@tensorflow/tfjs-layers - npm Package Compare versions

Comparing version 0.7.2 to 0.7.3

dist/base_callbacks.d.ts

@@ -7,2 +7,3 @@ import { Tensor } from '@tensorflow/tfjs-core';
 };
+export declare type YieldEveryOptions = 'auto' | 'batch' | 'epoch' | 'never';
 export declare abstract class BaseCallback {
@@ -34,7 +35,25 @@ validationData: Tensor | Tensor[];
 }
+export declare class ModelTrainingYielder {
+    static readonly SKIP_FIRST_BATCHES: number;
+    static readonly DECISION_BATCH_COUNT: number;
+    static readonly THRESHOLD_MILLIS: number;
+    private yieldEvery;
+    private batchCount;
+    private lastYieldBatchCount;
+    private batchStartMillis;
+    private batchDurationsMillis;
+    private autoYieldEveryBatches;
+    constructor(yieldEvery: YieldEveryOptions);
+    private resolveOneTensorInLogs(logs);
+    maybeYieldOnBatch(logs: UnresolvedLogs): Promise<void>;
+    maybeYieldOnEpoch(): Promise<void>;
+}
 export declare class BaseLogger extends BaseCallback {
     private seen;
     private totals;
-    constructor();
-    onEpochBegin(epoch: number, logs?: UnresolvedLogs): Promise<void>;
+    private autoYielder;
+    private yieldEvery;
+    constructor(yieldEvery?: YieldEveryOptions);
+    onTrainBegin(logs?: UnresolvedLogs): Promise<void>;
+    onEpochBegin(epoch: number): Promise<void>;
     onBatchEnd(batch: number, logs?: UnresolvedLogs): Promise<void>;
@@ -41,0 +60,0 @@ onEpochEnd(epoch: number, logs?: UnresolvedLogs): Promise<void>;

