@tensorflow/tfjs-core
Version:
Hardware-accelerated JavaScript library for machine intelligence
907 lines • 68.3 kB
JavaScript
"use strict";
/**
* @license
* Copyright 2017 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.
* =============================================================================
*/
var __awaiter = (this && this.__awaiter) || function (thisArg, _arguments, P, generator) {
return new (P || (P = Promise))(function (resolve, reject) {
function fulfilled(value) { try { step(generator.next(value)); } catch (e) { reject(e); } }
function rejected(value) { try { step(generator["throw"](value)); } catch (e) { reject(e); } }
function step(result) { result.done ? resolve(result.value) : new P(function (resolve) { resolve(result.value); }).then(fulfilled, rejected); }
step((generator = generator.apply(thisArg, _arguments || [])).next());
});
};
var __generator = (this && this.__generator) || function (thisArg, body) {
var _ = { label: 0, sent: function() { if (t[0] & 1) throw t[1]; return t[1]; }, trys: [], ops: [] }, f, y, t, g;
return g = { next: verb(0), "throw": verb(1), "return": verb(2) }, typeof Symbol === "function" && (g[Symbol.iterator] = function() { return this; }), g;
function verb(n) { return function (v) { return step([n, v]); }; }
function step(op) {
if (f) throw new TypeError("Generator is already executing.");
while (_) try {
if (f = 1, y && (t = op[0] & 2 ? y["return"] : op[0] ? y["throw"] || ((t = y["return"]) && t.call(y), 0) : y.next) && !(t = t.call(y, op[1])).done) return t;
if (y = 0, t) op = [op[0] & 2, t.value];
switch (op[0]) {
case 0: case 1: t = op; break;
case 4: _.label++; return { value: op[1], done: false };
case 5: _.label++; y = op[1]; op = [0]; continue;
case 7: op = _.ops.pop(); _.trys.pop(); continue;
default:
if (!(t = _.trys, t = t.length > 0 && t[t.length - 1]) && (op[0] === 6 || op[0] === 2)) { _ = 0; continue; }
if (op[0] === 3 && (!t || (op[1] > t[0] && op[1] < t[3]))) { _.label = op[1]; break; }
if (op[0] === 6 && _.label < t[1]) { _.label = t[1]; t = op; break; }
if (t && _.label < t[2]) { _.label = t[2]; _.ops.push(op); break; }
if (t[2]) _.ops.pop();
_.trys.pop(); continue;
}
op = body.call(thisArg, _);
} catch (e) { op = [6, e]; y = 0; } finally { f = t = 0; }
if (op[0] & 5) throw op[1]; return { value: op[0] ? op[1] : void 0, done: true };
}
};
var _this = this;
Object.defineProperty(exports, "__esModule", { value: true });
var tf = require("../index");
var jasmine_util_1 = require("../jasmine_util");
var test_util_1 = require("../test_util");
jasmine_util_1.describeWithFlags('batchNorm4D', jasmine_util_1.ALL_ENVS, function () {
it('simple batchnorm4D, no offset or scale, 2x1x1x2', function () { return __awaiter(_this, void 0, void 0, function () {
var xT, meanT, varianceT, varianceEpsilon, result, x, mean, variance, _a;
return __generator(this, function (_b) {
switch (_b.label) {
case 0:
xT = tf.tensor4d([2, 4, 9, 23], [2, 1, 1, 2]);
meanT = tf.tensor1d([1, 2]);
varianceT = tf.tensor1d([2, 3]);
varianceEpsilon = .001;
result = tf.batchNorm4d(xT, meanT, varianceT, undefined, undefined, varianceEpsilon);
return [4 /*yield*/, xT.array()];
case 1:
x = _b.sent();
return [4 /*yield*/, meanT.array()];
case 2:
mean = _b.sent();
return [4 /*yield*/, varianceT.array()];
case 3:
variance = _b.sent();
_a = test_util_1.expectArraysClose;
return [4 /*yield*/, result.data()];
case 4:
_a.apply(void 0, [_b.sent(), [
(x[0][0][0][0] - mean[0]) * 1 / Math.sqrt(variance[0] + varianceEpsilon),
(x[0][0][0][1] - mean[1]) * 1 / Math.sqrt(variance[1] + varianceEpsilon),
(x[1][0][0][0] - mean[0]) * 1 / Math.sqrt(variance[0] + varianceEpsilon),
(x[1][0][0][1] - mean[1]) * 1 / Math.sqrt(variance[1] + varianceEpsilon)
]]);
return [2 /*return*/];
}
});
}); });
it('simple batchnorm4D, no offset, 2x1x1x2', function () { return __awaiter(_this, void 0, void 0, function () {
var xT, meanT, varianceT, scaleT, varianceEpsilon, result, x, mean, variance, scale, _a;
return __generator(this, function (_b) {
switch (_b.label) {
case 0:
xT = tf.tensor4d([2, 4, 9, 23], [2, 1, 1, 2]);
