Huge News!Announcing our $40M Series B led by Abstract Ventures.Learn More →
Socket
Sign inDemoInstall
Socket
Sign inDemoInstall

@tensorflow/tfjs-layers

Package Overview
Dependencies
Maintainers
11
Versions
159
Alerts
File Explorer

Advanced tools

License
Socket logo

Install Socket

Detect and block malicious and high-risk dependencies

Install

@tensorflow/tfjs-layers - npm Package Compare versions

Comparing version 0.4.1 to 0.5.0

2

dist/backend/tfjs_backend.d.ts

@@ -6,2 +6,3 @@ import { Scalar, Tensor, Tensor1D, Tensor2D } from '@tensorflow/tfjs-core';

import { epsilon as common_epsilon } from './common';
export declare function disposeScalarCache(): void;
export declare function setBackend(requestedBackend: 'cpu' | 'webgl'): void;

@@ -33,2 +34,3 @@ export declare function getBackend(): 'cpu' | 'webgl';

export declare function sliceAlongLastAxis(array: Tensor, start: number, size: number): Tensor;
export declare function normalizeBatchInTraining(x: Tensor, gamma: Tensor, beta: Tensor, reductionAxes: number[], epsilon?: number): [Tensor, Tensor, Tensor];
export declare function concatenate(tensors: Tensor[], axis?: number): Tensor;

@@ -35,0 +37,0 @@ export declare function concatAlongFirstAxis(a: Tensor, b: Tensor): Tensor;

89

dist/backend/tfjs_backend.js

@@ -5,3 +5,2 @@ "use strict";

var tfjs_core_1 = require("@tensorflow/tfjs-core");
var _ = require("underscore");
var common_1 = require("../common");

@@ -24,2 +23,3 @@ var errors_1 = require("../errors");

}
exports.disposeScalarCache = disposeScalarCache;
function setBackend(requestedBackend) {

@@ -206,4 +206,4 @@ tfc.setBackend(requestedBackend);

default:
throw new errors_1.ValueError("sliceAlongFirstAxis() received an unsupported subtype of Tensor: " +
("" + array.constructor.name));
throw new errors_1.ValueError("sliceAlongFirstAxis() received an unsupported tensor rank: " +
("" + array.rank));
}

@@ -223,7 +223,51 @@ }

default:
throw new errors_1.ValueError("sliceAlongLastAxis() received an unsupported subtype of Tensor: " +
("" + array.constructor.name));
throw new errors_1.ValueError("sliceAlongLastAxis() received an unsupported tensor rank: " +
("" + array.rank));
}
}
exports.sliceAlongLastAxis = sliceAlongLastAxis;
function regularNormalizeBatchInTraining(x, gamma, beta, reductionAxes, epsilon) {
if (epsilon === void 0) { epsilon = 1e-3; }
return tfjs_core_1.tidy(function () {
var meanAndVariance = tfc.moments(x, reductionAxes);
var mean = meanAndVariance.mean;
var variance = meanAndVariance.variance;
var normed = batchNormalization(x, mean, variance, beta, gamma, epsilon);
return [normed, mean, variance];
});
}
function broadcastNormalizeBatchInTraining(x, gamma, beta, reductionAxes, epsilon) {
if (epsilon === void 0) { epsilon = 1e-3; }
return tfjs_core_1.tidy(function () {
var meanAndVariance = tfc.moments(x, reductionAxes);
var mean = meanAndVariance.mean;
var variance = meanAndVariance.variance;
var targetShape = [];
for (var _i = 0, _a = math_utils.range(0, ndim(x)); _i < _a.length; _i++) {
var axis = _a[_i];
if (reductionAxes.indexOf(axis) !== -1) {
targetShape.push(1);
}
else {
targetShape.push(x.shape[axis]);
}
}
var broadcastMean = reshape(mean, targetShape);
var broadcastVariance = reshape(variance, targetShape);
var broadcastGamma = gamma == null ? null : reshape(gamma, targetShape);
var broadcastBeta = beta == null ? null : reshape(beta, targetShape);
var normed = batchNormalization(x, broadcastMean, broadcastVariance, broadcastBeta, broadcastGamma, epsilon);
return [normed, mean, variance];
});
}
function normalizeBatchInTraining(x, gamma, beta, reductionAxes, epsilon) {
if (epsilon === void 0) { epsilon = 1e-3; }
if (tfjs_core_1.util.arraysEqual(reductionAxes.slice().sort(), math_utils.range(0, ndim(x) - 1))) {
return regularNormalizeBatchInTraining(x, gamma, beta, reductionAxes, epsilon);
}
else {
return broadcastNormalizeBatchInTraining(x, gamma, beta, reductionAxes, epsilon);
}
}
exports.normalizeBatchInTraining = normalizeBatchInTraining;
function concatenate(tensors, axis) {

@@ -258,4 +302,4 @@ if (axis === void 0) { axis = -1; }

default:
throw new errors_1.ValueError('concatAlongFirstAxis() received an unsupported subtype of ' +
'Tensor: ' + a.constructor.name);
throw new errors_1.ValueError('concatAlongFirstAxis() received an unsupported tensor rank: ' +
a.rank);
}

@@ -497,3 +541,4 @@ }

}
return tfc.oneHot(indices, numClasses);
indices = indices.toInt();
return tfc.oneHot(indices, numClasses).toFloat();
}

@@ -513,4 +558,7 @@ exports.oneHot = oneHot;

if (Array.isArray(indices)) {
indices = tfjs_core_1.tensor1d(indices);
indices = tfjs_core_1.tensor1d(indices, 'int32');
}
else {
indices = indices.toInt();
}
return tfc.gather(reference, indices, axis);

@@ -597,9 +645,9 @@ }

if (ndim(x) === 2) {
out = tfc.batchNormalization2d(x, mean, variance, epsilon);
out = tfc.batchNormalization2d(x, mean, variance, epsilon, gamma, beta);
}
else if (ndim(x) === 3) {
out = tfc.batchNormalization3d(x, mean, variance, epsilon);
out = tfc.batchNormalization3d(x, mean, variance, epsilon, gamma, beta);
}
else if (ndim(x) === 4) {
out = tfc.batchNormalization4d(x, mean, variance, epsilon);
out = tfc.batchNormalization4d(x, mean, variance, epsilon, gamma, beta);
}

@@ -610,8 +658,2 @@ else {

}
if (gamma != null) {
out = multiply(out, gamma);
}
if (beta != null) {
out = add(out, beta);
}
return out;

@@ -724,3 +766,3 @@ }

function dropout(x, level, noiseShape, seed) {
if (noiseShape != null && !_.isEqual(x.shape, noiseShape)) {
if (noiseShape != null && !tfjs_core_1.util.arraysEqual(x.shape, noiseShape)) {
throw new errors_1.NotImplementedError('Non-default noise shape is not implemented yet: ' +

@@ -919,3 +961,3 @@ JSON.stringify(noiseShape));

output = clip(output, exports.epsilon(), 1 - exports.epsilon());
return tfc.neg(tfc.sum(tfc.mul(target, tfc.log(output)), shape(output).length - 1));
return tfc.neg(tfc.sum(tfc.mul(target.toFloat(), tfc.log(output)), shape(output).length - 1));
}

@@ -925,3 +967,3 @@ exports.categoricalCrossentropy = categoricalCrossentropy;

if (fromLogits === void 0) { fromLogits = false; }
var flatTarget = tfc.floor(flatten(target));
var flatTarget = tfc.floor(flatten(target)).toInt();
var outputShape = shape(output);

@@ -974,3 +1016,3 @@ var oneHotTarget = reshape(tfc.oneHot(flatTarget, outputShape[outputShape.length - 1]), outputShape);

}
var axes = [1, 0].concat(_.range(2, ndim));
var axes = [1, 0].concat(math_utils.range(2, ndim));
inputs = transpose(inputs, axes);

@@ -1011,3 +1053,4 @@ if (mask != null) {

lastOutput,
transpose(outputs, [1, 0].concat(_.range(2, outputs.shape.length))), states
transpose(outputs, [1, 0].concat(math_utils.range(2, outputs.shape.length))),
states
];

@@ -1014,0 +1057,0 @@ }

8

dist/constraints.d.ts
import { Tensor } from '@tensorflow/tfjs-core';
import { ConfigDict, ConfigDictValue } from './types';
export declare abstract class Constraint {
import { ConfigDict, ConfigDictValue, Serializable } from './types';
export declare abstract class Constraint extends Serializable {
abstract apply(w: Tensor): Tensor;

@@ -18,2 +18,3 @@ getConfig(): ConfigDict;

apply(w: Tensor): Tensor;
getClassName(): string;
getConfig(): ConfigDict;

@@ -29,2 +30,3 @@ }

apply(w: Tensor): Tensor;
getClassName(): string;
getConfig(): ConfigDict;

@@ -34,2 +36,3 @@ }

apply(w: Tensor): Tensor;
getClassName(): string;
}

@@ -53,2 +56,3 @@ export interface MinMaxNormConfig {

apply(w: Tensor): Tensor;
getClassName(): string;
getConfig(): ConfigDict;

@@ -55,0 +59,0 @@ }

19

dist/constraints.js

@@ -21,2 +21,3 @@ "use strict";

var K = require("./backend/tfjs_backend");
var types_1 = require("./types");
var generic_utils_1 = require("./utils/generic_utils");

@@ -26,4 +27,6 @@ function calcL2Norms(w, axis) {

}
var Constraint = (function () {
var Constraint = (function (_super) {
__extends(Constraint, _super);
function Constraint() {
return _super !== null && _super.apply(this, arguments) || this;
}

@@ -37,3 +40,3 @@ Constraint.prototype.getConfig = function () {

return Constraint;
}());
}(types_1.Serializable));
exports.Constraint = Constraint;

@@ -56,2 +59,5 @@ var MaxNorm = (function (_super) {

};
MaxNorm.prototype.getClassName = function () {
return 'MaxNorm';
};
MaxNorm.prototype.getConfig = function () {

@@ -75,2 +81,5 @@ return { maxValue: this.maxValue, axis: this.axis };

