@genai-fi/nanogpt 0.3.2 → 0.4.1

package/dist/training/AdamExt.js

@@ -1,4 +1,4 @@
-import { A as r, b as c, f as h, s as g, e as o } from "../index-pWA4_lUh.js";
+import { A as r, b as c, f as h, s as g, e as o } from "../index-XjBAhiFO.js";
 class u extends r {
   constructor(t, e, s, a, i) {
     super(t, e, s, a), this.config = i, this.startLearningRate = t;

package/dist/training/DatasetBuilder.js

@@ -1,7 +1,7 @@
-import { a7 as $, a8 as m, a9 as M, a as R, aa as f, ab as v, ac as z, j as _, t as x } from "../index-pWA4_lUh.js";
+import { a9 as $, aa as m, ab as M, a as R, ac as f, ad as v, ae as z, j as _, t as x } from "../index-XjBAhiFO.js";
 import { s as E } from "../index-C4L8Cm77.js";
-import { s as P } from "../stack-BMm-efee.js";
-import { t as D } from "../tensor-DJVbYhh1.js";
+import { s as P } from "../stack-DaIMO5iX.js";
+import { t as D } from "../tensor-Cu1fU7H7.js";
 import "../index-Tf7vU29b.js";
 /**
  * @license

package/dist/training/FullTrainer.js

@@ -1,7 +1,7 @@
 import { generateText as v } from "../utilities/generate.js";
 import L from "./Trainer.js";
 import x from "./Evaluator.js";
-import { a as h } from "../index-pWA4_lUh.js";
+import { a as h } from "../index-XjBAhiFO.js";
 const D = {
   desiredLoss: 0.01,
   logInterval: 1,

package/dist/training/Trainer.js

@@ -1,12 +1,13 @@
-import { DatasetBuilder as p } from "./DatasetBuilder.js";
+import { DatasetBuilder as d } from "./DatasetBuilder.js";
 import h from "./AdamExt.js";
-import { t as g, v as u, a as o } from "../index-pWA4_lUh.js";
-import { m as y, a as f, b as S, n as z } from "../moments-DfcpfwKi.js";
-import { m as b } from "../sum-C7Mgy9Bw.js";
-import { z as n } from "../zeros-CCy9C3uU.js";
-class M {
+import { t as g, v as u, a as o } from "../index-XjBAhiFO.js";
+import { m as y, n as f } from "../norm-BzY929B_.js";
+import { m as S, a as z } from "../moments-CrWRPcR3.js";
+import { m as b } from "../max-DJvEiCAJ.js";
+import { z as n } from "../zeros-CmJFiC84.js";
+class G {
   constructor(t, s, e = 1e-3) {
-    this.tokenizer = s, this.model = t, this.learningRate = e, this.resetOptimizer(), this.datasetBuilder = new p(s, t.config.blockSize);
+    this.tokenizer = s, this.model = t, this.learningRate = e, this.resetOptimizer(), this.datasetBuilder = new d(s, t.config.gpt.blockSize);
   }
   model;
   optimizer;
@@ -45,21 +46,21 @@ class M {
   printGradients(t) {
     Object.keys(t).forEach((s) => {
       const e = t[s];
-      console.log(`${s}:`), console.log(`  Shape: ${e.shape}`), console.log(`  Mean: ${y(e).dataSync()[0]}`), console.log(`  Std: ${f(e).variance.sqrt().dataSync()[0]}`), console.log(`  Min: ${S(e).dataSync()[0]}`), console.log(`  Max: ${b(e).dataSync()[0]}`), console.log(`  Norm: ${z(e).dataSync()[0]}`);
+      console.log(`${s}:`), console.log(`  Shape: ${e.shape}`), console.log(`  Mean: ${S(e).dataSync()[0]}`), console.log(`  Std: ${z(e).variance.sqrt().dataSync()[0]}`), console.log(`  Min: ${y(e).dataSync()[0]}`), console.log(`  Max: ${b(e).dataSync()[0]}`), console.log(`  Norm: ${f(e).dataSync()[0]}`);
     });
   }
   trainStep(t, s = !1, e = !1) {
     return g(() => {
       this.model.getProfiler()?.startMemory();
       const { xs: a, ys: r } = t, l = () => {
-        const { loss: d, logits: m } = this.model.forward(a, r, !0);
-        return m.dispose(), d;
+        const { loss: m, logits: p } = this.model.forward(a, r, !0);
+        return p.dispose(), m;
       }, { value: c, grads: i } = u(l);
       return s ? this.model.getProfiler()?.endMemory("Training") : (e && (console.log("-------"), this.printGradients(i), console.log("-------")), this.optimizer.applyGradients(i), this.model.getProfiler()?.endMemory("Training"), o(i)), c;
     });
   }
   dummyPass() {
-    const t = n([1, this.model.config.blockSize], "int32"), s = n([1, this.model.config.blockSize], "int32");
+    const t = n([1, this.model.config.gpt.blockSize], "int32"), s = n([1, this.model.config.gpt.blockSize], "int32");
     try {
       const e = this.trainStep({ xs: t, ys: s }, !0);
       e.dataSync(), e.dispose();
@@ -94,5 +95,5 @@ class M {
   }
 }
 export {
-  M as default
+  G as default
 };