meanT = tf.tensor1d([1, 2]);
varianceT = tf.tensor1d([2, 3]);
scaleT = tf.tensor1d([4, 5]);
varianceEpsilon = .001;
result = tf.batchNorm4d(xT, meanT, varianceT, undefined, scaleT, varianceEpsilon);
return [4 /*yield*/, xT.buffer()];
case 1:
x = _b.sent();
return [4 /*yield*/, meanT.buffer()];
case 2:
mean = _b.sent();
return [4 /*yield*/, varianceT.buffer()];
case 3:
variance = _b.sent();
return [4 /*yield*/, scaleT.buffer()];
case 4:
scale = _b.sent();
_a = test_util_1.expectArraysClose;
return [4 /*yield*/, result.data()];
case 5:
_a.apply(void 0, [_b.sent(), [
(x.get(0, 0, 0, 0) - mean.get(0)) * scale.get(0) /
Math.sqrt(variance.get(0) + varianceEpsilon),
(x.get(0, 0, 0, 1) - mean.get(1)) * scale.get(1) /
Math.sqrt(variance.get(1) + varianceEpsilon),
(x.get(1, 0, 0, 0) - mean.get(0)) * scale.get(0) /
Math.sqrt(variance.get(0) + varianceEpsilon),
(x.get(1, 0, 0, 1) - mean.get(1)) * scale.get(1) /
Math.sqrt(variance.get(1) + varianceEpsilon)
]]);
return [2 /*return*/];
}
});
}); });
it('simple batchnorm4D, no scale, 2x1x1x2', function () { return __awaiter(_this, void 0, void 0, function () {
var xT, meanT, varianceT, offsetT, varianceEpsilon, result, x, mean, variance, offset, _a;
return __generator(this, function (_b) {
switch (_b.label) {
case 0:
xT = tf.tensor4d([2, 4, 9, 23], [2, 1, 1, 2]);
meanT = tf.tensor1d([1, 2]);
varianceT = tf.tensor1d([2, 3]);
offsetT = tf.tensor1d([4, 5]);
varianceEpsilon = .001;
result = tf.batchNorm4d(xT, meanT, varianceT, offsetT, undefined, varianceEpsilon);
return [4 /*yield*/, xT.buffer()];
case 1:
x = _b.sent();
return [4 /*yield*/, meanT.buffer()];
case 2:
mean = _b.sent();
return [4 /*yield*/, varianceT.buffer()];
case 3:
variance = _b.sent();
return [4 /*yield*/, offsetT.buffer()];
case 4:
offset = _b.sent();
_a = test_util_1.expectArraysClose;
return [4 /*yield*/, result.data()];
case 5:
_a.apply(void 0, [_b.sent(), [
offset.get(0) +
(x.get(0, 0, 0, 0) - mean.get(0)) * 1 /
Math.sqrt(variance.get(0) + varianceEpsilon),
offset.get(1) +
(x.get(0, 0, 0, 1) - mean.get(1)) * 1 /
Math.sqrt(variance.get(1) + varianceEpsilon),
offset.get(0) +
(x.get(1, 0, 0, 0) - mean.get(0)) * 1 /
Math.sqrt(variance.get(0) + varianceEpsilon),
offset.get(1) +
(x.get(1, 0, 0, 1) - mean.get(1)) * 1 /
Math.sqrt(variance.get(1) + varianceEpsilon)
]]);
return [2 /*return*/];
}
});
}); });
it('simple batchnorm4D, 2x1x1x2', function () { return __awaiter(_this, void 0, void 0, function () {
var xT, meanT, varianceT, offsetT, scaleT, varianceEpsilon, result, x, mean, variance, scale, offset, _a;
return __generator(this, function (_b) {
switch (_b.label) {
case 0:
xT = tf.tensor4d([2, 4, 9, 23], [2, 1, 1, 2]);
meanT = tf.tensor1d([1, 2]);
varianceT = tf.tensor1d([2, 3]);
offsetT = tf.tensor1d([3, 4]);
scaleT = tf.tensor1d([4, 5]);
varianceEpsilon = .001;
result = tf.batchNorm4d(xT, meanT, varianceT, offsetT, scaleT, varianceEpsilon);
return [4 /*yield*/, xT.buffer()];
case 1:
x = _b.sent();
return [4 /*yield*/, meanT.buffer()];
case 2:
mean = _b.sent();
return [4 /*yield*/, varianceT.buffer()];
case 3:
variance = _b.sent();
return [4 /*yield*/, scaleT.buffer()];
case 4:
scale = _b.sent();
return [4 /*yield*/, offsetT.buffer()];
case 5:
offset = _b.sent();
_a = test_util_1.expectArraysClose;
return [4 /*yield*/, result.data()];
case 6:
_a.apply(void 0, [_b.sent(), [
offset.get(0) +
(x.get(0, 0, 0, 0) - mean.get(0)) * scale.get(0) /
Math.sqrt(variance.get(0) + varianceEpsilon),
offset.get(1) +
(x.get(0, 0, 0, 1) - mean.get(1)) * scale.get(1) /
Math.sqrt(variance.get(1) + varianceEpsilon),
offset.get(0) +
(x.get(1, 0, 0, 0) - mean.get(0)) * scale.get(0) /
Math.sqrt(variance.get(0) + varianceEpsilon),
offset.get(1) +
(x.get(1, 0, 0, 1) - mean.get(1)) * scale.get(1) /
Math.sqrt(variance.get(1) + varianceEpsilon)
]]);
return [2 /*return*/];
}
});
}); });
it('accepts a tensor-like object', function () { return __awaiter(_this, void 0, void 0, function () {
var x, mean, variance, offset, scale, varianceEpsilon, result, _a;
return __generator(this, function (_b) {
switch (_b.label) {
case 0:
x = [[[[2, 4]]], [[[9, 23]]]];
