@genai-fi/nanogpt 0.10.1 → 0.10.3

package/dist/tensor-DdQUJZlz.js

@@ -1,909 +0,0 @@
-import { g as v } from "./index-D5v913EJ.js";
-import { p as B } from "./index-xuotMAFm.js";
-function At(t) {
-  let e = t.length, r = 0;
-  for (; e > 0; )
-    r = Math.random() * e | 0, e--, J(t, e, r);
-}
-function St(t, e, r) {
-  return Math.max(t, Math.min(e, r));
-}
-function Tt(t) {
-  return t % 2 === 0 ? t : t + 1;
-}
-function J(t, e, r) {
-  const n = t[e];
-  t[e] = t[r], t[r] = n;
-}
-function $t(t) {
-  let e = 0;
-  for (let r = 0; r < t.length; r++)
-    e += t[r];
-  return e;
-}
-function m(t, e) {
-  if (!t)
-    throw new Error(typeof e == "string" ? e : e());
-}
-function It(t, e, r = "") {
-  m(j(t, e), () => r + ` Shapes ${t} and ${e} must match`);
-}
-function Ft(t) {
-  m(t != null, () => "The input to the tensor constructor must be a non-null value.");
-}
-function R(t) {
-  if (t.length === 0)
-    return 1;
-  let e = t[0];
-  for (let r = 1; r < t.length; r++)
-    e *= t[r];
-  return e;
-}
-function Et(t) {
-  return t.length === 0;
-}
-function j(t, e) {
-  if (t === e)
-    return !0;
-  if (t == null || e == null || t.length !== e.length)
-    return !1;
-  for (let r = 0; r < t.length; r++)
-    if (t[r] !== e[r])
-      return !1;
-  return !0;
-}
-function W(t) {
-  return t % 1 === 0;
-}
-function kt(t) {
-  const e = Math.ceil(Math.sqrt(t));
-  return [e, Math.ceil(t / e)];
-}
-function X(t, e) {
-  return e <= t.length ? t : t + " ".repeat(e - t.length);
-}
-function Ut(t, e = (s) => 0, r, n) {
-  return new Promise((s, i) => {
-    let a = 0;
-    const o = () => {
-      if (t()) {
-        s();
-        return;
-      }
-      a++;
-      const u = e(a);
-      if (r != null && a >= r) {
-        i();
-        return;
-      }
-      n != null ? n(o, u) : setTimeout(o, u);
-    };
-    o();
-  });
-}
-function xt(t, e) {
-  let r = 1, n = -1;
-  for (let i = 0; i < t.length; ++i)
-    if (t[i] >= 0)
-      r *= t[i];
-    else if (t[i] === -1) {
-      if (n !== -1)
-        throw Error(`Shapes can only have 1 implicit size. Found -1 at dim ${n} and dim ${i}`);
-      n = i;
-    } else if (t[i] < 0)
-      throw Error(`Shapes can not be < 0. Found ${t[i]} at dim ${i}`);
-  if (n === -1) {
-    if (e > 0 && e !== r)
-      throw Error(`Size(${e}) must match the product of shape ${t}`);
-    return t;
-  }
-  if (r === 0)
-    throw Error(`Cannot infer the missing size in [${t}] when there are 0 elements`);
-  if (e % r !== 0)
-    throw Error(`The implicit shape can't be a fractional number. Got ${e} / ${r}`);
-  const s = t.slice();
-  return s[n] = e / r, s;
-}
-function K(t, e) {
-  const r = e.length;
-  return t = t == null ? e.map((n, s) => s) : [].concat(t), m(t.every((n) => n >= -r && n < r), () => `All values in axis param must be in range [-${r}, ${r}) but got axis ${t}`), m(t.every((n) => W(n)), () => `All values in axis param must be integers but got axis ${t}`), t.map((n) => n < 0 ? r + n : n);
-}
-function Dt(t, e) {
-  const r = [], n = [], s = e != null && Array.isArray(e) && e.length === 0, i = e == null || s ? null : K(e, t).sort();
