@tensorflow/tfjs-layers - npm Package Compare versions

Comparing version 0.0.3 to 0.0.4

dist/activations.d.ts

@@ -1,2 +0,2 @@
-import { Tensor } from 'deeplearn';
+import { Tensor } from '@tensorflow/tfjs-core';
 import { ConfigDictValue } from './types';
@@ -3,0 +3,0 @@ export declare type ActivationFn = (tensor: Tensor, axis?: number) => Tensor;

dist/activations.js

 "use strict";
 Object.defineProperty(exports, "__esModule", { value: true });
-var deeplearn_1 = require("deeplearn");
+var tfjs_core_1 = require("@tensorflow/tfjs-core");
 var K = require("./backend/deeplearnjs_backend");
@@ -62,3 +62,3 @@ var errors_1 = require("./errors");
 function relu6(x) {
-    return K.minimum(deeplearn_1.scalar(6.0), K.relu(x));
+    return K.minimum(tfjs_core_1.scalar(6.0), K.relu(x));
 }
@@ -65,0 +65,0 @@ exports.relu6 = relu6;

dist/backend/deeplearnjs_backend.d.ts

@@ -1,2 +0,2 @@
-import { Scalar, Tensor, Tensor1D } from 'deeplearn';
+import { Scalar, Tensor, Tensor1D } from '@tensorflow/tfjs-core';
 import { DataFormat, PaddingMode, PoolMode } from '../common';
@@ -3,0 +3,0 @@ import { Constraint } from '../constraints';