mean = [1, 2];
variance = [2, 3];
offset = [3, 4];
scale = [4, 5];
varianceEpsilon = .001;
result = tf.batchNorm4d(x, mean, variance, offset, scale, varianceEpsilon);
_a = test_util_1.expectArraysClose;
return [4 /*yield*/, result.data()];
case 1:
_a.apply(void 0, [_b.sent(), [
offset[0] +
(x[0][0][0][0] - mean[0]) * scale[0] /
Math.sqrt(variance[0] + varianceEpsilon),
offset[1] +
(x[0][0][0][1] - mean[1]) * scale[1] /
Math.sqrt(variance[1] + varianceEpsilon),
offset[0] +
(x[1][0][0][0] - mean[0]) * scale[0] /
Math.sqrt(variance[0] + varianceEpsilon),
offset[1] +
(x[1][0][0][1] - mean[1]) * scale[1] /
Math.sqrt(variance[1] + varianceEpsilon)
]]);
return [2 /*return*/];
}
});
}); });
it('simple batchnorm4D gradients, 2x1x1x2', function () { return __awaiter(_this, void 0, void 0, function () {
var x, mean, variance, offset, scale, varianceEpsilon, dy, gradX, _a, gradMean, _b, gradVariance, _c, gradOffset, _d, _e, gradScale, _f;
return __generator(this, function (_g) {
switch (_g.label) {
case 0:
x = tf.tensor4d([2, 4, 9, 23], [2, 1, 1, 2]);
mean = tf.tensor1d([1, 2]);
variance = tf.tensor1d([2, 3]);
offset = tf.tensor1d([3, 4]);
scale = tf.tensor1d([2, 5]);
varianceEpsilon = .001;
dy = tf.tensor4d([-1, -1, -1, -1], [2, 1, 1, 2]);
gradX = tf.grad(function (x) { return tf.batchNorm4d(x, mean, variance, offset, scale, varianceEpsilon); })(x, dy);
_a = test_util_1.expectArraysClose;
return [4 /*yield*/, gradX.data()];
case 1:
_a.apply(void 0, [_g.sent(), [-1.414, -2.887, -1.414, -2.887]]);
expect(gradX.shape).toEqual([2, 1, 1, 2]);
gradMean = tf.grad(function (mean) { return tf.batchNorm4d(x, mean, variance, offset, scale, varianceEpsilon); })(mean, dy);
_b = test_util_1.expectArraysClose;
return [4 /*yield*/, gradMean.data()];
case 2:
_b.apply(void 0, [_g.sent(), [2.828, 5.773]]);
expect(gradMean.shape).toEqual([2]);
gradVariance = tf.grad(function (variance) { return tf.batchNorm4d(x, mean, variance, offset, scale, varianceEpsilon); })(variance, dy);
_c = test_util_1.expectArraysClose;
return [4 /*yield*/, gradVariance.data()];
case 3:
_c.apply(void 0, [_g.sent(), [3.180, 11.060]]);
expect(gradVariance.shape).toEqual([2]);
gradOffset = tf.grad(function (offset) { return tf.batchNorm4d(x, mean, variance, offset, scale, varianceEpsilon); })(offset, dy);
_d = test_util_1.expectArraysClose;
return [4 /*yield*/, gradOffset.data()];
case 4:
_e = [_g.sent()];
return [4 /*yield*/, dy.sum([0, 1, 2]).data()];
case 5:
_d.apply(void 0, _e.concat([_g.sent()]));
expect(gradOffset.shape).toEqual([2]);
gradScale = tf.grad(function (scale) { return tf.batchNorm4d(x, mean, variance, offset, scale, varianceEpsilon); })(scale, dy);
_f = test_util_1.expectArraysClose;
return [4 /*yield*/, gradScale.data()];
case 6:
_f.apply(void 0, [_g.sent(), [-6.362, -13.277]]);
expect(gradScale.shape).toEqual([2]);
return [2 /*return*/];
}
});
}); });
it('batchnorm4D gradients, same shapes in x, mean and variance', function () { return __awaiter(_this, void 0, void 0, function () {
var x, mean, variance, scale, offset, varianceEpsilon, dy, gradX, _a, gradMean, _b, gradVariance, _c, gradOffset, _d, _e, gradScale, _f;
return __generator(this, function (_g) {
switch (_g.label) {
case 0:
x = tf.tensor4d([10, 20, 30, 40], [2, 1, 1, 2]);
mean = tf.tensor4d([0, 5, 10, 15], [2, 1, 1, 2]);
variance = tf.tensor4d([2, 4, 6, 8], [2, 1, 1, 2]);
scale = tf.tensor4d([2, 5, 2, 5], [2, 1, 1, 2]);
offset = tf.tensor4d([0, 0, 0, 0], [2, 1, 1, 2]);
varianceEpsilon = .001;
dy = tf.tensor4d([-1, -1, -1, -1], [2, 1, 1, 2]);
gradX = tf.grad(function (x) { return tf.batchNorm4d(x, mean, variance, offset, scale, varianceEpsilon); })(x, dy);
_a = test_util_1.expectArraysClose;
return [4 /*yield*/, gradX.data()];
case 1:
_a.apply(void 0, [_g.sent(), [-1.414, -2.500, -0.816, -1.768]]);
expect(gradX.shape).toEqual([2, 1, 1, 2]);
gradMean = tf.grad(function (mean) { return tf.batchNorm4d(x, mean, variance, offset, scale, varianceEpsilon); })(mean, dy);
_b = test_util_1.expectArraysClose;
return [4 /*yield*/, gradMean.data()];
case 2:
_b.apply(void 0, [_g.sent(), [1.414, 2.500, 0.816, 1.768]]);