package/dist/training/sparseCrossEntropy.js

@@ -1,26 +1,27 @@
 import { gatherSub as L } from "../ops/gatherSub.js";
 import { scatterSub as y } from "../ops/scatterSub.js";
-import { e as u, c as i, z as S, t as f, s as G } from "../index-pWA4_lUh.js";
-import { s as v, l as z } from "../softmax-Be_lsqUc.js";
-import { m as k } from "../sum-C7Mgy9Bw.js";
+import { e as u, c as i, z as S, t as f, s as G } from "../index-XjBAhiFO.js";
+import { s as v } from "../softmax-4DOn6cPq.js";
+import { m as z } from "../max-DJvEiCAJ.js";
+import { l as k } from "../log_sum_exp-DJPkVZZn.js";
 function F(a, s) {
   return f(() => {
-    const e = a.shape[a.shape.length - 1], n = a.shape.slice(0, -1).reduce((d, c) => d * c, 1), p = a.shape.length > 2 ? a.reshape([n, e]) : a, o = s.shape.length > 1 ? s.reshape([n]).cast("int32") : s.cast("int32"), t = k(p, -1, !0), r = G(p, t), h = z(r, -1);
-    return L(h, o, r);
+    const e = a.shape[a.shape.length - 1], o = a.shape.slice(0, -1).reduce((d, c) => d * c, 1), p = a.shape.length > 2 ? a.reshape([o, e]) : a, n = s.shape.length > 1 ? s.reshape([o]).cast("int32") : s.cast("int32"), t = z(p, -1, !0), r = G(p, t), h = k(r, -1);
+    return L(h, n, r);
   });
 }
-function $() {
+function j() {
   return u().backendName === "tensorflow" ? i((s, e, m) => {
-    const n = s.shape.length > 2 ? s.reshape([-1, s.shape[s.shape.length - 1]]) : s, p = e.shape.length > 1 ? e.reshape([-1]).cast("int32") : e.cast("int32"), [o, t] = u().runKernel(
+    const o = s.shape.length > 2 ? s.reshape([-1, s.shape[s.shape.length - 1]]) : s, p = e.shape.length > 1 ? e.reshape([-1]).cast("int32") : e.cast("int32"), [n, t] = u().runKernel(
       "NativeSparseSoftmaxCrossEntropy",
-      { logits: n, labels: p },
+      { logits: o, labels: p },
       {}
     );
-    return m([t.reshape(s.shape)]), { value: o, gradFunc: (r, h) => [h[0], S(e)] };
+    return m([t.reshape(s.shape)]), { value: n, gradFunc: (r, h) => [h[0], S(e)] };
   }) : i(
     // @ts-expect-error Invalid params
     (s, e, m) => {
-      const n = s.shape[s.shape.length - 1], o = s.shape.slice(0, -1).reduce((d, c) => d * c, 1), t = s.reshape([o, n]), r = e.reshape([o]).cast("int32"), h = F(t, r);
+      const o = s.shape[s.shape.length - 1], n = s.shape.slice(0, -1).reduce((d, c) => d * c, 1), t = s.reshape([n, o]), r = e.reshape([n]).cast("int32"), h = F(t, r);
       return m([t, r]), t.dispose(), r.dispose(), { value: h, gradFunc: (d, c) => f(() => {
         const l = c[0], b = c[1], x = v(l), C = y(x, b, d), E = S(e);
         return [C.reshape(s.shape), E];
@@ -29,6 +30,6 @@ function $() {
   );
 }
 export {
-  $ as createSoftmaxCrossEntropyWithGrad,
+  j as createSoftmaxCrossEntropyWithGrad,
   F as sparseSoftmaxCrossEntropy
 };

