@tensorflow/tfjs-core
Version:
Hardware-accelerated JavaScript library for machine intelligence
547 lines • 24.2 kB
JavaScript
;
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 backend_cpu_1 = require("./kernels/backend_cpu");
var backend_webgl_1 = require("./kernels/backend_webgl");
var test_util_1 = require("./test_util");
jasmine_util_1.describeWithFlags('fromPixels + regular math op', test_util_1.WEBGL_ENVS, function () {
it('debug mode does not error when no nans', function () {
var pixels = new ImageData(2, 2);
for (var i = 0; i < 8; i++) {
pixels.data[i] = 100;
}
for (var i = 8; i < 16; i++) {
pixels.data[i] = 250;
}
var a = tf.fromPixels(pixels, 4);
var b = tf.scalar(20, 'int32');
var res = tf.add(a, b);
test_util_1.expectArraysEqual(res, [
120, 120, 120, 120, 120, 120, 120, 120, 270, 270, 270, 270, 270, 270, 270,
270
]);
});
});
jasmine_util_1.describeWithFlags('gradients', test_util_1.ALL_ENVS, function () {
it('matmul + relu', function () {
var a = tf.tensor2d([-1, 2, -3, 10, -20, 30], [2, 3]);
var b = tf.tensor2d([2, -3, 4, -1, 2, -3], [3, 2]);
var _a = tf.grads(function (a, b) {
var m = tf.matMul(a, b);
var y = tf.relu(m);
return tf.sum(y);
})([a, b]), da = _a[0], db = _a[1];
var dedm = tf.step(tf.matMul(a, b));
expect(da.shape).toEqual(a.shape);
var transposeA = false;
var transposeB = true;
test_util_1.expectArraysClose(da, tf.matMul(dedm, b, transposeA, transposeB));
expect(db.shape).toEqual(b.shape);
transposeA = true;
transposeB = false;
test_util_1.expectArraysClose(db, tf.matMul(a, dedm, transposeA, transposeB));
});
it('grad(f)', function () {
var grad = tf.grad(function (x) { return x.square(); });
var result = grad(tf.tensor1d([.1, .2]));
test_util_1.expectArraysClose(result, [.2, .4]);
});
it('calling grad(f) twice works', function () {
var grad = tf.grad(function (x) { return x.square(); });
var result = grad(tf.tensor1d([.1, .2]));
var result2 = grad(tf.tensor1d([.1, .4]));
test_util_1.expectArraysClose(result, [.2, .4]);
test_util_1.expectArraysClose(result2, [.2, .8]);
});
it('grads(f)', function () {
var grads = tf.grads(function (x) { return x.square(); });
var result = grads([tf.tensor1d([.1, .2])]);
test_util_1.expectArraysClose(result[0], [.2, .4]);
});
it('calling grads(f) twice works', function () {
var grads = tf.grads(function (x) { return x.square(); });
var result = grads([tf.tensor1d([.1, .2])]);
var result2 = grads([tf.tensor1d([.1, .4])]);
test_util_1.expectArraysClose(result[0], [.2, .4]);
test_util_1.expectArraysClose(result2[0], [.2, .8]);
});
it('works with reshape', function () {
var a = tf.tensor2d([1, 2, 3, 4], [2, 2]);
var exponent = tf.tensor1d([2, 2, 2, 2], 'int32');
var da = tf.grad(function (a) {
var b = a.flatten();
var m = tf.pow(b, exponent);
return tf.sum(m);
})(a);
expect(da.shape).toEqual([2, 2]);
test_util_1.expectArraysClose(da, [2, 4, 6, 8]);
});
it('reshape outside tf.grads() throws error', function () {
var a = tf.tensor2d([1, 2, 3, 4], [2, 2]);
var b = a.flatten();
var exponent = tf.tensor1d([2, 2, 2, 2], 'int32');
var f = function () {
tf.grads(function (a, b) {
var m = tf.pow(b, exponent);
return tf.sum(m);
})([a, b]);
};
expect(f).toThrowError();
});
it('does not error if irrelevant (pruned) ops are missing grads', function () {
var a = tf.tensor1d([true, true], 'bool');
var b = tf.tensor1d([false, true], 'bool');
var da = tf.grad(function (a) {
a.logicalAnd(b);
return a.sum();