-  let a = 0;
-  for (let o = 0; o < t.length; ++o) {
-    if (i != null) {
-      if (i[a] === o && t[o] !== 1)
-        throw new Error(`Can't squeeze axis ${o} since its dim '${t[o]}' is not 1`);
-      (i[a] == null || i[a] > o) && t[o] === 1 && (r.push(t[o]), n.push(o)), i[a] <= o && a++;
-    }
-    t[o] !== 1 && (r.push(t[o]), n.push(o));
-  }
-  return { newShape: r, keptDims: n };
-}
-function Mt(t, e) {
-  return q(t, e);
-}
-function q(t, e) {
-  let r = null;
-  if (t == null || t === "float32")
-    r = new Float32Array(e);
-  else if (t === "int32")
-    r = new Int32Array(e);
-  else if (t === "bool")
-    r = new Uint8Array(e);
-  else if (t === "string")
-    r = new Array(e);
-  else
-    throw new Error(`Unknown data type ${t}`);
-  return r;
-}
-function Y(t, e) {
-  for (let r = 0; r < t.length; r++) {
-    const n = t[r];
-    if (isNaN(n) || !isFinite(n))
-      throw Error(`A tensor of type ${e} being uploaded contains ${n}.`);
-  }
-}
-function Rt(t) {
-  return t === "bool" || t === "complex64" || t === "float32" || t === "int32" || t === "string";
-}
-function Pt(t, e) {
-  return !(e === "complex64" || e === "float32" && t !== "complex64" || e === "int32" && t !== "float32" && t !== "complex64" || e === "bool" && t === "bool");
-}
-function _t(t) {
-  if (t === "float32" || t === "int32")
-    return 4;
-  if (t === "complex64")
-    return 8;
-  if (t === "bool")
-    return 1;
-  throw new Error(`Unknown dtype ${t}`);
-}
-function Lt(t) {
-  if (t == null)
-    return 0;
-  let e = 0;
-  return t.forEach((r) => e += r.length), e;
-}
-function z(t) {
-  return typeof t == "string" || t instanceof String;
-}
-function tt(t) {
-  return typeof t == "boolean";
-}
-function et(t) {
-  return typeof t == "number";
-}
-function rt(t) {
-  return Array.isArray(t) ? rt(t[0]) : t instanceof Float32Array ? "float32" : t instanceof Int32Array || t instanceof Uint8Array || t instanceof Uint8ClampedArray ? "int32" : et(t) ? "float32" : z(t) ? "string" : tt(t) ? "bool" : "float32";
-}
-function vt(t) {
-  return !!(t && t.constructor && t.call && t.apply);
-}
-function Bt(t, e) {
-  for (let r = e; r < t; ++r)
-    if (t % r === 0)
-      return r;
-  return t;
-}
-function P(t) {
-  const e = t.length;
-  if (e < 2)
-    return [];
-  const r = new Array(e - 1);
-  r[e - 2] = t[e - 1];
-  for (let n = e - 3; n >= 0; --n)
-    r[n] = r[n + 1] * t[n + 1];
-  return r;
-}
-function N(t, e, r, n = !1) {
-  const s = new Array();
-  if (e.length === 1) {
-    const i = e[0] * (n ? 2 : 1);
-    for (let a = 0; a < i; a++)
-      s[a] = r[t + a];
-  } else {
-    const i = e[0], a = e.slice(1), o = a.reduce((u, g) => u * g) * (n ? 2 : 1);
-    for (let u = 0; u < i; u++)
-      s[u] = N(t + u * o, a, r, n);
-  }
-  return s;
-}
-function $(t, e, r = !1) {
-  if (t.length === 0)
-    return e[0];
-  const n = t.reduce((s, i) => s * i) * (r ? 2 : 1);
-  if (n === 0)
-    return [];
-  if (n !== e.length)
-    throw new Error(`[${t}] does not match the input size ${e.length}${r ? " for a complex tensor" : ""}.`);
-  return N(0, t, e, r);
-}
-function Gt(t, e) {