};
UnitNorm.prototype.getClassName = function () {
return 'UnitNorm';
};
UnitNorm.prototype.getConfig = function () {

@@ -91,2 +100,5 @@ return { axis: this.axis };

};
NonNeg.prototype.getClassName = function () {
return 'NonNeg';
};
return NonNeg;

@@ -117,2 +129,5 @@ }(Constraint));

};
MinMaxNorm.prototype.getClassName = function () {
return 'MinMaxNorm';
};
MinMaxNorm.prototype.getConfig = function () {

@@ -119,0 +134,0 @@ return {

7

dist/engine/topology.d.ts

@@ -5,3 +5,3 @@ import { Scalar, Tensor } from '@tensorflow/tfjs-core';

import { Regularizer } from '../regularizers';
import { ConfigDict, DType, JsonDict, LayerVariable, NamedTensorMap, RegularizerFn, Shape, SymbolicTensor, TensorInterface } from '../types';
import { ConfigDict, DType, JsonDict, LayerVariable, NamedTensorMap, RegularizerFn, Serializable, Shape, SymbolicTensor, TensorInterface } from '../types';
import * as generic_utils from '../utils/generic_utils';

@@ -70,3 +70,3 @@ export declare type Op = (x: LayerVariable) => LayerVariable;

export declare type CallHook = (inputs: Tensor | Tensor[], kwargs: any) => void;
export declare class Layer {
export declare abstract class Layer extends Serializable {
name: string;

@@ -136,2 +136,3 @@ inputSpec: InputSpec[];

apply(inputs: Tensor | Tensor[] | SymbolicTensor | SymbolicTensor[], kwargs?: any): Tensor | Tensor[] | SymbolicTensor;
getClassName(): string;
getConfig(): ConfigDict;

@@ -152,3 +153,3 @@ }

}
export declare class Container extends Layer {
export declare abstract class Container extends Layer {
inputs: SymbolicTensor[];

@@ -155,0 +156,0 @@ outputs: SymbolicTensor[];

5

dist/engine/training.d.ts

@@ -56,2 +56,4 @@ import { Optimizer, Scalar, Tensor, Tensor1D } from '@tensorflow/tfjs-core';

[outputName: string]: string;
} | LossOrMetricFn | LossOrMetricFn[] | {
[outputName: string]: LossOrMetricFn;
};

@@ -66,2 +68,4 @@ metrics?: string[] | {

[outputName: string]: string;
} | LossOrMetricFn | LossOrMetricFn[] | {
[outputName: string]: LossOrMetricFn;
};

@@ -80,2 +84,3 @@ lossFunctions: LossOrMetricFn[];

constructor(config: ContainerConfig);
getClassName(): string;
compile(config: ModelCompileConfig): void;

@@ -82,0 +87,0 @@ private checkTrainableWeightsConsistency();

34

dist/engine/training.js

@@ -56,3 +56,2 @@ "use strict";

var tfjs_core_1 = require("@tensorflow/tfjs-core");
var _ = require("underscore");
var K = require("../backend/tfjs_backend");

@@ -65,2 +64,3 @@ var callbacks_1 = require("../callbacks");

var generic_utils_1 = require("../utils/generic_utils");
var math_utils_1 = require("../utils/math_utils");
var executor_1 = require("./executor");

@@ -176,5 +176,5 @@ var topology_1 = require("./topology");

function checkArrayLengths(inputs, targets, weights) {
var setX = _.unique(inputs.map(function (input) { return input.shape[0]; }));
var setX = generic_utils_1.unique(inputs.map(function (input) { return input.shape[0]; }));
setX.sort();
var setY = _.unique(targets.map(function (target) { return target.shape[0]; }));
var setY = generic_utils_1.unique(targets.map(function (target) { return target.shape[0]; }));
setY.sort();

@@ -191,3 +191,3 @@ if (setX.length > 1) {

}
if (setX.length > 0 && setY.length > 0 && !_.isEqual(setX, setY)) {
if (setX.length > 0 && setY.length > 0 && !tfjs_core_1.util.arraysEqual(setX, setY)) {
throw new errors_1.ValueError("Input Tensors should have the same number of samples as target " +

@@ -219,3 +219,3 @@ ("Tensors. Found " + setX[0] + " input sample(s) and " + setY[0] + " target ") +

}
if (_.contains(keyLosses, loss)) {
if (keyLosses.indexOf(loss) !== -1) {
var slicedYShape = y.shape.slice(1);

@@ -269,3 +269,3 @@ var slicedShape = shape.slice(1);

else {
return K.gather(arrays, indices);
return K.gather(arrays, indices.dtype === 'int32' ? indices : indices.toInt());
}

@@ -357,2 +357,5 @@ }

}
Model.prototype.getClassName = function () {
return 'Model';
};
Model.prototype.compile = function (config) {

@@ -374,6 +377,7 @@ var _this = this;

var lossFunctions = [];
if (!Array.isArray(config.loss) && typeof config.loss !== 'string') {
if (!Array.isArray(config.loss) && typeof config.loss !== 'string' &&
typeof config.loss !== 'function') {
config.loss = config.loss;
for (var name_3 in config.loss) {
if (!_.contains(this.outputNames, name_3)) {
if (this.outputNames.indexOf(name_3) === -1) {
throw new errors_1.ValueError("Unknown entry in loss dictionary: \"" + name_3 + "\". Only expect the " +

@@ -398,3 +402,4 @@ ("following keys: " + this.outputNames));

}
lossFunctions = config.loss.map(function (l) { return losses.get(l); });
var theLosses = config.loss;
lossFunctions = theLosses.map(function (l) { return losses.get(l); });
}

@@ -679,3 +684,3 @@ else {

if (numTrainSamples != null) {
indexArray = _.range(numTrainSamples);
indexArray = math_utils_1.range(0, numTrainSamples);
}

@@ -706,3 +711,3 @@ this.history = new callbacks_1.History();

_loop_4 = function (epoch) {
var epochLogs, epochIndexArray, epochIndexArray1D_1, batches_1, _loop_5, batchIndex;
var epochLogs, epochIndexArray1D_1, batches_1, _loop_5, batchIndex;
return __generator(this, function (_a) {

@@ -714,3 +719,2 @@ switch (_a.label) {

epochLogs = {};
epochIndexArray = indexArray;
if (!(stepsPerEpoch != null)) return [3, 2];

@@ -723,5 +727,5 @@ throw new errors_1.NotImplementedError('stepsPerEpoch mode is not implemented yet.');

else if (shuffle) {
epochIndexArray = _.shuffle(indexArray);
tfjs_core_1.util.shuffle(indexArray);
}
epochIndexArray1D_1 = tfjs_core_1.tensor1d(epochIndexArray);
epochIndexArray1D_1 = tfjs_core_1.tensor1d(indexArray);
batches_1 = makeBatches(numTrainSamples, batchSize);

@@ -826,3 +830,3 @@ _loop_5 = function (batchIndex) {

var batches = makeBatches(numSamples, batchSize);
var indexArray = tfjs_core_1.tensor1d(_.range(numSamples));
var indexArray = tfjs_core_1.tensor1d(math_utils_1.range(0, numSamples));
for (var batchIndex = 0; batchIndex < batches.length; ++batchIndex) {

@@ -829,0 +833,0 @@ var batchStart = batches[batchIndex][0];

28

dist/exports.d.ts

@@ -0,1 +1,2 @@

import { Tensor } from '@tensorflow/tfjs-core';
import { Constraint, MaxNormConfig, MinMaxNormConfig, UnitNormConfig } from './constraints';

@@ -8,3 +9,3 @@ import { ContainerConfig, InputConfig, InputLayerConfig, Layer, LayerConfig } from './engine/topology';

import { DepthwiseConv2DLayerConfig } from './layers/convolutional_depthwise';
import { ActivationLayerConfig, DenseLayerConfig, DropoutLayerConfig, RepeatVectorLayerConfig } from './layers/core';
import { ActivationLayerConfig, DenseLayerConfig, DropoutLayerConfig, RepeatVectorLayerConfig, ReshapeLayerConfig } from './layers/core';
import { EmbeddingLayerConfig } from './layers/embeddings';

@@ -16,3 +17,3 @@ import { ConcatenateLayerConfig } from './layers/merge';

import { GRUCellLayerConfig, GRULayerConfig, LSTMCellLayerConfig, LSTMLayerConfig, RNNCell, RNNLayerConfig, SimpleRNNCellLayerConfig, SimpleRNNLayerConfig, StackedRNNCellsConfig } from './layers/recurrent';
import { BidirectionalLayerConfig, WrapperLayerConfig } from './layers/wrappers';
import { BidirectionalLayerConfig, Wrapper, WrapperLayerConfig } from './layers/wrappers';
import { Sequential, SequentialConfig } from './models';

@@ -35,3 +36,3 @@ import { L1Config, L1L2Config, L2Config, Regularizer } from './regularizers';

static softmax(config?: SoftmaxLayerConfig): Layer;
static thresohldedReLU(config?: ThresholdedReLULayerConfig): Layer;
static thresholdedReLU(config?: ThresholdedReLULayerConfig): Layer;
static conv1d(config: ConvLayerConfig): Layer;

@@ -47,6 +48,7 @@ static conv2d(config: ConvLayerConfig): Layer;

static repeatVector(config: RepeatVectorLayerConfig): Layer;
static reshape(config: ReshapeLayerConfig): Layer;
static embedding(config: EmbeddingLayerConfig): Layer;
static add(config?: LayerConfig): Layer;
static average(config?: LayerConfig): Layer;
static concatenate(config: ConcatenateLayerConfig): Layer;
static concatenate(config?: ConcatenateLayerConfig): Layer;
static maximum(config?: LayerConfig): Layer;

@@ -56,3 +58,3 @@ static minimum(config?: LayerConfig): Layer;

static batchNormalization(config: BatchNormalizationLayerConfig): Layer;
static zeroPadding2d(config: ZeroPadding2DLayerConfig): Layer;
static zeroPadding2d(config?: ZeroPadding2DLayerConfig): Layer;
static averagePooling1d(config: Pooling1DLayerConfig): Layer;