package/dist/utilities/dummy.js

@@ -1,14 +1,14 @@
-import "../index-pWA4_lUh.js";
-import { z as n } from "../zeros-CCy9C3uU.js";
-async function c(s) {
-  const o = n([1, s.config.blockSize], "int32"), { logits: t, loss: i } = s.forward(o, void 0, !1);
+import "../index-XjBAhiFO.js";
+import { z as n } from "../zeros-CmJFiC84.js";
+async function a(s) {
+  const o = n([1, s.config.gpt.blockSize], "int32"), { logits: t, loss: i } = s.forward(o, void 0, !1);
   await t.data(), t.dispose(), i && i.dispose(), o.dispose();
 }
-function f(s) {
-  const o = n([1, s.config.blockSize], "int32"), { logits: t, loss: i } = s.forward(o, void 0, !1);
+function c(s) {
+  const o = n([1, s.config.gpt.blockSize], "int32"), { logits: t, loss: i } = s.forward(o, void 0, !1);
   t.dispose(), i && i.dispose(), o.dispose();
 }
 export {
-  f as dummyPass,
-  c as dummyPassAsync
+  c as dummyPass,
+  a as dummyPassAsync
 };

package/dist/utilities/generate.js

@@ -1,20 +1,20 @@
-import { t as y } from "../index-pWA4_lUh.js";
-import { t as x } from "../tensor2d-ZuQSh2D-.js";
-import { c as g } from "../concat-BIZS_td9.js";
+import { t as y } from "../index-XjBAhiFO.js";
+import { t as x } from "../tensor2d-D0CKdG6B.js";
+import { c as f } from "../concat-DS_qH7MI.js";
 async function A(o, r, a, c, T) {
   if (c <= 0)
     throw new Error("Length must be a positive integer");
   if (a.length === 0)
     throw new Error("Prompt cannot be an empty string");
-  const d = await o.tokenise([a], !0), s = r.config.useRope ? new Array(r.config.nLayer).fill(void 0) : void 0, p = y(() => {
-    let e = x(d, [1, d[0].length], "int32"), t = e;
-    for (let f = 0; f < c; f++) {
-      const { output: n } = r.generate(e, s, T), l = e, m = t;
-      t = g([t, n], 1), e = s ? n : g([e, n], 1), l.dispose(), m.dispose(), s || n.dispose();
+  const p = await o.tokenise([a], !0), s = r.config.gpt.useRope ? new Array(r.config.gpt.nLayer).fill(void 0) : void 0, d = y(() => {
+    let t = x(p, [1, p[0].length], "int32"), e = t;
+    for (let g = 0; g < c; g++) {
+      const { output: n } = r.generate(t, s, T), l = t, m = e;
+      e = f([e, n], 1), t = s ? n : f([t, n], 1), l.dispose(), m.dispose(), s || n.dispose();
     }
-    return t;
-  }), w = await p.array();
-  p.dispose();
+    return e;
+  }), w = await d.array();
+  d.dispose();
   const i = w[0], u = i.indexOf(o.eosToken);
   return u !== -1 && i.splice(u), await o.decode(i);
 }