-  if (Array.isArray(t))
-    return t;
-  if (e === "float32")
-    return t instanceof Float32Array ? t : new Float32Array(t);
-  if (e === "int32")
-    return t instanceof Int32Array ? t : new Int32Array(t);
-  if (e === "bool" || e === "string")
-    return Uint8Array.from(new Int32Array(t));
-  throw new Error(`Unknown dtype ${e}`);
-}
-function Ot(t, e) {
-  const r = nt(t, e);
-  for (let n = 0; n < r.length; n++)
-    r[n] = 1;
-  return r;
-}
-function nt(t, e) {
-  if (e == null || e === "float32" || e === "complex64")
-    return new Float32Array(t);
-  if (e === "int32")
-    return new Int32Array(t);
-  if (e === "bool")
-    return new Uint8Array(t);
-  throw new Error(`Unknown data type ${e}`);
-}
-function Vt(t, e) {
-  const r = t.reduce((n, s) => n * s, 1);
-  if (e == null || e === "float32")
-    return $(t, new Float32Array(r));
-  if (e === "int32")
-    return $(t, new Int32Array(r));
-  if (e === "bool")
-    return $(t, new Uint8Array(r));
-  throw new Error(`Unknown data type ${e}`);
-}
-function Ct(t) {
-  t.forEach((e) => {
-    m(Number.isInteger(e) && e >= 0, () => `Tensor must have a shape comprised of positive integers but got shape [${t}].`);
-  });
-}
-function jt(t, e, r) {
-  if (e === 0)
-    return 0;
-  if (e === 1)
-    return t[0];
-  let n = t[t.length - 1];
-  for (let s = 0; s < t.length - 1; ++s)
-    n += r[s] * t[s];
-  return n;
-}
-function qt(t, e, r) {
-  if (e === 0)
-    return [];
-  if (e === 1)
-    return [t];
-  const n = new Array(e);
-  for (let s = 0; s < n.length - 1; ++s)
-    n[s] = Math.floor(t / r[s]), t -= n[s] * r[s];
-  return n[n.length - 1] = t, n;
-}
-function H(t) {
-  return t && t.then && typeof t.then == "function";
-}
-const G = "tfjsflags";
-class zt {
-  // tslint:disable-next-line: no-any
-  constructor(e) {
-    this.global = e, this.flags = {}, this.flagRegistry = {}, this.urlFlags = {}, this.getQueryParams = st, this.populateURLFlags();
-  }
-  setPlatform(e, r) {
-    this.platform != null && (y().getBool("IS_TEST") || y().getBool("PROD") || console.warn(`Platform ${this.platformName} has already been set. Overwriting the platform with ${e}.`)), this.platformName = e, this.platform = r;
-  }
-  registerFlag(e, r, n) {
-    if (this.flagRegistry[e] = { evaluationFn: r, setHook: n }, this.urlFlags[e] != null) {
-      const s = this.urlFlags[e];
-      y().getBool("IS_TEST") || y().getBool("PROD") || console.warn(`Setting feature override from URL ${e}: ${s}.`), this.set(e, s);
-    }
-  }
-  async getAsync(e) {
-    return e in this.flags ? this.flags[e] : (this.flags[e] = await this.evaluateFlag(e), this.flags[e]);
-  }
-  get(e) {
-    if (e in this.flags)
-      return this.flags[e];
-    const r = this.evaluateFlag(e);
-    if (H(r))
-      throw new Error(`Flag ${e} cannot be synchronously evaluated. Please use getAsync() instead.`);
-    return this.flags[e] = r, this.flags[e];
-  }
-  getNumber(e) {
-    return this.get(e);
-  }
-  getBool(e) {
-    return this.get(e);
-  }
-  getString(e) {
-    return this.get(e);
-  }
-  getFlags() {
-    return this.flags;
-  }
-  // For backwards compatibility.