@@ -78,3 +80,3 @@ static avgPool1d(config: Pooling1DLayerConfig): Layer;

static stackedRNNCells(config: StackedRNNCellsConfig): RNNCell;
static bidirectional(config: BidirectionalLayerConfig): Layer;
static bidirectional(config: BidirectionalLayerConfig): Wrapper;
static timeDistributed(config: WrapperLayerConfig): Layer;

@@ -103,2 +105,16 @@ }

}
export declare class MetricExports {
static binaryAccuracy(yTrue: Tensor, yPred: Tensor): Tensor;
static binaryCrossentropy(yTrue: Tensor, yPred: Tensor): Tensor;
static categoricalAccuracy(yTrue: Tensor, yPred: Tensor): Tensor;
static categoricalCrossentropy(yTrue: Tensor, yPred: Tensor): Tensor;
static cosineProximity(yTrue: Tensor, yPred: Tensor): Tensor;
meanAbsoluteError(yTrue: Tensor, yPred: Tensor): Tensor;
meanAbsolutePercentageError(yTrue: Tensor, yPred: Tensor): Tensor;
MAPE(yTrue: Tensor, yPred: Tensor): Tensor;
mape(yTrue: Tensor, yPred: Tensor): Tensor;
static meanSquaredError(yTrue: Tensor, yPred: Tensor): Tensor;
static MSE(yTrue: Tensor, yPred: Tensor): Tensor;
static mse(yTrue: Tensor, yPred: Tensor): Tensor;
}
export declare class RegularizerExports {

@@ -105,0 +121,0 @@ static l1l2(config?: L1L2Config): Regularizer;

112

dist/exports.js

@@ -25,2 +25,4 @@ "use strict";

var wrappers_1 = require("./layers/wrappers");
var losses_1 = require("./losses");
var metrics_1 = require("./metrics");
var models_1 = require("./models");

@@ -82,3 +84,3 @@ var regularizers_1 = require("./regularizers");

};
LayerExports.thresohldedReLU = function (config) {
LayerExports.thresholdedReLU = function (config) {
return new advanced_activations_1.ThresholdedReLU(config);

@@ -116,2 +118,5 @@ };

};
LayerExports.reshape = function (config) {
return new core_1.Reshape(config);
};
LayerExports.embedding = function (config) {

@@ -257,3 +262,3 @@ return new embeddings_1.Embedding(config);

})
], LayerExports, "thresohldedReLU", null);
], LayerExports, "thresholdedReLU", null);
__decorate([

@@ -354,2 +359,11 @@ tfjs_core_1.doc({

namespace: 'layers',
useDocsFrom: 'Reshape',
configParamIndices: [0]
})
], LayerExports, "reshape", null);
__decorate([
tfjs_core_1.doc({
heading: 'Layers',
subheading: 'Basic',
namespace: 'layers',
useDocsFrom: 'Embedding',

@@ -784,2 +798,96 @@ configParamIndices: [0]

exports.InitializerExports = InitializerExports;
var MetricExports = (function () {
function MetricExports() {
}
MetricExports.binaryAccuracy = function (yTrue, yPred) {
return metrics_1.binaryAccuracy(yTrue, yPred);
};
MetricExports.binaryCrossentropy = function (yTrue, yPred) {
return metrics_1.binaryCrossentropy(yTrue, yPred);
};
MetricExports.categoricalAccuracy = function (yTrue, yPred) {
return metrics_1.categoricalAccuracy(yTrue, yPred);
};
MetricExports.categoricalCrossentropy = function (yTrue, yPred) {
return losses_1.categoricalCrossentropy(yTrue, yPred);
};
MetricExports.cosineProximity = function (yTrue, yPred) {
return losses_1.cosineProximity(yTrue, yPred);
};
MetricExports.prototype.meanAbsoluteError = function (yTrue, yPred) {
return losses_1.meanAbsoluteError(yTrue, yPred);
};
MetricExports.prototype.meanAbsolutePercentageError = function (yTrue, yPred) {
return losses_1.meanAbsolutePercentageError(yTrue, yPred);
};
MetricExports.prototype.MAPE = function (yTrue, yPred) {
return losses_1.meanAbsolutePercentageError(yTrue, yPred);
};
MetricExports.prototype.mape = function (yTrue, yPred) {
return losses_1.meanAbsolutePercentageError(yTrue, yPred);
};
MetricExports.meanSquaredError = function (yTrue, yPred) {
return losses_1.meanSquaredError(yTrue, yPred);
};
MetricExports.MSE = function (yTrue, yPred) {
return losses_1.meanSquaredError(yTrue, yPred);
};
MetricExports.mse = function (yTrue, yPred) {
return losses_1.meanSquaredError(yTrue, yPred);
};
__decorate([
tfjs_core_1.doc({
heading: 'Metrics',
namespace: 'metrics',
useDocsFrom: 'meanAbsoluteError'
})
], MetricExports.prototype, "meanAbsoluteError", null);
__decorate([
tfjs_core_1.doc({
heading: 'Metrics',
namespace: 'metrics',
useDocsFrom: 'meanAbsolutePercentageError'
})
], MetricExports.prototype, "meanAbsolutePercentageError", null);
__decorate([
tfjs_core_1.doc({ heading: 'Metrics', namespace: 'metrics', useDocsFrom: 'binaryAccuracy' })
], MetricExports, "binaryAccuracy", null);
__decorate([
tfjs_core_1.doc({
heading: 'Metrics',
namespace: 'metrics',
useDocsFrom: 'binaryCrossentropy'
})
], MetricExports, "binaryCrossentropy", null);
__decorate([
tfjs_core_1.doc({
heading: 'Metrics',
namespace: 'metrics',
useDocsFrom: 'categoricalAccuracy'
})
], MetricExports, "categoricalAccuracy", null);
__decorate([
tfjs_core_1.doc({
heading: 'Metrics',
namespace: 'metrics',
useDocsFrom: 'categoricalCrossentropy'
})
], MetricExports, "categoricalCrossentropy", null);
__decorate([
tfjs_core_1.doc({
heading: 'Metrics',
namespace: 'metrics',
useDocsFrom: 'cosineProximity'
})
], MetricExports, "cosineProximity", null);
__decorate([
tfjs_core_1.doc({
heading: 'Metrics',
namespace: 'metrics',
useDocsFrom: 'meanSquaredError'
})
], MetricExports, "meanSquaredError", null);
return MetricExports;
}());
exports.MetricExports = MetricExports;
var RegularizerExports = (function () {

@@ -786,0 +894,0 @@ function RegularizerExports() {

5

dist/index.d.ts
import * as backend from './backend/tfjs_backend';
import { ConstraintExports, InitializerExports, LayerExports, ModelExports, RegularizerExports } from './exports';
import { ConstraintExports, InitializerExports, LayerExports, MetricExports, ModelExports, RegularizerExports } from './exports';
export { Callback, CallbackList, CustomCallback, CustomCallbackConfig, Logs } from './callbacks';

@@ -7,3 +7,3 @@ export { Model, ModelCompileConfig, ModelEvaluateConfig, ModelFitConfig, ModelPredictConfig } from './engine/training';

export { ModelAndWeightsConfig, Sequential, SequentialConfig } from './models';
export { SymbolicTensor } from './types';
export { Shape, SymbolicTensor } from './types';
export { version as version_layers } from './version';

@@ -18,2 +18,3 @@ export { backend };

export declare const initializers: typeof InitializerExports;
export declare const metrics: typeof MetricExports;
export declare const regularizers: typeof RegularizerExports;

1

dist/index.js

@@ -27,2 +27,3 @@ "use strict";

exports.initializers = exports_1.InitializerExports;
exports.metrics = exports_1.MetricExports;
exports.regularizers = exports_1.RegularizerExports;

13

dist/initializers.d.ts
import { Tensor } from '@tensorflow/tfjs-core';
import { DType, Shape } from './types';
import { DType, Serializable, Shape } from './types';
import { ConfigDict, ConfigDictValue } from './types';

@@ -11,3 +11,3 @@ import { Constructor } from './utils/generic_utils';

export declare function checkDistribution(value?: string): void;
export declare abstract class Initializer {
export declare abstract class Initializer extends Serializable {
static fromConfig<T>(cls: Constructor<T>, config: ConfigDict): T;

@@ -19,5 +19,7 @@ fromConfigUsesCustomObjects(): boolean;

export declare class Zeros extends Initializer {
getClassName(): string;
apply(shape: Shape, dtype?: DType): Tensor;
}
export declare class Ones extends Initializer {
getClassName(): string;
apply(shape: Shape, dtype?: DType): Tensor;

@@ -32,2 +34,3 @@ }

apply(shape: Shape, dtype?: DType): Tensor;
getClassName(): string;
getConfig(): ConfigDict;

@@ -48,2 +51,3 @@ }

apply(shape: Shape, dtype?: DType): Tensor;
getClassName(): string;
getConfig(): ConfigDict;

@@ -64,2 +68,3 @@ }

apply(shape: Shape, dtype?: DType): Tensor;
getClassName(): string;
getConfig(): ConfigDict;

@@ -80,2 +85,3 @@ }

apply(shape: Shape, dtype?: DType): Tensor;
getClassName(): string;
getConfig(): ConfigDict;

@@ -90,2 +96,3 @@ }

apply(shape: Shape, dtype?: DType): Tensor;
getClassName(): string;
getConfig(): ConfigDict;

@@ -106,2 +113,3 @@ }

apply(shape: Shape, dtype?: DType): Tensor;
getClassName(): string;
getConfig(): ConfigDict;

@@ -133,2 +141,3 @@ }

apply(shape: Shape, dtype?: DType): Tensor;
getClassName(): string;
getConfig(): ConfigDict;

@@ -135,0 +144,0 @@ }

36

dist/initializers.js

@@ -20,3 +20,2 @@ "use strict";

var tfjs_core_1 = require("@tensorflow/tfjs-core");
var _ = require("underscore");
var K = require("./backend/tfjs_backend");