dist/base_callbacks.js

@@ -277,10 +277,131 @@ "use strict";
 exports.CallbackList = CallbackList;
+var ModelTrainingYielder = (function () {
+    function ModelTrainingYielder(yieldEvery) {
+        this.yieldEvery = yieldEvery;
+        this.batchCount = 0;
+        this.batchDurationsMillis = [];
+        this.autoYieldEveryBatches = null;
+        this.batchStartMillis = tfjs_core_1.util.now();
+    }
+    ModelTrainingYielder.prototype.resolveOneTensorInLogs = function (logs) {
+        return __awaiter(this, void 0, void 0, function () {
+            var _a, _b, _i, key, value;
+            return __generator(this, function (_c) {
+                switch (_c.label) {
+                    case 0:
+                        _a = [];
+                        for (_b in logs)
+                            _a.push(_b);
+                        _i = 0;
+                        _c.label = 1;
+                    case 1:
+                        if (!(_i < _a.length)) return [3, 4];
+                        key = _a[_i];
+                        value = logs[key];
+                        if (!(typeof value !== 'number')) return [3, 3];
+                        return [4, value.data()];
+                    case 2:
+                        _c.sent();
+                        return [3, 4];
+                    case 3:
+                        _i++;
+                        return [3, 1];
+                    case 4: return [2];
+                }
+            });
+        });
+    };
+    ModelTrainingYielder.prototype.maybeYieldOnBatch = function (logs) {
+        return __awaiter(this, void 0, void 0, function () {
+            var t, meanBatchDuration;
+            return __generator(this, function (_a) {
+                switch (_a.label) {
+                    case 0:
+                        if (!(this.yieldEvery === 'auto')) return [3, 7];
+                        this.batchCount++;
+                        if (!(this.autoYieldEveryBatches == null)) return [3, 3];
+                        return [4, this.resolveOneTensorInLogs(logs)];
+                    case 1:
+                        _a.sent();
+                        t = tfjs_core_1.util.now();
+                        return [4, tfjs_core_1.nextFrame()];
+                    case 2:
+                        _a.sent();
+                        if (this.batchCount > ModelTrainingYielder.SKIP_FIRST_BATCHES) {
+                            this.batchDurationsMillis.push(t - this.batchStartMillis);
+                            if (this.batchDurationsMillis.length >=
+                                ModelTrainingYielder.DECISION_BATCH_COUNT) {
+                                meanBatchDuration = this.batchDurationsMillis.reduce(function (dur, prev) { return dur + prev; }) /
+                                    this.batchDurationsMillis.length;
+                                this.autoYieldEveryBatches = Math.round(ModelTrainingYielder.THRESHOLD_MILLIS / meanBatchDuration);
+                                if (this.autoYieldEveryBatches < 1) {
+                                    this.autoYieldEveryBatches = 1;
+                                }
+                            }
+                        }
+                        this.batchStartMillis = tfjs_core_1.util.now();
+                        this.lastYieldBatchCount = this.batchCount;
+                        return [3, 6];
+                    case 3:
+                        if (!(this.batchCount - this.lastYieldBatchCount >=
+                            this.autoYieldEveryBatches)) return [3, 6];
+                        return [4, tfjs_core_1.nextFrame()];
+                    case 4:
+                        _a.sent();
+                        return [4, this.resolveOneTensorInLogs(logs)];
+                    case 5:
+                        _a.sent();
+                        this.lastYieldBatchCount = this.batchCount;
+                        _a.label = 6;
+                    case 6: return [3, 9];
+                    case 7:
+                        if (!(this.yieldEvery === 'batch')) return [3, 9];
+                        return [4, tfjs_core_1.nextFrame()];
+                    case 8:
+                        _a.sent();
+                        _a.label = 9;
+                    case 9: return [2];
+                }
+            });
+        });
+    };
+    ModelTrainingYielder.prototype.maybeYieldOnEpoch = function () {
+        return __awaiter(this, void 0, void 0, function () {
+            return __generator(this, function (_a) {
+                switch (_a.label) {
+                    case 0:
+                        if (!(this.yieldEvery === 'epoch')) return [3, 2];
+                        return [4, tfjs_core_1.nextFrame()];
+                    case 1:
+                        _a.sent();
+                        _a.label = 2;
+                    case 2: return [2];
+                }
+            });
+        });
+    };
+    ModelTrainingYielder.SKIP_FIRST_BATCHES = 1;
+    ModelTrainingYielder.DECISION_BATCH_COUNT = 2;
+    ModelTrainingYielder.THRESHOLD_MILLIS = 16;
+    return ModelTrainingYielder;
+}());
+exports.ModelTrainingYielder = ModelTrainingYielder;
 var BaseLogger = (function (_super) {
     __extends(BaseLogger, _super);
-    function BaseLogger() {
-        return _super.call(this) || this;
+    function BaseLogger(yieldEvery) {
+        var _this = _super.call(this) || this;
+        _this.yieldEvery = yieldEvery || 'auto';
+        return _this;
     }
-    BaseLogger.prototype.onEpochBegin = function (epoch, logs) {
+    BaseLogger.prototype.onTrainBegin = function (logs) {
         return __awaiter(this, void 0, void 0, function () {
             return __generator(this, function (_a) {
+                this.autoYielder = new ModelTrainingYielder(this.yieldEvery);
+                return [2];
+            });
+        });
+    };
+    BaseLogger.prototype.onEpochBegin = function (epoch) {
+        return __awaiter(this, void 0, void 0, function () {
+            return __generator(this, function (_a) {
                 this.seen = 0;
@@ -297,34 +418,39 @@ this.totals = {};
             return __generator(this, function (_a) {
-                if (logs == null) {
-                    logs = {};
-                }
-                batchSize = logs['size'] == null ? 0 : logs['size'];
-                this.seen += batchSize;
-                _loop_1 = function (key) {
-                    var value = logs[key];
-                    if (typeof value === 'number') {
-                        if (!this_1.totals.hasOwnProperty(key)) {
-                            this_1.totals[key] = 0;
+                switch (_a.label) {
+                    case 0: return [4, this.autoYielder.maybeYieldOnBatch(logs)];
+                    case 1:
+                        _a.sent();
+                        if (logs == null) {
+                            logs = {};
                         }
-                        this_1.totals[key] = this_1.totals[key] + value * batchSize;
-                    }
-                    else {
-                        var oldTotalsToDispose = void 0;
-                        if (key in this_1.totals) {
-                            oldTotalsToDispose = this_1.totals[key];
+                        batchSize = logs['size'] == null ? 0 : logs['size'];
+                        this.seen += batchSize;
+                        _loop_1 = function (key) {
+                            var value = logs[key];
+                            if (typeof value === 'number') {
+                                if (!this_1.totals.hasOwnProperty(key)) {
+                                    this_1.totals[key] = 0;
+                                }
+                                this_1.totals[key] = this_1.totals[key] + value * batchSize;
+                            }
+                            else {
+                                var oldTotalsToDispose = void 0;
+                                if (key in this_1.totals) {
+                                    oldTotalsToDispose = this_1.totals[key];
+                                }
+                                else {
+                                    this_1.totals[key] = state_1.getScalar(0);
+                                }
+                                this_1.totals[key] = tfjs_core_1.tidy(function () { return tfjs_core_1.add(_this.totals[key], tfjs_core_1.mul(value, state_1.getScalar(batchSize))); });
+                                if (oldTotalsToDispose != null) {
+                                    oldTotalsToDispose.dispose();
+                                }
+                            }
+                        };
+                        this_1 = this;
+                        for (key in logs) {
+                            _loop_1(key);
                         }
-                        else {
-                            this_1.totals[key] = state_1.getScalar(0);
-                        }
-                        this_1.totals[key] = tfjs_core_1.tidy(function () { return tfjs_core_1.add(_this.totals[key], tfjs_core_1.mul(value, state_1.getScalar(batchSize))); });
-                        if (oldTotalsToDispose != null) {
-                            oldTotalsToDispose.dispose();
-                        }
-                    }
-                };
-                this_1 = this;
-                for (key in logs) {
-                    _loop_1(key);
+                        return [2];
                 }
-                return [2];
             });
@@ -338,25 +464,30 @@ });
             return __generator(this, function (_b) {
-                if (logs != null) {
-                    _loop_2 = function (key) {
-                        if (this_2.totals[key] == null) {
-                            return "continue";
+                switch (_b.label) {
+                    case 0: return [4, this.autoYielder.maybeYieldOnEpoch()];
+                    case 1:
+                        _b.sent();
+                        if (logs != null) {
+                            _loop_2 = function (key) {
+                                if (this_2.totals[key] == null) {
+                                    return "continue";
+                                }
+                                if (typeof this_2.totals[key] === 'number') {
+                                    logs[key] = this_2.totals[key] / this_2.seen;
+                                }
+                                else {
+                                    tfjs_core_1.tidy(function () {
+                                        logs[key] = tfjs_core_1.mul(tfjs_core_1.div(state_1.getScalar(1), state_1.getScalar(_this.seen)), _this.totals[key]);
+                                        _this.totals[key].dispose();
+                                        tfjs_core_1.keep(logs[key]);
+                                    });
+                                }
+                            };
+                            this_2 = this;
+                            for (_i = 0, _a = this.params['metrics']; _i < _a.length; _i++) {
+                                key = _a[_i];
+                                _loop_2(key);
+                            }
                         }
-                        if (typeof this_2.totals[key] === 'number') {
-                            logs[key] = this_2.totals[key] / this_2.seen;
-                        }
-                        else {
-                            tfjs_core_1.tidy(function () {
-                                logs[key] = tfjs_core_1.mul(tfjs_core_1.div(state_1.getScalar(1), state_1.getScalar(_this.seen)), _this.totals[key]);
-                                _this.totals[key].dispose();
-                                tfjs_core_1.keep(logs[key]);
-                            });
-                        }
-                    };
-                    this_2 = this;
-                    for (_i = 0, _a = this.params['metrics']; _i < _a.length; _i++) {
-                        key = _a[_i];
-                        _loop_2(key);
-                    }
+                        return [2];
                 }
-                return [2];
             });
@@ -363,0 +494,0 @@ });