dist/backend/deeplearnjs_backend.js

 "use strict";
 Object.defineProperty(exports, "__esModule", { value: true });
-var dl = require("deeplearn");
-var deeplearn_1 = require("deeplearn");
+var tfc = require("@tensorflow/tfjs-core");
+var tfjs_core_1 = require("@tensorflow/tfjs-core");
 var _ = require("underscore");
@@ -24,3 +24,3 @@ var common_1 = require("../common");
 function setBackend(requestedBackend) {
-    dl.setBackend(requestedBackend);
+    tfc.setBackend(requestedBackend);
     backend = requestedBackend;
@@ -35,3 +35,3 @@ disposeScalarCache();
 function keep(x) {
-    return dl.keep(x);
+    return tfc.keep(x);
 }
@@ -48,4 +48,4 @@ exports.keep = keep;
     if (scalarCache[dtype][value] == null) {
-        scalarCache[dtype][value] = deeplearn_1.scalar(value, dtype);
-        dl.keep(scalarCache[dtype][value]);
+        scalarCache[dtype][value] = tfjs_core_1.scalar(value, dtype);
+        tfc.keep(scalarCache[dtype][value]);
     }
@@ -73,3 +73,3 @@ return scalarCache[dtype][value];
 function dtype(x) {
-    return (x instanceof deeplearn_1.Tensor) ? DEFAULT_DTYPE : x.dtype;
+    return (x instanceof tfjs_core_1.Tensor) ? DEFAULT_DTYPE : x.dtype;
 }
@@ -107,3 +107,3 @@ exports.dtype = dtype;
 function transpose(x, perm) {
-    return dl.transpose(x, perm);
+    return tfc.transpose(x, perm);
 }
@@ -113,3 +113,3 @@ exports.transpose = transpose;
 function reverse(x, axes) {
-    return dl.reverse(x, axes);
+    return tfc.reverse(x, axes);
 }
@@ -128,3 +128,3 @@ exports.reverse = reverse;
 function squeeze(x, axis) {
-    return dl.squeeze(x, [axis]);
+    return tfc.squeeze(x, [axis]);
 }
@@ -157,9 +157,9 @@ exports.squeeze = squeeze;
         case 1:
-            return dl.slice1d(array, start, size);
+            return tfc.slice1d(array, start, size);
         case 2:
-            return dl.slice2d(array, [start, 0], [size, array.shape[1]]);
+            return tfc.slice2d(array, [start, 0], [size, array.shape[1]]);
         case 3:
-            return dl.slice3d(array, [start, 0, 0], [size, array.shape[1], array.shape[2]]);
+            return tfc.slice3d(array, [start, 0, 0], [size, array.shape[1], array.shape[2]]);
         case 4:
-            return dl.slice4d(array, [start, 0, 0, 0], [size, array.shape[1], array.shape[2], array.shape[3]]);
+            return tfc.slice4d(array, [start, 0, 0, 0], [size, array.shape[1], array.shape[2], array.shape[3]]);
         default:
@@ -174,9 +174,9 @@ throw new errors_1.ValueError("sliceAlongFirstAxis() received an unsupported subtype of Tensor: " +
         case 1:
-            return dl.slice1d(array, start, size);
+            return tfc.slice1d(array, start, size);
         case 2:
-            return dl.slice2d(array, [0, start], [array.shape[0], size]);
+            return tfc.slice2d(array, [0, start], [array.shape[0], size]);
         case 3:
-            return dl.slice3d(array, [0, 0, start], [array.shape[0], array.shape[1], size]);
+            return tfc.slice3d(array, [0, 0, start], [array.shape[0], array.shape[1], size]);
         case 4:
-            return dl.slice4d(array, [0, 0, 0, start], [array.shape[0], array.shape[1], array.shape[2], size]);
+            return tfc.slice4d(array, [0, 0, 0, start], [array.shape[0], array.shape[1], array.shape[2], size]);
         default:
@@ -203,3 +203,3 @@ throw new errors_1.ValueError("sliceAlongLastAxis() received an unsupported subtype of Tensor: " +
     }
-    return dl.concat(tensors, axis);
+    return tfc.concat(tensors, axis);
 }
@@ -210,9 +210,9 @@ exports.concatenate = concatenate;
         case 1:
-            return dl.concat1d([a, b]);
+            return tfc.concat1d([a, b]);
         case 2:
-            return dl.concat2d([a, b], 0);
+            return tfc.concat2d([a, b], 0);
         case 3:
-            return dl.concat3d([a, b], 0);
+            return tfc.concat3d([a, b], 0);
         case 4:
-            return dl.concat4d([a, b], 0);
+            return tfc.concat4d([a, b], 0);
         default:
@@ -232,3 +232,3 @@ throw new errors_1.ValueError('concatAlongFirstAxis() received an unsupported subtype of ' +
     }
-    return dl.tile(x, n);
+    return tfc.tile(x, n);
 }
@@ -252,3 +252,3 @@ exports.tile = tile;
 function zeros(shape, dtype) {
-    return dl.zeros(shape);
+    return tfc.zeros(shape);
 }
@@ -261,11 +261,11 @@ exports.zeros = zeros;
 function zerosLike(x, dtype, name) {
-    return new types_1.LayerVariable(dl.zerosLike(x), dtype, name);
+    return new types_1.LayerVariable(tfc.zerosLike(x), dtype, name);
 }
 exports.zerosLike = zerosLike;
 function ones(shape, dtype) {
-    return dl.ones(shape);
+    return tfc.ones(shape);
 }
 exports.ones = ones;
 function onesVariable(shape, dtype, name) {
-    var allocated = dl.ones(shape);
+    var allocated = tfc.ones(shape);
     return new types_1.LayerVariable(allocated, dtype, name);
@@ -275,3 +275,3 @@ }
 function onesLike(x, dtype, name) {
-    var allocated = dl.onesLike(x);
+    var allocated = tfc.onesLike(x);
     return new types_1.LayerVariable(allocated, dtype, name);
@@ -291,3 +291,3 @@ }
     }
-    return deeplearn_1.tensor2d(buffer, [size, size]);
+    return tfjs_core_1.tensor2d(buffer, [size, size]);
 }
@@ -300,31 +300,31 @@ exports.eye = eye;
 function neg(x) {
-    return dl.neg(x);
+    return tfc.neg(x);
 }
 exports.neg = neg;
 function add(x, y) {
-    return dl.add(x, y);
+    return tfc.add(x, y);
 }
 exports.add = add;
 function subtract(x, y) {
-    return dl.sub(x, y);
+    return tfc.sub(x, y);
 }
 exports.subtract = subtract;
 function multiply(x, y) {
-    return dl.mul(x, y);
+    return tfc.mul(x, y);
 }
 exports.multiply = multiply;
 function divide(x, y) {
-    return dl.div(x, y);
+    return tfc.div(x, y);
 }
 exports.divide = divide;
 function scalarTimesArray(c, x) {
-    return dl.mul(c, x);
+    return tfc.mul(c, x);
 }
 exports.scalarTimesArray = scalarTimesArray;
 function scalarPlusArray(c, x) {
-    return dl.add(c, x);
+    return tfc.add(c, x);
 }
 exports.scalarPlusArray = scalarPlusArray;
 function randomUniform(shape, minval, maxval, dtype, seed) {
-    return dl.randomUniform(shape, minval, maxval);
+    return tfc.randomUniform(shape, minval, maxval);
 }
@@ -340,3 +340,3 @@ exports.randomUniform = randomUniform;
     if (stddev === void 0) { stddev = 1.0; }
-    return dl.truncatedNormal(shape, mean, stddev);
+    return tfc.truncatedNormal(shape, mean, stddev);
 }
@@ -358,3 +358,3 @@ exports.truncatedNormal = truncatedNormal;
     var dtypeString = (dtype === types_1.DType.float32) ? 'float32' : 'int32';
-    return dl.randomNormal(shape, mean, stddev, dtypeString, seed);
+    return tfc.randomNormal(shape, mean, stddev, dtypeString, seed);
 }
@@ -374,7 +374,7 @@ exports.randomNormal = randomNormal;
 function updateAdd(x, increment) {
-    return x.write(dl.add(x.read(), increment));
+    return x.write(tfc.add(x.read(), increment));
 }
 exports.updateAdd = updateAdd;
 function updateSub(x, decrement) {
-    return x.write(dl.sub(x.read(), decrement));
+    return x.write(tfc.sub(x.read(), decrement));
 }
@@ -389,3 +389,3 @@ exports.updateSub = updateSub;
         if (ndim(x) === 2) {
-            return dl.matMul(x, y);
+            return tfc.matMul(x, y);
         }
@@ -397,3 +397,3 @@ else if (ndim(x) === 3) {
             x = x.reshape([xShape0 * xShape1, xShape2]);
-            return dl.matMul(x, y).reshape([
+            return tfc.matMul(x, y).reshape([
                 xShape0, xShape1, y.shape[1]
@@ -414,3 +414,3 @@ ]);
     }
-    return dl.oneHot(indices, numClasses);
+    return tfc.oneHot(indices, numClasses);
 }
@@ -420,3 +420,3 @@ exports.oneHot = oneHot;
     axis = normalizeAxis(x, axis);
-    return dl.mean(x, axis, keepDims);
+    return tfc.mean(x, axis, keepDims);
 }
@@ -426,3 +426,3 @@ exports.mean = mean;
     if (axis === void 0) { axis = -1; }
-    return dl.argMax(x, axis);
+    return tfc.argMax(x, axis);
 }
@@ -432,41 +432,41 @@ exports.argmax = argmax;
     if (Array.isArray(indices)) {
-        indices = deeplearn_1.tensor1d(indices);
+        indices = tfjs_core_1.tensor1d(indices);
     }
-    return dl.gather(reference, indices, axis);
+    return tfc.gather(reference, indices, axis);
 }
 exports.gather = gather;
 function max(x, axis, keepDims) {
-    return dl.max(x, axis, keepDims);
+    return tfc.max(x, axis, keepDims);
 }
 exports.max = max;
 function min(x, axis, keepDims) {
-    return dl.min(x, axis, keepDims);
+    return tfc.min(x, axis, keepDims);
 }
 exports.min = min;
 function minimum(x, y) {
-    return dl.minimum(x, y);
+    return tfc.minimum(x, y);
 }
 exports.minimum = minimum;
 function sum(x, axis, keepDims) {
-    return dl.sum(x, axis, keepDims);
+    return tfc.sum(x, axis, keepDims);
 }
 exports.sum = sum;
 function abs(x) {
-    return dl.abs(x);
+    return tfc.abs(x);
 }
 exports.abs = abs;
 function square(x) {
-    return dl.mulStrict(x, x);
+    return tfc.mulStrict(x, x);
 }
 exports.square = square;
 function sqrt(x) {
-    return dl.sqrt(x);
+    return tfc.sqrt(x);
 }
 exports.sqrt = sqrt;
 function exp(x) {
-    return dl.exp(x);
+    return tfc.exp(x);
 }
 exports.exp = exp;
 function log(x) {
-    return dl.log(x);
+    return tfc.log(x);
 }
@@ -476,3 +476,3 @@ exports.log = log;
     if (typeof (a) === 'number') {
-        a = deeplearn_1.scalar(Math.round(a), 'int32');
+        a = tfjs_core_1.scalar(Math.round(a), 'int32');
     }
@@ -482,31 +482,31 @@ if (a.dtype !== 'int32') {
     }
-    return dl.pow(x, a);
+    return tfc.pow(x, a);
 }
 exports.pow = pow;
 function clip(x, minValue, maxValue) {
-    return dl.clipByValue(x, minValue, maxValue);
+    return tfc.clipByValue(x, minValue, maxValue);
 }
 exports.clip = clip;
 function equal(x, y) {
-    return dl.equal(x, y);
+    return tfc.equal(x, y);
 }
 exports.equal = equal;
 function greater(x, y) {
-    return dl.greater(x, y);
+    return tfc.greater(x, y);
 }
 exports.greater = greater;
 function greaterEqual(x, y) {
-    return dl.greaterEqual(x, y);
+    return tfc.greaterEqual(x, y);
 }
 exports.greaterEqual = greaterEqual;
 function maximum(x, y) {
-    return dl.maximum(x, y);
+    return tfc.maximum(x, y);
 }
 exports.maximum = maximum;
 function sin(x) {
-    return dl.sin(x.value());
+    return tfc.sin(x.value());
 }
 exports.sin = sin;
 function cos(x) {
-    return dl.cos(x.value());
+    return tfc.cos(x.value());
 }
@@ -518,9 +518,9 @@ exports.cos = cos;
     if (ndim(x) === 2) {
-        out = dl.batchNormalization2d(x, mean, variance, epsilon);
+        out = tfc.batchNormalization2d(x, mean, variance, epsilon);
     }
     else if (ndim(x) === 3) {
-        out = dl.batchNormalization3d(x, mean, variance, epsilon);
+        out = tfc.batchNormalization3d(x, mean, variance, epsilon);
     }
     else if (ndim(x) === 4) {
-        out = dl.batchNormalization4d(x, mean, variance, epsilon);
+        out = tfc.batchNormalization4d(x, mean, variance, epsilon);
     }
@@ -548,3 +548,3 @@ else {
     }
-    return dl.add(x, bias);
+    return tfc.add(x, bias);
 }
@@ -558,23 +558,23 @@ exports.biasAdd = biasAdd;
     }
-    return dl.elu(x);
+    return tfc.elu(x);
 }
 exports.elu = elu;
 function selu(x) {
-    return dl.selu(x);
+    return tfc.selu(x);
 }
 exports.selu = selu;
 function relu(x) {
-    return dl.relu(x);
+    return tfc.relu(x);
 }
 exports.relu = relu;
 function softplus(x) {
-    return dl.log(dl.add(getScalar(1), dl.exp(x)));
+    return tfc.log(tfc.add(getScalar(1), tfc.exp(x)));
 }
 exports.softplus = softplus;
 function softsign(x) {
-    return dl.div(x, dl.add(getScalar(1), dl.abs(x)));
+    return tfc.div(x, tfc.add(getScalar(1), tfc.abs(x)));
 }
 exports.softsign = softsign;
 function tanh(x) {
-    return dl.tanh(x);
+    return tfc.tanh(x);
 }
@@ -590,5 +590,5 @@ exports.tanh = tanh;
     }
-    var multiplier = dl.step(dl.add(neg(level), randomUniform(x.shape, 0, 1, types_1.DType.float32)));
-    multiplier = dl.mul(divide(getScalar(1), subtract(getScalar(1), level)), multiplier);
-    return dl.mul(x, multiplier);
+    var multiplier = tfc.step(tfc.add(neg(level), randomUniform(x.shape, 0, 1, types_1.DType.float32)));
+    multiplier = tfc.mul(divide(getScalar(1), subtract(getScalar(1), level)), multiplier);
+    return tfc.mul(x, multiplier);
 }
@@ -598,3 +598,3 @@ exports.dropout = dropout;
     var squareSum = sum(square(x), axis, true);
-    var epsilonTensor = scalarTimesArray(deeplearn_1.scalar(exports.epsilon()), dl.onesLike(x));
+    var epsilonTensor = scalarTimesArray(tfjs_core_1.scalar(exports.epsilon()), tfc.onesLike(x));
     var norm = sqrt(maximum(squareSum, epsilonTensor));
@@ -606,3 +606,3 @@ return divide(x, norm);
     if (dataFormat === common_1.DataFormat.CHANNEL_FIRST) {
-        return dl.transpose(x, [0, 2, 3, 1]);
+        return tfc.transpose(x, [0, 2, 3, 1]);
     }
@@ -643,3 +643,3 @@ else {
     }
-    var y = dl.conv1d(x, kernel, strides, padding === common_1.PaddingMode.SAME ? 'same' : 'valid');
+    var y = tfc.conv1d(x, kernel, strides, padding === common_1.PaddingMode.SAME ? 'same' : 'valid');
     if (bias != null) {
@@ -686,3 +686,3 @@ y = biasAdd(y, bias);
     }
-    y = dl.conv2d(y, kernel, strides, padding === common_1.PaddingMode.SAME ? 'same' : 'valid');
+    y = tfc.conv2d(y, kernel, strides, padding === common_1.PaddingMode.SAME ? 'same' : 'valid');
     if (bias != null) {
@@ -692,3 +692,3 @@ y = biasAdd(y, bias);
     if (dataFormat === common_1.DataFormat.CHANNEL_FIRST) {
-        y = dl.transpose(y, [0, 3, 1, 2]);
+        y = tfc.transpose(y, [0, 3, 1, 2]);
     }
@@ -713,5 +713,5 @@ return y;
     }
-    y = dl.depthwiseConv2d(y, depthwiseKernel, strides, padding === common_1.PaddingMode.SAME ? 'same' : 'valid', dilationRate);
+    y = tfc.depthwiseConv2d(y, depthwiseKernel, strides, padding === common_1.PaddingMode.SAME ? 'same' : 'valid', dilationRate);
     if (dataFormat === common_1.DataFormat.CHANNEL_FIRST) {
-        y = dl.transpose(y, [0, 3, 1, 2]);
+        y = tfc.transpose(y, [0, 3, 1, 2]);
     }
@@ -738,9 +738,9 @@ return y;
     if (poolMode === common_1.PoolMode.MAX) {
-        y = dl.maxPool(x, poolSize, strides, paddingString);
+        y = tfc.maxPool(x, poolSize, strides, paddingString);
     }
     else {
-        y = dl.avgPool(x, poolSize, strides, paddingString);
+        y = tfc.avgPool(x, poolSize, strides, paddingString);
     }
     if (dataFormat === common_1.DataFormat.CHANNEL_FIRST) {
-        y = dl.transpose(y, [0, 3, 1, 2]);
+        y = tfc.transpose(y, [0, 3, 1, 2]);
     }
@@ -770,3 +770,3 @@ return y;
     if (axis === void 0) { axis = -1; }
-    return dl.softmax(x, axis);
+    return tfc.softmax(x, axis);
 }
@@ -784,3 +784,3 @@ exports.softmax = softmax;
     output = clip(output, exports.epsilon(), 1 - exports.epsilon());
-    return dl.neg(dl.sum(dl.mul(target, dl.log(output)), shape(output).length - 1));
+    return tfc.neg(tfc.sum(tfc.mul(target, tfc.log(output)), shape(output).length - 1));
 }
@@ -790,5 +790,5 @@ exports.categoricalCrossentropy = categoricalCrossentropy;
     if (fromLogits === void 0) { fromLogits = false; }
-    var flatTarget = dl.floor(flatten(target));
+    var flatTarget = tfc.floor(flatten(target));
     var outputShape = shape(output);
-    var oneHotTarget = reshape(dl.oneHot(flatTarget, outputShape[outputShape.length - 1]), outputShape);
+    var oneHotTarget = reshape(tfc.oneHot(flatTarget, outputShape[outputShape.length - 1]), outputShape);
     return categoricalCrossentropy(oneHotTarget, output, fromLogits);
@@ -802,3 +802,3 @@ }
         y = clip(output, exports.epsilon(), 1 - exports.epsilon());
-        y = log(divide(y, subtract(dl.onesLike(y), y)));
+        y = log(divide(y, subtract(tfc.onesLike(y), y)));
     }
@@ -812,6 +812,6 @@ else {
 function sigmoidCrossEntropyWithLogits(target, output) {
-    var maxOutput = dl.maximum(output, dl.zerosLike(output));
-    var outputXTarget = dl.mul(output, target);
-    var sigmoidOutput = dl.log(dl.add(getScalar(1), dl.exp(dl.neg(dl.abs(output)))));
-    var result = dl.add(dl.sub(maxOutput, outputXTarget), sigmoidOutput);
+    var maxOutput = tfc.maximum(output, tfc.zerosLike(output));
+    var outputXTarget = tfc.mul(output, target);
+    var sigmoidOutput = tfc.log(tfc.add(getScalar(1), tfc.exp(tfc.neg(tfc.abs(output)))));
+    var result = tfc.add(tfc.sub(maxOutput, outputXTarget), sigmoidOutput);
     return result;
@@ -821,7 +821,7 @@ }
 function sigmoid(x) {
-    return dl.sigmoid(x);
+    return tfc.sigmoid(x);
 }
 exports.sigmoid = sigmoid;
 function hardSigmoid(x) {
-    var y = scalarPlusArray(deeplearn_1.scalar(0.5), scalarTimesArray(deeplearn_1.scalar(0.2), x));
+    var y = scalarPlusArray(tfjs_core_1.scalar(0.5), scalarTimesArray(tfjs_core_1.scalar(0.2), x));
     return clip(y, 0, 1);
@@ -884,5 +884,5 @@ }
     var variableList = variables.map(function (variable) { return variable.read(); });
-    var valudAndGrads = deeplearn_1.variableGrads(lossFn, variableList);
+    var valudAndGrads = tfjs_core_1.variableGrads(lossFn, variableList);
     return variables.map(function (variable) { return valudAndGrads.grads[variable.name]; });
 }
 exports.gradients = gradients;