})(a);
test_util_1.expectArraysClose(da, [1, 1]);
});
it('errors if relevant ops are missing grads', function () {
var a = tf.tensor1d([true, true], 'bool');
var b = tf.tensor1d([false, true], 'bool');
var dfda = tf.grad(function (a) {
return a.logicalAnd(b);
});
expect(function () { return dfda(a); }).toThrowError();
});
it('works with asType', function () {
var a = tf.tensor2d([1, 2, 3, 4], [2, 2], 'int32');
var exponent = tf.tensor2d([2, 2, 2, 2], [2, 2], 'int32');
var da = tf.grad(function (a) {
var b = a.toFloat();
var m = tf.pow(b, exponent);
return tf.sum(m);
})(a);
expect(da.shape).toEqual([2, 2]);
expect(da.dtype).toEqual('float32');
test_util_1.expectArraysClose(da, [2, 4, 6, 8]);
});
it('asType outside of tf.grads() throws error', function () {
var a = tf.tensor2d([1, 2, 3, 4], [2, 2], 'int32');
var b = a.toFloat();
var exponent = tf.tensor2d([2, 2, 2, 2], [2, 2], 'int32');
var f = function () {
tf.grad(function (a) {
var m = tf.pow(b, exponent);
return tf.sum(m);
})(a);
};
expect(f).toThrowError();
});
});
jasmine_util_1.describeWithFlags('valueAndGradients', test_util_1.ALL_ENVS, function () {
it('matmul + relu', function () {
var a = tf.tensor2d([-1, 2, -3, 10, -20, 30], [2, 3]);
var b = tf.tensor2d([2, -3, 4, -1, 2, -3], [3, 2]);
var _a = tf.valueAndGrads(function (a, b) {
var m = tf.matMul(a, b);
var y = tf.relu(m);
return tf.sum(y);
})([a, b]), value = _a.value, grads = _a.grads;
test_util_1.expectNumbersClose(value.get(), 10);
var dedm = tf.step(tf.matMul(a, b));
var da = grads[0], db = grads[1];
var transposeA = false;
var transposeB = true;
test_util_1.expectArraysClose(da, tf.matMul(dedm, b, transposeA, transposeB));
transposeA = true;
transposeB = false;
test_util_1.expectArraysClose(db, tf.matMul(a, dedm, transposeA, transposeB));
});
it('matmul + relu + inner tidy', function () {
var a = tf.tensor2d([-1, 2, -3, 10, -20, 30], [2, 3]);
var b = tf.tensor2d([2, -3, 4, -1, 2, -3], [3, 2]);
var _a = tf.valueAndGrads(function (a, b) {
var m = tf.matMul(a, b);
return tf.tidy(function () {
var y = tf.relu(m);
return tf.sum(y);
});
})([a, b]), value = _a.value, grads = _a.grads;
test_util_1.expectNumbersClose(value.get(), 10);
var dedm = tf.step(tf.matMul(a, b));
var da = grads[0], db = grads[1];
var transposeA = false;
var transposeB = true;
test_util_1.expectArraysClose(da, tf.matMul(dedm, b, transposeA, transposeB));
transposeA = true;
transposeB = false;
test_util_1.expectArraysClose(db, tf.matMul(a, dedm, transposeA, transposeB));
});
});
jasmine_util_1.describeWithFlags('higher-order gradients', test_util_1.ALL_ENVS, function () {
it('grad(grad(f))', function () {
var gradgrad = tf.grad(tf.grad(function (x) { return x.mul(x).mul(x); }));
var result = gradgrad(tf.tensor1d([.1, .2]));
test_util_1.expectArraysClose(result, [.6, 1.2]);
});
it('grads(grads(f))', function () {
var grads = tf.grads(function (x) { return x.mul(x).mul(x); });
var gradsgrads = tf.grads(function (x) { return grads([x])[0]; });
var result = gradsgrads([tf.tensor1d([.1, .2])]);
test_util_1.expectArraysClose(result[0], [.6, 1.2]);
});
});
jasmine_util_1.describeWithFlags('customGradient', test_util_1.ALL_ENVS, function () {
it('basic', function () {
var a = tf.scalar(3);
var b = tf.scalar(2, 'int32');
var dy = tf.scalar(4);
var customPow = tf.customGrad(function (a) {
var value = tf.pow(a, b);
var gradFunc = function (dy) { return dy.mul(tf.scalar(0.1)); };