dist/engine/container.d.ts

 import { Scalar, serialization, Tensor } from '@tensorflow/tfjs-core';
 import { JsonDict, Kwargs, NamedTensorMap, Shape } from '../types';
 import { LayerVariable } from '../variables';
-import { Layer, Node, SymbolicTensor } from './topology';
+import { Layer, Node, SymbolicTensor, DisposeResult } from './topology';
 export declare function loadWeightsFromJson(weightsJSON: JsonDict, layers: Layer[], skipMismatch?: boolean): void;
-export declare function loadWeightsFromNamedTensorMap(weights: NamedTensorMap, layers: Layer[]): void;
+export declare function loadWeightsFromNamedTensorMap(weights: NamedTensorMap, layers: Layer[], strict?: boolean): void;
 export interface ContainerConfig {
@@ -37,6 +37,8 @@ inputs: SymbolicTensor | SymbolicTensor[];
     constructor(config: ContainerConfig);
+    protected assertNotDisposed(): void;
+    dispose(): DisposeResult;
     readonly trainableWeights: LayerVariable[];
     readonly nonTrainableWeights: LayerVariable[];
     readonly weights: LayerVariable[];
-    loadWeights(weightsJSON: JsonDict | NamedTensorMap, skipMismatch?: boolean, isNamedTensorMap?: boolean): void;
+    loadWeights(weightsJSON: JsonDict | NamedTensorMap, skipMismatch?: boolean, isNamedTensorMap?: boolean, strict?: boolean): void;
     private updatedConfig();
@@ -43,0 +45,0 @@ toJSON(unused?: any, returnString?: boolean): string | JsonDict;