expect(gradMean.shape).toEqual([2, 1, 1, 2]);
gradVariance = tf.grad(function (variance) { return tf.batchNorm4d(x, mean, variance, offset, scale, varianceEpsilon); })(variance, dy);
_c = test_util_1.expectArraysClose;
return [4 /*yield*/, gradVariance.data()];
case 3:
_c.apply(void 0, [_g.sent(), [3.533, 4.686, 1.360, 2.762]]);
expect(gradVariance.shape).toEqual([2, 1, 1, 2]);
gradOffset = tf.grad(function (offset) { return tf.batchNorm4d(x, mean, variance, offset, scale, varianceEpsilon); })(offset, dy);
_d = test_util_1.expectArraysClose;
return [4 /*yield*/, gradOffset.data()];
case 4:
_e = [_g.sent()];
return [4 /*yield*/, dy.data()];
case 5:
_d.apply(void 0, _e.concat([_g.sent()]));
expect(gradOffset.shape).toEqual([2, 1, 1, 2]);
gradScale = tf.grad(function (scale) { return tf.batchNorm4d(x, mean, variance, offset, scale, varianceEpsilon); })(scale, dy);
_f = test_util_1.expectArraysClose;
return [4 /*yield*/, gradScale.data()];
case 6:
_f.apply(void 0, [_g.sent(), [-7.069, -7.499, -8.164, -8.838]]);
expect(gradScale.shape).toEqual([2, 1, 1, 2]);
return [2 /*return*/];
}
});
}); });
});
jasmine_util_1.describeWithFlags('batchNorm3D', jasmine_util_1.ALL_ENVS, function () {
it('simple batchnorm3D, no offset or scale, 2x1x2', function () { return __awaiter(_this, void 0, void 0, function () {
var xT, meanT, varianceT, varianceEpsilon, result, x, mean, variance, _a;
return __generator(this, function (_b) {
switch (_b.label) {
case 0:
xT = tf.tensor3d([2, 4, 9, 23], [2, 1, 2]);
meanT = tf.tensor1d([1, 2]);
varianceT = tf.tensor1d([2, 3]);
varianceEpsilon = .001;
result = tf.batchNorm3d(xT, meanT, varianceT, undefined, undefined, varianceEpsilon);
return [4 /*yield*/, xT.buffer()];
case 1:
x = _b.sent();
return [4 /*yield*/, meanT.buffer()];
case 2:
mean = _b.sent();
return [4 /*yield*/, varianceT.buffer()];
case 3:
variance = _b.sent();
_a = test_util_1.expectArraysClose;
return [4 /*yield*/, result.data()];
case 4:
_a.apply(void 0, [_b.sent(), [
(x.get(0, 0, 0) - mean.get(0)) * 1 /
Math.sqrt(variance.get(0) + varianceEpsilon),
(x.get(0, 0, 1) - mean.get(1)) * 1 /
Math.sqrt(variance.get(1) + varianceEpsilon),
(x.get(1, 0, 0) - mean.get(0)) * 1 /
Math.sqrt(variance.get(0) + varianceEpsilon),
(x.get(1, 0, 1) - mean.get(1)) * 1 /
Math.sqrt(variance.get(1) + varianceEpsilon)
]]);
return [2 /*return*/];
}
});
}); });
it('simple batchnorm3D, no offset, 2x1x2', function () { return __awaiter(_this, void 0, void 0, function () {
var xT, meanT, varianceT, scaleT, varianceEpsilon, result, x, mean, variance, scale, _a;
return __generator(this, function (_b) {
switch (_b.label) {
case 0:
xT = tf.tensor3d([2, 4, 9, 23], [2, 1, 2]);
meanT = tf.tensor1d([1, 2]);
varianceT = tf.tensor1d([2, 3]);
scaleT = tf.tensor1d([4, 5]);
varianceEpsilon = .001;
result = tf.batchNorm3d(xT, meanT, varianceT, undefined, scaleT, varianceEpsilon);
return [4 /*yield*/, xT.buffer()];
case 1:
x = _b.sent();
return [4 /*yield*/, meanT.buffer()];
case 2:
mean = _b.sent();
return [4 /*yield*/, varianceT.buffer()];
case 3:
variance = _b.sent();
return [4 /*yield*/, scaleT.buffer()];
case 4:
scale = _b.sent();
_a = test_util_1.expectArraysClose;
return [4 /*yield*/, result.data()];
case 5:
_a.apply(void 0, [_b.sent(), [
(x.get(0, 0, 0) - mean.get(0)) * scale.get(0) /
Math.sqrt(variance.get(0) + varianceEpsilon),
(x.get(0, 0, 1) - mean.get(1)) * scale.get(1) /
Math.sqrt(variance.get(1) + varianceEpsilon),
(x.get(1, 0, 0) - mean.get(0)) * scale.get(0) /
Math.sqrt(variance.get(0) + varianceEpsilon),
(x.get(1, 0, 1) - mean.get(1)) * scale.get(1) /
Math.sqrt(variance.get(1) + varianceEpsilon)
]]);
return [2 /*return*/];
}
});
}); });
it('simple batchnorm3D, no scale, 2x1x2', function () { return __awaiter(_this, void 0, void 0, function () {
var xT, meanT, varianceT, offsetT, varianceEpsilon, result, x, mean, variance, offset, _a;
return __generator(this, function (_b) {
switch (_b.label) {
case 0:
xT = tf.tensor3d([2, 4, 9, 23], [2, 1, 2]);
meanT = tf.tensor1d([1, 2]);
varianceT = tf.tensor1d([2, 3]);
offsetT = tf.tensor1d([4, 5]);
varianceEpsilon = .001;