@@ -50,4 +49,6 @@ var common_1 = require("./common");

exports.checkDistribution = checkDistribution;
var Initializer = (function () {
var Initializer = (function (_super) {
__extends(Initializer, _super);
function Initializer() {
return _super !== null && _super.apply(this, arguments) || this;
}

@@ -67,3 +68,3 @@ Initializer.fromConfig = function (cls, config) {

return Initializer;
}());
}(types_1.Serializable));
exports.Initializer = Initializer;

@@ -75,2 +76,5 @@ var Zeros = (function (_super) {

}
Zeros.prototype.getClassName = function () {
return 'Zeros';
};
Zeros.prototype.apply = function (shape, dtype) {

@@ -88,2 +92,5 @@ return K.zeros(shape, dtype);

}
Ones.prototype.getClassName = function () {
return 'Ones';
};
Ones.prototype.apply = function (shape, dtype) {

@@ -106,2 +113,5 @@ return K.ones(shape, dtype);

};
Constant.prototype.getClassName = function () {
return 'Constant';
};
Constant.prototype.getConfig = function () {

@@ -130,2 +140,5 @@ return {

};
RandomUniform.prototype.getClassName = function () {
return 'RandomUniform';
};
RandomUniform.prototype.getConfig = function () {

@@ -152,2 +165,5 @@ return { minval: this.minval, maxval: this.maxval, seed: this.seed };

};
RandomNormal.prototype.getClassName = function () {
return 'RandomNormal';
};
RandomNormal.prototype.getConfig = function () {

@@ -174,2 +190,5 @@ return { mean: this.mean, stddev: this.stddev, seed: this.seed };

};
TruncatedNormal.prototype.getClassName = function () {
return 'TruncatedNormal';
};
TruncatedNormal.prototype.getConfig = function () {

@@ -198,2 +217,5 @@ return { mean: this.mean, stddev: this.stddev, seed: this.seed };

};
Identity.prototype.getClassName = function () {
return 'Identity';
};
Identity.prototype.getConfig = function () {

@@ -215,3 +237,3 @@ return { gain: this.gain.get() };

}
else if (_.contains([3, 4, 5], shape.length)) {
else if ([3, 4, 5].indexOf(shape.length) !== -1) {
if (dataFormat === 'channelsFirst') {

@@ -273,2 +295,5 @@ var receptiveFieldSize = math_utils_1.arrayProd(shape, 2);

};
VarianceScaling.prototype.getClassName = function () {
return 'VarianceScaling';
};
VarianceScaling.prototype.getConfig = function () {

@@ -359,2 +384,5 @@ return {

};
Orthogonal.prototype.getClassName = function () {
return 'Orthogonal';
};
Orthogonal.prototype.getConfig = function () {

@@ -361,0 +389,0 @@ return {

4

dist/layers/advanced_activations.d.ts

@@ -13,2 +13,3 @@ import { Tensor } from '@tensorflow/tfjs-core';

computeOutputShape(inputShape: Shape | Shape[]): Shape | Shape[];
getClassName(): string;
getConfig(): ConfigDict;

@@ -25,2 +26,3 @@ }

computeOutputShape(inputShape: Shape | Shape[]): Shape | Shape[];
getClassName(): string;
getConfig(): ConfigDict;

@@ -38,2 +40,3 @@ }

computeOutputShape(inputShape: Shape | Shape[]): Shape | Shape[];
getClassName(): string;
getConfig(): ConfigDict;

@@ -50,3 +53,4 @@ }

computeOutputShape(inputShape: Shape | Shape[]): Shape | Shape[];
getClassName(): string;
getConfig(): ConfigDict;
}

12

dist/layers/advanced_activations.js

@@ -39,2 +39,5 @@ "use strict";

};
LeakyReLU.prototype.getClassName = function () {
return 'LeakyReLU';
};
LeakyReLU.prototype.getConfig = function () {

@@ -72,2 +75,5 @@ var config = { alpha: this.alpha };

};
ELU.prototype.getClassName = function () {
return 'ELU';
};
ELU.prototype.getConfig = function () {

@@ -102,2 +108,5 @@ var config = { alpha: this.alpha };

};
ThresholdedReLU.prototype.getClassName = function () {
return 'ThresholdedReLU';
};
ThresholdedReLU.prototype.getConfig = function () {

@@ -131,2 +140,5 @@ var config = { theta: this.theta };

};
Softmax.prototype.getClassName = function () {
return 'Softmax';
};
Softmax.prototype.getConfig = function () {

@@ -133,0 +145,0 @@ var config = { axis: this.axis };

1

dist/layers/convolutional_depthwise.d.ts

@@ -21,2 +21,3 @@ import { Tensor } from '@tensorflow/tfjs-core';

constructor(config: DepthwiseConv2DLayerConfig);
getClassName(): string;
build(inputShape: Shape | Shape[]): void;

@@ -23,0 +24,0 @@ call(inputs: Tensor | Tensor[], kwargs: any): Tensor | Tensor[];

3

dist/layers/convolutional_depthwise.js

@@ -34,2 +34,5 @@ "use strict";

}
DepthwiseConv2D.prototype.getClassName = function () {
return 'DepthwiseConv2D';
};
DepthwiseConv2D.prototype.build = function (inputShape) {

@@ -36,0 +39,0 @@ inputShape = generic_utils_1.getExactlyOneShape(inputShape);

5

dist/layers/convolutional.d.ts

@@ -55,2 +55,3 @@ import { Tensor } from '@tensorflow/tfjs-core';

constructor(config: ConvLayerConfig);
getClassName(): string;
getConfig(): ConfigDict;

@@ -61,2 +62,3 @@ }

constructor(config: ConvLayerConfig);
getClassName(): string;
build(inputShape: Shape | Shape[]): void;

@@ -91,2 +93,3 @@ call(inputs: Tensor | Tensor[], kwargs: any): Tensor | Tensor[];

call(inputs: Tensor | Tensor[], kwargs: any): Tensor | Tensor[];
getClassName(): string;
getConfig(): ConfigDict;

@@ -96,6 +99,8 @@ }

constructor(config?: SeparableConvLayerConfig);
getClassName(): string;
}
export declare class Conv1D extends Conv {
constructor(config: ConvLayerConfig);
getClassName(): string;
getConfig(): ConfigDict;
}

15

dist/layers/convolutional.js

@@ -163,2 +163,5 @@ "use strict";

}
Conv2D.prototype.getClassName = function () {
return 'Conv2D';
};
Conv2D.prototype.getConfig = function () {

@@ -184,2 +187,5 @@ var config = _super.prototype.getConfig.call(this);

}
Conv2DTranspose.prototype.getClassName = function () {
return 'Conv2DTranspose';
};
Conv2DTranspose.prototype.build = function (inputShape) {

@@ -379,2 +385,5 @@ inputShape = generic_utils.getExactlyOneShape(inputShape);

};
SeparableConv.prototype.getClassName = function () {
return 'SeparableConv';
};
SeparableConv.prototype.getConfig = function () {

@@ -408,2 +417,5 @@ var config = _super.prototype.getConfig.call(this);

}
SeparableConv2D.prototype.getClassName = function () {
return 'SeparableConv2D';
};
return SeparableConv2D;

@@ -420,2 +432,5 @@ }(SeparableConv));

}
Conv1D.prototype.getClassName = function () {
return 'Conv1D';
};
Conv1D.prototype.getConfig = function () {

@@ -422,0 +437,0 @@ var config = _super.prototype.getConfig.call(this);

6

dist/layers/core.d.ts

@@ -22,2 +22,3 @@ import { Tensor } from '@tensorflow/tfjs-core';

call(inputs: Tensor | Tensor[], kwargs: any): Tensor | Tensor[];
getClassName(): string;
getConfig(): ConfigDict;

@@ -56,2 +57,3 @@ }

call(inputs: Tensor | Tensor[], kwargs: any): Tensor | Tensor[];
getClassName(): string;
getConfig(): ConfigDict;

@@ -62,2 +64,3 @@ }

computeOutputShape(inputShape: Shape | Shape[]): Shape | Shape[];
getClassName(): string;
call(inputs: Tensor | Tensor[], kwargs: any): Tensor | Tensor[];

@@ -71,2 +74,3 @@ }

constructor(config: ActivationLayerConfig);
getClassName(): string;
call(inputs: Tensor | Tensor[], kwargs: any): Tensor | Tensor[];

@@ -85,2 +89,3 @@ }

call(inputs: Tensor | Tensor[], kwargs: any): Tensor | Tensor[];
getClassName(): string;
getConfig(): ConfigDict;

@@ -94,3 +99,4 @@ }

computeOutputShape(inputShape: Shape): Shape;
getClassName(): string;
call(inputs: Tensor | Tensor[], kwargs: any): Tensor | Tensor[];
}

23

dist/layers/core.js

@@ -13,3 +13,3 @@ "use strict";

Object.defineProperty(exports, "__esModule", { value: true });
var _ = require("underscore");
var tfjs_core_1 = require("@tensorflow/tfjs-core");
var activations_1 = require("../activations");

@@ -55,3 +55,4 @@ var K = require("../backend/tfjs_backend");

var input = generic_utils.getExactlyOneTensor(inputs);
if (this.noiseShape != null && !_.isEqual(input.shape, this.noiseShape)) {
if (this.noiseShape != null &&
!tfjs_core_1.util.arraysEqual(input.shape, this.noiseShape)) {
throw new errors_1.NotImplementedError('Non-default noise shape is not implemented in Dropout layer yet: ' +

@@ -68,2 +69,5 @@ JSON.stringify(this.noiseShape));

};
Dropout.prototype.getClassName = function () {
return 'Dropout';
};
Dropout.prototype.getConfig = function () {

@@ -148,2 +152,5 @@ var config = {

};
Dense.prototype.getClassName = function () {
return 'Dense';
};
Dense.prototype.getConfig = function () {

@@ -190,2 +197,5 @@ var config = {

};
Flatten.prototype.getClassName = function () {
return 'Flatten';
};
Flatten.prototype.call = function (inputs, kwargs) {

@@ -207,2 +217,5 @@ this.invokeCallHook(inputs, kwargs);