dist/callbacks.d.ts

@@ -1,6 +0,9 @@
-import { Scalar, Tensor } from 'deeplearn';
+import { Scalar, Tensor } from '@tensorflow/tfjs-core';
 import { Model } from './engine/training';
-export declare type Logs = {
+export declare type UnresolvedLogs = {
     [key: string]: number | Scalar;
 };
+export declare type Logs = {
+    [key: string]: number;
+};
 export declare type Params = {
@@ -15,8 +18,8 @@ [key: string]: number | string | boolean | number[] | string[] | boolean[];
     setModel(model: Model): void;
-    onEpochBegin(epoch: number, logs?: Logs): Promise<void>;
-    onEpochEnd(epoch: number, logs?: Logs): Promise<void>;
-    onBatchBegin(batch: number, logs?: Logs): Promise<void>;
-    onBatchEnd(batch: number, logs?: Logs): Promise<void>;
-    onTrainBegin(logs?: Logs): Promise<void>;
-    onTrainEnd(logs?: Logs): Promise<void>;
+    onEpochBegin(epoch: number, logs?: UnresolvedLogs): Promise<void>;
+    onEpochEnd(epoch: number, logs?: UnresolvedLogs): Promise<void>;
+    onBatchBegin(batch: number, logs?: UnresolvedLogs): Promise<void>;
+    onBatchEnd(batch: number, logs?: UnresolvedLogs): Promise<void>;
+    onTrainBegin(logs?: UnresolvedLogs): Promise<void>;
+    onTrainEnd(logs?: UnresolvedLogs): Promise<void>;
 }
@@ -30,8 +33,8 @@ export declare class CallbackList {
     setModel(model: Model): void;
-    onEpochBegin(epoch: number, logs?: Logs): Promise<void>;
-    onEpochEnd(epoch: number, logs?: Logs): Promise<void>;
-    onBatchBegin(batch: number, logs?: Logs): Promise<void>;
-    onBatchEnd(batch: number, logs?: Logs): Promise<void>;
-    onTrainBegin(logs?: Logs): Promise<void>;
-    onTrainEnd(logs?: Logs): Promise<void>;
+    onEpochBegin(epoch: number, logs?: UnresolvedLogs): Promise<void>;
+    onEpochEnd(epoch: number, logs?: UnresolvedLogs): Promise<void>;
+    onBatchBegin(batch: number, logs?: UnresolvedLogs): Promise<void>;
+    onBatchEnd(batch: number, logs?: UnresolvedLogs): Promise<void>;
+    onTrainBegin(logs?: UnresolvedLogs): Promise<void>;
+    onTrainEnd(logs?: UnresolvedLogs): Promise<void>;
 }
@@ -43,6 +46,7 @@ export declare class BaseLogger extends Callback {
     constructor();
-    onEpochBegin(epoch: number, logs?: Logs): Promise<void>;
-    onBatchEnd(batch: number, logs?: Logs): Promise<void>;
-    onEpochEnd(epoch: number, logs?: Logs): Promise<void>;
+    onEpochBegin(epoch: number, logs?: UnresolvedLogs): Promise<void>;
+    onBatchEnd(batch: number, logs?: UnresolvedLogs): Promise<void>;
+    onEpochEnd(epoch: number, logs?: UnresolvedLogs): Promise<void>;
 }
+export declare function resolveScalarsInLogs(logs: UnresolvedLogs): Promise<void>;
 export declare class History extends Callback {
@@ -53,4 +57,4 @@ epoch: number[];
     };
-    onTrainBegin(logs?: Logs): Promise<void>;
-    onEpochEnd(epoch: number, logs?: Logs): Promise<void>;
+    onTrainBegin(logs?: UnresolvedLogs): Promise<void>;
+    onEpochEnd(epoch: number, logs?: UnresolvedLogs): Promise<void>;
     syncData(): Promise<void>;
@@ -74,9 +78,9 @@ }
     constructor(config: CustomCallbackConfig);
-    onEpochBegin(epoch: number, logs?: Logs): Promise<void>;
-    onEpochEnd(epoch: number, logs?: Logs): Promise<void>;
-    onBatchBegin(batch: number, logs?: Logs): Promise<void>;
-    onBatchEnd(batch: number, logs?: Logs): Promise<void>;
-    onTrainBegin(logs?: Logs): Promise<void>;
-    onTrainEnd(logs?: Logs): Promise<void>;
+    onEpochBegin(epoch: number, logs?: UnresolvedLogs): Promise<void>;
+    onEpochEnd(epoch: number, logs?: UnresolvedLogs): Promise<void>;
+    onBatchBegin(batch: number, logs?: UnresolvedLogs): Promise<void>;
+    onBatchEnd(batch: number, logs?: UnresolvedLogs): Promise<void>;
+    onTrainBegin(logs?: UnresolvedLogs): Promise<void>;
+    onTrainEnd(logs?: UnresolvedLogs): Promise<void>;
 }
 export declare function standardizeCallbacks(callbacks: Callback | Callback[] | CustomCallbackConfig | CustomCallbackConfig[]): Callback[];

dist/callbacks.js

@@ -48,3 +48,3 @@ "use strict";
 Object.defineProperty(exports, "__esModule", { value: true });
-var deeplearn_1 = require("deeplearn");
+var tfjs_core_1 = require("@tensorflow/tfjs-core");
 var K = require("./backend/deeplearnjs_backend");
@@ -317,3 +317,3 @@ var generic_utils = require("./utils/generic_utils");
                         }
-                        deeplearn_1.tidy(function () {
+                        tfjs_core_1.tidy(function () {
                             _this.totals[key] =
@@ -347,3 +347,3 @@ K.scalarPlusArray(_this.totals[key], K.multiply(value, K.getScalar(batchSize)));
                         else {
-                            deeplearn_1.tidy(function () {
+                            tfjs_core_1.tidy(function () {
                                 logs[key] =
@@ -368,2 +368,33 @@ K.scalarTimesArray(K.divide(K.getScalar(1), K.getScalar(_this.seen)), _this.totals[key]);
 exports.BaseLogger = BaseLogger;
+function resolveScalarsInLogs(logs) {
+    return __awaiter(this, void 0, void 0, function () {
+        var promises, keys, key, value, valueScalar, values, i;
+        return __generator(this, function (_a) {
+            switch (_a.label) {
+                case 0:
+                    if (logs == null) {
+                        return [2];
+                    }
+                    promises = [];
+                    keys = [];
+                    for (key in logs) {
+                        value = logs[key];
+                        if (typeof value !== 'number') {
+                            valueScalar = value;
+                            promises.push(valueScalar.data());
+                            keys.push(key);
+                        }
+                    }
+                    return [4, Promise.all(promises)];
+                case 1:
+                    values = _a.sent();
+                    for (i = 0; i < values.length; ++i) {
+                        logs[keys[i]] = values[i][0];
+                    }
+                    return [2];
+            }
+        });
+    });
+}
+exports.resolveScalarsInLogs = resolveScalarsInLogs;
 var History = (function (_super) {
@@ -453,8 +484,11 @@ __extends(History, _super);
                     case 0:
-                        if (!(this.epochBegin != null)) return [3, 2];
-                        return [4, this.epochBegin(epoch, logs)];
+                        if (!(this.epochBegin != null)) return [3, 3];
+                        return [4, resolveScalarsInLogs(logs)];
                     case 1:
                         _a.sent();
-                        _a.label = 2;
-                    case 2: return [2];
+                        return [4, this.epochBegin(epoch, logs)];
+                    case 2:
+                        _a.sent();
+                        _a.label = 3;
+                    case 3: return [2];
                 }
@@ -469,8 +503,11 @@ });
                     case 0:
-                        if (!(this.epochEnd != null)) return [3, 2];
-                        return [4, this.epochEnd(epoch, logs)];
+                        if (!(this.epochEnd != null)) return [3, 3];
+                        return [4, resolveScalarsInLogs(logs)];
                     case 1:
                         _a.sent();
-                        _a.label = 2;
-                    case 2: return [2];
+                        return [4, this.epochEnd(epoch, logs)];
+                    case 2:
+                        _a.sent();
+                        _a.label = 3;
+                    case 3: return [2];
                 }
@@ -485,8 +522,11 @@ });
                     case 0:
-                        if (!(this.batchBegin != null)) return [3, 2];
-                        return [4, this.batchBegin(batch, logs)];
+                        if (!(this.batchBegin != null)) return [3, 3];
+                        return [4, resolveScalarsInLogs(logs)];
                     case 1:
                         _a.sent();
-                        _a.label = 2;
-                    case 2: return [2];
+                        return [4, this.batchBegin(batch, logs)];
+                    case 2:
+                        _a.sent();
+                        _a.label = 3;
+                    case 3: return [2];
                 }
@@ -501,8 +541,11 @@ });
                     case 0:
-                        if (!(this.batchEnd != null)) return [3, 2];
-                        return [4, this.batchEnd(batch, logs)];
+                        if (!(this.batchEnd != null)) return [3, 3];
+                        return [4, resolveScalarsInLogs(logs)];
                     case 1:
                         _a.sent();
-                        _a.label = 2;
-                    case 2: return [2];
+                        return [4, this.batchEnd(batch, logs)];
+                    case 2:
+                        _a.sent();
+                        _a.label = 3;
+                    case 3: return [2];
                 }
@@ -517,8 +560,11 @@ });
                     case 0:
-                        if (!(this.trainBegin != null)) return [3, 2];
-                        return [4, this.trainBegin(logs)];
+                        if (!(this.trainBegin != null)) return [3, 3];
+                        return [4, resolveScalarsInLogs(logs)];
                     case 1:
                         _a.sent();
-                        _a.label = 2;
-                    case 2: return [2];
+                        return [4, this.trainBegin(logs)];
+                    case 2:
+                        _a.sent();
+                        _a.label = 3;
+                    case 3: return [2];
                 }
@@ -533,8 +579,11 @@ });
                     case 0:
-                        if (!(this.trainEnd != null)) return [3, 2];
-                        return [4, this.trainEnd(logs)];
+                        if (!(this.trainEnd != null)) return [3, 3];
+                        return [4, resolveScalarsInLogs(logs)];
                     case 1:
                         _a.sent();
-                        _a.label = 2;
-                    case 2: return [2];
+                        return [4, this.trainEnd(logs)];
+                    case 2:
+                        _a.sent();
+                        _a.label = 3;
+                    case 3: return [2];
                 }
@@ -541,0 +590,0 @@ });

dist/constraints.d.ts

@@ -1,2 +0,2 @@
-import { Tensor } from 'deeplearn';
+import { Tensor } from '@tensorflow/tfjs-core';
 import { ConfigDict, ConfigDictValue } from './types';
@@ -3,0 +3,0 @@ export declare abstract class Constraint {

dist/constraints.js

@@ -19,3 +19,3 @@ "use strict";
 Object.defineProperty(exports, "__esModule", { value: true });
-var deeplearn_1 = require("deeplearn");
+var tfjs_core_1 = require("@tensorflow/tfjs-core");
 var K = require("./backend/deeplearnjs_backend");
@@ -33,3 +33,3 @@ var generic_utils_1 = require("./utils/generic_utils");
     Constraint = __decorate([
-        deeplearn_1.doc({ heading: 'Constraints', subheading: 'Classes', namespace: 'constraints' })
+        tfjs_core_1.doc({ heading: 'Constraints', subheading: 'Classes', namespace: 'constraints' })
     ], Constraint);
@@ -36,0 +36,0 @@ return Constraint;

dist/engine/executor.d.ts

@@ -1,2 +0,2 @@
-import { Tensor } from 'deeplearn';
+import { Tensor } from '@tensorflow/tfjs-core';
 import { SymbolicTensor } from '../types';
@@ -3,0 +3,0 @@ export interface Feed {

dist/engine/topology.d.ts

@@ -1,2 +0,2 @@
-import { Tensor } from 'deeplearn';
+import { Tensor } from '@tensorflow/tfjs-core';
 import { Constraint } from '../constraints';
@@ -3,0 +3,0 @@ import { Initializer } from '../initializers';