result = tf.batchNorm3d(xT, meanT, varianceT, offsetT, undefined, varianceEpsilon);
return [4 /*yield*/, xT.buffer()];
case 1:
x = _b.sent();
return [4 /*yield*/, meanT.buffer()];
case 2:
mean = _b.sent();
return [4 /*yield*/, varianceT.buffer()];
case 3:
variance = _b.sent();
return [4 /*yield*/, offsetT.buffer()];
case 4:
offset = _b.sent();
_a = test_util_1.expectArraysClose;
return [4 /*yield*/, result.data()];
case 5:
_a.apply(void 0, [_b.sent(), [
offset.get(0) +
(x.get(0, 0, 0) - mean.get(0)) * 1 /
Math.sqrt(variance.get(0) + varianceEpsilon),
offset.get(1) +
(x.get(0, 0, 1) - mean.get(1)) * 1 /
Math.sqrt(variance.get(1) + varianceEpsilon),
offset.get(0) +
(x.get(1, 0, 0) - mean.get(0)) * 1 /
Math.sqrt(variance.get(0) + varianceEpsilon),
offset.get(1) +
(x.get(1, 0, 1) - mean.get(1)) * 1 /
Math.sqrt(variance.get(1) + varianceEpsilon)
]]);
return [2 /*return*/];
}
});
}); });
it('simple batchnorm3D, 2x1x2', function () { return __awaiter(_this, void 0, void 0, function () {
var xT, meanT, varianceT, offsetT, scaleT, varianceEpsilon, result, x, mean, variance, offset, scale, _a;
return __generator(this, function (_b) {
switch (_b.label) {
case 0:
xT = tf.tensor3d([2, 4, 9, 23], [2, 1, 2]);
meanT = tf.tensor1d([1, 2]);
varianceT = tf.tensor1d([2, 3]);
offsetT = tf.tensor1d([3, 4]);
scaleT = tf.tensor1d([4, 5]);
varianceEpsilon = .001;
result = tf.batchNorm3d(xT, meanT, varianceT, offsetT, scaleT, varianceEpsilon);
return [4 /*yield*/, xT.buffer()];
case 1:
x = _b.sent();
return [4 /*yield*/, meanT.buffer()];
case 2:
mean = _b.sent();
return [4 /*yield*/, varianceT.buffer()];
case 3:
variance = _b.sent();
return [4 /*yield*/, offsetT.buffer()];
case 4:
offset = _b.sent();
return [4 /*yield*/, scaleT.buffer()];
case 5:
scale = _b.sent();
_a = test_util_1.expectArraysClose;
return [4 /*yield*/, result.data()];
case 6:
_a.apply(void 0, [_b.sent(), [
offset.get(0) +
(x.get(0, 0, 0) - mean.get(0)) * scale.get(0) /
Math.sqrt(variance.get(0) + varianceEpsilon),
offset.get(1) +
(x.get(0, 0, 1) - mean.get(1)) * scale.get(1) /
Math.sqrt(variance.get(1) + varianceEpsilon),
offset.get(0) +
(x.get(1, 0, 0) - mean.get(0)) * scale.get(0) /
Math.sqrt(variance.get(0) + varianceEpsilon),
offset.get(1) +
(x.get(1, 0, 1) - mean.get(1)) * scale.get(1) /
Math.sqrt(variance.get(1) + varianceEpsilon)
]]);
return [2 /*return*/];
}
});
}); });
it('accepts a tensor-like object', function () { return __awaiter(_this, void 0, void 0, function () {
var x, mean, variance, offset, scale, varianceEpsilon, result, _a;
return __generator(this, function (_b) {
switch (_b.label) {
case 0:
x = [[[2, 4]], [[9, 23]]];
mean = [1, 2];
variance = [2, 3];
offset = [3, 4];
scale = [4, 5];
varianceEpsilon = .001;
result = tf.batchNorm3d(x, mean, variance, offset, scale, varianceEpsilon);
_a = test_util_1.expectArraysClose;
return [4 /*yield*/, result.data()];
case 1:
_a.apply(void 0, [_b.sent(), [
offset[0] +
(x[0][0][0] - mean[0]) * scale[0] /
Math.sqrt(variance[0] + varianceEpsilon),
offset[1] +
(x[0][0][1] - mean[1]) * scale[1] /
Math.sqrt(variance[1] + varianceEpsilon),
offset[0] +
(x[1][0][0] - mean[0]) * scale[0] /
Math.sqrt(variance[0] + varianceEpsilon),
offset[1] +
(x[1][0][1] - mean[1]) * scale[1] /
Math.sqrt(variance[1] + varianceEpsilon)
]]);
return [2 /*return*/];
}
});
}); });
it('batchnorm3D, x,mean,var,offset,scale are all 3D', function () { return __awaiter(_this, void 0, void 0, function () {
var shape, xT, meanT, varianceT, offsetT, scaleT, varianceEpsilon, result, x, mean, variance, offset, scale, _a;
return __generator(this, function (_b) {
switch (_b.label) {
case 0:
shape = [2, 1, 2];
xT = tf.tensor3d([2, 4, 9, 23], shape);
meanT = tf.tensor3d([1, 2, 3, 4], shape);
varianceT = tf.tensor3d([2, 3, 4, 5], shape);
offsetT = tf.tensor3d([3, 4, 5, 6], shape);
scaleT = tf.tensor3d([4, 5, 6, 7], shape);
varianceEpsilon = .001;
result = tf.batchNorm3d(xT, meanT, varianceT, offsetT, scaleT, varianceEpsilon);
return [4 /*yield*/, xT.buffer()];
case 1:
x = _b.sent();
return [4 /*yield*/, meanT.buffer()];
case 2:
mean = _b.sent();
return [4 /*yield*/, varianceT.buffer()];