}
Activation.prototype.getClassName = function () {
return 'Activation';
};
Activation.prototype.call = function (inputs, kwargs) {

@@ -232,2 +245,5 @@ this.invokeCallHook(inputs, kwargs);

};
RepeatVector.prototype.getClassName = function () {
return 'RepeatVector';
};
RepeatVector.prototype.getConfig = function () {

@@ -306,2 +322,5 @@ var config = {

};
Reshape.prototype.getClassName = function () {
return 'Reshape';
};
Reshape.prototype.call = function (inputs, kwargs) {

@@ -308,0 +327,0 @@ this.invokeCallHook(inputs, kwargs);

1

dist/layers/embeddings.d.ts

@@ -33,3 +33,4 @@ import { Tensor } from '@tensorflow/tfjs-core';

call(inputs: Tensor | Tensor[], kwargs: any): Tensor | Tensor[];
getClassName(): string;
getConfig(): ConfigDict;
}

9

dist/layers/embeddings.js

@@ -13,3 +13,2 @@ "use strict";

Object.defineProperty(exports, "__esModule", { value: true });
var _ = require("underscore");
var K = require("../backend/tfjs_backend");

@@ -71,4 +70,5 @@ var constraints_1 = require("../constraints");

var i = 0;
for (var _i = 0, _a = _.zip(inLens, inputShape.slice(1)); _i < _a.length; _i++) {
var _b = _a[_i], s1 = _b[0], s2 = _b[1];
for (var k = 0; k < inLens.length; ++k) {
var s1 = inLens[k];
var s2 = inputShape[k + 1];
if ((s1 != null) && (s2 != null) && (s1 !== s2)) {

@@ -95,2 +95,5 @@ throw new errors_1.ValueError("\"inputLength\" is " + this.inputLength + ", but received " +

};
Embedding.prototype.getClassName = function () {
return 'Embedding';
};
Embedding.prototype.getConfig = function () {

@@ -97,0 +100,0 @@ var config = {

7

dist/layers/merge.d.ts

@@ -7,2 +7,3 @@ import { Tensor } from '@tensorflow/tfjs-core';

constructor(config?: LayerConfig);
getClassName(): string;
protected mergeFunction(inputs: Tensor[]): Tensor;

@@ -16,2 +17,3 @@ private computeElementwiseOpOutputShape(shape1, shape2);

constructor(config?: LayerConfig);
getClassName(): string;
protected mergeFunction(inputs: Tensor[]): Tensor;

@@ -22,2 +24,3 @@ }

constructor(config?: LayerConfig);
getClassName(): string;
protected mergeFunction(inputs: Tensor[]): Tensor;

@@ -28,2 +31,3 @@ }

constructor(config?: LayerConfig);
getClassName(): string;
protected mergeFunction(inputs: Tensor[]): Tensor;

@@ -34,2 +38,3 @@ }

constructor(config?: LayerConfig);
getClassName(): string;
protected mergeFunction(inputs: Tensor[]): Tensor;

@@ -40,2 +45,3 @@ }

constructor(config?: LayerConfig);
getClassName(): string;
protected mergeFunction(inputs: Tensor[]): Tensor;

@@ -51,2 +57,3 @@ }

constructor(config?: ConcatenateLayerConfig);
getClassName(): string;
build(inputShape: Shape | Shape[]): void;

@@ -53,0 +60,0 @@ protected mergeFunction(inputs: Tensor[]): Tensor;

40

dist/layers/merge.js

@@ -13,3 +13,3 @@ "use strict";

Object.defineProperty(exports, "__esModule", { value: true });
var _ = require("underscore");
var tfjs_core_1 = require("@tensorflow/tfjs-core");
var K = require("../backend/tfjs_backend");

@@ -27,2 +27,5 @@ var topology_1 = require("../engine/topology");

}
Merge.prototype.getClassName = function () {
return 'Merge';
};
Merge.prototype.mergeFunction = function (inputs) {

@@ -80,3 +83,3 @@ throw new errors_1.NotImplementedError();

}
batchSizes = _.uniq(batchSizes);
batchSizes = generic_utils.unique(batchSizes);
if (batchSizes.length > 1) {

@@ -92,3 +95,4 @@ throw new errors_1.ValueError("Can not merge tensors with different batch sizes. " +

var allRanks = inputShape.map(function (shape) { return shape.length; });
if (!_.contains(inputShape, null) && _.uniq(allRanks).length === 1) {
if (inputShape.indexOf(null) === -1 &&
generic_utils.unique(allRanks).length === 1) {
this.reshapeRequired = false;

@@ -105,4 +109,4 @@ }

var inputDims = inputs.map(function (input) { return K.ndim(input); });
if (!_.contains(inputDims, null)) {
var maxNDim = _.max(inputDims);
if (inputDims.indexOf(null) === -1) {
var maxNDim = mathUtils.max(inputDims);
for (var _i = 0, inputs_1 = inputs; _i < inputs_1.length; _i++) {

@@ -134,3 +138,3 @@ var x = inputs_1[_i];

else if (xNDim > 1) {
var dims = _.range(1, xNDim).concat([0]);
var dims = mathUtils.range(1, xNDim).concat([0]);
reshapedInputs.push(K.permuteDimensions(x, dims));

@@ -154,3 +158,3 @@ transposed = true;

else if (yNDim > 1) {
var dims = [yNDim - 1].concat(_.range(0, yNDim - 1));
var dims = [yNDim - 1].concat(mathUtils.range(0, yNDim - 1));
y = K.permuteDimensions(y, dims);

@@ -186,3 +190,3 @@ }

}
batchSizes = _.uniq(batchSizes);
batchSizes = generic_utils.unique(batchSizes);
if (batchSizes.length === 1) {

@@ -204,2 +208,5 @@ outputShape = batchSizes.concat(outputShape);

}
Add.prototype.getClassName = function () {
return 'Add';
};
Add.prototype.mergeFunction = function (inputs) {

@@ -232,2 +239,5 @@ var output = K.zeros(inputs[0].shape);

}
Multiply.prototype.getClassName = function () {
return 'Multiply';
};
Multiply.prototype.mergeFunction = function (inputs) {

@@ -260,2 +270,5 @@ var output = K.ones(inputs[0].shape);

}
Average.prototype.getClassName = function () {
return 'Average';
};
Average.prototype.mergeFunction = function (inputs) {

@@ -288,2 +301,5 @@ var output = K.zeros(inputs[0].shape);

}
Maximum.prototype.getClassName = function () {
return 'Maximum';
};
Maximum.prototype.mergeFunction = function (inputs) {

@@ -315,2 +331,5 @@ var output = inputs[0];

}
Minimum.prototype.getClassName = function () {
return 'Minimum';
};
Minimum.prototype.mergeFunction = function (inputs) {

@@ -350,2 +369,5 @@ var output = inputs[0];

}
Concatenate.prototype.getClassName = function () {
return 'Concatenate';
};
Concatenate.prototype.build = function (inputShape) {

@@ -376,3 +398,3 @@ if (!(Array.isArray(inputShape) && Array.isArray(inputShape[0])) ||

var shape = shapeSet_1[_a];
if (_.isEqual(shape, shapeWithoutConcatAxis)) {
if (tfjs_core_1.util.arraysEqual(shape, shapeWithoutConcatAxis)) {
exists = true;

@@ -379,0 +401,0 @@ break;

2

dist/layers/normalization.d.ts

@@ -41,6 +41,8 @@ import { Tensor } from '@tensorflow/tfjs-core';

private movingVariance;
private stepCount;
constructor(config: BatchNormalizationLayerConfig);
build(inputShape: Shape | Shape[]): void;
call(inputs: Tensor | Tensor[], kwargs: any): Tensor | Tensor[];
getClassName(): string;
getConfig(): ConfigDict;
}

75

dist/layers/normalization.js

@@ -13,3 +13,3 @@ "use strict";

Object.defineProperty(exports, "__esModule", { value: true });
var _ = require("underscore");
var tfjs_core_1 = require("@tensorflow/tfjs-core");
var K = require("../backend/tfjs_backend");

@@ -22,2 +22,3 @@ var constraints_1 = require("../constraints");

var generic_utils = require("../utils/generic_utils");
var math_utils_1 = require("../utils/math_utils");
var BatchNormalization = (function (_super) {

@@ -43,2 +44,3 @@ __extends(BatchNormalization, _super);

_this.gammaRegularizer = regularizers_1.getRegularizer(config.gammaRegularizer);
_this.stepCount = 0;
return _this;