-  get features() {
-    return this.flags;
-  }
-  set(e, r) {
-    if (this.flagRegistry[e] == null)
-      throw new Error(`Cannot set flag ${e} as it has not been registered.`);
-    this.flags[e] = r, this.flagRegistry[e].setHook != null && this.flagRegistry[e].setHook(r);
-  }
-  evaluateFlag(e) {
-    if (this.flagRegistry[e] == null)
-      throw new Error(`Cannot evaluate flag '${e}': no evaluation function found.`);
-    return this.flagRegistry[e].evaluationFn();
-  }
-  setFlags(e) {
-    this.flags = Object.assign({}, e);
-  }
-  reset() {
-    this.flags = {}, this.urlFlags = {}, this.populateURLFlags();
-  }
-  populateURLFlags() {
-    if (typeof this.global > "u" || typeof this.global.location > "u" || typeof this.global.location.search > "u")
-      return;
-    const e = this.getQueryParams(this.global.location.search);
-    G in e && e[G].split(",").forEach((n) => {
-      const [s, i] = n.split(":");
-      this.urlFlags[s] = ot(s, i);
-    });
-  }
-}
-function st(t) {
-  const e = {};
-  return t.replace(/[?&]([^=?&]+)(?:=([^&]*))?/g, (r, ...n) => (it(e, n[0], n[1]), n.join("="))), e;
-}
-function it(t, e, r) {
-  t[decodeURIComponent(e)] = decodeURIComponent(r || "");
-}
-function ot(t, e) {
-  const r = e.toLowerCase();
-  return r === "true" || r === "false" ? r === "true" : `${+r}` === r ? +r : e;
-}
-function y() {
-  return Q;
-}
-let Q = null;
-function Nt(t) {
-  Q = t;
-}
-let x;
-function at() {
-  if (x == null) {
-    let t;
-    if (typeof window < "u")
-      t = window;
-    else if (typeof v < "u")
-      t = v;
-    else if (typeof B < "u")
-      t = B;
-    else if (typeof self < "u")
-      t = self;
-    else
-      throw new Error("Could not find a global object");
-    x = t;
-  }
-  return x;
-}
-function lt() {
-  const t = at();
-  return t._tfGlobals == null && (t._tfGlobals = /* @__PURE__ */ new Map()), t._tfGlobals;
-}
-function ut(t, e) {
-  const r = lt();
-  if (r.has(t))
-    return r.get(t);
-  {
-    const n = e();
-    return r.set(t, n), r.get(t);
-  }
-}
-function ft(t) {
-  return t instanceof Float32Array || t instanceof Int32Array || t instanceof Uint8Array || t instanceof Uint8ClampedArray;
-}
-function Ht(t, e) {
-  return e === "string" ? gt(t) : ct([t], e);
-}
-function ht(t, e) {
-  return t instanceof Float32Array && e === "float32" || t instanceof Int32Array && e === "int32" || t instanceof Uint8Array && e === "bool";
-}
-function ct(t, e) {
-  if (e === "string")
-    throw new Error("Cannot convert a string[] to a TypedArray");
-  if (Array.isArray(t) && (t = M(t)), y().getBool("DEBUG") && Y(t, e), ht(t, e))
-    return t;
-  if (e == null || e === "float32" || e === "complex64")
-    return new Float32Array(t);
-  if (e === "int32")
-    return new Int32Array(t);
-  if (e === "bool") {
-    const r = new Uint8Array(t.length);
-    for (let n = 0; n < r.length; ++n)
-      Math.round(t[n]) !== 0 && (r[n] = 1);
-    return r;
-  } else
-    throw new Error(`Unknown data type ${e}`);
-}
-function Qt() {
-  return y().platform.now();
-}
-function gt(t, e = "utf-8") {
-  return e = e || "utf-8", y().platform.encode(t, e);
-}
-function O(t, e = "utf-8") {
-  return e = e || "utf-8", y().platform.decode(t, e);
-}
-function V(t) {
-  return y().platform.isTypedArray != null ? y().platform.isTypedArray(t) : ft(t);
-}
-function M(t, e = [], r = !1) {
-  if (e == null && (e = []), typeof t == "boolean" || typeof t == "number" || typeof t == "string" || H(t) || t == null || V(t) && r)
-    e.push(t);
-  else if (Array.isArray(t) || V(t))
-    for (let n = 0; n < t.length; ++n)
-      M(t[n], e, r);
-  else {
-    let n = -1;