dist/engine/training.d.ts

@@ -1,2 +0,2 @@
-import { Optimizer, Scalar, Tensor, Tensor1D } from 'deeplearn';
+import { Optimizer, Scalar, Tensor, Tensor1D } from '@tensorflow/tfjs-core';
 import { Callback, CustomCallbackConfig, History } from '../callbacks';
@@ -27,9 +27,13 @@ import * as optimizers from '../optimizers';
 }
+export interface ModelPredictConfig {
+    batchSize?: number;
+    verbose?: boolean;
+}
+export interface ModelEvaluateConfig {
+    batchSize?: number;
+    verbose?: ModelLoggingVerbosity;
+    sampleWeight?: Tensor;
+    steps?: number;
+}
 export interface ModelFitConfig {
-    x: Tensor | Tensor[] | {
-        [inputName: string]: Tensor;
-    };
-    y: Tensor | Tensor[] | {
-        [inputName: string]: Tensor;
-    };
     batchSize?: number;
@@ -78,6 +82,6 @@ epochs?: number;
     private checkTrainableWeightsConsistency();
-    evaluate(x: Tensor | Tensor[], y: Tensor | Tensor[], batchSize?: number, verbose?: ModelLoggingVerbosity, sampleWeight?: Tensor, steps?: number): Scalar | Scalar[];
+    evaluate(x: Tensor | Tensor[], y: Tensor | Tensor[], config?: ModelEvaluateConfig): Promise<Scalar | Scalar[]>;
     private checkNumSamples(ins, batchSize?, steps?, stepsName?);
     private predictLoop(ins, batchSize?, verbose?);
-    predict(x: Tensor | Tensor[], batchSize?: number, verbose?: boolean): Tensor | Tensor[];
+    predict(x: Tensor | Tensor[], config?: ModelPredictConfig): Promise<Tensor | Tensor[]>;
     predictOnBatch(x: Tensor): Tensor | Tensor[];
@@ -93,3 +97,7 @@ protected standardizeUserData(x: Tensor | Tensor[] | {
     private makeTestFunction();
-    fit(config: ModelFitConfig): Promise<History>;
+    fit(x: Tensor | Tensor[] | {
+        [inputName: string]: Tensor;
+    }, y: Tensor | Tensor[] | {
+        [inputName: string]: Tensor;
+    }, config?: ModelFitConfig): Promise<History>;
 }

dist/engine/training.js

@@ -54,4 +54,4 @@ "use strict";
 Object.defineProperty(exports, "__esModule", { value: true });
-var dl = require("deeplearn");
-var deeplearn_1 = require("deeplearn");
+var tfc = require("@tensorflow/tfjs-core");
+var tfjs_core_1 = require("@tensorflow/tfjs-core");
 var _ = require("underscore");
@@ -68,3 +68,3 @@ var K = require("../backend/deeplearnjs_backend");
 function isDataTensor(x) {
-    return x instanceof deeplearn_1.Tensor;
+    return x instanceof tfjs_core_1.Tensor;
 }
@@ -519,13 +519,16 @@ exports.isDataTensor = isDataTensor;
     };
-    Model.prototype.evaluate = function (x, y, batchSize, verbose, sampleWeight, steps) {
-        if (batchSize === void 0) { batchSize = 32; }
-        if (batchSize == null) {
-            batchSize = 32;
-        }
-        var standardizedOuts = this.standardizeUserData(x, y, true, batchSize);
-        var ins = standardizedOuts[0].concat(standardizedOuts[1]);
-        this.makeTestFunction();
-        var f = this.testFunction;
-        var testOuts = this.testLoop(f, ins, batchSize, verbose, steps);
-        return generic_utils_1.singletonOrArray(testOuts);
+    Model.prototype.evaluate = function (x, y, config) {
+        if (config === void 0) { config = {}; }
+        return __awaiter(this, void 0, void 0, function () {
+            var batchSize, standardizedOuts, ins, f, testOuts;
+            return __generator(this, function (_a) {
+                batchSize = config.batchSize == null ? 32 : config.batchSize;
+                standardizedOuts = this.standardizeUserData(x, y, true, batchSize);
+                ins = standardizedOuts[0].concat(standardizedOuts[1]);
+                this.makeTestFunction();
+                f = this.testFunction;
+                testOuts = this.testLoop(f, ins, batchSize, config.verbose, config.steps);
+                return [2, generic_utils_1.singletonOrArray(testOuts)];
+            });
+        });
     };
@@ -567,3 +570,3 @@ Model.prototype.checkNumSamples = function (ins, batchSize, steps, stepsName) {
         var _loop_3 = function (batchIndex) {
-            var batchOuts = dl.tidy(function () {
+            var batchOuts = tfc.tidy(function () {
                 var batchStart = batches[batchIndex][0];
@@ -601,7 +604,12 @@ var batchEnd = batches[batchIndex][1];
     };
-    Model.prototype.predict = function (x, batchSize, verbose) {
-        if (batchSize === void 0) { batchSize = 32; }
-        if (verbose === void 0) { verbose = false; }
-        checkInputData(x, this.inputNames, this.feedInputShapes, false);
-        return this.predictLoop(x, batchSize);
+    Model.prototype.predict = function (x, config) {
+        if (config === void 0) { config = {}; }
+        return __awaiter(this, void 0, void 0, function () {
+            var batchSize;
+            return __generator(this, function (_a) {
+                checkInputData(x, this.inputNames, this.feedInputShapes, false);
+                batchSize = config.batchSize == null ? 32 : config.batchSize;
+                return [2, this.predictLoop(x, batchSize)];
+            });
+        });
     };
@@ -643,3 +651,2 @@ Model.prototype.predictOnBatch = function (x) {
     Model.prototype.fitLoop = function (f, ins, outLabels, batchSize, epochs, verbose, callbacks, valF, valIns, shuffle, callbackMetrics, initialEpoch, stepsPerEpoch, validationSteps) {
-        if (epochs === void 0) { epochs = 100; }
         if (initialEpoch === void 0) { initialEpoch = 0; }
@@ -652,2 +659,8 @@ return __awaiter(this, void 0, void 0, function () {
                     case 0:
+                        if (batchSize == null) {
+                            batchSize = 32;
+                        }
+                        if (epochs == null) {
+                            epochs = 100;
+                        }
                         if (shuffle == null) {
@@ -715,3 +728,3 @@ shuffle = true;
                                         }
-                                        epochIndexArray1D_1 = deeplearn_1.tensor1d(epochIndexArray);
+                                        epochIndexArray1D_1 = tfjs_core_1.tensor1d(epochIndexArray);
                                         batches_1 = makeBatches(numTrainSamples, batchSize);
@@ -727,3 +740,3 @@ _loop_5 = function (batchIndex) {
                                                         _a.sent();
-                                                        dl.tidy(function () {
+                                                        tfc.tidy(function () {
                                                             var batchStart = batches_1[batchIndex][0];
@@ -814,3 +827,3 @@ var batchEnd = batches_1[batchIndex][1];
             var batches = makeBatches(numSamples, batchSize);
-            var indexArray = deeplearn_1.tensor1d(_.range(numSamples));
+            var indexArray = tfjs_core_1.tensor1d(_.range(numSamples));
             for (var batchIndex = 0; batchIndex < batches.length; ++batchIndex) {
@@ -856,3 +869,3 @@ var batchStart = batches[batchIndex][0];
         this.testFunction = function (data) {
-            return dl.tidy(function () {
+            return tfc.tidy(function () {
                 var valOutputs = [];
@@ -889,17 +902,10 @@ var totalLoss;
     };
-    Model.prototype.fit = function (config) {
+    Model.prototype.fit = function (x, y, config) {
+        if (config === void 0) { config = {}; }
         return __awaiter(this, void 0, void 0, function () {
             var _this = this;
-            var standardizedOuts, inputs, targets, doValidation, valX, valY, valIns, valStandardized, splitAt, originalBatchSize, ins, trainFunction, outLabels, valFunction, callbackMetrics, callbacks;
+            var batchSize, standardizedOuts, inputs, targets, doValidation, valX, valY, valIns, valStandardized, splitAt, originalBatchSize, ins, trainFunction, outLabels, valFunction, callbackMetrics, callbacks;
             return __generator(this, function (_a) {
-                if (config.batchSize == null) {
-                    config.batchSize = 32;
-                }
-                if (config.x == null) {
-                    throw new errors_1.ValueError('x data is missing from ModelFitConfig.');
-                }
-                if (config.y == null) {
-                    throw new errors_1.ValueError('y data is missing from ModelFitConfig.');
-                }
-                standardizedOuts = this.standardizeUserData(config.x, config.y, false, config.batchSize);
+                batchSize = config.batchSize == null ? 32 : config.batchSize;
+                standardizedOuts = this.standardizeUserData(x, y, false, batchSize);
                 inputs = standardizedOuts[0];
@@ -922,3 +928,3 @@ targets = standardizedOuts[1];
                     }
-                    valStandardized = this.standardizeUserData(valX, valY, true, config.batchSize);
+                    valStandardized = this.standardizeUserData(valX, valY, true, batchSize);
                     valX = valStandardized[0];
@@ -1003,3 +1009,3 @@ valY = valStandardized[1];
                 callbacks = callbacks_1.standardizeCallbacks(config.callbacks);
-                return [2, this.fitLoop(trainFunction, ins, outLabels, config.batchSize, config.epochs, config.verbose, callbacks, valFunction, valIns, config.shuffle, callbackMetrics, null, null, null)];
+                return [2, this.fitLoop(trainFunction, ins, outLabels, batchSize, config.epochs, config.verbose, callbacks, valFunction, valIns, config.shuffle, callbackMetrics, null, null, null)];
             });
@@ -1009,18 +1015,18 @@ });
     __decorate([
-        deeplearn_1.doc({ heading: 'Models', subheading: 'Classes' })
+        tfjs_core_1.doc({ heading: 'Models', subheading: 'Classes' })
     ], Model.prototype, "compile", null);
     __decorate([
-        deeplearn_1.doc({ heading: 'Models', subheading: 'Classes' })
+        tfjs_core_1.doc({ heading: 'Models', subheading: 'Classes', configParamIndices: [2] })
     ], Model.prototype, "evaluate", null);
     __decorate([
-        deeplearn_1.doc({ heading: 'Models', subheading: 'Classes' })
+        tfjs_core_1.doc({ heading: 'Models', subheading: 'Classes', configParamIndices: [1] })
     ], Model.prototype, "predict", null);
     __decorate([
-        deeplearn_1.doc({ heading: 'Models', subheading: 'Classes' })
+        tfjs_core_1.doc({ heading: 'Models', subheading: 'Classes' })
     ], Model.prototype, "predictOnBatch", null);
     __decorate([
-        deeplearn_1.doc({ heading: 'Models', subheading: 'Classes' })
+        tfjs_core_1.doc({ heading: 'Models', subheading: 'Classes', configParamIndices: [2] })
     ], Model.prototype, "fit", null);
     Model = __decorate([
-        deeplearn_1.doc({ heading: 'Models', subheading: 'Classes' })
+        tfjs_core_1.doc({ heading: 'Models', subheading: 'Classes' })
     ], Model);
@@ -1027,0 +1033,0 @@ return Model;

dist/exports.d.ts

@@ -12,3 +12,3 @@ import { MaxNorm, MaxNormConfig, MinMaxNorm, MinMaxNormConfig, NonNeg, UnitNorm, UnitNormConfig } from './constraints';
 import { GlobalPooling2DLayerConfig, Pooling1DLayerConfig, Pooling2DLayerConfig } from './layers/pooling';
-import { GRUCellLayerConfig, GRULayerConfig, LSTMCellLayerConfig, LSTMLayerConfig, RNNCell, SimpleRNNCellLayerConfig, SimpleRNNLayerConfig } from './layers/recurrent';
+import { GRUCellLayerConfig, GRULayerConfig, LSTMCellLayerConfig, LSTMLayerConfig, RNN, RNNCell, RNNLayerConfig, SimpleRNNCellLayerConfig, SimpleRNNLayerConfig, StackedRNNCellsConfig } from './layers/recurrent';
 import { BidirectionalLayerConfig, WrapperLayerConfig } from './layers/wrappers';
@@ -56,2 +56,4 @@ import { Sequential, SequentialConfig } from './models';
     static simpleRNNCell(config: SimpleRNNCellLayerConfig): RNNCell;
+    static rnn(config: RNNLayerConfig): RNN;
+    static stackedRNNCells(config: StackedRNNCellsConfig): RNNCell;
     static bidirectional(config: BidirectionalLayerConfig): Layer;
@@ -58,0 +60,0 @@ static timeDistributed(config: WrapperLayerConfig): Layer;