dist/engine/container.js

@@ -95,3 +95,4 @@ "use strict";
 exports.loadWeightsFromJson = loadWeightsFromJson;
-function loadWeightsFromNamedTensorMap(weights, layers) {
+function loadWeightsFromNamedTensorMap(weights, layers, strict) {
+    if (strict === void 0) { strict = true; }
     var nameToWeight = {};
@@ -112,13 +113,20 @@ var totalWeightsCount = 0;
     for (var name_2 in weights) {
-        weightValueTuples.push([nameToWeight[name_2], weights[name_2]]);
+        if (nameToWeight[name_2] != null) {
+            weightValueTuples.push([nameToWeight[name_2], weights[name_2]]);
+        }
+        else if (strict) {
+            throw new errors_1.ValueError("Provided weight data has no target variable: " + name_2);
+        }
         delete nameToWeight[name_2];
     }
-    var unsetNames = [];
-    for (var name_3 in nameToWeight) {
-        unsetNames.push(name_3);
+    if (strict) {
+        var unsetNames = [];
+        for (var name_3 in nameToWeight) {
+            unsetNames.push(name_3);
+        }
+        if (unsetNames.length > 0) {
+            throw new errors_1.ValueError(unsetNames.length + " of " + totalWeightsCount + " weights are not set: " +
+                ("" + unsetNames));
+        }
     }
-    if (unsetNames.length > 0) {
-        throw new errors_1.ValueError(unsetNames.length + " of " + totalWeightsCount + " weights are not set: " +
-            ("" + unsetNames));
-    }
     variables_1.batchSetValue(weightValueTuples);
@@ -381,4 +389,25 @@ }
         _this.built = true;
+        _this._refCount = 1;
         return _this;
     }
+    Container.prototype.assertNotDisposed = function () {
+        if (this._refCount === 0) {
+            throw new Error("Container '" + this.name + "' is already disposed.");
+        }
+    };
+    Container.prototype.dispose = function () {
+        this.assertNotDisposed();
+        var result = {
+            refCountAfterDispose: null,
+            numDisposedVariables: 0
+        };
+        if (--this._refCount === 0) {
+            for (var _i = 0, _a = this.layers; _i < _a.length; _i++) {
+                var layer = _a[_i];
+                result.numDisposedVariables += layer.dispose().numDisposedVariables;
+            }
+        }
+        result.refCountAfterDispose = this._refCount;
+        return result;
+    };
     Object.defineProperty(Container.prototype, "trainableWeights", {
@@ -432,7 +461,8 @@ get: function () {
     });
-    Container.prototype.loadWeights = function (weightsJSON, skipMismatch, isNamedTensorMap) {
+    Container.prototype.loadWeights = function (weightsJSON, skipMismatch, isNamedTensorMap, strict) {
         if (skipMismatch === void 0) { skipMismatch = false; }
         if (isNamedTensorMap === void 0) { isNamedTensorMap = false; }
+        if (strict === void 0) { strict = true; }
         if (isNamedTensorMap) {
-            loadWeightsFromNamedTensorMap(weightsJSON, this.layers);
+            loadWeightsFromNamedTensorMap(weightsJSON, this.layers, strict);
         }
@@ -439,0 +469,0 @@ else {

dist/engine/input_layer.d.ts

 import { DataType, serialization, Tensor } from '@tensorflow/tfjs-core';
 import { Kwargs, Shape } from '../types';
-import { Layer, SymbolicTensor } from './topology';
+import { Layer, SymbolicTensor, DisposeResult } from './topology';
 export interface InputLayerConfig {
@@ -17,2 +17,3 @@ inputShape?: Shape;
     apply(inputs: Tensor | Tensor[] | SymbolicTensor | SymbolicTensor[], kwargs?: Kwargs): Tensor | Tensor[] | SymbolicTensor;
+    dispose(): DisposeResult;
     getConfig(): serialization.ConfigDict;
@@ -19,0 +20,0 @@ }

dist/engine/input_layer.js

@@ -78,2 +78,8 @@ "use strict";
     };
+    InputLayer.prototype.dispose = function () {
+        return {
+            refCountAfterDispose: this._refCount,
+            numDisposedVariables: 0
+        };
+    };
     InputLayer.prototype.getConfig = function () {
@@ -80,0 +86,0 @@ return {

dist/engine/topology.d.ts

@@ -56,2 +56,6 @@ import { DataType, Scalar, serialization, Tensor } from '@tensorflow/tfjs-core';
 }
+export interface DisposeResult {
+    refCountAfterDispose: number;
+    numDisposedVariables: number;
+}
 export declare class Node {
@@ -106,2 +110,3 @@ callArgs: Kwargs;
     protected _stateful: boolean;
+    protected _refCount: number | null;
     constructor(config: LayerConfig);
@@ -140,3 +145,6 @@ protected static nodeKey(layer: Layer, nodeIndex: number): string;
     getConfig(): serialization.ConfigDict;
+    protected disposeWeights(): number;
+    protected assertNotDisposed(): void;
+    dispose(): DisposeResult;
 }
 export declare function getSourceInputs(tensor: SymbolicTensor, layer?: Layer, nodeIndex?: number): SymbolicTensor[];

dist/engine/topology.js

@@ -155,2 +155,3 @@ "use strict";
         }
+        _this._refCount = null;
         return _this;
@@ -372,2 +373,3 @@ }
         kwargs = kwargs || {};
+        this.assertNotDisposed();
         var inputsList = generic_utils.toList(inputs);
@@ -407,2 +409,5 @@ var allAreSymbolic = true;
                 }
+                if (_this._refCount === null && noneAreSymbolic) {
+                    _this._refCount = 1;
+                }
             }
@@ -444,2 +449,3 @@ _this.assertInputCompatibility(inputs);
                 _this.addInboundNode(inputs, output, null, null, inputShape, outputShape, kwargs);
+                _this._refCount++;
                 if (_this.activityRegularizer != null) {
@@ -657,2 +663,30 @@ throw new errors_1.NotImplementedError('Layer invocation in the presence of activity ' +
     };
+    Layer.prototype.disposeWeights = function () {
+        this.weights.forEach(function (weight) { return weight.dispose(); });
+        return this.weights.length;
+    };
+    Layer.prototype.assertNotDisposed = function () {
+        if (this._refCount === 0) {
+            throw new Error("Layer '" + this.name + "' is already disposed.");
+        }
+    };
+    Layer.prototype.dispose = function () {
+        if (!this.built) {
+            throw new Error("Cannot dispose Layer " + this.name + " because it has not been " +
+                "built yet.");
+        }
+        if (this._refCount === null) {
+            throw new Error("Cannot dispose Layer " + this.name + " because it has not been used " +
+                "yet.");
+        }
+        this.assertNotDisposed();
+        var numDisposedVariables = 0;
+        if (--this._refCount === 0) {
+            numDisposedVariables = this.disposeWeights();
+        }
+        return {
+            refCountAfterDispose: this._refCount,
+            numDisposedVariables: numDisposedVariables
+        };
+    };
     return Layer;
@@ -659,0 +693,0 @@ }(tfjs_core_1.serialization.Serializable));