-    for (const s of Object.keys(t))
-      /^([1-9]+[0-9]*|0)$/.test(s) && (n = Math.max(n, Number(s)));
-    for (let s = 0; s <= n; s++)
-      M(t[s], e, r);
-  }
-  return e;
-}
-const C = 20, A = 3, D = 7;
-function dt(t, e, r, n) {
-  const s = P(e), i = pt(t, e, r, s), a = e.length, o = I(t, e, r, s, i), u = ["Tensor"];
-  return n && (u.push(`  dtype: ${r}`), u.push(`  rank: ${a}`), u.push(`  shape: [${e}]`), u.push("  values:")), u.push(o.map((g) => "    " + g).join(`
-`)), u.join(`
-`);
-}
-function pt(t, e, r, n) {
-  const s = R(e), i = n[n.length - 1], a = new Array(i).fill(0), o = e.length, u = r === "complex64" ? T(t) : t;
-  if (o > 1)
-    for (let g = 0; g < s / i; g++) {
-      const b = g * i;
-      for (let h = 0; h < i; h++)
-        a[h] = Math.max(a[h], S(u[b + h], 0, r).length);
-    }
-  return a;
-}
-function S(t, e, r) {
-  let n;
-  return Array.isArray(t) ? n = `${parseFloat(t[0].toFixed(D))} + ${parseFloat(t[1].toFixed(D))}j` : z(t) ? n = `'${t}'` : r === "bool" ? n = Z(t) : n = parseFloat(t.toFixed(D)).toString(), X(n, e);
-}
-function Z(t) {
-  return t === 0 ? "false" : "true";
-}
-function I(t, e, r, n, s, i = !0) {
-  const a = r === "complex64" ? 2 : 1, o = e[0], u = e.length;
-  if (u === 0) {
-    if (r === "complex64") {
-      const l = T(t);
-      return [S(l[0], 0, r)];
-    }
-    return r === "bool" ? [Z(t[0])] : [t[0].toString()];
-  }
-  if (u === 1) {
-    if (o > C) {
-      const c = A * a;
-      let d = Array.from(t.slice(0, c)), E = Array.from(t.slice((o - A) * a, o * a));
-      return r === "complex64" && (d = T(d), E = T(E)), [
-        "[" + d.map((k, U) => S(k, s[U], r)).join(", ") + ", ..., " + E.map((k, U) => S(k, s[o - A + U], r)).join(", ") + "]"
-      ];
-    }
-    return [
-      "[" + (r === "complex64" ? T(t) : Array.from(t)).map((c, d) => S(c, s[d], r)).join(", ") + "]"
-    ];
-  }
-  const g = e.slice(1), b = n.slice(1), h = n[0] * a, f = [];
-  if (o > C) {
-    for (let l = 0; l < A; l++) {
-      const c = l * h, d = c + h;
-      f.push(...I(
-        t.slice(c, d),
-        g,
-        r,
-        b,
-        s,
-        !1
-        /* isLast */
-      ));
-    }
-    f.push("...");
-    for (let l = o - A; l < o; l++) {
-      const c = l * h, d = c + h;
-      f.push(...I(
-        t.slice(c, d),
-        g,
-        r,
-        b,
-        s,
-        l === o - 1
-        /* isLast */
-      ));
-    }
-  } else
-    for (let l = 0; l < o; l++) {
-      const c = l * h, d = c + h;
-      f.push(...I(
-        t.slice(c, d),
-        g,
-        r,
-        b,
-        s,
-        l === o - 1
-        /* isLast */
-      ));
-    }
-  const _ = u === 2 ? "," : "";
-  f[0] = "[" + (o > 0 ? f[0] + _ : "");
-  for (let l = 1; l < f.length - 1; l++)
-    f[l] = " " + f[l] + _;
-  let L = `,
-`;
-  for (let l = 2; l < u; l++)
-    L += `
-`;
-  return f[f.length - 1] = " " + f[f.length - 1] + "]" + (i ? "" : L), f;
-}
-function T(t) {
-  const e = [];
-  for (let r = 0; r < t.length; r += 2)
-    e.push([t[r], t[r + 1]]);
-  return e;
-}
-class Zt {
-  constructor(e, r, n) {
-    if (this.dtype = r, this.shape = e.slice(), this.size = R(e), n != null) {
-      const s = n.length;
-      m(s === this.size, () => `Length of values '${s}' does not match the size inferred by the shape '${this.size}'.`);
-    }
-    if (r === "complex64")
-      throw new Error("complex64 dtype TensorBuffers are not supported. Please create a TensorBuffer for the real and imaginary parts separately and call tf.complex(real, imag).");
-    this.values = n || q(r, this.size), this.strides = P(e);
-  }
-  /**
-   * Sets a value in the buffer at a given location.