dist/exports.js

@@ -9,3 +9,3 @@ "use strict";
 Object.defineProperty(exports, "__esModule", { value: true });
-var deeplearn_1 = require("deeplearn");
+var tfjs_core_1 = require("@tensorflow/tfjs-core");
 var constraints_1 = require("./constraints");
@@ -45,3 +45,3 @@ var topology_1 = require("./engine/topology");
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Models',
@@ -54,3 +54,3 @@ subheading: 'Creation',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Models',
@@ -63,3 +63,3 @@ subheading: 'Creation',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Models',
@@ -71,3 +71,3 @@ subheading: 'Loading',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Models',
@@ -80,3 +80,3 @@ subheading: 'Inputs',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Models',
@@ -184,2 +184,8 @@ subheading: 'Inputs',
     };
+    LayerExports.rnn = function (config) {
+        return new recurrent_1.RNN(config);
+    };
+    LayerExports.stackedRNNCells = function (config) {
+        return new recurrent_1.StackedRNNCells(config);
+    };
     LayerExports.bidirectional = function (config) {
@@ -192,3 +198,3 @@ return new wrappers_1.Bidirectional(config);
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Layers',
@@ -202,3 +208,3 @@ subheading: 'Convolutional',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Layers',
@@ -212,3 +218,3 @@ subheading: 'Convolutional',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Layers',
@@ -222,3 +228,3 @@ subheading: 'Convolutional',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Layers',
@@ -232,3 +238,3 @@ subheading: 'Core',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Layers',
@@ -242,3 +248,3 @@ subheading: 'Core',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Layers',
@@ -252,3 +258,3 @@ subheading: 'Core',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Layers',
@@ -262,3 +268,3 @@ subheading: 'Core',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Layers',
@@ -272,3 +278,3 @@ subheading: 'Core',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Layers',
@@ -282,3 +288,3 @@ subheading: 'Core',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Layers',
@@ -292,3 +298,3 @@ subheading: 'Merge',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Layers',
@@ -302,3 +308,3 @@ subheading: 'Merge',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Layers',
@@ -312,3 +318,3 @@ subheading: 'Merge',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Layers',
@@ -322,3 +328,3 @@ subheading: 'Merge',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Layers',
@@ -332,3 +338,3 @@ subheading: 'Merge',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Layers',
@@ -342,3 +348,3 @@ subheading: 'Merge',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Layers',
@@ -352,3 +358,3 @@ subheading: 'Normalization',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Layers',
@@ -362,3 +368,3 @@ subheading: 'Pooling',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Layers',
@@ -372,3 +378,3 @@ subheading: 'Pooling',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Layers',
@@ -382,3 +388,3 @@ subheading: 'Pooling',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Layers',
@@ -392,3 +398,3 @@ subheading: 'Pooling',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Layers',
@@ -402,3 +408,3 @@ subheading: 'Pooling',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Layers',
@@ -412,3 +418,3 @@ subheading: 'Pooling',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Layers',
@@ -422,3 +428,3 @@ subheading: 'Pooling',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Layers',
@@ -432,3 +438,3 @@ subheading: 'Pooling',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Layers',
@@ -442,3 +448,3 @@ subheading: 'Recurrent',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Layers',
@@ -452,3 +458,3 @@ subheading: 'Recurrent',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Layers',
@@ -462,3 +468,3 @@ subheading: 'Recurrent',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Layers',
@@ -472,3 +478,3 @@ subheading: 'Recurrent',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Layers',
@@ -482,3 +488,3 @@ subheading: 'Recurrent',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Layers',
@@ -492,4 +498,22 @@ subheading: 'Recurrent',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Layers',
+            subheading: 'Recurrent',
+            namespace: 'layers',
+            useDocsFrom: 'RNN',
+            configParamIndices: [0]
+        })
+    ], LayerExports, "rnn", null);
+    __decorate([
+        tfjs_core_1.doc({
+            heading: 'Layers',
+            subheading: 'Recurrent',
+            namespace: 'layers',
+            useDocsFrom: 'RNN',
+            configParamIndices: [0]
+        })
+    ], LayerExports, "stackedRNNCells", null);
+    __decorate([
+        tfjs_core_1.doc({
+            heading: 'Layers',
             subheading: 'Wrapper',
@@ -502,3 +526,3 @@ namespace: 'layers',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Layers',
@@ -530,3 +554,3 @@ subheading: 'Wrapper',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Constraints',
@@ -539,3 +563,3 @@ namespace: 'constraints',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Constraints',
@@ -548,6 +572,6 @@ namespace: 'constraints',
     __decorate([
-        deeplearn_1.doc({ heading: 'Constraints', namespace: 'constraints', useDocsFrom: 'NonNeg' })
+        tfjs_core_1.doc({ heading: 'Constraints', namespace: 'constraints', useDocsFrom: 'NonNeg' })
     ], ConstraintExports, "nonNeg", null);
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Constraints',
@@ -602,3 +626,3 @@ namespace: 'constraints',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Initializers',
@@ -610,6 +634,6 @@ namespace: 'initializers',
     __decorate([
-        deeplearn_1.doc({ heading: 'Initializers', namespace: 'initializers', useDocsFrom: 'Ones' })
+        tfjs_core_1.doc({ heading: 'Initializers', namespace: 'initializers', useDocsFrom: 'Ones' })
     ], InitializerExports, "ones", null);
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Initializers',
@@ -622,3 +646,3 @@ namespace: 'initializers',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Initializers',
@@ -631,3 +655,3 @@ namespace: 'initializers',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Initializers',
@@ -640,3 +664,3 @@ namespace: 'initializers',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Initializers',
@@ -649,3 +673,3 @@ namespace: 'initializers',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Initializers',
@@ -658,3 +682,3 @@ namespace: 'initializers',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Initializers',
@@ -667,3 +691,3 @@ namespace: 'initializers',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Initializers',
@@ -676,3 +700,3 @@ namespace: 'initializers',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Initializers',
@@ -685,3 +709,3 @@ namespace: 'initializers',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Initializers',
@@ -694,3 +718,3 @@ namespace: 'initializers',
     __decorate([
-        deeplearn_1.doc({
+        tfjs_core_1.doc({
             heading: 'Initializers',
@@ -718,9 +742,9 @@ namespace: 'initializers',
     __decorate([
-        deeplearn_1.doc({ heading: 'Regularizers', namespace: 'regularizers', useDocsFrom: 'L1L2' })
+        tfjs_core_1.doc({ heading: 'Regularizers', namespace: 'regularizers', useDocsFrom: 'L1L2' })
     ], RegularizerExports, "l1l2", null);
     __decorate([
-        deeplearn_1.doc({ heading: 'Regularizers', namespace: 'regularizers', useDocsFrom: 'L1L2' })
+        tfjs_core_1.doc({ heading: 'Regularizers', namespace: 'regularizers', useDocsFrom: 'L1L2' })
     ], RegularizerExports, "l1", null);
     __decorate([
-        deeplearn_1.doc({ heading: 'Regularizers', namespace: 'regularizers', useDocsFrom: 'L1L2' })
+        tfjs_core_1.doc({ heading: 'Regularizers', namespace: 'regularizers', useDocsFrom: 'L1L2' })
     ], RegularizerExports, "l2", null);
@@ -727,0 +751,0 @@ return RegularizerExports;

dist/index.d.ts

@@ -1,7 +0,9 @@
-import * as dl from 'deeplearn';
+import * as dl from '@tensorflow/tfjs-core';
 import * as backend from './backend/deeplearnjs_backend';
 import { ConstraintExports, InitializerExports, LayerExports, ModelExports, RegularizerExports } from './exports';
 export { Callback, CallbackList, CustomCallback, CustomCallbackConfig, Logs } from './callbacks';
-export { Model } from './engine/training';
-export { ModelAndWeightsConfig, Sequential } from './models';
+export { Layer } from './engine/topology';
+export { Model, ModelCompileConfig, ModelEvaluateConfig, ModelFitConfig, ModelPredictConfig } from './engine/training';
+export { GRUCellLayerConfig, GRULayerConfig, LSTMCellLayerConfig, LSTMLayerConfig, RNN, RNNCell, RNNLayerConfig, SimpleRNNCellLayerConfig, SimpleRNNLayerConfig } from './layers/recurrent';
+export { ModelAndWeightsConfig, Sequential, SequentialConfig } from './models';
 export { SymbolicTensor } from './types';
@@ -8,0 +10,0 @@ export { version as version_layers } from './version';

dist/index.js

 "use strict";
 Object.defineProperty(exports, "__esModule", { value: true });
-var dl = require("deeplearn");
+var dl = require("@tensorflow/tfjs-core");
 exports.dl = dl;
@@ -12,4 +12,9 @@ var backend = require("./backend/deeplearnjs_backend");
 exports.CustomCallback = callbacks_1.CustomCallback;
+var topology_1 = require("./engine/topology");
+exports.Layer = topology_1.Layer;
 var training_1 = require("./engine/training");
 exports.Model = training_1.Model;
+var recurrent_1 = require("./layers/recurrent");
+exports.RNN = recurrent_1.RNN;
+exports.RNNCell = recurrent_1.RNNCell;
 var models_1 = require("./models");
@@ -16,0 +21,0 @@ exports.Sequential = models_1.Sequential;

dist/initializers.d.ts

@@ -1,2 +0,2 @@
-import { Tensor } from 'deeplearn';
+import { Tensor } from '@tensorflow/tfjs-core';
 import { DType, Shape } from './types';
@@ -3,0 +3,0 @@ import { ConfigDict, ConfigDictValue } from './types';

dist/initializers.js

@@ -19,3 +19,3 @@ "use strict";
 Object.defineProperty(exports, "__esModule", { value: true });
-var deeplearn_1 = require("deeplearn");
+var tfjs_core_1 = require("@tensorflow/tfjs-core");
 var _ = require("underscore");
@@ -57,3 +57,3 @@ var K = require("./backend/deeplearnjs_backend");
     Initializer = __decorate([
-        deeplearn_1.doc({ heading: 'Initializers', subheading: 'Classes', namespace: 'initializers' })
+        tfjs_core_1.doc({ heading: 'Initializers', subheading: 'Classes', namespace: 'initializers' })
     ], Initializer);
@@ -95,3 +95,3 @@ return Initializer;
     Constant.prototype.apply = function (shape, dtype) {
-        return K.scalarTimesArray(deeplearn_1.scalar(this.value), K.ones(shape, dtype));
+        return K.scalarTimesArray(tfjs_core_1.scalar(this.value), K.ones(shape, dtype));
     };
@@ -174,3 +174,3 @@ Constant.prototype.getConfig = function () {
         var _this = _super.call(this) || this;
-        _this.gain = config.gain != null ? deeplearn_1.scalar(config.gain) : K.getScalar(1.0);
+        _this.gain = config.gain != null ? tfjs_core_1.scalar(config.gain) : K.getScalar(1.0);
         return _this;
@@ -177,0 +177,0 @@ }

dist/layers/convolutional_depthwise.d.ts

@@ -1,2 +0,2 @@
-import { Tensor } from 'deeplearn';
+import { Tensor } from '@tensorflow/tfjs-core';
 import { Constraint, ConstraintIdentifier } from '../constraints';
@@ -3,0 +3,0 @@ import { Initializer, InitializerIdentifier } from '../initializers';

