UNPKG

@tensorflow/tfjs-core

Version:

Hardware-accelerated JavaScript library for machine intelligence

101 lines 4.75 kB
/** * @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('AdamOptimizer', ALL_ENVS, () => { it('basic', async () => { const learningRate = .1; const beta1 = .8; const beta2 = .9; const optimizer = tf.train.adam(learningRate, beta1, beta2); const x = tf.tensor1d([2, 4]).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); // new_first_m = [ // beta1 * old_first_m_w1 + (1-beta1) * grad_w1, // beta1 * old_first_m_w2 + (1-beta1) * grad_w2 // ] = [.8, 1.6] // new_second_m = [ // beta2 * old_second_m_w1 + (1-beta2) * grad_w1**2, // beta2 * old_second_m_w2 + (1-beta2) * grad_w2**2 // ] = [1.6, 6.4] // m = [new_first_m/(1-acc_beta1)] = [4, 8] // v = [new_second_m/(1-acc_beta2)] = [16, 64] // x = [x - lr * m / sqrt(v)] = [1.9, 3.9] // expectArraysClose(await x.data(), [1.9, 3.9]); cost.dispose(); numTensors = tf.memory().numTensors; cost = optimizer.minimize(f, /* returnCost */ false); // new_first_m = [ // beta1 * old_first_m_w1 + (1-beta1) * grad_w1, // beta1 * old_first_m_w2 + (1-beta1) * grad_w2 // ] = [1.4, 2.84] // new_second_m = [ // beta2 * old_second_m_w1 + (1-beta2) * grad_w1**2, // beta2 * old_second_m_w2 + (1-beta2) * grad_w2**2 // ] = [2.884, 11.884] // m = [new_first_m/(1-acc_beta1)] = [3.888888, 7.88889] // v = [new_second_m/(1-acc_beta2)] = [15.1789, 62.5473] // x = [x - lr * m / sqrt(v)] = [1.8000001, 3.8002] // expectArraysClose(await x.data(), [1.8000001, 3.8002]); // 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 should be the argument to variable(). expect(tf.memory().numTensors).toBe(1); }); it('Continue training after loading weights', async () => { const learningRate = .1; const beta1 = .8; const beta2 = .9; const optimizer1 = tf.train.adam(learningRate, beta1, beta2); const x = tf.tensor1d([2, 4]).variable(); const f = () => x.square().sum(); let cost = optimizer1.minimize(f, /* returnCost */ true); expect(optimizer1.iterations).toEqual(1); expectArraysClose(await cost.data(), 20); const weights = await optimizer1.getWeights(); expect(weights.length).toEqual(3); expect(weights[0].name).toEqual('iter'); expect(weights[1].name).toEqual(`${x.name}/m`); expect(weights[2].name).toEqual(`${x.name}/v`); const optimizer2 = tf.train.adam(learningRate, beta1, beta2); await optimizer2.setWeights(weights); cost = optimizer2.minimize(f, /* returnCost */ true); expectArraysClose(await cost.data(), 18.82); expect(optimizer2.iterations).toEqual(2); const optimizer3 = tf.train.adam(learningRate, beta1, beta2); await optimizer3.setWeights(await optimizer2.getWeights()); cost = optimizer2.minimize(f, /* returnCost */ true); expectArraysClose(await cost.data(), 17.681284); expect(optimizer3.iterations).toEqual(2); }); it('serialization round-trip', () => { const originalOpt = tf.train.adam(0.1, 0.2, 0.3, 2e-8); const reserialized = tf.AdamOptimizer.fromConfig(tf.AdamOptimizer, originalOpt.getConfig()); expect(reserialized.getConfig()).toEqual(originalOpt.getConfig()); }); }); //# sourceMappingURL=adam_optimizer_test.js.map