package/dist/utilities/load.js

@@ -3,7 +3,7 @@ import { importWeights as b } from "./weights.js";
 import u from "../tokeniser/CharTokeniser.js";
 import F from "../NanoGPTModel.js";
 import { dummyPassAsync as j } from "./dummy.js";
-import { d as T } from "../index-pWA4_lUh.js";
+import { d as T } from "../index-XjBAhiFO.js";
 import E from "../tokeniser/bpe.js";
 async function A(t) {
   const o = await fetch(t);

package/dist/utilities/profile.js

@@ -1,4 +1,4 @@
-import { m as s } from "../index-pWA4_lUh.js";
+import { m as s } from "../index-XjBAhiFO.js";
 const m = 1024 * 1024;
 class i {
   log = /* @__PURE__ */ new Map();

package/dist/utilities/weights.js

@@ -1,5 +1,5 @@
-import "../index-pWA4_lUh.js";
-import { t as p } from "../tensor-DJVbYhh1.js";
+import "../index-XjBAhiFO.js";
+import { t as p } from "../tensor-Cu1fU7H7.js";
 function h(n) {
   const e = n.reduce((s, o) => s + o.length, 0), a = new Float32Array(e);
   let t = 0;

package/dist/{variable-Dl_ub3pk.js → variable-BS4AKqNU.js}

@@ -1,4 +1,4 @@
-import { E as i } from "./index-pWA4_lUh.js";
+import { E as i } from "./index-XjBAhiFO.js";
 /**
  * @license
  * Copyright 2018 Google LLC. All Rights Reserved.

package/dist/{zeros-CCy9C3uU.js → zeros-CmJFiC84.js}

@@ -1,4 +1,4 @@
-import { o as m, h as r, M as l, E as c, N as i, k as p, O as u, n as f } from "./index-pWA4_lUh.js";
+import { o as m, h as r, U as l, E as c, V as i, k as p, W as u, n as f } from "./index-XjBAhiFO.js";
 /**
  * @license
  * Copyright 2020 Google LLC. All Rights Reserved.

package/package.json

@@ -1,6 +1,6 @@
 {
     "name": "@genai-fi/nanogpt",
-    "version": "0.3.2",
+    "version": "0.4.1",
     "type": "module",
     "main": "dist/main.js",
     "types": "dist/main.d.ts",

package/dist/exports_layers-tbTBcwMM.js

@@ -1,25 +0,0 @@
-import { E as e, D as r, a as o } from "./random_width-oeUIlUZj.js";
-import "./index-pWA4_lUh.js";
-/**
- * @license
- * Copyright 2018 Google LLC
- *
- * Use of this source code is governed by an MIT-style
- * license that can be found in the LICENSE file or at
- * https://opensource.org/licenses/MIT.
- * =============================================================================
- */
-function s(n) {
-  return new o(n);
-}
-function u(n) {
-  return new r(n);
-}
-function d(n) {
-  return new e(n);
-}
-export {
-  u as a,
-  s as d,
-  d as e
-};

package/dist/layers/LayerNorm.d.ts

@@ -1,13 +0,0 @@
-import { default as TF } from '@tensorflow/tfjs';
-export default class LayerNorm {
-    private gamma;
-    private epsilon;
-    private tf;
-    constructor(tf: typeof TF, shape: number[], epsilon?: number, name?: string);
-    get trainableWeights(): TF.Variable[];
-    set trainable(value: boolean);
-    getWeights(): TF.Tensor[];
-    setWeights(weights: TF.Tensor[]): void;
-    apply(x: TF.Tensor): TF.Tensor;
-    dispose(): void;
-}