dist/engine/training.d.ts

 import * as tfc from '@tensorflow/tfjs-core';
 import { io, ModelPredictConfig, Optimizer, Scalar, Tensor, Tensor1D } from '@tensorflow/tfjs-core';
-import { BaseCallback, CustomCallbackConfig, History } from '../base_callbacks';
+import { BaseCallback, CustomCallbackConfig, History, YieldEveryOptions } from '../base_callbacks';
 import { LossOrMetricFn, NamedTensorMap, Shape } from '../types';
@@ -48,3 +48,3 @@ import { Container, ContainerConfig } from './container';
     validationSteps?: number;
-    yieldEvery?: 'batch' | 'epoch' | 'never';
+    yieldEvery?: YieldEveryOptions;
 }
@@ -98,3 +98,3 @@ export interface ModelCompileConfig {
     }, checkBatchAxis?: boolean, batchSize?: number): [Tensor[], Tensor[], Tensor[]];
-    private fitLoop(f, ins, outLabels?, batchSize?, epochs?, verbose?, callbacks?, valF?, valIns?, shuffle?, callbackMetrics?, initialEpoch?, stepsPerEpoch?, validationSteps?);
+    private fitLoop(f, ins, outLabels?, batchSize?, epochs?, verbose?, callbacks?, valF?, valIns?, shuffle?, callbackMetrics?, initialEpoch?, stepsPerEpoch?, validationSteps?, yieldEvery?);
     private testLoop(f, ins, batchSize?, verbose?, steps?);
@@ -101,0 +101,0 @@ private getDedupedMetricsNames();

dist/engine/training.js

@@ -720,6 +720,6 @@ "use strict";
     };
-    Model.prototype.fitLoop = function (f, ins, outLabels, batchSize, epochs, verbose, callbacks, valF, valIns, shuffle, callbackMetrics, initialEpoch, stepsPerEpoch, validationSteps) {
+    Model.prototype.fitLoop = function (f, ins, outLabels, batchSize, epochs, verbose, callbacks, valF, valIns, shuffle, callbackMetrics, initialEpoch, stepsPerEpoch, validationSteps, yieldEvery) {
         return __awaiter(this, void 0, void 0, function () {
             var _this = this;
-            var doValidation, numTrainSamples, indexArray, callbackList, _loop_4, this_1, epoch, state_2;
+            var doValidation, numTrainSamples, indexArray, baseLogger, callbackList, _loop_4, this_1, epoch, state_2;
             return __generator(this, function (_a) {
@@ -756,7 +756,8 @@ switch (_a.label) {
                         this.history = new base_callbacks_1.History();
+                        baseLogger = new base_callbacks_1.BaseLogger(yieldEvery);
                         if (callbacks == null) {
-                            callbacks = [new base_callbacks_1.BaseLogger()];
+                            callbacks = [baseLogger];
                         }
                         else {
-                            callbacks = [new base_callbacks_1.BaseLogger()].concat(callbacks);
+                            callbacks = [baseLogger].concat(callbacks);
                         }
@@ -1093,3 +1094,3 @@ callbacks = callbacks.concat([this.history]);
                         callbacks = base_callbacks_1.standardizeCallbacks(config.callbacks);
-                        return [4, this.fitLoop(trainFunction, ins, outLabels, batchSize, config.epochs, config.verbose, callbacks, valFunction, valIns, config.shuffle, callbackMetrics, config.initialEpoch, null, null)];
+                        return [4, this.fitLoop(trainFunction, ins, outLabels, batchSize, config.epochs, config.verbose, callbacks, valFunction, valIns, config.shuffle, callbackMetrics, config.initialEpoch, null, null, config.yieldEvery)];
                     case 1:
@@ -1096,0 +1097,0 @@ out = _a.sent();