dist/layers/convolutional.d.ts

@@ -1,2 +0,2 @@
-import { Tensor } from 'deeplearn';
+import { Tensor } from '@tensorflow/tfjs-core';
 import { ActivationFn } from '../activations';
@@ -3,0 +3,0 @@ import { DataFormat, PaddingMode } from '../common';

dist/layers/core.d.ts

@@ -1,2 +0,2 @@
-import { Tensor } from 'deeplearn';
+import { Tensor } from '@tensorflow/tfjs-core';
 import { ActivationFn, ActivationIdentifier } from '../activations';
@@ -3,0 +3,0 @@ import { Constraint, ConstraintIdentifier } from '../constraints';

dist/layers/embeddings.d.ts

@@ -1,2 +0,2 @@
-import { Tensor } from 'deeplearn';
+import { Tensor } from '@tensorflow/tfjs-core';
 import { Constraint, ConstraintIdentifier } from '../constraints';
@@ -3,0 +3,0 @@ import { Layer, LayerConfig } from '../engine/topology';

dist/layers/merge.d.ts

@@ -1,2 +0,2 @@
-import { Tensor } from 'deeplearn';
+import { Tensor } from '@tensorflow/tfjs-core';
 import { Layer, LayerConfig } from '../engine/topology';
@@ -3,0 +3,0 @@ import { Shape } from '../types';

dist/layers/normalization.d.ts

@@ -1,2 +0,2 @@
-import { Tensor } from 'deeplearn';
+import { Tensor } from '@tensorflow/tfjs-core';
 import { Constraint, ConstraintIdentifier } from '../constraints';
@@ -3,0 +3,0 @@ import { Layer, LayerConfig } from '../engine/topology';

dist/layers/pooling.d.ts

@@ -1,2 +0,2 @@
-import { Tensor } from 'deeplearn';
+import { Tensor } from '@tensorflow/tfjs-core';
 import { DataFormat, PaddingMode } from '../common';
@@ -3,0 +3,0 @@ import { Layer, LayerConfig } from '../engine/topology';

dist/layers/recurrent.d.ts

@@ -1,2 +0,2 @@
-import { Tensor } from 'deeplearn';
+import { Tensor } from '@tensorflow/tfjs-core';
 import { ActivationFn, ActivationIdentifier } from '../activations';
@@ -262,1 +262,17 @@ import { Constraint, ConstraintIdentifier } from '../constraints';
 }
+export interface StackedRNNCellsConfig extends LayerConfig {
+    cells: RNNCell[];
+}
+export declare class StackedRNNCells extends RNNCell {
+    protected cells: RNNCell[];
+    constructor(config: StackedRNNCellsConfig);
+    readonly stateSize: number[];
+    call(inputs: Tensor | Tensor[], kwargs: any): Tensor | Tensor[];
+    build(inputShape: Shape | Shape[]): void;
+    getConfig(): ConfigDict;
+    static fromConfig<T>(cls: generic_utils.Constructor<T>, config: ConfigDict, customObjects?: ConfigDict): T;
+    readonly trainableWeights: LayerVariable[];
+    readonly nonTrainableWeights: LayerVariable[];
+    getWeights(): Tensor[];
+    setWeights(weights: Tensor[]): void;
+}

dist/layers/recurrent.js

@@ -19,3 +19,3 @@ "use strict";
 Object.defineProperty(exports, "__esModule", { value: true });
-var deeplearn_1 = require("deeplearn");
+var tfjs_core_1 = require("@tensorflow/tfjs-core");
 var _ = require("underscore");
@@ -33,2 +33,3 @@ var activations_1 = require("../activations");
 var math_utils = require("../utils/math_utils");
+var serialization_1 = require("./serialization");
 var RNN = (function (_super) {
@@ -38,2 +39,3 @@ __extends(RNN, _super);
         var _this = _super.call(this, config) || this;
+        var cell;
         if (config.cell == null) {
@@ -43,9 +45,12 @@ throw new errors_1.ValueError('cell property is missing for the constructor of RNN.');
         else if (Array.isArray(config.cell)) {
-            throw new errors_1.NotImplementedError('StackedRNNCells has not been implemented yet.');
+            cell = new StackedRNNCells({ cells: config.cell });
         }
-        if (config.cell.stateSize == null) {
+        else {
+            cell = config.cell;
+        }
+        if (cell.stateSize == null) {
             throw new errors_1.ValueError('The RNN cell should have an attribute `stateSize` (tuple of ' +
                 'integers, one integer per RNN state).');
         }
-        _this.cell = config.cell;
+        _this.cell = cell;
         _this.returnSequences =
@@ -380,3 +385,3 @@ config.returnSequences == null ? false : config.returnSequences;
     RNNCell = __decorate([
-        deeplearn_1.doc({ heading: 'Layers', subheading: 'Classes' })
+        tfjs_core_1.doc({ heading: 'Layers', subheading: 'Classes' })
     ], RNNCell);
@@ -1240,1 +1245,161 @@ return RNNCell;
 generic_utils.ClassNameMap.register('LSTM', LSTM);
+var StackedRNNCells = (function (_super) {
+    __extends(StackedRNNCells, _super);
+    function StackedRNNCells(config) {
+        var _this = _super.call(this, config) || this;
+        _this.cells = config.cells;
+        return _this;
+    }
+    Object.defineProperty(StackedRNNCells.prototype, "stateSize", {
+        get: function () {
+            var stateSize = [];
+            for (var _i = 0, _a = this.cells.slice().reverse(); _i < _a.length; _i++) {
+                var cell = _a[_i];
+                if (Array.isArray(cell.stateSize)) {
+                    stateSize.push.apply(stateSize, cell.stateSize);
+                }
+                else {
+                    stateSize.push(cell.stateSize);
+                }
+            }
+            return stateSize;
+        },
+        enumerable: true,
+        configurable: true
+    });
+    StackedRNNCells.prototype.call = function (inputs, kwargs) {
+        inputs = inputs;
+        var states = inputs.slice(1);
+        var nestedStates = [];
+        for (var _i = 0, _a = this.cells.slice().reverse(); _i < _a.length; _i++) {
+            var cell = _a[_i];
+            if (Array.isArray(cell.stateSize)) {
+                nestedStates.push(states.splice(0, cell.stateSize.length));
+            }
+            else {
+                nestedStates.push(states.splice(0, 1));
+            }
+        }
+        nestedStates.reverse();
+        var newNestedStates = [];
+        var callInputs;
+        for (var i = 0; i < this.cells.length; ++i) {
+            var cell = this.cells[i];
+            states = nestedStates[i];
+            if (i === 0) {
+                callInputs = [inputs[0]].concat(states);
+            }
+            else {
+                callInputs = [callInputs[0]].concat(states);
+            }
+            callInputs = cell.call(callInputs, kwargs);
+            newNestedStates.push(callInputs.slice(1));
+        }
+        states = [];
+        for (var _b = 0, _c = newNestedStates.slice().reverse(); _b < _c.length; _b++) {
+            var cellStates = _c[_b];
+            states.push.apply(states, cellStates);
+        }
+        return [callInputs[0]].concat(states);
+    };
+    StackedRNNCells.prototype.build = function (inputShape) {
+        if (generic_utils.isArrayOfShapes(inputShape)) {
+            inputShape = inputShape[0];
+        }
+        inputShape = inputShape;
+        var outputDim;
+        for (var _i = 0, _a = this.cells; _i < _a.length; _i++) {
+            var cell = _a[_i];
+            cell.build(inputShape);
+            if (Array.isArray(cell.stateSize)) {
+                outputDim = cell.stateSize[0];
+            }
+            else {
+                outputDim = cell.stateSize;
+            }
+            inputShape = [inputShape[0], outputDim];
+        }
+        this.built = true;
+    };
+    StackedRNNCells.prototype.getConfig = function () {
+        var cellConfigs = [];
+        for (var _i = 0, _a = this.cells; _i < _a.length; _i++) {
+            var cell = _a[_i];
+            cellConfigs.push({
+                'className': this.constructor.name,
+                'config': cell.getConfig(),
+            });
+        }
+        var config = { 'cells': cellConfigs };
+        var baseConfig = _super.prototype.getConfig.call(this);
+        Object.assign(config, baseConfig);
+        return config;
+    };
+    StackedRNNCells.fromConfig = function (cls, config, customObjects) {
+        if (customObjects === void 0) { customObjects = {}; }
+        var cells = [];
+        for (var _i = 0, _a = config['cells']; _i < _a.length; _i++) {
+            var cellConfig = _a[_i];
+            cells.push(serialization_1.deserialize(cellConfig, customObjects));
+        }
+        return new cls({ cells: cells });
+    };
+    Object.defineProperty(StackedRNNCells.prototype, "trainableWeights", {
+        get: function () {
+            if (!this.trainable) {
+                return [];
+            }
+            var weights = [];
+            for (var _i = 0, _a = this.cells; _i < _a.length; _i++) {
+                var cell = _a[_i];
+                weights.push.apply(weights, cell.trainableWeights);
+            }
+            return weights;
+        },
+        enumerable: true,
+        configurable: true
+    });
+    Object.defineProperty(StackedRNNCells.prototype, "nonTrainableWeights", {
+        get: function () {
+            var weights = [];
+            for (var _i = 0, _a = this.cells; _i < _a.length; _i++) {
+                var cell = _a[_i];
+                weights.push.apply(weights, cell.nonTrainableWeights);
+            }
+            if (!this.trainable) {
+                var trainableWeights = [];
+                for (var _b = 0, _c = this.cells; _b < _c.length; _b++) {
+                    var cell = _c[_b];
+                    trainableWeights.push.apply(trainableWeights, cell.trainableWeights);
+                }
+                return trainableWeights.concat(weights);
+            }
+            return weights;
+        },
+        enumerable: true,
+        configurable: true
+    });
+    StackedRNNCells.prototype.getWeights = function () {
+        var weights = [];
+        for (var _i = 0, _a = this.cells; _i < _a.length; _i++) {
+            var cell = _a[_i];
+            weights.push.apply(weights, cell.weights);
+        }
+        return K.batchGetValue(weights);
+    };
+    StackedRNNCells.prototype.setWeights = function (weights) {
+        var tuples = [];
+        for (var _i = 0, _a = this.cells; _i < _a.length; _i++) {
+            var cell = _a[_i];
+            var numParams = cell.weights.length;
+            var inputWeights = weights.splice(numParams);
+            for (var i = 0; i < cell.weights.length; ++i) {
+                tuples.push([cell.weights[i], inputWeights[i]]);
+            }
+        }
+        K.batchSetValue(tuples);
+    };
+    return StackedRNNCells;
+}(RNNCell));
+exports.StackedRNNCells = StackedRNNCells;
+generic_utils.ClassNameMap.register('StackedRNNCells', StackedRNNCells);

dist/layers/wrappers.d.ts

@@ -1,2 +0,2 @@
-import { Tensor } from 'deeplearn';
+import { Tensor } from '@tensorflow/tfjs-core';
 import { Layer, LayerConfig } from '../engine/topology';
@@ -3,0 +3,0 @@ import { Shape, TensorInterface } from '../types';

dist/losses.d.ts

@@ -1,2 +0,2 @@
-import { Tensor } from 'deeplearn';
+import { Tensor } from '@tensorflow/tfjs-core';
 import { LossOrMetricFn } from './types';
@@ -3,0 +3,0 @@ export declare function meanSquaredError(yTrue: Tensor, yPred: Tensor): Tensor;

dist/metrics.d.ts

@@ -1,2 +0,2 @@
-import { Tensor } from 'deeplearn';
+import { Tensor } from '@tensorflow/tfjs-core';
 import { cosineProximity, meanAbsoluteError, meanAbsolutePercentageError, meanSquaredError } from './losses';
@@ -3,0 +3,0 @@ import { LossOrMetricFn } from './types';