package/dist/layers/LayerNorm.js

@@ -1,33 +0,0 @@
-class h {
-  gamma;
-  //private beta: TF.Variable;
-  epsilon;
-  tf;
-  constructor(a, s, t = 1e-5, e = "") {
-    this.tf = a, this.epsilon = t, this.gamma = a.variable(a.ones(s), !0, `${e}_gamma`, "float32");
-  }
-  get trainableWeights() {
-    return [this.gamma];
-  }
-  set trainable(a) {
-    this.gamma.trainable = a;
-  }
-  getWeights() {
-    return [this.gamma];
-  }
-  setWeights(a) {
-    this.gamma.assign(a[0]);
-  }
-  apply(a) {
-    return this.tf.tidy(() => {
-      const s = a.mean(-1, !0), t = a.sub(s), i = t.square().mean(-1, !0).add(this.epsilon).rsqrt();
-      return t.mul(i).mul(this.gamma);
-    });
-  }
-  dispose() {
-    this.gamma.dispose();
-  }
-}
-export {
-  h as default
-};

package/dist/moments-DfcpfwKi.js

@@ -1,132 +0,0 @@
-import { o as i, h as m, E as y, a2 as A, a0 as E, a3 as o, a4 as h, J as b, f as S, a5 as v, a6 as _, s as k, U as q } from "./index-pWA4_lUh.js";
-import { e as w, s as u, m as l } from "./sum-C7Mgy9Bw.js";
-import { r as c } from "./reshape-C8CR_Bad.js";
-/**
- * @license
- * Copyright 2020 Google Inc. 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 K(r, e = null, n = !1) {
-  const t = { x: m(r, "x", "min") }, a = { axis: e, keepDims: n };
-  return y.runKernel(A, t, a);
-}
-const d = /* @__PURE__ */ i({ min_: K });
-/**
- * @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(r, e = "euclidean", n = null, s = !1) {
-  r = m(r, "x", "norm");
-  const t = $(r, e, n);
-  let a = t.shape;
-  if (s) {
-    const f = E(n, r.shape);
-    a = w(t.shape, f);
-  }
-  return c(t, a);
-}
-function $(r, e, n = null) {
-  if (r.rank === 0)
-    return o(r);
-  if (r.rank !== 1 && n === null)
-    return $(c(r, [-1]), e, n);
-  if (r.rank === 1 || typeof n == "number" || Array.isArray(n) && n.length === 1) {
-    if (e === 1)
-      return u(o(r), n);
-    if (e === 1 / 0)
-      return l(o(r), n);
-    if (e === -1 / 0)
-      return d(o(r), n);
-    if (e === "euclidean" || e === 2)
-      return h(u(b(o(r), S(2, "int32")), n));
-    throw new Error(`Error in norm: invalid ord value: ${e}`);
-  }
-  if (Array.isArray(n) && n.length === 2) {
-    if (e === 1)
-      return l(u(o(r), n[0]), n[1] - 1);
-    if (e === 1 / 0)
-      return l(u(o(r), n[1]), n[0]);
-    if (e === -1 / 0)
-      return d(u(o(r), n[1]), n[0]);
-    if (e === "fro" || e === "euclidean")
-      return h(u(v(r), n));
-    throw new Error(`Error in norm: invalid ord value: ${e}`);
-  }
-  throw new Error(`Error in norm: invalid axis: ${n}`);
-}
-const J = /* @__PURE__ */ i({ norm_: M });
-/**
- * @license
- * Copyright 2020 Google Inc. 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 T(r, e = null, n = !1) {
-  const t = { x: m(r, "x", "mean") }, a = { axis: e, keepDims: n };
-  return y.runKernel(_, t, a);
-}
-const p = /* @__PURE__ */ i({ mean_: T });
-/**
- * @license
- * Copyright 2020 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 g(r, e = null, n = !1) {
-  r = m(r, "x", "moments");
-  const s = E(e, r.shape), t = p(r, s, n);
-  let a = t.shape;
-  n || (a = w(t.shape, s));
-  const f = v(k(q(r, "float32"), c(t, a))), I = p(f, s, n);
-  return { mean: t, variance: I };
-}
-const P = /* @__PURE__ */ i({ moments_: g });
-export {
-  P as a,
-  d as b,
-  p as m,
-  J as n
-};