return { value: value, gradFunc: gradFunc };
});
var _a = tf.valueAndGrad(function (a) { return customPow(a); })(a, dy), value = _a.value, grad = _a.grad;
expect(value.shape).toEqual(a.shape);
test_util_1.expectArraysClose(value, [9]);
expect(grad.shape).toEqual(a.shape);
test_util_1.expectArraysClose(grad, [.4]);
});
it('second order derivative through customGradient', function () {
var a = tf.scalar(3);
var b = tf.scalar(2, 'int32');
var dy = tf.scalar(5);
var customPow = tf.customGrad(function (a) {
var value = tf.pow(a, b);
var gradFunc = function (dy) { return dy.mul(a); };
return { value: value, gradFunc: gradFunc };
});
var dda = tf.grad(tf.grad(function (a) { return customPow(a); }))(a, dy);
expect(dda.shape).toEqual(a.shape);
test_util_1.expectArraysClose(dda, dy);
});
it('calling gradient of custom op twice works', function () {
var customOp = tf.customGrad(function (x) {
return { value: x.square(), gradFunc: function (dy) { return dy.mul(x.abs()); } };
});
var x = tf.tensor1d([-1, -2, 3]);
var grad = tf.grad(function (x) { return customOp(x); });
test_util_1.expectArraysClose(grad(x), [1, 2, 3]);
test_util_1.expectArraysClose(grad(x), [1, 2, 3]);
});
});
jasmine_util_1.describeWithFlags('memory', test_util_1.ALL_ENVS, function () {
it('Sum(float)', function () {
expect(tf.memory().numTensors).toBe(0);
expect(tf.memory().numBytes).toBe(0);
var sum = tf.tidy(function () {
var a = tf.tensor1d([1, 2, 3, 4]);
expect(tf.memory().numTensors).toBe(1);
expect(tf.memory().numBytes).toBe(4 * 4);
return a.sum();
});
expect(tf.memory().numTensors).toBe(1);
expect(tf.memory().numBytes).toBe(4);
test_util_1.expectArraysClose(sum, [1 + 2 + 3 + 4]);
});
it('Sum(bool)', function () {
var sum = tf.tidy(function () {
var a = tf.tensor1d([true, true, false, true], 'bool');
expect(tf.memory().numTensors).toBe(1);
expect(tf.memory().numBytes).toBe(4);
return a.sum();
});
expect(tf.memory().numTensors).toBe(1);
expect(tf.memory().numBytes).toBe(4);
expect(sum.dtype).toBe('int32');
test_util_1.expectArraysClose(sum, [1 + 1 + 0 + 1]);
});
it('Sum(int32)', function () {
var sum = tf.tidy(function () {
var a = tf.tensor1d([1, 1, 0, 1], 'int32');
expect(tf.memory().numTensors).toBe(1);
expect(tf.memory().numBytes).toBe(4 * 4);
return a.sum();
});
expect(tf.memory().numTensors).toBe(1);
expect(tf.memory().numBytes).toBe(4);
expect(sum.dtype).toBe('int32');
test_util_1.expectArraysClose(sum, [1 + 1 + 0 + 1]);
});
it('string tensor', function () {
var a = tf.tensor([['a', 'bb'], ['c', 'd']]);
expect(tf.memory().numTensors).toBe(1);
expect(tf.memory().numBytes).toBe(10);
a.dispose();
expect(tf.memory().numTensors).toBe(0);
expect(tf.memory().numBytes).toBe(0);
});
it('unreliable is true for string tensors', function () {
tf.tensor('a');
var mem = tf.memory();
expect(mem.unreliable).toBe(true);
var expectedReason = 'Memory usage by string tensors is approximate ' +
'(2 bytes per character)';
expect(mem.reasons.indexOf(expectedReason) >= 0).toBe(true);
});
});
jasmine_util_1.describeWithFlags('memory webgl', test_util_1.WEBGL_ENVS, function () {
it('unreliable is falsy/not present when all tensors are numeric', function () {
tf.tensor(1);
var mem = tf.memory();
expect(mem.numTensors).toBe(1);
expect(mem.numDataBuffers).toBe(1);
expect(mem.numBytes).toBe(4);
expect(mem.unreliable).toBeFalsy();
});
});
jasmine_util_1.describeWithFlags('memory cpu', test_util_1.CPU_ENVS, function () {