-   *
-   * @param value The value to set.
-   * @param locs  The location indices.
-   *
-   * @doc {heading: 'Tensors', subheading: 'Creation'}
-   */
-  set(e, ...r) {
-    r.length === 0 && (r = [0]), m(r.length === this.rank, () => `The number of provided coordinates (${r.length}) must match the rank (${this.rank})`);
-    const n = this.locToIndex(r);
-    this.values[n] = e;
-  }
-  /**
-   * Returns the value in the buffer at the provided location.
-   *
-   * @param locs The location indices.
-   *
-   * @doc {heading: 'Tensors', subheading: 'Creation'}
-   */
-  get(...e) {
-    e.length === 0 && (e = [0]);
-    let r = 0;
-    for (const s of e) {
-      if (s < 0 || s >= this.shape[r]) {
-        const i = `Requested out of range element at ${e}.   Buffer shape=${this.shape}`;
-        throw new Error(i);
-      }
-      r++;
-    }
-    let n = e[e.length - 1];
-    for (let s = 0; s < e.length - 1; ++s)
-      n += this.strides[s] * e[s];
-    return this.values[n];
-  }
-  locToIndex(e) {
-    if (this.rank === 0)
-      return 0;
-    if (this.rank === 1)
-      return e[0];
-    let r = e[e.length - 1];
-    for (let n = 0; n < e.length - 1; ++n)
-      r += this.strides[n] * e[n];
-    return r;
-  }
-  indexToLoc(e) {
-    if (this.rank === 0)
-      return [];
-    if (this.rank === 1)
-      return [e];
-    const r = new Array(this.shape.length);
-    for (let n = 0; n < r.length - 1; ++n)
-      r[n] = Math.floor(e / this.strides[n]), e -= r[n] * this.strides[n];
-    return r[r.length - 1] = e, r;
-  }
-  get rank() {
-    return this.shape.length;
-  }
-  /**
-   * Creates an immutable `tf.Tensor` object from the buffer.
-   *
-   * @doc {heading: 'Tensors', subheading: 'Creation'}
-   */
-  toTensor() {
-    return p().makeTensor(this.values, this.shape, this.dtype);
-  }
-}
-let p = null, w = null;
-function Jt(t) {
-  p = t;
-}
-function Wt(t) {
-  w = t;
-}
-class F {
-  constructor(e, r, n, s) {
-    this.kept = !1, this.isDisposedInternal = !1, this.shape = e.slice(), this.dtype = r || "float32", this.size = R(e), this.strides = P(e), this.dataId = n, this.id = s, this.rankType = this.rank < 5 ? this.rank.toString() : "higher";
-  }
-  get rank() {
-    return this.shape.length;
-  }
-  /**
-   * Returns a promise of `tf.TensorBuffer` that holds the underlying data.
-   *
-   * @doc {heading: 'Tensors', subheading: 'Classes'}
-   */
-  async buffer() {
-    const e = await this.data();
-    return w.buffer(this.shape, this.dtype, e);
-  }
-  /**
-   * Returns a `tf.TensorBuffer` that holds the underlying data.