dist/exports_layers.d.ts

@@ -7,3 +7,3 @@ import { InputLayerConfig } from './engine/input_layer';
 import { DepthwiseConv2DLayerConfig } from './layers/convolutional_depthwise';
-import { ActivationLayerConfig, DenseLayerConfig, DropoutLayerConfig, RepeatVectorLayerConfig, ReshapeLayerConfig } from './layers/core';
+import { ActivationLayerConfig, DenseLayerConfig, DropoutLayerConfig, PermuteLayerConfig, RepeatVectorLayerConfig, ReshapeLayerConfig } from './layers/core';
 import { EmbeddingLayerConfig } from './layers/embeddings';
@@ -14,3 +14,3 @@ import { ConcatenateLayerConfig } from './layers/merge';
 import { GlobalPooling2DLayerConfig, Pooling1DLayerConfig, Pooling2DLayerConfig } from './layers/pooling';
-import { GRUCellLayerConfig, GRULayerConfig, LSTMCellLayerConfig, LSTMLayerConfig, RNNCell, RNNLayerConfig, SimpleRNNCellLayerConfig, SimpleRNNLayerConfig, StackedRNNCellsConfig } from './layers/recurrent';
+import { GRUCellLayerConfig, GRULayerConfig, LSTMCellLayerConfig, LSTMLayerConfig, RNN, RNNCell, RNNLayerConfig, SimpleRNNCellLayerConfig, SimpleRNNLayerConfig, StackedRNNCellsConfig } from './layers/recurrent';
 import { BidirectionalLayerConfig, Wrapper, WrapperLayerConfig } from './layers/wrappers';
@@ -35,2 +35,3 @@ export declare function inputLayer(config: InputLayerConfig): Layer;
 export declare function reshape(config: ReshapeLayerConfig): Layer;
+export declare function permute(config: PermuteLayerConfig): Layer;
 export declare function embedding(config: EmbeddingLayerConfig): Layer;
@@ -71,2 +72,2 @@ export declare function add(config?: LayerConfig): Layer;
 export declare const maxPool2d: typeof maxPooling2d;
-export { Layer, input };
+export { Layer, RNN, RNNCell, input };

dist/exports_layers.js

@@ -18,2 +18,4 @@ "use strict";
 var recurrent_1 = require("./layers/recurrent");
+exports.RNN = recurrent_1.RNN;
+exports.RNNCell = recurrent_1.RNNCell;
 var wrappers_1 = require("./layers/wrappers");
@@ -92,2 +94,6 @@ function inputLayer(config) {
 exports.reshape = reshape;
+function permute(config) {
+    return new core_1.Permute(config);
+}
+exports.permute = permute;
 function embedding(config) {
@@ -94,0 +100,0 @@ return new embeddings_1.Embedding(config);

dist/exports.d.ts

@@ -9,3 +9,3 @@ import { io } from '@tensorflow/tfjs-core';
 export declare function sequential(config?: SequentialConfig): Sequential;
-export declare function loadModel(pathOrIOHandler: string | io.IOHandler): Promise<Model>;
+export declare function loadModel(pathOrIOHandler: string | io.IOHandler, strict?: boolean): Promise<Model>;
 export declare function input(config: InputConfig): SymbolicTensor;

dist/exports.js

@@ -14,4 +14,5 @@ "use strict";
 exports.sequential = sequential;
-function loadModel(pathOrIOHandler) {
-    return models_1.loadModelInternal(pathOrIOHandler);
+function loadModel(pathOrIOHandler, strict) {
+    if (strict === void 0) { strict = true; }
+    return models_1.loadModelInternal(pathOrIOHandler, strict);
 }
@@ -18,0 +19,0 @@ exports.loadModel = loadModel;

dist/index.d.ts

@@ -6,3 +6,3 @@ import * as constraints from './exports_constraints';
 import * as regularizers from './exports_regularizers';
-export { CallbackList, CustomCallback, CustomCallbackConfig } from './base_callbacks';
+export { CallbackList, CustomCallback, CustomCallbackConfig, History } from './base_callbacks';
 export { Callback } from './callbacks';
@@ -9,0 +9,0 @@ export { SymbolicTensor } from './engine/topology';

dist/index.js

@@ -16,2 +16,3 @@ "use strict";
 exports.CustomCallback = base_callbacks_1.CustomCallback;
+exports.History = base_callbacks_1.History;
 var callbacks_1 = require("./callbacks");
@@ -18,0 +19,0 @@ exports.Callback = callbacks_1.Callback;

dist/layers/core.d.ts

@@ -98,1 +98,13 @@ import { serialization, Tensor } from '@tensorflow/tfjs-core';
 }
+export interface PermuteLayerConfig extends LayerConfig {
+    dims: number[];
+}
+export declare class Permute extends Layer {
+    static className: string;
+    readonly dims: number[];
+    private readonly dimsIncludingBatch;
+    constructor(config: PermuteLayerConfig);
+    computeOutputShape(inputShape: Shape | Shape[]): Shape | Shape[];
+    call(inputs: Tensor | Tensor[], kwargs: Kwargs): Tensor | Tensor[];
+    getConfig(): serialization.ConfigDict;
+}