dist/metrics.js

 "use strict";
 Object.defineProperty(exports, "__esModule", { value: true });
-var deeplearn_1 = require("deeplearn");
+var tfjs_core_1 = require("@tensorflow/tfjs-core");
 var K = require("./backend/deeplearnjs_backend");
@@ -8,3 +8,3 @@ var errors_1 = require("./errors");
 function binaryAccuracy(yTrue, yPred) {
-    var threshold = K.scalarTimesArray(K.getScalar(0.5), deeplearn_1.onesLike(yPred));
+    var threshold = K.scalarTimesArray(K.getScalar(0.5), tfjs_core_1.onesLike(yPred));
     var yPredThresholded = K.cast(K.greater(yPred, threshold), yTrue.dtype);
@@ -11,0 +11,0 @@ return K.mean(K.equal(yTrue, yPredThresholded), -1);

dist/models.d.ts

@@ -1,5 +0,5 @@
-import { Scalar, Tensor, WeightsManifestConfig } from 'deeplearn';
+import { Scalar, Tensor, WeightsManifestConfig } from '@tensorflow/tfjs-core';
 import { History } from './callbacks';
 import { Layer } from './engine/topology';
-import { Model, ModelCompileConfig, ModelFitConfig, ModelLoggingVerbosity } from './engine/training';
+import { Model, ModelCompileConfig, ModelEvaluateConfig, ModelFitConfig, ModelPredictConfig } from './engine/training';
 import { Shape } from './types';
@@ -29,8 +29,12 @@ import { ConfigDict, JsonDict } from './types';
     updatable: boolean;
-    evaluate(x: Tensor | Tensor[], y: Tensor | Tensor[], batchSize?: number, verbose?: ModelLoggingVerbosity, sampleWeight?: Tensor, steps?: number): Scalar | Scalar[];
-    predict(x: Tensor | Tensor[], batchSize?: number, verbose?: boolean): Tensor | Tensor[];
+    evaluate(x: Tensor | Tensor[], y: Tensor | Tensor[], config?: ModelEvaluateConfig): Promise<Scalar | Scalar[]>;
+    predict(x: Tensor | Tensor[], config?: ModelPredictConfig): Promise<Tensor | Tensor[]>;
     predictOnBatch(x: Tensor): Tensor | Tensor[];
     compile(config: ModelCompileConfig): void;
-    fit(config: ModelFitConfig): Promise<History>;
+    fit(x: Tensor | Tensor[] | {
+        [inputName: string]: Tensor;
+    }, y: Tensor | Tensor[] | {
+        [inputName: string]: Tensor;
+    }, config?: ModelFitConfig): Promise<History>;
     static fromConfig<T>(cls: generic_utils.Constructor<T>, config: ConfigDict): T;
 }

dist/models.js

@@ -54,3 +54,3 @@ "use strict";
 Object.defineProperty(exports, "__esModule", { value: true });
-var deeplearn_1 = require("deeplearn");
+var tfjs_core_1 = require("@tensorflow/tfjs-core");
 var K = require("./backend/deeplearnjs_backend");
@@ -76,3 +76,3 @@ var topology_1 = require("./engine/topology");
                     if (!(modelAndWeightsConfig.weightsManifest != null)) return [3, 2];
-                    return [4, deeplearn_1.loadWeights(modelAndWeightsConfig.weightsManifest, modelAndWeightsConfig.pathPrefix, model.weights.map(function (weight) { return weight.name; }))];
+                    return [4, tfjs_core_1.loadWeights(modelAndWeightsConfig.weightsManifest, modelAndWeightsConfig.pathPrefix, model.weights.map(function (weight) { return weight.name; }))];
                 case 1:
@@ -259,16 +259,23 @@ weightValues = _a.sent();
     });
-    Sequential.prototype.evaluate = function (x, y, batchSize, verbose, sampleWeight, steps) {
-        if (batchSize === void 0) { batchSize = 32; }
-        if (!this.built) {
-            throw new errors_1.RuntimeError('The model needs to be compiled before being used.');
-        }
-        return this.model.evaluate(x, y, batchSize, verbose, sampleWeight, steps);
+    Sequential.prototype.evaluate = function (x, y, config) {
+        if (config === void 0) { config = {}; }
+        return __awaiter(this, void 0, void 0, function () {
+            return __generator(this, function (_a) {
+                if (!this.built) {
+                    throw new errors_1.RuntimeError('The model needs to be compiled before being used.');
+                }
+                return [2, this.model.evaluate(x, y, config)];
+            });
+        });
     };
-    Sequential.prototype.predict = function (x, batchSize, verbose) {
-        if (batchSize === void 0) { batchSize = 32; }
-        if (verbose === void 0) { verbose = false; }
-        if (this.model == null) {
-            this.build();
-        }
-        return this.model.predict(x, batchSize, verbose);
+    Sequential.prototype.predict = function (x, config) {
+        if (config === void 0) { config = {}; }
+        return __awaiter(this, void 0, void 0, function () {
+            return __generator(this, function (_a) {
+                if (this.model == null) {
+                    this.build();
+                }
+                return [2, this.model.predict(x, config)];
+            });
+        });
     };
@@ -290,3 +297,4 @@ Sequential.prototype.predictOnBatch = function (x) {
     };
-    Sequential.prototype.fit = function (config) {
+    Sequential.prototype.fit = function (x, y, config) {
+        if (config === void 0) { config = {}; }
         return __awaiter(this, void 0, void 0, function () {
@@ -298,3 +306,3 @@ return __generator(this, function (_a) {
                 }
-                return [2, this.model.fit(config)];
+                return [2, this.model.fit(x, y, config)];
             });
@@ -322,15 +330,15 @@ });
     __decorate([
-        deeplearn_1.doc({ heading: 'Models', subheading: 'Classes' })
+        tfjs_core_1.doc({ heading: 'Models', subheading: 'Classes' })
     ], Sequential.prototype, "add", null);
     __decorate([
-        deeplearn_1.doc({ heading: 'Models', subheading: 'Classes' })
+        tfjs_core_1.doc({ heading: 'Models', subheading: 'Classes', configParamIndices: [2] })
     ], Sequential.prototype, "evaluate", null);
     __decorate([
-        deeplearn_1.doc({ heading: 'Models', subheading: 'Classes' })
+        tfjs_core_1.doc({ heading: 'Models', subheading: 'Classes', configParamIndices: [1] })
     ], Sequential.prototype, "predict", null);
     __decorate([
-        deeplearn_1.doc({ heading: 'Models', subheading: 'Classes' })
+        tfjs_core_1.doc({ heading: 'Models', subheading: 'Classes', configParamIndices: [2] })
     ], Sequential.prototype, "fit", null);
     Sequential = Sequential_1 = __decorate([
-        deeplearn_1.doc({ heading: 'Models', subheading: 'Classes' })
+        tfjs_core_1.doc({ heading: 'Models', subheading: 'Classes' })
     ], Sequential);
@@ -337,0 +345,0 @@ return Sequential;

dist/optimizers.d.ts

@@ -1,2 +0,2 @@
-import { AdagradOptimizer, AdamOptimizer, Optimizer as CoreOptimizer, RMSPropOptimizer, Scalar, SGDOptimizer } from 'deeplearn';
+import { AdagradOptimizer, AdamOptimizer, Optimizer as CoreOptimizer, RMSPropOptimizer, Scalar, SGDOptimizer } from '@tensorflow/tfjs-core';
 import { ConfigDict, LayerVariable } from './types';
@@ -3,0 +3,0 @@ import { Constructor } from './utils/generic_utils';

dist/optimizers.js

@@ -13,3 +13,3 @@ "use strict";
 Object.defineProperty(exports, "__esModule", { value: true });
-var deeplearn_1 = require("deeplearn");
+var tfjs_core_1 = require("@tensorflow/tfjs-core");
 var K = require("./backend/deeplearnjs_backend");
@@ -20,3 +20,3 @@ var errors_1 = require("./errors");
     function LayersOptimizer(config) {
-        if (config instanceof deeplearn_1.Optimizer) {
+        if (config instanceof tfjs_core_1.Optimizer) {
             this.createdFromCoreOptimizer = true;
@@ -84,6 +84,6 @@ this.constructFromCoreOptimizer(config);
         }
-        this.optimizer = deeplearn_1.train.sgd(this.lr);
+        this.optimizer = tfjs_core_1.train.sgd(this.lr);
     };
     SGD.prototype.constructFromCoreOptimizer = function (optimizer) {
-        if (!(optimizer instanceof deeplearn_1.SGDOptimizer)) {
+        if (!(optimizer instanceof tfjs_core_1.SGDOptimizer)) {
             throw new errors_1.ValueError('Cannot construct SGD from a non-SGD core optimizer');
@@ -126,6 +126,6 @@ }
         }
-        this.optimizer = deeplearn_1.train.adam(this.lr, this.beta1, this.beta2, this.epsilon);
+        this.optimizer = tfjs_core_1.train.adam(this.lr, this.beta1, this.beta2, this.epsilon);
     };
     Adam.prototype.constructFromCoreOptimizer = function (optimizer) {
-        if (!(optimizer instanceof deeplearn_1.AdamOptimizer)) {
+        if (!(optimizer instanceof tfjs_core_1.AdamOptimizer)) {
             throw new errors_1.ValueError('Cannot construct Adam from a non-Adam core optimizer');
@@ -164,6 +164,6 @@ }
         }
-        this.optimizer = deeplearn_1.train.rmsprop(this.lr, this.rho, null, this.epsilon);
+        this.optimizer = tfjs_core_1.train.rmsprop(this.lr, this.rho, null, this.epsilon);
     };
     RMSProp.prototype.constructFromCoreOptimizer = function (optimizer) {
-        if (!(optimizer instanceof deeplearn_1.RMSPropOptimizer)) {
+        if (!(optimizer instanceof tfjs_core_1.RMSPropOptimizer)) {
             throw new errors_1.ValueError('Cannot construct RMSProp from a non-RMSProp core optimizer');
@@ -200,6 +200,6 @@ }
         }
-        this.optimizer = deeplearn_1.train.adagrad(this.lr);
+        this.optimizer = tfjs_core_1.train.adagrad(this.lr);
     };
     Adagrad.prototype.constructFromCoreOptimizer = function (optimizer) {
-        if (!(optimizer instanceof deeplearn_1.AdagradOptimizer)) {
+        if (!(optimizer instanceof tfjs_core_1.AdagradOptimizer)) {
             throw new errors_1.ValueError('Cannot construct Adagrad from a non-Adagrad core optimizer');
@@ -206,0 +206,0 @@ }

dist/regularizers.d.ts

@@ -1,2 +0,2 @@
-import { Tensor } from 'deeplearn';
+import { Tensor } from '@tensorflow/tfjs-core';
 import { LayerVariable } from './types';
@@ -3,0 +3,0 @@ import { ConfigDict, ConfigDictValue } from './types';

dist/regularizers.js

@@ -19,3 +19,3 @@ "use strict";
 Object.defineProperty(exports, "__esModule", { value: true });
-var deeplearn_1 = require("deeplearn");
+var tfjs_core_1 = require("@tensorflow/tfjs-core");
 var K = require("./backend/deeplearnjs_backend");
@@ -43,3 +43,3 @@ var generic_utils_1 = require("./utils/generic_utils");
     L1L2.prototype.apply = function (x) {
-        var regularization = deeplearn_1.zeros([1]);
+        var regularization = tfjs_core_1.zeros([1]);
         if (this.hasL1) {
@@ -60,3 +60,3 @@ regularization = K.add(regularization, K.sum(K.scalarTimesArray(this.l1, K.abs(x.read()))));
     L1L2 = L1L2_1 = __decorate([
-        deeplearn_1.doc({ heading: 'Regularizers', namespace: 'regularizers' })
+        tfjs_core_1.doc({ heading: 'Regularizers', namespace: 'regularizers' })
     ], L1L2);
@@ -63,0 +63,0 @@ return L1L2;