case 3:
variance = _b.sent();
return [4 /*yield*/, offsetT.buffer()];
case 4:
offset = _b.sent();
return [4 /*yield*/, scaleT.buffer()];
case 5:
scale = _b.sent();
_a = test_util_1.expectArraysClose;
return [4 /*yield*/, result.data()];
case 6:
_a.apply(void 0, [_b.sent(), [
offset.get(0, 0, 0) +
(x.get(0, 0, 0) - mean.get(0, 0, 0)) * scale.get(0, 0, 0) /
Math.sqrt(variance.get(0, 0, 0) + varianceEpsilon),
offset.get(0, 0, 1) +
(x.get(0, 0, 1) - mean.get(0, 0, 1)) * scale.get(0, 0, 1) /
Math.sqrt(variance.get(0, 0, 1) + varianceEpsilon),
offset.get(1, 0, 0) +
(x.get(1, 0, 0) - mean.get(1, 0, 0)) * scale.get(1, 0, 0) /
Math.sqrt(variance.get(1, 0, 0) + varianceEpsilon),
offset.get(1, 0, 1) +
(x.get(1, 0, 1) - mean.get(1, 0, 1)) * scale.get(1, 0, 1) /
Math.sqrt(variance.get(1, 0, 1) + varianceEpsilon)
]]);
return [2 /*return*/];
}
});
}); });
it('simple batchnorm3D gradients, 2x1x2', function () { return __awaiter(_this, void 0, void 0, function () {
var x, mean, variance, offset, scale, varianceEpsilon, dy, gradX, _a, gradMean, _b, gradVariance, _c, gradOffset, _d, gradScale, _e;
return __generator(this, function (_f) {
switch (_f.label) {
case 0:
x = tf.tensor3d([2, 4, 9, 23], [2, 1, 2]);
mean = tf.tensor1d([1, 2]);
variance = tf.tensor1d([2, 3]);
offset = tf.tensor1d([3, 4]);
scale = tf.tensor1d([2, 5]);
varianceEpsilon = .001;
dy = tf.tensor3d([1, 1, 1, 1], [2, 1, 2]);
gradX = tf.grad(function (x) { return tf.batchNorm3d(x, mean, variance, offset, scale, varianceEpsilon); })(x, dy);
_a = test_util_1.expectArraysClose;
return [4 /*yield*/, gradX.data()];
case 1:
_a.apply(void 0, [_f.sent(), [1.414, 2.887, 1.414, 2.887]]);
expect(gradX.shape).toEqual([2, 1, 2]);
gradMean = tf.grad(function (mean) { return tf.batchNorm3d(x, mean, variance, offset, scale, varianceEpsilon); })(mean, dy);
_b = test_util_1.expectArraysClose;
return [4 /*yield*/, gradMean.data()];
case 2:
_b.apply(void 0, [_f.sent(), [-2.828, -5.773]]);
expect(gradMean.shape).toEqual([2]);
gradVariance = tf.grad(function (variance) { return tf.batchNorm3d(x, mean, variance, offset, scale, varianceEpsilon); })(variance, dy);
_c = test_util_1.expectArraysClose;
return [4 /*yield*/, gradVariance.data()];
case 3:
_c.apply(void 0, [_f.sent(), [-3.180, -11.060]]);
expect(gradVariance.shape).toEqual([2]);
gradOffset = tf.grad(function (offset) { return tf.batchNorm3d(x, mean, variance, offset, scale, varianceEpsilon); })(offset, dy);
_d = test_util_1.expectArraysClose;
return [4 /*yield*/, gradOffset.data()];
case 4:
_d.apply(void 0, [_f.sent(), [2, 2]]);
expect(gradOffset.shape).toEqual([2]);
gradScale = tf.grad(function (scale) { return tf.batchNorm3d(x, mean, variance, offset, scale, varianceEpsilon); })(scale, dy);
_e = test_util_1.expectArraysClose;
return [4 /*yield*/, gradScale.data()];
case 5:
_e.apply(void 0, [_f.sent(), [6.362, 13.277]]);
expect(gradScale.shape).toEqual([2]);
return [2 /*return*/];
}
});
}); });
it('batchnorm3D gradients, same shapes in x, mean and variance', function () { return __awaiter(_this, void 0, void 0, function () {
var x, mean, variance, scale, offset, varianceEpsilon, dy, gradX, _a, gradMean, _b, gradVariance, _c, gradOffset, _d, gradScale, _e;
return __generator(this, function (_f) {
switch (_f.label) {
case 0:
x = tf.tensor3d([10, 20, 30, 40], [2, 1, 2]);
mean = tf.tensor3d([0, 5, 10, 15], [2, 1, 2]);
variance = tf.tensor3d([2, 4, 6, 8], [2, 1, 2]);
scale = tf.tensor3d([2, 5, 2, 5], [2, 1, 2]);
offset = tf.tensor3d([0, 0, 0, 0], [2, 1, 2]);
varianceEpsilon = .001;
dy = tf.tensor3d([1, 1, 1, 1], [2, 1, 2]);
gradX = tf.grad(function (x) { return tf.batchNorm3d(x, mean, variance, offset, scale, varianceEpsilon); })(x, dy);
_a = test_util_1.expectArraysClose;
return [4 /*yield*/, gradX.data()];
case 1:
_a.apply(void 0, [_f.sent(), [1.414, 2.500, 0.816, 1.768]]);
expect(gradX.shape).toEqual([2, 1, 2]);
gradMean = tf.grad(function (mean) { return tf.batchNorm3d(x, mean, variance, offset, scale, varianceEpsilon); })(mean, dy);
_b = test_util_1.expectArraysClose;
return [4 /*yield*/, gradMean.data()];
case 2:
_b.apply(void 0, [_f.sent(), [-1.414, -2.500, -0.816, -1.768]]);
expect(gradMean.shape).toEqual([2, 1, 2]);
gradVariance = tf.grad(function (variance) { return tf.batchNorm3d(x, mean, variance, offset, scale, varianceEpsilon); })(variance, dy);
_c = test_util_1.expectArraysClose;
return [4 /*yield*/, gradVariance.data()];
case 3:
_c.apply(void 0, [_f.sent(), [-3.533, -4.686, -1.360, -2.762]]);
expect(gradVariance.shape).toEqual([2, 1, 2]);
gradOffset = tf.grad(function (offset) { return tf.batchNorm3d(x, mean, variance, offset, scale, varianceEpsilon); })(offset, dy);
_d = test_util_1.expectArraysClose;
return [4 /*yield*/, gradOffset.data()];
case 4:
_d.apply(void 0, [_f.sent(), [1, 1, 1, 1]]);
expect(gradOffset.shape).toEqual([2, 1, 2]);
gradScale = tf.grad(function (scale) { return tf.batchNorm3d(x, mean, variance, offset, scale, varianceEpsilon); })(scale, dy);
_e = test_util_1.expectArraysClose;
return [4 /*yield*/, gradScale.data()];
case 5:
_e.apply(void 0, [_f.sent(), [7.069, 7.499, 8.164, 8.838]]);
expect(gradScale.shape).toEqual([2, 1, 2]);
return [2 /*return*/];
}
});
}); });
it('batchnorm matches tensorflow, 2x3x3', function () { return __awaiter(_this, void 0, void 0, function () {
var x, mean, variance, offset, scale, varianceEpsilon, result, _a;
return __generator(this, function (_b) {
switch (_b.label) {
case 0:
x = tf.tensor3d([
0.49955603, 0.04158615, -1.09440524, 2.03854165, -0.61578344,
2.87533573, 1.18105987, 0.807462, 1.87888837, 2.26563962, -0.37040935,
1.35848753, -0.75347094, 0.15683117, 0.91925946, 0.34121279,
0.92717143, 1.89683965
], [2, 3, 3]);
mean = tf.tensor1d([0.39745062, -0.48062894, 0.4847822]);
variance = tf.tensor1d([0.32375343, 0.67117643, 1.08334653]);
offset = tf.tensor1d([0.69398749, -1.29056387, 0.9429723]);
scale = tf.tensor1d([-0.5607271, 0.9878457, 0.25181573]);
varianceEpsilon = .001;
result = tf.batchNorm3d(x, mean, variance, offset, scale, varianceEpsilon);
_a = test_util_1.expectArraysClose;
return [4 /*yield*/, result.data()];
case 1:
_a.apply(void 0, [_b.sent(), [
0.59352049, -0.66135202, 0.5610874, -0.92077015, -1.45341019, 1.52106473,
-0.07704776, 0.26144429, 1.28010017, -1.14422404, -1.15776136, 1.15425493,
1.82644104, -0.52249442, 1.04803919, 0.74932291, 0.40568101, 1.2844412
]]);
return [2 /*return*/];
}
});
}); });
});
jasmine_util_1.describeWithFlags('batchNorm2D', jasmine_util_1.ALL_ENVS, function () {
it('simple batchnorm2D, no offset or scale, 2x2', function () { return __awaiter(_this, void 0, void 0, function () {
var xT, meanT, varianceT, varianceEpsilon, result, x, mean, variance, _a;
return __generator(this, function (_b) {
switch (_b.label) {
case 0:
xT = tf.tensor2d([2, 4, 9, 23], [2, 2]);
meanT = tf.tensor1d([1, 2]);
varianceT = tf.tensor1d([2, 3]);
varianceEpsilon = .001;
result = tf.batchNorm2d(xT, meanT, varianceT, undefined, undefined, varianceEpsilon);
return [4 /*yield*/, xT.buffer()];
case 1:
x = _b.sent();
return [4 /*yield*/, meanT.buffer()];
case 2:
mean = _b.sent();
return [4 /*yield*/, varianceT.buffer()];
case 3:
variance = _b.sent();
_a = test_util_1.expectArraysClose;
return [4 /*yield*/, result.data()];
case 4:
_a.apply(void 0, [_b.sent(), [
(x.get(0, 0) - mean.get(0)) * 1 /
Math.sqrt(variance.get(0) + varianceEpsilon),
(x.get(0, 1) - mean.get(1)) * 1 /
Math.sqrt(variance.get(1) + varianceEpsilon),
(x.get(1, 0) - mean.get(0)) * 1 /
Math.sqrt(variance.get(0) + varianceEpsilon),
(x.get(1, 1) - mean.get(1)) * 1 /
Math.sqrt(variance.get(1) + varianceEpsilon)
]]);
return [2 /*return*/];
}
});
}); });
it('simple batchnorm2D, no offset, 2x2', function () { return __awaiter(_this, void 0, void 0, function () {
var xT, meanT, varianceT, scaleT, varianceEpsilon, result, x, mean, variance, scale, _a;
return __generator(this, function (_b) {
switch (_b.label) {
case 0:
xT = tf.tensor2d([2, 4, 9, 23], [2, 2]);
meanT = tf.tensor1d([1, 2]);
varianceT = tf.tensor1d([2, 3]);
scaleT = tf.tensor1d([4, 5]);
varianceEpsilon = .001;
result = tf.batchNorm2d(xT, meanT, varianceT, undefined, scaleT, varianceEpsilon);
return [4 /*yield*/, xT.buffer()];
case 1:
x = _b.sent();
return [4 /*yield*/, meanT.buffer()];
case 2:
mean = _b.sent();
return [4 /*yield*/, varianceT.buffer()];
case 3:
variance = _b.sent();
return [4 /*yield*/, scaleT.buffer()];
case 4:
scale = _b.sent();
_a = test_util_1.expectArraysClose;
return [4 /*yield*/, result.data()];
case 5:
_a.apply(void 0, [_b.sent(), [
(x.get(0, 0) - mean.get(0)) * scale.get(0) /
Math.sqrt(variance.get(0) + varianceEpsilon),
(x.get(0, 1) - mean.get(1)) * scale.get(1) /
Math.sqrt(variance.get(1) + varianceEpsilon),
(x.get(1, 0) - mean.get(0)) * scale.get(0) /
Math.sqrt(variance.get(0) + varianceEpsilon),
(x.get(1, 1) - mean.get(1)) * scale.get(1) /
Math.sqrt(variance.get(1) + varianceEpsilon)
]]);
return [2 /*return*/];
}
});
}); });
it('simple batchnorm2D, no scale, 2x2', function () { return __awaiter(_this, void 0, void 0, function () {
var xT, meanT, varianceT, offsetT, varianceEpsilon, result, offset, mean, variance, x, _a;
return __generator(this, function (_b) {
switch (_b.label) {
case 0:
xT = tf.tensor2d([2, 4, 9, 23], [2, 2]);
meanT = tf.tensor1d([1, 2]);
varianceT = tf.tensor1d([2, 3]);
offsetT = tf.tensor1d([4, 5]);
varianceEpsilon = .001;
result = tf.batchNorm2d(xT, meanT, varianceT, offsetT, undefined, varianceEpsilon);
return [4 /*yield*/, offsetT.array()];
case 1:
offset = _b.sent();
return [4 /*yield*/, meanT.array()];
case 2:
mean = _b.sent();
return [4 /*yield*/, varianceT.array()];
case 3:
variance = _b.sent();
return [4 /*yield*/, xT.array()];
case 4:
x = _b.sent();
_a = test_util_1.expectArraysClose;
return [4 /*yield*/, result.data()];
case 5:
_a.apply(void 0, [_b.sent(), [
offset[0] +
(x[0][0] - mean[0]) * 1 / Math.sqrt(variance[0] + varianceEpsilon),
offset[1] +
(x[0][1] - mean[1]) * 1 / Math.sqrt(variance[1] + varianceEpsilon),
offset[0] +
(x[1][0] - mean[0]) * 1 / Math.sqrt(variance[0] + varianceEpsilon),
offset[1] +
(x[1][1] - mean[1]) * 1 / Math.sqrt(variance[1] + varianceEpsilon)
]]);
return [2 /*return*/];
}
});
}); });
it('simple batchnorm2D, 2x2', function () { return __awaiter(_this, void 0, void 0, function () {
var xT, meanT, varianceT, offsetT, scaleT, varianceEpsilon, result, offset, mean, variance, scale, x, _a;
return __generator(this, function (_b) {
switch (_b.label) {
case 0:
xT = tf.tensor2d([2, 4, 9, 23], [2, 2]);
meanT = tf.tensor1d([1, 2]);
varianceT = tf.tensor1d([2, 3]);
offsetT = tf.tensor1d([3, 4]);
scaleT = tf.tensor1d([4, 5]);
varianceEpsilon = .001;
result = tf.batchNorm2d(xT, meanT, varianceT, offsetT, scaleT, varianceEpsilon);
return [4 /*yield*/, offsetT.array()];
case 1:
offset = _b.sent();
return [4 /*yield*/, meanT.array()];
case 2:
mean = _b.sent();
return [4 /*yield*/, varianceT.array()];
case 3:
variance = _b.sent();
return [4 /*yield*/, scaleT.array()];
case 4:
scale = _b.sent();
return [4 /*yield*/, xT.array()];
case 5:
x = _b.sent();
_a = test_util_1.expectArraysClose;
return [4 /*yield*/, result.data()];
case 6:
_a.apply(void 0, [_b.sent(), [
offset[0] +
(x[0][0] - mean[0]) * scale[0] /
Math.sqrt(variance[0] + varianceEpsilon),
offset[1] +
(x[0][1] - mean[1]) * scale[1] /
Math.sqrt(variance[1] + varianceEpsilon),
offset[0] +
(x[1][0] - mean[0]) * scale[0] /
Math.sqrt(variance[0] + varianceEpsilon),
offset[1] +
(x[1][1] - mean[1]) * scale[1] /
Math.sqrt(variance[1] + varianceEpsilon)
]]);
return [2 /*return*/];
}
});
}); });
it('simple batchnorm2D gradients, 2x2', function () { return __awaiter(_this, void 0, void 0, function () {
var x, mean, variance, offset, scale, varianceEpsilon, dy, _a, gradX, gradMean, gradVariance, gradOffset, gradScale, _b, _c, _d, _e, _f;
return __generator(this, function (_g) {
switch (_g.label) {
case 0:
x = tf.tensor2d([2, 4, 9, 23], [2, 2]);
mean = tf.tensor1d([1, 2]);
variance = tf.tensor1d([2, 3]);
offset = tf.tensor1d([3, 4]);
scale = tf.tensor1d([2, 5]);
varianceEpsilon = .001;
dy = tf.tensor2d([1, 1, 1, 1], [2, 2]);
_a = tf.grads(function (x, mean, variance, offset, scale) {
return tf.batchNorm2d(x, mean, variance, offset, scale, varianceEpsilon);
})([x, mean, variance, offset, scale], dy), gradX = _a[0], gradMean = _a[1], gradV