dist/layers/core.js

@@ -15,10 +15,10 @@ "use strict";
 var activations_1 = require("../activations");
+var state_1 = require("../backend/state");
 var K = require("../backend/tfjs_backend");
 var constraints_1 = require("../constraints");
 var topology_1 = require("../engine/topology");
-var state_1 = require("../backend/state");
 var errors_1 = require("../errors");
 var initializers_1 = require("../initializers");
 var regularizers_1 = require("../regularizers");
-var math_utils = require("../utils/math_utils");
+var math_utils_1 = require("../utils/math_utils");
 var types_utils_1 = require("../utils/types_utils");
@@ -193,3 +193,3 @@ var Dropout = (function (_super) {
         }
-        return [inputShape[0], math_utils.arrayProd(inputShape, 1)];
+        return [inputShape[0], math_utils_1.arrayProd(inputShape, 1)];
     };
@@ -300,3 +300,3 @@ Flatten.prototype.call = function (inputs, kwargs) {
         }
-        var originalSize = math_utils.arrayProd(inputShape);
+        var originalSize = math_utils_1.arrayProd(inputShape);
         if (unknown !== null) {
@@ -351,2 +351,48 @@ if (known === 0 || originalSize % known !== 0) {
 tfjs_core_1.serialization.SerializationMap.register(Reshape);
+var Permute = (function (_super) {
+    __extends(Permute, _super);
+    function Permute(config) {
+        var _this = _super.call(this, config) || this;
+        if (config.dims == null) {
+            throw new Error('Required configuration field `dims` is missing during Permute ' +
+                'constructor call.');
+        }
+        if (!Array.isArray(config.dims)) {
+            throw new Error('Permute constructor requires `dims` to be an Array, but received ' +
+                (config.dims + " instead."));
+        }
+        var expectedSortedIndices = math_utils_1.range(1, config.dims.length + 1);
+        if (!tfjs_core_1.util.arraysEqual(config.dims.slice().sort(), expectedSortedIndices)) {
+            throw new Error('Invalid permutation `dims`: ' + JSON.stringify(config.dims) +
+                ' `dims` must contain consecutive integers starting from 1.');
+        }
+        _this.dims = config.dims;
+        _this.dimsIncludingBatch = [0].concat(_this.dims);
+        _this.inputSpec = [new topology_1.InputSpec({ ndim: _this.dims.length + 1 })];
+        return _this;
+    }
+    Permute.prototype.computeOutputShape = function (inputShape) {
+        inputShape = types_utils_1.getExactlyOneShape(inputShape);
+        var outputShape = inputShape.slice();
+        this.dims.forEach(function (dim, i) {
+            outputShape[i + 1] = inputShape[dim];
+        });
+        return outputShape;
+    };
+    Permute.prototype.call = function (inputs, kwargs) {
+        return tfjs_core_1.transpose(types_utils_1.getExactlyOneTensor(inputs), this.dimsIncludingBatch);
+    };
+    Permute.prototype.getConfig = function () {
+        var config = {
+            dims: this.dims,
+        };
+        var baseConfig = _super.prototype.getConfig.call(this);
+        Object.assign(config, baseConfig);
+        return config;
+    };
+    Permute.className = 'Permute';
+    return Permute;
+}(topology_1.Layer));
+exports.Permute = Permute;
+tfjs_core_1.serialization.SerializationMap.register(Permute);
 //# sourceMappingURL=core.js.map

dist/models.d.ts

@@ -17,4 +17,4 @@ import { io, Scalar, serialization, Tensor } from '@tensorflow/tfjs-core';
 }
-export declare function loadModelInternal(pathOrIOHandler: string | io.IOHandler): Promise<Model>;
-export declare function loadModelFromIOHandler(handler: io.IOHandler, customObjects?: serialization.ConfigDict): Promise<Model>;
+export declare function loadModelInternal(pathOrIOHandler: string | io.IOHandler, strict?: boolean): Promise<Model>;
+export declare function loadModelFromIOHandler(handler: io.IOHandler, customObjects?: serialization.ConfigDict, strict?: boolean): Promise<Model>;
 export interface SequentialConfig {
@@ -21,0 +21,0 @@ layers?: Layer[];

dist/models.js

@@ -90,3 +90,4 @@ "use strict";
 exports.modelFromJSON = modelFromJSON;
-function loadModelInternal(pathOrIOHandler) {
+function loadModelInternal(pathOrIOHandler, strict) {
+    if (strict === void 0) { strict = true; }
     return __awaiter(this, void 0, void 0, function () {
@@ -106,3 +107,3 @@ var handlers;
             }
-            return [2, loadModelFromIOHandler(pathOrIOHandler)];
+            return [2, loadModelFromIOHandler(pathOrIOHandler, undefined, strict)];
         });
@@ -112,3 +113,4 @@ });
 exports.loadModelInternal = loadModelInternal;
-function loadModelFromIOHandler(handler, customObjects) {
+function loadModelFromIOHandler(handler, customObjects, strict) {
+    if (strict === void 0) { strict = true; }
     return __awaiter(this, void 0, void 0, function () {
@@ -138,3 +140,3 @@ var artifacts, modelTopology, model, skipMismatch, isNamedTensorMap;
                         isNamedTensorMap = true;
-                        model.loadWeights(tfjs_core_1.io.decodeWeights(artifacts.weightData, artifacts.weightSpecs), skipMismatch, isNamedTensorMap);
+                        model.loadWeights(tfjs_core_1.io.decodeWeights(artifacts.weightData, artifacts.weightSpecs), skipMismatch, isNamedTensorMap, strict);
                     }
@@ -141,0 +143,0 @@ return [2, model];