dist/types.d.ts

@@ -1,3 +0,3 @@
-import * as dl from 'deeplearn';
-import { Tensor } from 'deeplearn';
+import * as tfc from '@tensorflow/tfjs-core';
+import { Tensor } from '@tensorflow/tfjs-core';
 import { Constraint } from './constraints';
@@ -43,3 +43,3 @@ import { Layer } from './engine/topology';
     readonly trainable: boolean;
-    protected readonly val: dl.Variable;
+    protected readonly val: tfc.Variable;
     readonly constraint: Constraint;
@@ -46,0 +46,0 @@ constructor(val: Tensor | ConcreteTensor, dtype?: DType, name?: string, trainable?: boolean, constraint?: Constraint);

dist/types.js

@@ -9,3 +9,3 @@ "use strict";
 Object.defineProperty(exports, "__esModule", { value: true });
-var deeplearn_1 = require("deeplearn");
+var tfjs_core_1 = require("@tensorflow/tfjs-core");
 var common_1 = require("./common");
@@ -33,3 +33,3 @@ var DType;
     SymbolicTensor = __decorate([
-        deeplearn_1.doc({ heading: 'Models', 'subheading': 'Classes' })
+        tfjs_core_1.doc({ heading: 'Models', 'subheading': 'Classes' })
     ], SymbolicTensor);
@@ -79,3 +79,3 @@ return SymbolicTensor;
         this.val =
-            deeplearn_1.variable(getValueTensor(val), this.trainable, this.name, this.dtype);
+            tfjs_core_1.variable(getValueTensor(val), this.trainable, this.name, this.dtype);
     }
@@ -82,0 +82,0 @@ LayerVariable.prototype.read = function () {

dist/utils/generic_utils.d.ts

@@ -1,2 +0,2 @@
-import { Tensor } from 'deeplearn';
+import { Tensor } from '@tensorflow/tfjs-core';
 import { ConfigDict, ConfigDictValue, DType, Shape } from '../types';
@@ -3,0 +3,0 @@ export declare function pyListRepeat(value: any, numValues: number): any[];

dist/utils/math_utils.js

 "use strict";
 Object.defineProperty(exports, "__esModule", { value: true });
-var dl = require("deeplearn");
-var deeplearn_1 = require("deeplearn");
+var tfc = require("@tensorflow/tfjs-core");
+var tfjs_core_1 = require("@tensorflow/tfjs-core");
 function isInteger(x) {
@@ -25,14 +25,14 @@ return x === parseInt(x.toString(), 10);
     array = Array.isArray(array) ? new Float32Array(array) : array;
-    return deeplearn_1.tensor1d(array);
+    return tfjs_core_1.tensor1d(array);
 }
 function min(array) {
-    return dl.min(toArray1D(array)).dataSync()[0];
+    return tfc.min(toArray1D(array)).dataSync()[0];
 }
 exports.min = min;
 function max(array) {
-    return dl.max(toArray1D(array)).dataSync()[0];
+    return tfc.max(toArray1D(array)).dataSync()[0];
 }
 exports.max = max;
 function sum(array) {
-    return dl.sum(toArray1D(array)).dataSync()[0];
+    return tfc.sum(toArray1D(array)).dataSync()[0];
 }
@@ -45,6 +45,6 @@ exports.sum = sum;
 function variance(array) {
-    var demeaned = dl.sub(toArray1D(array), deeplearn_1.scalar(mean(array)));
-    var sumSquare = dl.sum(dl.mulStrict(demeaned, demeaned)).dataSync()[0];
+    var demeaned = tfc.sub(toArray1D(array), tfjs_core_1.scalar(mean(array)));
+    var sumSquare = tfc.sum(tfc.mulStrict(demeaned, demeaned)).dataSync()[0];
     return sumSquare / array.length;
 }
 exports.variance = variance;

dist/utils/test_utils.d.ts

@@ -1,2 +0,2 @@
-import { Tensor } from 'deeplearn';
+import { Tensor } from '@tensorflow/tfjs-core';
 export declare function expectTensorsClose(actual: Tensor | number[], expected: Tensor | number[], epsilon?: number): void;
@@ -3,0 +3,0 @@ export declare function expectTensorsValuesInRange(actual: Tensor, low: number, high: number): void;

dist/utils/test_utils.js

 "use strict";
 Object.defineProperty(exports, "__esModule", { value: true });
-var deeplearn_1 = require("deeplearn");
+var tfjs_core_1 = require("@tensorflow/tfjs-core");
 var deeplearnjs_backend_1 = require("../backend/deeplearnjs_backend");
@@ -18,3 +18,3 @@ var errors_1 = require("../errors");
     }
-    deeplearn_1.test_util.expectArraysClose(actual, expected, epsilon);
+    tfjs_core_1.test_util.expectArraysClose(actual, expected, epsilon);
 }
@@ -26,3 +26,3 @@ exports.expectTensorsClose = expectTensorsClose;
     }
-    deeplearn_1.test_util.expectValuesInRange(actual.dataSync(), low, high);
+    tfjs_core_1.test_util.expectValuesInRange(actual.dataSync(), low, high);
 }
@@ -36,3 +36,3 @@ exports.expectTensorsValuesInRange = expectTensorsValuesInRange;
 function describeMathCPU(testName, tests) {
-    deeplearn_1.test_util.describeWithFlags(testName, deeplearn_1.test_util.CPU_ENVS, function () {
+    tfjs_core_1.test_util.describeWithFlags(testName, tfjs_core_1.test_util.CPU_ENVS, function () {
         beforeEach(function () {
@@ -46,3 +46,3 @@ deeplearnjs_backend_1.setBackend('cpu');
 function describeMathGPU(testName, tests) {
-    deeplearn_1.test_util.describeWithFlags(testName, webgl2Features, function () {
+    tfjs_core_1.test_util.describeWithFlags(testName, webgl2Features, function () {
         beforeEach(function () {
@@ -49,0 +49,0 @@ deeplearnjs_backend_1.setBackend('webgl');

dist/version.d.ts

@@ -1,2 +0,2 @@
-declare const version = "0.0.3";
+declare const version = "0.0.4";
 export { version };

dist/version.js

 "use strict";
 Object.defineProperty(exports, "__esModule", { value: true });
-var version = '0.0.3';
+var version = '0.0.4';
 exports.version = version;

package.json

 {
   "name": "@tensorflow/tfjs-layers",
-  "version": "0.0.3",
+  "version": "0.0.4",
   "description": "TensorFlow layers API in JavaScript",
@@ -9,2 +9,3 @@ "private": false,
   "devDependencies": {
+    "@tensorflow/tfjs-core": "0.0.1",
     "@types/jasmine": "~2.5.53",
@@ -15,3 +16,2 @@ "@types/underscore": "^1.8.7",
     "cross-spawn": "~5.1.0",
-    "deeplearn": "next",
     "deeplearn-src": "PAIR-code/deeplearnjs",
@@ -48,4 +48,4 @@ "http-server": "~0.10.0",
   "peerDependencies": {
-    "deeplearn": "next"
+    "@tensorflow/tfjs-core": "0.0.1"
   }
 }

README.md

@@ -1,9 +0,94 @@
-# TensorFlow.js Layers: easy open source machine learning in the browser
+# TensorFlow.js Layers: High-Level Machine Learning Model API
 
-TensorFlow.js Layers lets users build, train and execute machine learning models
-in the browser, using the established and familiar [Keras](https://keras.io/)
-API, without needing to install any software.
+A part of the TensorFlow.js ecosystem, TensorFlow.js Layers is a high-level
+API built on [TensorFlow.js Core](https://github.com/tensorflow/tfjs-core),
+enabling users to build, train and execute deep learning models in the browser.
+TensorFlow.js Layers is modeled after
+[Keras](https://keras.io/) and
+[tf.keras](https://www.tensorflow.org/api_docs/python/tf/keras) and can
+load models saved from those libraries.
 
-TFJS Layers provides higher-level building blocks for neural networks, executing
-using the hardware-accelerated
-[deeplearn.js](https://github.com/PAIR-code/deeplearnjs) library.
+## Importing
+
+There are three ways to import TensorFlow.js Layers
+
+1. You can access TensorFlow.js Layers through the union package
+   between the TensorFlow.js Core and Layers:
+   [@tensorflow/tfjs](https://www.npmjs.com/package/@tensorflow/tfjs)
+2. You can get [TensorFlow.js] Layers as a module:
+   [@tensorflow/tfjs-layers](https://www.npmjs.com/package/@tensorflow/tfjs-layers).
+   Note that `tfjs-layers` has peer dependency on core, so if you import
+   `@tensorflow/tfjs-layers`, you also need to import
+   `@tensorflow/tfjs-core`.
+3. As a standalone through [unpkg](https://unpkg.com/).
+
+Option 1 is the most convenient, but leads to a larger bundle size (we will be
+adding more packages to it in the future). Use option 2 if you care about bundle
+size.
+
+## Getting started
+
+### Building, training and executing a model
+
+The following example shows how to build a toy model with only one `dense` layer
+to perform linear regression.
+
+```js
+import * as tf from '@tensorflow/tfjs';
+
+// A sequential model is a container which you can add layers to.
+const model = tf.sequential();
+
+// Add a dense layer with 1 output unit.
+model.add(tf.layers.dense({units: 1, inputShape: [1]}));
+
+// Specify the loss type and optimizer for training.
+model.compile({loss: 'meanSquaredError', optimizer: 'SGD'});
+
+// Generate some synthetic data for training.
+const xs = tf.tensor2d([[1], [2], [3], [4]], [4, 1]);
+const ys = tf.tensor2d([[1], [3], [5], [7]], [4, 1]);
+
+// Train the model.
+await model.fit(xs, ys, {epochs: 50});
+
+// Ater the training, perform inference.
+const output = await model.predict(tf.tensor2d([[5]], [1, 1]));
+output.print();
+```
+
+### Loading a pretrained Keras model
+
+You can also load a model previously trained and saved from elsewhere (e.g.,
+from Python Keras) and use it for inference or transfer learning in the browser.
+
+For example, in Python, save your Keras model using
+[tensorflowjs](https://pypi.org/project/tensorflowjs/),
+which can be installed using `pip install tensorflowjs`.
+
+
+```python
+import tensorflowjs as tfjs
+
+# ... Create and train your Keras model.
+
+# Save your Keras model in TensorFlow.js format.
+tfjs.converter.save_keras_model(model, '/path/to/tfjs_artifacts/')
+
+# Then use your favorite web server to serve the directory at a URL, say
+#   http://foo.bar/tfjs_artifacts/model.json
+```
+
+To load the model with TensorFlow.js Layers:
+
+```js
+import * as tf from '@tensorflow/tfjs';
+
+const model = await tf.loadModel('http://foo.bar/tfjs_artifacts/model.json');
+// Now the model is ready for inference, evaluation or re-training.
+```
+
+## For more information
+
+- [TensorFlow.js API documentation](https://js.tensorflow.org/api/index.html)
+- [TensorFlow.js Tutorials](https://js.tensorflow.org/tutorials/)

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

New author

Supply chain risk

A new npm collaborator published a version of the package for the first time. New collaborators are usually benign additions to a project, but do indicate a change to the security surface area of a package.

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

2268525

increased by1.31%

Lines of code

38745

increased by1.45%

Number of lines in readme file

95

increased by850%

Worsened metrics

Number of medium supply chain risk alerts

19

increased by5.56%

No dependency changes