@tensorflow/tfjs-node/dist/kernels/FusedBatchNorm.js

"use strict";
/**
 * @license
 * Copyright 2020 Google LLC. All Rights Reserved.
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at
 *
 * http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 * =============================================================================
 */
Object.defineProperty(exports, "__esModule", { value: true });
var tfjs_1 = require("@tensorflow/tfjs");
var nodejs_kernel_backend_1 = require("../nodejs_kernel_backend");
exports.fusedBatchNormConfig = {
    kernelName: tfjs_1.FusedBatchNorm,
    backendName: 'tensorflow',
    kernelFunc: function (args) {
        var _a = args.inputs, x = _a.x, mean = _a.mean, variance = _a.variance;
        var _b = args.inputs, scale = _b.scale, offset = _b.offset;
        var backend = args.backend;
        var varianceEpsilon = args.attrs.varianceEpsilon;
        return tfjs_1.tidy(function () {
            if (mean.rank > 1) {
                // Fused batch norm doesn't work with high-dim mean/var/scale/offset.
                var inv = tfjs_1.rsqrt(tfjs_1.add(variance, tfjs_1.scalar(varianceEpsilon)));
                if (scale != null) {
                    inv = tfjs_1.mul(inv, scale);
                }
                var xNorm = tfjs_1.mul(tfjs_1.sub(x, mean), inv);
                return offset != null ? tfjs_1.add(xNorm, offset) : xNorm;
            }
            var dataFormat = 'NHWC';
            var depth = x.shape[3];
            var opAttrs = [
                nodejs_kernel_backend_1.createTensorsTypeOpAttr('T', x.dtype),
                {
                    name: 'epsilon',
                    type: backend.binding.TF_ATTR_FLOAT,
                    value: varianceEpsilon
                },
                {
                    name: 'data_format',
                    type: backend.binding.TF_ATTR_STRING,
                    value: dataFormat
                },
                { name: 'is_training', type: backend.binding.TF_ATTR_BOOL, value: false },
            ];
            var numOutputs = 5;
            if (scale == null) {
                scale = tfjs_1.fill([depth], 1);
            }
            if (offset == null) {
                offset = tfjs_1.fill([depth], 0);
            }
            return backend.executeMultipleOutputs(tfjs_1.FusedBatchNorm, opAttrs, [x, scale, offset, mean, variance], numOutputs)[0];
        });
    }
};