dist/utils/layer_utils.js

@@ -41,2 +41,11 @@ "use strict";
     model.checkTrainableWeightsConsistency();
+    var trainableCount = countTrainableParams(model);
+    var nonTrainableCount = variable_utils_1.countParamsInWeights(model.nonTrainableWeights);
+    printFn("Total params: " + (trainableCount + nonTrainableCount));
+    printFn("Trainable params: " + trainableCount);
+    printFn("Non-trainable params: " + nonTrainableCount);
+    printFn('_'.repeat(lineLength));
+}
+exports.printSummary = printSummary;
+function countTrainableParams(model) {
     var trainableCount;
@@ -50,9 +59,4 @@ if (model.collectedTrainableWeights != null) {
     }
-    var nonTrainableCount = variable_utils_1.countParamsInWeights(model.nonTrainableWeights);
-    printFn("Total params: " + (trainableCount + nonTrainableCount));
-    printFn("Trainable params: " + trainableCount);
-    printFn("Non-trainable params: " + nonTrainableCount);
-    printFn('_'.repeat(lineLength));
+    return trainableCount;
 }
-exports.printSummary = printSummary;
 function isModelSequentialLike(model) {
@@ -59,0 +63,0 @@ var sequentialLike = true;

dist/variables.d.ts

@@ -17,2 +17,4 @@ import * as tfc from '@tensorflow/tfjs-core';
     write(newVal: Tensor): this;
+    dispose(): void;
+    protected assertNotDisposed(): void;
 }
@@ -19,0 +21,0 @@ export declare function variable(x: Tensor, dtype?: DataType, name?: string, constraint?: Constraint): LayerVariable;

dist/variables.js

@@ -26,5 +26,7 @@ "use strict";
     LayerVariable.prototype.read = function () {
+        this.assertNotDisposed();
         return this.val;
     };
     LayerVariable.prototype.write = function (newVal) {
+        this.assertNotDisposed();
         checkShapesMatch(this.val, newVal);
@@ -37,2 +39,11 @@ this.val.assign(newVal);
     };
+    LayerVariable.prototype.dispose = function () {
+        this.assertNotDisposed();
+        this.val.dispose();
+    };
+    LayerVariable.prototype.assertNotDisposed = function () {
+        if (this.val.isDisposed) {
+            throw new Error("LayersVariable " + this.name + " is already disposed.");
+        }
+    };
     return LayerVariable;
@@ -39,0 +50,0 @@ }());

dist/version.d.ts

@@ -1,2 +0,2 @@
-declare const version = "0.7.2";
+declare const version = "0.7.3";
 export { version };

dist/version.js

 "use strict";
 Object.defineProperty(exports, "__esModule", { value: true });
-var version = '0.7.2';
+var version = '0.7.3';
 exports.version = version;
 //# sourceMappingURL=version.js.map

package.json

 {
   "name": "@tensorflow/tfjs-layers",
-  "version": "0.7.2",
+  "version": "0.7.3",
   "description": "TensorFlow layers API in JavaScript",
@@ -13,3 +13,3 @@ "private": false,
   "devDependencies": {
-    "@tensorflow/tfjs-core": "~0.12.8",
+    "@tensorflow/tfjs-core": "~0.12.10",
     "@types/jasmine": "~2.5.53",
@@ -31,3 +31,3 @@ "clang-format": "~1.2.2",
     "rollup-plugin-uglify": "~3.0.0",
-    "tslint": "~5.6.0",
+    "tslint": "~5.11.0",
     "tslint-no-circular-imports": "^0.5.0",
@@ -47,7 +47,7 @@ "typescript": "2.8.3",
     "test-travis": "karma start --browsers='bs_firefox_mac,bs_chrome_mac' --singleRun --reporters='dots,karma-typescript'",
-    "lint": "tslint -p . --type-check -t verbose"
+    "lint": "tslint -p . -t verbose"
   },
   "peerDependencies": {
-    "@tensorflow/tfjs-core": "~0.12.8"
+    "@tensorflow/tfjs-core": "~0.12.10"
   }
 }

New alerts

License Policy Violation

License

This package is not allowed per your license policy. Review the package's license to ensure compliance.

Found 1 instance in 1 package

Fixed alerts

License Policy Violation

License

This package is not allowed per your license policy. Review the package's license to ensure compliance.

Found 1 instance in 1 package

Improved metrics

Total package byte prevSize

5490709

increased by2.41%

Lines of code

26604

increased by2.42%

No dependency changes