@@ -71,32 +73,47 @@ }

var _this = this;
var training = kwargs['training'] == null ? false : kwargs['training'];
var input = generic_utils.getExactlyOneTensor(inputs);
var inputShape = K.shape(input);
var ndim = inputShape.length;
var reductionAxes = _.range(ndim);
var axis = this.axis >= 0 ? this.axis : (this.axis + ndim);
reductionAxes.splice(axis, 1);
var broadcastShape = generic_utils.pyListRepeat(1, ndim);
broadcastShape[axis] = inputShape[axis];
var sortedReductionAxes = reductionAxes.slice();
sortedReductionAxes.sort();
var needsBroadcasting = !_.isEqual(sortedReductionAxes, _.range(ndim).slice(0, ndim - 1));
var normalizeInference = function () {
if (needsBroadcasting) {
var broadcastMovingMean = K.reshape(_this.movingMean.read(), broadcastShape);
var broadcastMovingVariance = K.reshape(_this.movingVariance.read(), broadcastShape);
var broadcastBeta = _this.center ? K.reshape(_this.beta.read(), broadcastShape) : null;
var broadcastGamma = _this.scale ? K.reshape(_this.gamma.read(), broadcastShape) : null;
return K.batchNormalization(input, broadcastMovingMean, broadcastMovingVariance, broadcastBeta, broadcastGamma, _this.epsilon);
return tfjs_core_1.tidy(function () {
var training = kwargs['training'] == null ? false : kwargs['training'];
var input = generic_utils.getExactlyOneTensor(inputs);
var inputShape = K.shape(input);
var ndim = inputShape.length;
var reductionAxes = math_utils_1.range(0, ndim);
var axis = _this.axis >= 0 ? _this.axis : (_this.axis + ndim);
reductionAxes.splice(axis, 1);
var broadcastShape = generic_utils.pyListRepeat(1, ndim);
broadcastShape[axis] = inputShape[axis];
var sortedReductionAxes = reductionAxes.slice();
sortedReductionAxes.sort();
var needsBroadcasting = !tfjs_core_1.util.arraysEqual(sortedReductionAxes, math_utils_1.range(0, ndim).slice(0, ndim - 1));
var normalizeInference = function () {
if (needsBroadcasting) {
var broadcastMovingMean = K.reshape(_this.movingMean.read(), broadcastShape);
var broadcastMovingVariance = K.reshape(_this.movingVariance.read(), broadcastShape);
var broadcastBeta = _this.center ? K.reshape(_this.beta.read(), broadcastShape) : null;
var broadcastGamma = _this.scale ? K.reshape(_this.gamma.read(), broadcastShape) : null;
return K.batchNormalization(input, broadcastMovingMean, broadcastMovingVariance, broadcastBeta, broadcastGamma, _this.epsilon);
}
else {
return K.batchNormalization(input, _this.movingMean.read(), _this.movingVariance.read(), _this.beta == null ? null : _this.beta.read(), _this.gamma == null ? null : _this.gamma.read(), _this.epsilon);
}
};
if (!training) {
return normalizeInference();
}
else {
return K.batchNormalization(input, _this.movingMean.read(), _this.movingVariance.read(), _this.beta == null ? null : _this.beta.read(), _this.gamma == null ? null : _this.gamma.read(), _this.epsilon);
}
};
if (!training) {
return normalizeInference();
}
throw new errors_1.NotImplementedError('BatchNormalization.call() has not been implemented for training ' +
'mode yet.');
var _a = K.normalizeBatchInTraining(input, _this.gamma.read(), _this.beta.read(), reductionAxes, _this.epsilon), normedTraining = _a[0], mean = _a[1], variance = _a[2];
var sampleSize = math_utils_1.arrayProd(reductionAxes.map(function (axis) { return input.shape[axis]; }));
var varianceDebiased = variance.mul(K.getScalar(sampleSize / (sampleSize - (1 + _this.epsilon))));
var updateMovingMeanAndVariance = function () {
_this.stepCount++;
var newMovingMean = tfjs_core_1.movingAverage(_this.movingMean.read(), mean, _this.momentum, _this.stepCount);
_this.movingMean.write(newMovingMean);
var newMovingVariance = tfjs_core_1.movingAverage(_this.movingVariance.read(), varianceDebiased, _this.momentum, _this.stepCount);
_this.movingVariance.write(newMovingVariance);
};
updateMovingMeanAndVariance();
return normedTraining;
});
};
BatchNormalization.prototype.getClassName = function () {
return 'BatchNormalization';
};
BatchNormalization.prototype.getConfig = function () {

@@ -103,0 +120,0 @@ var config = {

1

dist/layers/padding.d.ts

@@ -15,3 +15,4 @@ import { Tensor } from '@tensorflow/tfjs-core';

call(inputs: Tensor | Tensor[], kwargs: any): Tensor | Tensor[];
getClassName(): string;
getConfig(): ConfigDict;
}

3

dist/layers/padding.js

@@ -105,2 +105,5 @@ "use strict";

};
ZeroPadding2D.prototype.getClassName = function () {
return 'ZeroPadding2D';
};
ZeroPadding2D.prototype.getConfig = function () {

@@ -107,0 +110,0 @@ var config = {

8

dist/layers/pooling.d.ts

@@ -22,2 +22,3 @@ import { Tensor } from '@tensorflow/tfjs-core';

constructor(config: Pooling1DLayerConfig);
getClassName(): string;
protected poolingFunction(inputs: Tensor, poolSize: [number, number], strides: [number, number], padding: PaddingMode, dataFormat: DataFormat): Tensor;

@@ -27,2 +28,3 @@ }

constructor(config: Pooling1DLayerConfig);
getClassName(): string;
protected poolingFunction(inputs: Tensor, poolSize: [number, number], strides: [number, number], padding: PaddingMode, dataFormat: DataFormat): Tensor;

@@ -49,2 +51,3 @@ }

constructor(config: Pooling2DLayerConfig);
getClassName(): string;
protected poolingFunction(inputs: Tensor, poolSize: [number, number], strides: [number, number], padding: PaddingMode, dataFormat: DataFormat): Tensor;

@@ -54,2 +57,3 @@ }

constructor(config: Pooling2DLayerConfig);
getClassName(): string;
protected poolingFunction(inputs: Tensor, poolSize: [number, number], strides: [number, number], padding: PaddingMode, dataFormat: DataFormat): Tensor;

@@ -64,2 +68,3 @@ }

constructor(config: LayerConfig);
getClassName(): string;
call(inputs: Tensor | Tensor[], kwargs: any): Tensor | Tensor[];

@@ -69,2 +74,3 @@ }

constructor(config: LayerConfig);
getClassName(): string;
call(inputs: Tensor | Tensor[], kwargs: any): Tensor | Tensor[];

@@ -84,5 +90,7 @@ }

call(inputs: Tensor | Tensor[], kwargs: any): Tensor | Tensor[];
getClassName(): string;
}
export declare class GlobalMaxPooling2D extends GlobalPooling2D {
call(inputs: Tensor | Tensor[], kwargs: any): Tensor | Tensor[];
getClassName(): string;
}

24

dist/layers/pooling.js

@@ -64,2 +64,5 @@ "use strict";

}
MaxPooling1D.prototype.getClassName = function () {
return 'MaxPooling1D';
};
MaxPooling1D.prototype.poolingFunction = function (inputs, poolSize, strides, padding, dataFormat) {

@@ -79,2 +82,5 @@ common_1.checkDataFormat(dataFormat);

}
AveragePooling1D.prototype.getClassName = function () {
return 'AveragePooling1D';
};
AveragePooling1D.prototype.poolingFunction = function (inputs, poolSize, strides, padding, dataFormat) {

@@ -147,2 +153,5 @@ common_1.checkDataFormat(dataFormat);

}
MaxPooling2D.prototype.getClassName = function () {
return 'MaxPooling2D';
};
MaxPooling2D.prototype.poolingFunction = function (inputs, poolSize, strides, padding, dataFormat) {

@@ -162,2 +171,5 @@ common_1.checkDataFormat(dataFormat);

}
AveragePooling2D.prototype.getClassName = function () {
return 'AveragePooling2D';
};
AveragePooling2D.prototype.poolingFunction = function (inputs, poolSize, strides, padding, dataFormat) {

@@ -193,2 +205,5 @@ common_1.checkDataFormat(dataFormat);

}
GlobalAveragePooling1D.prototype.getClassName = function () {
return 'GlobalAveragePooling1D';
};
GlobalAveragePooling1D.prototype.call = function (inputs, kwargs) {

@@ -207,2 +222,5 @@ var input = generic_utils.getExactlyOneTensor(inputs);

}
GlobalMaxPooling1D.prototype.getClassName = function () {
return 'GlobalMaxPooling1D';
};
GlobalMaxPooling1D.prototype.call = function (inputs, kwargs) {

@@ -261,2 +279,5 @@ var input = generic_utils.getExactlyOneTensor(inputs);

};
GlobalAveragePooling2D.prototype.getClassName = function () {
return 'GlobalAveragePooling2D';
};
return GlobalAveragePooling2D;

@@ -280,2 +301,5 @@ }(GlobalPooling2D));

};
GlobalMaxPooling2D.prototype.getClassName = function () {
return 'GlobalMaxPooling2D';
};
return GlobalMaxPooling2D;

@@ -282,0 +306,0 @@ }(GlobalPooling2D));

8

dist/layers/recurrent.d.ts

@@ -50,2 +50,3 @@ import { Tensor } from '@tensorflow/tfjs-core';

readonly nonTrainableWeights: LayerVariable[];
getClassName(): string;
getConfig(): ConfigDict;

@@ -98,2 +99,3 @@ }

call(inputs: Tensor | Tensor[], kwargs: any): Tensor | Tensor[];
getClassName(): string;
getConfig(): ConfigDict;

@@ -134,2 +136,3 @@ }

readonly recurrentDropout: number;
getClassName(): string;
getConfig(): ConfigDict;

@@ -170,2 +173,3 @@ }

call(inputs: Tensor | Tensor[], kwargs: any): Tensor | Tensor[];
getClassName(): string;
getConfig(): ConfigDict;

@@ -194,2 +198,3 @@ }

readonly implementation: number;
getClassName(): string;
getConfig(): ConfigDict;

@@ -233,2 +238,3 @@ static fromConfig<T>(cls: generic_utils.Constructor<T>, config: ConfigDict): T;

call(inputs: Tensor | Tensor[], kwargs: any): Tensor | Tensor[];
getClassName(): string;
getConfig(): ConfigDict;

@@ -259,2 +265,3 @@ }

readonly implementation: number;
getClassName(): string;
getConfig(): ConfigDict;

@@ -272,2 +279,3 @@ static fromConfig<T>(cls: generic_utils.Constructor<T>, config: ConfigDict): T;

build(inputShape: Shape | Shape[]): void;
getClassName(): string;
getConfig(): ConfigDict;

@@ -274,0 +282,0 @@ static fromConfig<T>(cls: generic_utils.Constructor<T>, config: ConfigDict, customObjects?: ConfigDict): T;

38

dist/layers/recurrent.js

@@ -20,3 +20,2 @@ "use strict";

var tfjs_core_1 = require("@tensorflow/tfjs-core");
var _ = require("underscore");
var activations_1 = require("../activations");

@@ -69,3 +68,3 @@ var K = require("../backend/tfjs_backend");

var numStates = Array.isArray(this.cell.stateSize) ? this.cell.stateSize.length : 1;
return _.range(numStates).map(function (x) { return null; });
return math_utils.range(0, numStates).map(function (x) { return null; });
}

