const o = d(r, s), a = new $(t, e, null, n);
+  const o = f(r, s), a = new M(t, e, null, n);
   for (let i = 0; i < o.values.length; i++)
     o.values[i] = a.nextValue();
   return o.toTensor();
 }
-const D = /* @__PURE__ */ l({ randomUniform_: M });
+const D = /* @__PURE__ */ l({ randomUniform_: y });
 /**
  * @license
  * Copyright 2019 Google LLC. All Rights Reserved.

package/dist/{exports_initializers-DAKM8UO9.js → exports_initializers-DKk7-bsx.js}

@@ -1,4 +1,4 @@
-import { R as o } from "./random_width-D8Pwy_na.js";
+import { R as o } from "./random_width-DKGeiFuR.js";
 /**
  * @license
  * Copyright 2018 Google LLC

package/dist/{gather-C1siEkdp.js → gather-CPg6ZlQA.js}

@@ -1,4 +1,4 @@
-import { B as g, C as t, E as h, G as p } from "./index-CUQrfsw_.js";
+import { C as g, D as t, E as h, G as p } from "./index-DdmHGZjq.js";
 /**
  * @license
  * Copyright 2018 Google LLC. All Rights Reserved.

package/dist/{gelu-Bd3UBBxg.js → gelu-BkcmEEyD.js}

@@ -1,4 +1,4 @@
-import { i as t, e as n } from "./index-CUQrfsw_.js";
+import { i as t, e as n } from "./index-DdmHGZjq.js";
 import "./ops/cpu/gelu.js";
 import "./ops/webgl/gelu.js";
 const a = {

package/dist/{gpgpu_math-TFLxaLkw.js → gpgpu_math-D_ODOLix.js}

@@ -1,6 +1,6 @@
-import { aC as z, af as v, aD as oe, aE as ie, aF as ae, n as F, aG as me, aH as J, a2 as B, aI as I, ae as T, j as E, aJ as H, aK as $e, aL as Y, ad as Te } from "./index-CUQrfsw_.js";
+import { aD as z, ag as v, aE as oe, aF as ie, aG as ae, n as F, aH as me, aI as J, a3 as B, aJ as I, af as T, j as E, aK as H, aL as $e, aM as Y, ae as Te } from "./index-DdmHGZjq.js";
 import "./index-Tf7vU29b.js";
-import { b as Se } from "./backend_util-BJ-_jSeK.js";
+import { b as Se } from "./backend_util-yC3YH1jo.js";
 /**
  * @license
  * Copyright 2018 Google LLC. All Rights Reserved.