package/dist/softmax-Be_lsqUc.js

@@ -1,105 +0,0 @@
-import { o as r, h as e, E as c, _ as $, $ as h, a0 as E, s as S, Z as _, a1 as d } from "./index-pWA4_lUh.js";
-import { m as k, s as w, e as K } from "./sum-C7Mgy9Bw.js";
-import { r as l } from "./reshape-C8CR_Bad.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 T(n) {
-  const s = { x: e(n, "x", "exp") };
-  return c.runKernel($, s);
-}
-const b = /* @__PURE__ */ r({ exp_: T });
-/**
- * @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 L(n) {
-  const s = { x: e(n, "x", "log", "float32") };
-  return c.runKernel(h, s);
-}
-const N = /* @__PURE__ */ r({ log_: L });
-/**
- * @license
- * Copyright 2020 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 v(n, o = null, s = !1) {
-  const t = e(n, "x", "logSumExp"), a = E(o, t.shape), p = k(
-    t,
-    a,
-    !0
-    /* keepDims */
-  ), i = S(t, p), f = b(i), m = w(f, a), u = N(m), x = _(l(p, u.shape), u);
-  if (s) {
-    const g = K(x.shape, a);
-    return l(x, g);
-  }
-  return x;
-}
-const I = /* @__PURE__ */ r({ logSumExp_: v });
-/**
- * @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 y(n, o = -1) {
-  const s = e(n, "logits", "softmax", "float32");
-  if (o === -1 && (o = s.rank - 1), o !== s.rank - 1)
-    throw Error(`Softmax along a non-last dimension is not yet supported. Logits was rank ${s.rank} and dim was ${o}`);
-  const t = { logits: s }, a = { dim: o };
-  return c.runKernel(d, t, a);
-}
-const M = /* @__PURE__ */ r({ softmax_: y });
-export {
-  N as a,
-  b as e,
-  I as l,
-  M as s
-};

package/dist/training/LayerTrainer.d.ts

@@ -1,29 +0,0 @@
-import { ITokeniser } from '../tokeniser/type';
-import { default as NanoGPT } from '../NanoGPTModel';
-import { default as GPTTrainer, TrainingOptions } from './Trainer';
-import { Tensor } from '@tensorflow/tfjs-core';
-import { Dataset } from '@tensorflow/tfjs-data';
-interface LayerTrainingOptions extends TrainingOptions {
-    stepsPerLayer: number;
-    maxPasses: number;
-    onLayerChange?: (layer: number, pass: number, valLoss?: number) => Promise<void> | void;
-    onPassComplete?: (pass: number) => Promise<void> | void;
-}
-export default class LayerTrainer extends GPTTrainer {
-    private trainingPattern;
-    private startPass;
-    private startLayer;
-    constructor(model: NanoGPT, tokenizer: ITokeniser, learningRate?: number);
-    private applyTrainingPattern;
-    trainOnDataset(dataset: Dataset<{
-        xs: Tensor;
-        ys: Tensor;
-    }>, options: Partial<LayerTrainingOptions>, validationDataset?: Dataset<{
-        xs: Tensor;
-        ys: Tensor;
-    }>): Promise<{
-        losses: number[];
-        validationLosses: number[];
-    }>;
-}
-export {};