1 | "use strict";
|
2 |
|
3 |
|
4 |
|
5 |
|
6 |
|
7 |
|
8 |
|
9 |
|
10 |
|
11 |
|
12 |
|
13 |
|
14 |
|
15 |
|
16 |
|
17 |
|
18 | Object.defineProperty(exports, "__esModule", { value: true });
|
19 | var tfjs_1 = require("@tensorflow/tfjs");
|
20 | var nodejs_kernel_backend_1 = require("../nodejs_kernel_backend");
|
21 | exports.maxPool3DConfig = {
|
22 | kernelName: tfjs_1.MaxPool3D,
|
23 | backendName: 'tensorflow',
|
24 | kernelFunc: function (args) {
|
25 | var x = args.inputs.x;
|
26 | var backend = args.backend;
|
27 | var _a = args.attrs, filterSize = _a.filterSize, strides = _a.strides, pad = _a.pad, dataFormat = _a.dataFormat, dimRoundingMode = _a.dimRoundingMode;
|
28 | var convInfo = tfjs_1.backend_util.computePool3DInfo(x.shape, filterSize, strides, 1 , pad, dimRoundingMode, dataFormat);
|
29 | if (convInfo.padInfo.type !== 'VALID' && convInfo.padInfo.type !== 'SAME') {
|
30 | throw new Error("TF Backend supports only 'valid' and 'same' padding " +
|
31 | ("while padding was " + convInfo.padInfo.type));
|
32 | }
|
33 | var ksize = [
|
34 | 1, convInfo.filterDepth, convInfo.filterHeight, convInfo.filterWidth, 1
|
35 | ];
|
36 | var $strides = [
|
37 | 1, convInfo.strideDepth, convInfo.strideHeight, convInfo.strideWidth, 1
|
38 | ];
|
39 | var padding = convInfo.padInfo.type;
|
40 | var $dataFormat = convInfo.dataFormat === 'channelsLast' ? 'NDHWC' : 'NCDHW';
|
41 | var opAttrs = [
|
42 | nodejs_kernel_backend_1.createTensorsTypeOpAttr('T', x.dtype),
|
43 | { name: 'ksize', type: backend.binding.TF_ATTR_INT, value: ksize },
|
44 | { name: 'strides', type: backend.binding.TF_ATTR_INT, value: $strides },
|
45 | { name: 'padding', type: backend.binding.TF_ATTR_STRING, value: padding }, {
|
46 | name: 'data_format',
|
47 | type: backend.binding.TF_ATTR_STRING,
|
48 | value: $dataFormat
|
49 | }
|
50 | ];
|
51 | return backend.executeSingleOutput(tfjs_1.MaxPool3D, opAttrs, [x]);
|
52 | }
|
53 | };
|