@@ -138,3 +137,3 @@ else {

if (this.stateSpec != null) {
if (!_.isEqual(this.stateSpec.map(function (spec) { return spec.shape[spec.shape.length - 1]; }), stateSize)) {
if (!tfjs_core_1.util.arraysEqual(this.stateSpec.map(function (spec) { return spec.shape[spec.shape.length - 1]; }), stateSize)) {
throw new errors_1.ValueError("An initialState was passed that is not compatible with " +

@@ -197,3 +196,3 @@ ("cell.stateSize. Received stateSpec=" + this.stateSpec + "; ") +

var expectedShape = [batchSize, dim];
if (!_.isEqual(value.shape, expectedShape)) {
if (!tfjs_core_1.util.arraysEqual(value.shape, expectedShape)) {
throw new errors_1.ValueError("State " + index + " is incompatible with layer " + this.name + ": " +

@@ -355,2 +354,5 @@ ("expected shape=" + expectedShape + ", received shape=" + value.shape));

});
RNN.prototype.getClassName = function () {
return 'RNN';
};
RNN.prototype.getConfig = function () {

@@ -369,3 +371,3 @@ var config = {

config.cell = {
className: this.cell.constructor.name,
className: this.cell.getClassName(),
config: cellConfig,

@@ -454,2 +456,5 @@ };

};
SimpleRNNCell.prototype.getClassName = function () {
return 'SimpleRNNCell';
};
SimpleRNNCell.prototype.getConfig = function () {

@@ -593,2 +598,5 @@ var config = {

});
SimpleRNN.prototype.getClassName = function () {
return 'SimpleRNN';
};
SimpleRNN.prototype.getConfig = function () {

@@ -728,2 +736,5 @@ var config = {

};
GRUCell.prototype.getClassName = function () {
return 'GRUCell';
};
GRUCell.prototype.getConfig = function () {

@@ -879,2 +890,5 @@ var config = {

});
GRU.prototype.getClassName = function () {
return 'GRU';
};
GRU.prototype.getConfig = function () {

@@ -969,2 +983,5 @@ var config = {

};
CustomInit.prototype.getClassName = function () {
return 'CustomInit';
};
return CustomInit;

@@ -1053,2 +1070,5 @@ }(initializers_1.Initializer)))();

};
LSTMCell.prototype.getClassName = function () {
return 'LSTMCell';
};
LSTMCell.prototype.getConfig = function () {

@@ -1212,2 +1232,5 @@ var config = {

});
LSTM.prototype.getClassName = function () {
return 'LSTM';
};
LSTM.prototype.getConfig = function () {

@@ -1325,2 +1348,5 @@ var config = {

};
StackedRNNCells.prototype.getClassName = function () {
return 'StackedRNNCells';
};
StackedRNNCells.prototype.getConfig = function () {

@@ -1331,3 +1357,3 @@ var cellConfigs = [];

cellConfigs.push({
'className': this.constructor.name,
'className': this.getClassName(),
'config': cell.getConfig(),

@@ -1334,0 +1360,0 @@ });

4

dist/layers/serialization.d.ts

@@ -1,2 +0,2 @@

import { ConfigDict } from '../types';
export declare function deserialize(config: ConfigDict, customObjects?: ConfigDict): any;
import { ConfigDict, Serializable } from '../types';
export declare function deserialize(config: ConfigDict, customObjects?: ConfigDict): Serializable;

2

dist/layers/wrappers.d.ts

@@ -29,2 +29,3 @@ import { Tensor } from '@tensorflow/tfjs-core';

call(inputs: Tensor | Tensor[], kwargs: any): Tensor | Tensor[];
getClassName(): string;
}

@@ -59,2 +60,3 @@ export declare enum BidirectionalMergeMode {

readonly nonTrainableWeights: LayerVariable[];
getClassName(): string;
}

11

dist/layers/wrappers.js

@@ -82,3 +82,3 @@ "use strict";

'layer': {
'className': this.layer.constructor.name,
'className': this.layer.getClassName(),
'config': this.layer.getConfig(),

@@ -142,2 +142,5 @@ }

};
TimeDistributed.prototype.getClassName = function () {
return 'TimeDistributed';
};
return TimeDistributed;

@@ -168,3 +171,4 @@ }(Wrapper));

layerConfig['goBackwards'] === true ? false : true;
_this.backwardLayer = serialization_1.deserialize({ className: config.layer.constructor.name, config: layerConfig });
_this.backwardLayer =
serialization_1.deserialize({ className: config.layer.getClassName(), config: layerConfig });
_this.forwardLayer.name = 'forward_' + _this.forwardLayer.name;

@@ -338,2 +342,5 @@ _this.backwardLayer.name = 'backward_' + _this.backwardLayer.name;

});
Bidirectional.prototype.getClassName = function () {
return 'Bidirectional';
};
return Bidirectional;

@@ -340,0 +347,0 @@ }(Wrapper));

2

dist/losses.d.ts

@@ -28,2 +28,2 @@ import { Tensor } from '@tensorflow/tfjs-core';

export declare const cosine: typeof cosineProximity;
export declare function get(identifier: string): LossOrMetricFn;
export declare function get(identifierOrFn: string | LossOrMetricFn): LossOrMetricFn;

13

dist/losses.js

@@ -99,3 +99,3 @@ "use strict";

exports.cosine = cosineProximity;
function get(identifier) {
function get(identifierOrFn) {
var lossesMap = {

@@ -117,7 +117,12 @@ meanSquaredError: meanSquaredError,

};
if (identifier in lossesMap) {
return lossesMap[identifier];
if (typeof identifierOrFn === 'string') {
if (identifierOrFn in lossesMap) {
return lossesMap[identifierOrFn];
}
throw new errors_1.ValueError("Unknown loss " + identifierOrFn);
}
throw new errors_1.ValueError("Unknown loss " + identifier);
else {
return identifierOrFn;
}
}
exports.get = get;

6

dist/metrics.d.ts
import { Tensor } from '@tensorflow/tfjs-core';
import { cosineProximity, meanAbsoluteError, meanAbsolutePercentageError, meanSquaredError } from './losses';
import { categoricalCrossentropy as categoricalCrossentropyLoss, cosineProximity, meanAbsoluteError, meanAbsolutePercentageError, meanSquaredError, sparseCategoricalCrossentropy as sparseCategoricalCrossentropyLoss } from './losses';
import { LossOrMetricFn } from './types';

@@ -7,5 +7,3 @@ export declare function binaryAccuracy(yTrue: Tensor, yPred: Tensor): Tensor;

export declare function binaryCrossentropy(yTrue: Tensor, yPred: Tensor): Tensor;
export declare function categoricalCrossentropy(yTrue: Tensor, yPred: Tensor): Tensor;
export declare function sparseCategoricalAccuracy(yTrue: Tensor, yPred: Tensor): Tensor;
export declare function sparseCategoricalCrossentropy(yTrue: Tensor, yPred: Tensor): Tensor;
export declare function topKCategoricalAccuracy(yTrue: Tensor, yPred: Tensor): Tensor;

@@ -19,3 +17,5 @@ export declare function sparseTopKCategoricalAccuracy(yTrue: Tensor, yPred: Tensor): Tensor;

export declare const MAPE: typeof meanAbsolutePercentageError;
export declare const categoricalCrossentropy: typeof categoricalCrossentropyLoss;
export declare const cosine: typeof cosineProximity;
export declare const sparseCategoricalCrossentropy: typeof sparseCategoricalCrossentropyLoss;
export declare function get(identifier: string | LossOrMetricFn): LossOrMetricFn;

14

dist/metrics.js

@@ -18,9 +18,5 @@ "use strict";

function binaryCrossentropy(yTrue, yPred) {
throw new errors_1.NotImplementedError();
return K.mean(K.binaryCrossentropy(yTrue, yPred), -1);
}
exports.binaryCrossentropy = binaryCrossentropy;
function categoricalCrossentropy(yTrue, yPred) {
throw new errors_1.NotImplementedError();
}
exports.categoricalCrossentropy = categoricalCrossentropy;
function sparseCategoricalAccuracy(yTrue, yPred) {

@@ -30,6 +26,2 @@ throw new errors_1.NotImplementedError();

exports.sparseCategoricalAccuracy = sparseCategoricalAccuracy;
function sparseCategoricalCrossentropy(yTrue, yPred) {
throw new errors_1.NotImplementedError();
}
exports.sparseCategoricalCrossentropy = sparseCategoricalCrossentropy;
function topKCategoricalAccuracy(yTrue, yPred) {

@@ -49,3 +41,5 @@ throw new errors_1.NotImplementedError();

exports.MAPE = losses_1.meanAbsolutePercentageError;
exports.categoricalCrossentropy = losses_1.categoricalCrossentropy;
exports.cosine = losses_1.cosineProximity;
exports.sparseCategoricalCrossentropy = losses_1.sparseCategoricalCrossentropy;
function get(identifier) {

@@ -55,2 +49,4 @@ var metricsMap = {

categoricalAccuracy: categoricalAccuracy,
categoricalCrossentropy: exports.categoricalCrossentropy,
sparseCategoricalCrossentropy: exports.sparseCategoricalCrossentropy,
mse: exports.mse,

@@ -57,0 +53,0 @@ MSE: exports.MSE,

2

dist/models.d.ts

@@ -23,2 +23,3 @@ import { Scalar, Tensor, WeightsManifestConfig } from '@tensorflow/tfjs-core';

constructor(config?: SequentialConfig);
getClassName(): string;
add(layer: Layer): void;

@@ -40,2 +41,3 @@ pop(): void;

static fromConfig<T>(cls: generic_utils.Constructor<T>, config: ConfigDict): T;
getConfig(): any;
}

14

dist/models.js

@@ -138,2 +138,5 @@ "use strict";

Sequential_1 = Sequential;
Sequential.prototype.getClassName = function () {
return 'Sequential';
};
Sequential.prototype.add = function (layer) {

@@ -322,2 +325,13 @@ if (this.outputs.length === 0) {

};
Sequential.prototype.getConfig = function () {
var config = [];
for (var _i = 0, _a = this.layers; _i < _a.length; _i++) {
var layer = _a[_i];
config.push({
className: layer.getClassName(),
config: layer.getConfig(),
});
}
return config;
};
__decorate([

