@tensorflow/tfjs-core
Version:
Hardware-accelerated JavaScript library for machine intelligence
113 lines • 6.53 kB
JavaScript
/**
* @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.
* =============================================================================
*/
import * as tf from '../../index';
import { ALL_ENVS, describeWithFlags } from '../../jasmine_util';
import { expectArraysClose } from '../../test_util';
describeWithFlags('softmaxCrossEntropy', ALL_ENVS, () => {
it('All wrong', async () => {
const label = tf.tensor2d([[0, 0, 1], [1, 0, 0], [0, 1, 0]], [3, 3]);
const predictions = tf.tensor2d([[10.0, -10.0, -10.0], [-10.0, 10.0, -10.0], [-10.0, -10.0, 10.0]], [3, 3]);
const y = tf.losses.softmaxCrossEntropy(label, predictions);
expect(y.shape).toEqual([]);
expectArraysClose(await y.data(), 20);
});
it('All right', async () => {
const label = tf.tensor2d([[1, 0, 0], [0, 1, 0], [0, 0, 1]], [3, 3]);
const predictions = tf.tensor2d([[10.0, -10.0, -10.0], [-10.0, 10.0, -10.0], [-10.0, -10.0, 10.0]], [3, 3]);
const y = tf.losses.softmaxCrossEntropy(label, predictions);
expect(y.shape).toEqual([]);
expectArraysClose(await y.data(), 0);
});
it('Weighted - Reduction.SUM_BY_NONZERO_WEIGHTS', async () => {
const label = tf.tensor2d([[0, 0, 1], [1, 0, 0], [0, 1, 0]], [3, 3]);
const predictions = tf.tensor2d([[10.0, -10.0, -10.0], [-10.0, 10.0, -10.0], [-10.0, -10.0, 10.0]], [3, 3]);
const weights = tf.tensor2d([[0.1, 0.2, 0.3], [0.1, 0.2, 0.3], [0.1, 0.2, 0.3]]);
const y = tf.losses.softmaxCrossEntropy(label, predictions, weights);
expect(y.shape).toEqual([]);
expectArraysClose(await y.data(), 4);
});
it('Weighted - Reduction.NONE', async () => {
const label = tf.tensor2d([[0, 0, 1], [1, 0, 0], [0, 1, 0]], [3, 3]);
const predictions = tf.tensor2d([[10.0, -10.0, -10.0], [-10.0, 10.0, -10.0], [-10.0, -10.0, 10.0]], [3, 3]);
const weights = tf.tensor1d([0.1, 0.2, 0.3]);
const y = tf.losses.softmaxCrossEntropy(label, predictions, weights, undefined, tf.Reduction.NONE);
expect(y.shape).toEqual([3]);
expectArraysClose(await y.data(), [2, 4, 6]);
});
it('Reduction.MEAN', async () => {
const label = tf.tensor2d([[0, 0, 1], [1, 0, 0], [0, 1, 0]], [3, 3]);
const predictions = tf.tensor2d([[10.0, -10.0, -10.0], [-10.0, 10.0, -10.0], [-10.0, -10.0, 10.0]], [3, 3]);
const y = tf.losses.softmaxCrossEntropy(label, predictions, undefined, undefined, tf.Reduction.MEAN);
expect(y.shape).toEqual([]);
expectArraysClose(await y.data(), 20);
});
it('Weighted - Reduction.MEAN', async () => {
const label = tf.tensor2d([[0, 0, 1], [1, 0, 0], [0, 1, 0]], [3, 3]);
const predictions = tf.tensor2d([[10.0, -10.0, -10.0], [-10.0, 10.0, -10.0], [-10.0, -10.0, 10.0]], [3, 3]);
const weights = tf.tensor1d([0.1, 0.2, 0.3]);
const y = tf.losses.softmaxCrossEntropy(label, predictions, weights, undefined, tf.Reduction.MEAN);
expect(y.shape).toEqual([]);
expectArraysClose(await y.data(), 20);
});
it('Label Smoothing - Weighted - Reduction.MEAN', async () => {
const label = tf.tensor2d([[0, 0, 1], [1, 0, 0], [0, 1, 0]], [3, 3]);
const predictions = tf.tensor2d([[10.0, -10.0, -10.0], [-10.0, 10.0, -10.0], [-10.0, -10.0, 10.0]], [3, 3]);
const weights = tf.tensor2d([[0.1, 0.2, 0.3]]);
const labelSmoothing = 0.3;
const y = tf.losses.softmaxCrossEntropy(label, predictions, weights, labelSmoothing, tf.Reduction.MEAN);
expect(y.shape).toEqual([]);
expectArraysClose(await y.data(), 18);
});
it('throws when multiClassLabels and logits are of different shapes', () => {
const multiClassLabels = tf.tensor2d([10, 10, 10, 10, 10, 10, 10, 10, 10], [3, 3]);
const logits = tf.tensor2d([10, 10, 10, 10, 10, 10], [2, 3]);
const e = new RegExp('Error in softmaxCrossEntropy: Shapes 3,3 and 2,3 must match');
expect(() => tf.losses.softmaxCrossEntropy(multiClassLabels, logits))
.toThrowError(e);
});
it('throws when passed multiClassLabels as a non-tensor', () => {
const predictions = tf.tensor2d([[10.0, -10.0, -10.0], [-10.0, 10.0, -10.0], [-10.0, -10.0, 10.0]], [3, 3]);
const weights = tf.tensor2d([[0.1, 0.2, 0.3]]);
const e = new RegExp('Argument \'onehotLabels\' passed to \'softmaxCrossEntropy\' ' +
'must be a Tensor');
expect(() => tf.losses.softmaxCrossEntropy({}, predictions, weights, tf.Reduction.MEAN))
.toThrowError(e);
});
it('throws when passed logits as a non-tensor', () => {
const label = tf.tensor2d([[0, 0, 1], [1, 0, 0], [0, 1, 0]], [3, 3]);
const weights = tf.tensor2d([[0.1, 0.2, 0.3]]);
const e = new RegExp('Argument \'logits\' passed to \'softmaxCrossEntropy\' ' +
'must be a Tensor');
expect(() => tf.losses.softmaxCrossEntropy(label, {}, weights, tf.Reduction.MEAN))
.toThrowError(e);
});
it('throws when passed weights as a non-tensor', () => {
const label = tf.tensor2d([[0, 0, 1], [1, 0, 0], [0, 1, 0]], [3, 3]);
const predictions = tf.tensor2d([[10.0, -10.0, -10.0], [-10.0, 10.0, -10.0], [-10.0, -10.0, 10.0]], [3, 3]);
const e = /Argument 'weights' passed to 'softmaxCrossEntropy' must be a Tensor/;
expect(() => tf.losses.softmaxCrossEntropy(label, predictions, {}, tf.Reduction.MEAN))
.toThrowError(e);
});
it('accepts a tensor-like object', async () => {
const label = [[0, 0, 1], [1, 0, 0], [0, 1, 0]];
const predictions = [[10.0, -10.0, -10.0], [-10.0, 10.0, -10.0], [-10.0, -10.0, 10.0]];
const y = tf.losses.softmaxCrossEntropy(label, predictions);
expect(y.shape).toEqual([]);
expectArraysClose(await y.data(), 20);
});
});
//# sourceMappingURL=softmax_cross_entropy_test.js.map