it('unreliable is true due to auto gc', function () {
tf.tensor(1);
var mem = tf.memory();
expect(mem.numTensors).toBe(1);
expect(mem.numDataBuffers).toBe(1);
expect(mem.numBytes).toBe(4);
expect(mem.unreliable).toBe(true);
var expectedReason = 'The reported memory is an upper bound. Due to automatic garbage ' +
'collection, the true allocated memory may be less.';
expect(mem.reasons.indexOf(expectedReason) >= 0).toBe(true);
});
it('unreliable is true due to both auto gc and string tensors', function () {
tf.tensor(1);
tf.tensor('a');
var mem = tf.memory();
expect(mem.numTensors).toBe(2);
expect(mem.numDataBuffers).toBe(2);
expect(mem.numBytes).toBe(6);
expect(mem.unreliable).toBe(true);
var expectedReasonGC = 'The reported memory is an upper bound. Due to automatic garbage ' +
'collection, the true allocated memory may be less.';
expect(mem.reasons.indexOf(expectedReasonGC) >= 0).toBe(true);
var expectedReasonString = 'Memory usage by string tensors is approximate ' +
'(2 bytes per character)';
expect(mem.reasons.indexOf(expectedReasonString) >= 0).toBe(true);
});
});
jasmine_util_1.describeWithFlags('profile', test_util_1.ALL_ENVS, function () {
it('squaring', function () { return __awaiter(_this, void 0, void 0, function () {
var profile, result;
return __generator(this, function (_a) {
switch (_a.label) {
case 0: return [4, tf.profile(function () {
var x = tf.tensor1d([1, 2, 3]);
var x2 = x.square();
x2.dispose();
x2 = x.square();
x2.dispose();
return x;
})];
case 1:
profile = _a.sent();
result = profile.result;
expect(profile.newBytes).toBe(12);
expect(profile.peakBytes).toBe(24);
expect(profile.newTensors).toBe(1);
test_util_1.expectArraysClose(result, [1, 2, 3]);
expect(profile.kernels).toEqual([
{
'name': 'square',
'bytesAdded': 12,
'totalBytesSnapshot': 24,
'tensorsAdded': 1,
'totalTensorsSnapshot': 2,
'inputShapes': [[3]],
'outputShape': [3]
},
{
'name': 'square',
'bytesAdded': 12,
'totalBytesSnapshot': 24,
'tensorsAdded': 1,
'totalTensorsSnapshot': 2,
'inputShapes': [[3]],
'outputShape': [3]
}
]);
return [2];
}
});
}); });
it('squaring without disposing', function () { return __awaiter(_this, void 0, void 0, function () {
var profile, result;
return __generator(this, function (_a) {
switch (_a.label) {
case 0: return [4, tf.profile(function () {
var x = tf.tensor1d([1, 2, 3]);
var x2 = x.square();
return x2;
})];
case 1:
profile = _a.sent();
result = profile.result;
expect(profile.newBytes).toBe(24);
expect(profile.peakBytes).toBe(24);
expect(profile.newTensors).toBe(2);
test_util_1.expectArraysClose(result, [1, 4, 9]);
expect(profile.kernels).toEqual([{
'name': 'square',
'bytesAdded': 12,
'totalBytesSnapshot': 24,
'tensorsAdded': 1,
'totalTensorsSnapshot': 2,
'inputShapes': [[3]],
'outputShape': [3]
}]);
return [2];
}
});
}); });
});
jasmine_util_1.describeWithFlags('disposeVariables', test_util_1.ALL_ENVS, function () {
it('reuse same name variable', function () {
tf.tensor1d([1, 2, 3]).variable(true, 'v1');
tf.tensor1d([1, 2, 3]).variable(true, 'v2');
expect(function () {
tf.tensor1d([1, 2, 3]).variable(true, 'v1');
}).toThrowError();
tf.disposeVariables();
tf.tensor1d([1, 2, 3]).variable(true, 'v1');
tf.tensor1d([1, 2, 3]).variable(true, 'v2');
});
});
describe('Switching cpu backends', function () {
beforeEach(function () {
tf.ENV.registerBackend('cpu1', function () { return new backend_cpu_1.MathBackendCPU(); });