-   * @doc {heading: 'Tensors', subheading: 'Classes'}
-   */
-  bufferSync() {
-    return w.buffer(this.shape, this.dtype, this.dataSync());
-  }
-  /**
-   * Returns the tensor data as a nested array. The transfer of data is done
-   * asynchronously.
-   *
-   * @doc {heading: 'Tensors', subheading: 'Classes'}
-   */
-  async array() {
-    const e = await this.data();
-    return $(this.shape, e, this.dtype === "complex64");
-  }
-  /**
-   * Returns the tensor data as a nested array. The transfer of data is done
-   * synchronously.
-   *
-   * @doc {heading: 'Tensors', subheading: 'Classes'}
-   */
-  arraySync() {
-    return $(this.shape, this.dataSync(), this.dtype === "complex64");
-  }
-  /**
-   * Asynchronously downloads the values from the `tf.Tensor`. Returns a
-   * promise of `TypedArray` that resolves when the computation has finished.
-   *
-   * @doc {heading: 'Tensors', subheading: 'Classes'}
-   */
-  async data() {
-    this.throwIfDisposed();
-    const e = p().read(this.dataId);
-    if (this.dtype === "string") {
-      const r = await e;
-      try {
-        return r.map((n) => O(n));
-      } catch {
-        throw new Error("Failed to decode the string bytes into utf-8. To get the original bytes, call tensor.bytes().");
-      }
-    }
-    return e;
-  }
-  /**
-   * Copy the tensor's data to a new GPU resource. Comparing to the `dataSync()`
-   * and `data()`, this method prevents data from being downloaded to CPU.
-   *
-   * For WebGL backend, the data will be stored on a densely packed texture.
-   * This means that the texture will use the RGBA channels to store value.
-   *
-   * For WebGPU backend, the data will be stored on a buffer. There is no
-   * parameter, so can not use a user-defined size to create the buffer.
-   *
-   * @param options:
-   *     For WebGL,
-   *         - customTexShape: Optional. If set, will use the user defined
-   *     texture shape to create the texture.
-   *
-   * @returns For WebGL backend, a GPUData contains the new texture and
-   *     its information.
-   *     {
-   *        tensorRef: The tensor that is associated with this texture,
-   *        texture: WebGLTexture,
-   *        texShape: [number, number] // [height, width]
-   *     }
-   *
-   *     For WebGPU backend, a GPUData contains the new buffer.
-   *     {
-   *        tensorRef: The tensor that is associated with this buffer,
-   *        buffer: GPUBuffer,
-   *     }
-   *
-   *     Remember to dispose the GPUData after it is used by
-   *     `res.tensorRef.dispose()`.
-   *
-   * @doc {heading: 'Tensors', subheading: 'Classes'}
-   */
-  dataToGPU(e) {
-    return this.throwIfDisposed(), p().readToGPU(this.dataId, e);
-  }
-  /**
-   * Synchronously downloads the values from the `tf.Tensor`. This blocks the
-   * UI thread until the values are ready, which can cause performance issues.
-   *
-   * @doc {heading: 'Tensors', subheading: 'Classes'}
-   */
-  dataSync() {
-    this.throwIfDisposed();
-    const e = p().readSync(this.dataId);
-    if (this.dtype === "string")
-      try {
-        return e.map((r) => O(r));
-      } catch {
-        throw new Error("Failed to decode the string bytes into utf-8. To get the original bytes, call tensor.bytes().");
-      }
-    return e;
-  }
-  /** Returns the underlying bytes of the tensor's data. */
-  async bytes() {
-    this.throwIfDisposed();
-    const e = await p().read(this.dataId);
-    return this.dtype === "string" ? e : new Uint8Array(e.buffer);
-  }
-  /**
-   * Disposes `tf.Tensor` from memory.
