@tensorflow/tfjs-core
Version:
Hardware-accelerated JavaScript library for machine intelligence
88 lines • 4.01 kB
JavaScript
/**
* @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.
* =============================================================================
*/
import * as tf from '../index';
import { ALL_ENVS, describeWithFlags } from '../jasmine_util';
import { expectArraysClose } from '../test_util';
describeWithFlags('AdadeltaOptimizer', ALL_ENVS, () => {
it('basic', async () => {
const learningRate = .1;
const rho = .95;
const optimizer = tf.train.adadelta(learningRate, rho);
const x = tf.tensor1d([1, 2]).variable();
const f = () => x.square().sum();
let numTensors = tf.memory().numTensors;
let cost = optimizer.minimize(f, /* returnCost */ true);
// Cost & 2 accumulators should be the only additional arrays.
expect(tf.memory().numTensors).toBe(numTensors + 3);
// epsilon = 1-e8
// newAccumulatedGrad = rho * accumulatedGrad + (1 - rho) * grad ^ 2
// updates = -grad * sqrt(accumulatedUpdate + epsilon) /
// sqrt(accumulatedGrad + epsilon)
// newAccumulatedUpdate = rho * accumulatedUpdate + (1 - rho) * updates ^ 2
// x += learningRate * updates
//
// de/dx = [2, 4]
// accumulatedGrad = [0, 0]
// newAccumulatedGrad = [.2, .8]
// updates = [-2, -4]
// newAccumulatedUpdate = [.2, .8]
// x = [0.8, 1.6]
expectArraysClose(await x.data(), [0.8, 1.6]);
cost.dispose();
numTensors = tf.memory().numTensors;
cost = optimizer.minimize(f, /* returnCost */ false);
// de/dx = [1.6, 3.2]
// accumulatedGrad = [.2, .8]
// accumulatedUpdate = [.2, .8]
// newAccumulatedGrad = [0.318, 1.272]
// updates = [-1.6, -3.2]
// x = [0.64, 1.28]
expectArraysClose(await x.data(), [0.64, 1.28]);
// There should be no new additional Tensors.
expect(tf.memory().numTensors).toBe(numTensors);
expect(cost).toBe(null);
x.dispose();
optimizer.dispose();
// The only tensor remaining is the argument to variable().
expect(tf.memory().numTensors).toBe(1);
});
it('Save, load weights and continue training', async () => {
const learningRate = .1;
const rho = .95;
const optimizer1 = tf.train.adadelta(learningRate, rho);
const x = tf.tensor1d([1, 2]).variable();
const f = () => x.square().sum();
let cost = optimizer1.minimize(f, /* returnCost */ true);
expectArraysClose(await cost.data(), 5);
expectArraysClose(await x.data(), [0.8, 1.6]);
const weights = await optimizer1.getWeights();
expect(weights.length).toEqual(3);
expect(weights[0].name).toEqual('iter');
const optimizer2 = tf.train.adadelta(learningRate, rho);
await optimizer2.setWeights(weights);
cost = optimizer2.minimize(f, /* returnCost */ true);
expectArraysClose(await cost.data(), 3.2);
expectArraysClose(await x.data(), [0.64, 1.28]);
expect(optimizer2.iterations).toEqual(2);
});
it('serialization round-trip', () => {
const originalOpt = tf.train.adadelta(0.1, 0.2, 2e-8);
const reserialized = tf.AdadeltaOptimizer.fromConfig(tf.AdadeltaOptimizer, originalOpt.getConfig());
expect(reserialized.getConfig()).toEqual(originalOpt.getConfig());
});
});
//# sourceMappingURL=adadelta_optimizer_test.js.map