tf.ENV.registerBackend('cpu2', function () { return new backend_cpu_1.MathBackendCPU(); });
});
afterEach(function () {
tf.ENV.removeBackend('cpu1');
tf.ENV.removeBackend('cpu2');
});
it('Move data from cpu1 to cpu2 backend', function () {
tf.setBackend('cpu1');
var a = tf.scalar(5);
tf.setBackend('cpu2');
var b = tf.scalar(3);
expect(tf.memory().numDataBuffers).toBe(2);
expect(tf.memory().numTensors).toBe(2);
expect(tf.memory().numBytes).toBe(8);
test_util_1.expectArraysClose(a, [5]);
test_util_1.expectArraysClose(b, [3]);
tf.setBackend('cpu1');
test_util_1.expectArraysClose(a, [5]);
test_util_1.expectArraysClose(b, [3]);
tf.dispose([a, b]);
expect(tf.memory().numDataBuffers).toBe(0);
expect(tf.memory().numTensors).toBe(0);
expect(tf.memory().numBytes).toBe(0);
});
it('can execute op with data from mixed backends', function () {
tf.setBackend('cpu1');
var a = tf.scalar(5);
tf.setBackend('cpu2');
var b = tf.scalar(3);
tf.tidy(function () {
tf.setBackend('cpu1');
test_util_1.expectArraysClose(tf.add(a, b), [8]);
tf.setBackend('cpu2');
test_util_1.expectArraysClose(tf.add(a, b), [8]);
});
expect(tf.memory().numTensors).toBe(2);
expect(tf.memory().numDataBuffers).toBe(2);
tf.dispose([a, b]);
expect(tf.memory().numTensors).toBe(0);
expect(tf.memory().numDataBuffers).toBe(0);
});
});
jasmine_util_1.describeWithFlags('Switching WebGL + CPU backends', test_util_1.WEBGL_ENVS, function () {
beforeEach(function () {
tf.ENV.registerBackend('webgl1', function () { return new backend_webgl_1.MathBackendWebGL(); });
tf.ENV.registerBackend('webgl2', function () { return new backend_webgl_1.MathBackendWebGL(); });
tf.ENV.registerBackend('cpu1', function () { return new backend_cpu_1.MathBackendCPU(); });
});
afterEach(function () {
tf.ENV.removeBackend('webgl1');
tf.ENV.removeBackend('webgl2');
tf.ENV.removeBackend('cpu1');
});
it('can execute op with data from mixed backends', function () {
tf.setBackend('webgl1');
var a = tf.scalar(5);
tf.setBackend('webgl2');
var b = tf.scalar(3);
tf.setBackend('cpu1');
var c = tf.scalar(2);
tf.tidy(function () {
tf.setBackend('webgl1');
test_util_1.expectArraysClose(tf.addN([a, b, c]), [10]);
tf.setBackend('webgl2');
test_util_1.expectArraysClose(tf.addN([a, b, c]), [10]);
tf.setBackend('cpu1');
test_util_1.expectArraysClose(tf.addN([a, b, c]), [10]);
});
expect(tf.memory().numTensors).toBe(3);
expect(tf.memory().numDataBuffers).toBe(3);
tf.dispose([a, b, c]);
expect(tf.memory().numTensors).toBe(0);
expect(tf.memory().numDataBuffers).toBe(0);
});
it('fromPixels with mixed backends works', function () {
tf.setBackend('webgl1');
var a = tf.fromPixels(new ImageData(new Uint8ClampedArray([1, 2, 3, 4]), 1, 1));
tf.setBackend('webgl2');
var b = tf.fromPixels(new ImageData(new Uint8ClampedArray([5, 6, 7, 8]), 1, 1));
test_util_1.expectArraysClose(tf.add(a, b), [6, 8, 10]);
});
it('single tidy multiple backends', function () {
expect(tf.memory().numTensors).toBe(0);
tf.tidy(function () {
tf.setBackend('webgl1');
var a = tf.scalar(1);
a.square();
tf.setBackend('webgl2');
var b = tf.scalar(1);
b.square();
expect(tf.memory().numTensors).toBe(4);
});
expect(tf.memory().numTensors).toBe(0);
});
});
describe('Memory allocation outside a test scope', function () {
it('constructing a tensor works', function () {
tf.setBackend('cpu');
var a = tf.tensor1d([1, 2, 3]);
test_util_1.expectArraysClose(a, [1, 2, 3]);
a.dispose();
});
});
//# sourceMappingURL=engine_test.js.map