@@ -324,0 +338,0 @@ tfjs_core_1.doc({ heading: 'Models', subheading: 'Classes' })

5

dist/regularizers.d.ts
import { Scalar, Tensor } from '@tensorflow/tfjs-core';
import { ConfigDict, ConfigDictValue } from './types';
import { ConfigDict, ConfigDictValue, Serializable } from './types';
import * as generic_utils from './utils/generic_utils';
export declare abstract class Regularizer {
export declare abstract class Regularizer extends Serializable {
abstract apply(x: Tensor): Scalar;

@@ -24,2 +24,3 @@ }

apply(x: Tensor): Scalar;
getClassName(): string;
getConfig(): ConfigDict;

@@ -26,0 +27,0 @@ static fromConfig(cls: generic_utils.Constructor<L1L2>, config: ConfigDict): L1L2;

10

dist/regularizers.js

@@ -21,8 +21,11 @@ "use strict";

var K = require("./backend/tfjs_backend");
var types_1 = require("./types");
var generic_utils_1 = require("./utils/generic_utils");
var Regularizer = (function () {
var Regularizer = (function (_super) {
__extends(Regularizer, _super);
function Regularizer() {
return _super !== null && _super.apply(this, arguments) || this;
}
return Regularizer;
}());
}(types_1.Serializable));
exports.Regularizer = Regularizer;

@@ -53,2 +56,5 @@ var L1L2 = (function (_super) {

};
L1L2.prototype.getClassName = function () {
return 'L1L2';
};
L1L2.prototype.getConfig = function () {

@@ -55,0 +61,0 @@ return { 'l1': this.l1.dataSync()[0], 'l2': this.l2.dataSync()[0] };

4

dist/types.d.ts

@@ -70,1 +70,5 @@ import * as tfc from '@tensorflow/tfjs-core';

};
export declare abstract class Serializable {
abstract getClassName(): string;
abstract getConfig(): ConfigDict;
}

6

dist/types.js

@@ -96,1 +96,7 @@ "use strict";

exports.LayerVariable = LayerVariable;
var Serializable = (function () {
function Serializable() {
}
return Serializable;
}());
exports.Serializable = Serializable;

8

dist/utils/generic_utils.d.ts
import { Tensor } from '@tensorflow/tfjs-core';
import { ConfigDict, ConfigDictValue, DType, Shape } from '../types';
import { ConfigDict, ConfigDictValue, DType, Serializable, Shape } from '../types';
export declare function pyListRepeat(value: any, numValues: number): any[];

@@ -35,6 +35,5 @@ export declare function pyGetAttr<T>(obj: any, attrName: string, defaultValue?: T): T;

export declare function normalizeShapeList(x: Shape | Shape[]): Shape[];
export declare function isAllNullOrUndefined(iterableOrElement: {}): boolean;
export declare function toSnakeCase(name: string): string;
export declare function toCamelCase(identifier: string): string;
export declare function serializeKerasObject(instance: any): ConfigDictValue;
export declare function serializeKerasObject(instance: Serializable): ConfigDictValue;
export declare function deserializeKerasObject(identifier: string | ConfigDict, moduleObjects?: {

@@ -50,1 +49,4 @@ [objName: string]: any;

export declare function stringToDType(dtype: string): DType;
export declare function stringsEqual(xs: string[], ys: string[]): boolean;
export declare function unique<T>(xs: T[]): T[];
export declare function isObjectEmpty(obj: {}): boolean;

62

dist/utils/generic_utils.js

@@ -11,3 +11,2 @@ "use strict";

Object.defineProperty(exports, "__esModule", { value: true });
var _ = require("underscore");
var errors_1 = require("../errors");

@@ -34,3 +33,3 @@ var types_1 = require("../types");

}
if (_.isUndefined(defaultValue)) {
if (defaultValue === undefined) {
throw new errors_1.AttributeError('pyGetAttr: Attempting to get attribute ' + attrName +

@@ -172,6 +171,2 @@ 'with no default value defined');

exports.normalizeShapeList = normalizeShapeList;
function isAllNullOrUndefined(iterableOrElement) {
return _.every(toList(iterableOrElement), function (x) { return (_.isNull(x) || _.isUndefined(x)); });
}
exports.isAllNullOrUndefined = isAllNullOrUndefined;
function toSnakeCase(name) {

@@ -201,9 +196,3 @@ var intermediate = name.replace(/(.)([A-Z][a-z0-9]+)/g, '$1_$2');

}
if (instance.getConfig != null) {
return { className: instance.constructor.name, config: instance.getConfig() };
}
if (instance.name != null) {
return instance.name;
}
throw new errors_1.ValueError("Cannot serialize " + instance);
return { className: instance.getClassName(), config: instance.getConfig() };
}

@@ -241,9 +230,9 @@ exports.serializeKerasObject = serializeKerasObject;

var cls = void 0, fromConfig = void 0;
if (_.has(customObjects, className)) {
if (className in customObjects) {
_a = customObjects.get(className), cls = _a[0], fromConfig = _a[1];
}
else if (_.has(_GLOBAL_CUSTOM_OBJECTS, className)) {
else if (className in _GLOBAL_CUSTOM_OBJECTS) {
_b = _GLOBAL_CUSTOM_OBJECTS.className, cls = _b[0], fromConfig = _b[1];
}
else if (_.has(moduleObjects, className)) {
else if (className in moduleObjects) {
_c = moduleObjects[className], cls = _c[0], fromConfig = _c[1];

@@ -335,1 +324,42 @@ }

exports.stringToDType = stringToDType;
function stringsEqual(xs, ys) {
if (xs == null || ys == null) {
return xs === ys;
}
if (xs.length !== ys.length) {
return false;
}
for (var i = 0; i < xs.length; ++i) {
if (xs[i] !== ys[i]) {
return false;
}
}
return true;
}
exports.stringsEqual = stringsEqual;
function unique(xs) {
if (xs == null) {
return xs;
}
var out = [];
for (var _i = 0, xs_1 = xs; _i < xs_1.length; _i++) {
var x = xs_1[_i];
if (out.indexOf(x) === -1) {
out.push(x);
}
}
return out;
}
exports.unique = unique;
function isObjectEmpty(obj) {
if (obj == null) {
throw new errors_1.ValueError("Invalid value in obj: " + JSON.stringify(obj));
}
for (var key in obj) {
if (obj.hasOwnProperty(key)) {
return false;
}
}
return true;
}
exports.isObjectEmpty = isObjectEmpty;

1

dist/utils/math_utils.d.ts

@@ -9,1 +9,2 @@ export declare type ArrayTypes = Uint8Array | Int32Array | Float32Array;

export declare function variance(array: number[] | Float32Array): number;
export declare function range(begin: number, end: number): number[];

12

dist/utils/math_utils.js

@@ -5,2 +5,3 @@ "use strict";

var tfjs_core_1 = require("@tensorflow/tfjs-core");
var errors_1 = require("../errors");
function isInteger(x) {

@@ -50,1 +51,12 @@ return x === parseInt(x.toString(), 10);

exports.variance = variance;
function range(begin, end) {
if (end < begin) {
throw new errors_1.ValueError("end (" + end + ") < begin (" + begin + ") is forbidden.");
}
var out = [];
for (var i = begin; i < end; ++i) {
out.push(i);
}
return out;
}
exports.range = range;

11

dist/utils/test_utils.js

@@ -6,7 +6,2 @@ "use strict";

var errors_1 = require("../errors");
var webgl2Features = [{
'BACKEND': 'webgl',
'WEBGL_FLOAT_TEXTURE_ENABLED': true,
'WEBGL_VERSION': 2
}];
function expectTensorsClose(actual, expected, epsilon) {

@@ -37,3 +32,3 @@ if (actual == null) {

beforeEach(function () {
tfjs_backend_1.setBackend('cpu');
tfjs_backend_1.disposeScalarCache();
});

@@ -45,5 +40,5 @@ tests();

function describeMathGPU(testName, tests) {
tfjs_core_1.test_util.describeWithFlags(testName, webgl2Features, function () {
tfjs_core_1.test_util.describeWithFlags(testName, tfjs_core_1.test_util.WEBGL_ENVS, function () {
beforeEach(function () {
tfjs_backend_1.setBackend('webgl');
tfjs_backend_1.disposeScalarCache();
});

@@ -50,0 +45,0 @@ tests();

2

dist/version.d.ts

@@ -1,2 +0,2 @@

declare const version = "0.4.1";
declare const version = "0.5.0";
export { version };

2

dist/version.js
"use strict";
Object.defineProperty(exports, "__esModule", { value: true });
var version = '0.4.1';
var version = '0.5.0';
exports.version = version;

10

package.json
{
"name": "@tensorflow/tfjs-layers",
"version": "0.4.1",
"version": "0.5.0",
"description": "TensorFlow layers API in JavaScript",

@@ -9,5 +9,4 @@ "private": false,

"devDependencies": {
"@tensorflow/tfjs-core": "0.7.1",
"@tensorflow/tfjs-core": "0.8.1",
"@types/jasmine": "~2.5.53",
"@types/underscore": "^1.8.7",
"browserify": "~16.1.0",

@@ -42,8 +41,5 @@ "clang-format": "~1.2.2",

},
"dependencies": {
"underscore": "~1.8.3"
},
"peerDependencies": {
"@tensorflow/tfjs-core": "0.7.1"
"@tensorflow/tfjs-core": "0.8.1"
}
}
dist/engine/topology.js

Sorry, the diff of this file is too big to display

dist/tf-layers.js

Sorry, the diff of this file is too big to display

dist/tf-layers.min.js

Sorry, the diff of this file is too big to display

SocketSocket SOC 2 Logo

Product

  • Package Alerts
  • Integrations
  • Docs
  • Pricing
  • FAQ
  • Roadmap
  • Changelog

About

Packages

npm

Go

Maven

PyPI

Rubygems

Stay in touch

Get open source security insights delivered straight into your inbox.

  • Terms
  • Privacy
  • Security

Made with ⚡️ by Socket Inc