-   *
-   * @doc {heading: 'Tensors', subheading: 'Classes'}
-   */
-  dispose() {
-    this.isDisposed || (this.kerasMask && this.kerasMask.dispose(), p().disposeTensor(this), this.isDisposedInternal = !0);
-  }
-  get isDisposed() {
-    return this.isDisposedInternal;
-  }
-  throwIfDisposed() {
-    if (this.isDisposed)
-      throw new Error("Tensor is disposed.");
-  }
-  /**
-   * Prints the `tf.Tensor`. See `tf.print` for details.
-   *
-   * @param verbose Whether to print verbose information about the tensor,
-   *    including dtype and size.
-   *
-   * @doc {heading: 'Tensors', subheading: 'Classes'}
-   */
-  print(e = !1) {
-    return w.print(this, e);
-  }
-  /**
-   * Returns a copy of the tensor. See `tf.clone` for details.
-   * @doc {heading: 'Tensors', subheading: 'Classes'}
-   */
-  clone() {
-    return this.throwIfDisposed(), w.clone(this);
-  }
-  /**
-   * Returns a human-readable description of the tensor. Useful for logging.
-   *
-   * @doc {heading: 'Tensors', subheading: 'Classes'}
-   */
-  toString(e = !1) {
-    const r = this.dataSync();
-    return dt(r, this.shape, this.dtype, e);
-  }
-  cast(e) {
-    return this.throwIfDisposed(), w.cast(this, e);
-  }
-  variable(e = !0, r, n) {
-    return this.throwIfDisposed(), p().makeVariable(this, e, r, n);
-  }
-}
-Object.defineProperty(F, Symbol.hasInstance, {
-  value: (t) => !!t && t.data != null && t.dataSync != null && t.throwIfDisposed != null
-});
-function yt() {
-  return ut("Tensor", () => F);
-}
-yt();
-class mt extends F {
-  constructor(e, r, n, s) {
-    super(e.shape, e.dtype, e.dataId, s), this.trainable = r, this.name = n;
-  }
-  /**
-   * Assign a new `tf.Tensor` to this variable. The new `tf.Tensor` must have
-   * the same shape and dtype as the old `tf.Tensor`.
-   *
-   * @param newValue New tensor to be assigned to this variable.
-   *
-   * @doc {heading: 'Tensors', subheading: 'Classes'}
-   */
-  assign(e) {
-    if (e.dtype !== this.dtype)
-      throw new Error(`dtype of the new value (${e.dtype}) and previous value (${this.dtype}) must match`);
-    if (!j(e.shape, this.shape))
-      throw new Error(`shape of the new value (${e.shape}) and previous value (${this.shape}) must match`);
-    p().disposeTensor(this), this.dataId = e.dataId, p().incRef(
-      this,
-      null
-      /* backend */
-    );
-  }
-  dispose() {
-    p().disposeVariable(this), this.isDisposedInternal = !0;
-  }
-}
-Object.defineProperty(mt, Symbol.hasInstance, {
-  value: (t) => t instanceof F && t.assign != null && t.assign instanceof Function
-});
-export {
-  et as $,
-  O as A,
-  M as B,
-  $t as C,
-  V as D,
-  zt as E,
-  ut as F,
-  X as G,
-  W as H,
-  Ht as I,
-  kt as J,
-  Tt as K,
-  Bt as L,
-  yt as M,
-  rt as N,
-  ct as O,
-  H as P,
-  ft as Q,
-  Zt as R,
-  Rt as S,
-  F as T,
-  Wt as U,
-  mt as V,
-  qt as W,
-  jt as X,
-  Pt as Y,
-  q as Z,
-  J as _,
-  m as a,
-  At as a0,
-  $ as a1,
-  Vt as a2,
-  Ut as a3,
-  j as b,
-  Ft as c,
-  Ct as d,
-  y as e,
-  nt as f,
-  Mt as g,
-  Dt as h,
-  P as i,
-  It as j,
-  St as k,
-  Et as l,
-  Ot as m,
-  xt as n,
-  at as o,
-  K as p,
-  Nt as q,
-  z as r,
-  R as s,
-  gt as t,
-  Lt as u,
-  _t as v,
-  vt as w,
-  Qt as x,
-  Jt as y,
-  Gt as z
-};