@tensorflow/tfjs-core
Version:
Hardware-accelerated JavaScript library for machine intelligence
387 lines • 19.1 kB
JavaScript
/**
* @license
* Copyright 2017 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.
* =============================================================================
*/
import * as tf from '../index';
import { ALL_ENVS, describeWithFlags } from '../jasmine_util';
import { expectArraysClose } from '../test_util';
describeWithFlags('stridedSlice', ALL_ENVS, () => {
it('with ellipsisMask=1', async () => {
const t = tf.tensor2d([
[1, 2, 3, 4, 5],
[2, 3, 4, 5, 6],
[3, 4, 5, 6, 7],
[4, 5, 6, 7, 8],
[5, 6, 7, 8, 9],
[6, 7, 8, 9, 10],
[7, 8, 9, 10, 11],
[8, 8, 9, 10, 11],
[9, 8, 9, 10, 11],
[10, 8, 9, 10, 11],
]);
const begin = [0, 4];
const end = [0, 5];
const strides = [1, 1];
const beginMask = 0;
const endMask = 0;
const ellipsisMask = 1;
const output = t.stridedSlice(begin, end, strides, beginMask, endMask, ellipsisMask);
expect(output.shape).toEqual([10, 1]);
expectArraysClose(await output.data(), [5, 6, 7, 8, 9, 10, 11, 11, 11, 11]);
});
it('with ellipsisMask=1, begin / end masks and start / end normalization', async () => {
const t = tf.randomNormal([1, 6, 2006, 4]);
const output = tf.stridedSlice(t, [0, 0, 0], [0, 2004, 0], [1, 1, 1], 6, 4, 1);
expect(output.shape).toEqual([1, 6, 2004, 4]);
});
it('with ellipsisMask=1 and start / end normalization', async () => {
const t = tf.tensor3d([
[[1, 1, 1], [2, 2, 2]], [[3, 3, 3], [4, 4, 4]], [[5, 5, 5], [6, 6, 6]]
]);
const begin = [1, 0];
const end = [2, 1];
const strides = [1, 1];
const beginMask = 0;
const endMask = 0;
const ellipsisMask = 1;
const output = tf.stridedSlice(t, begin, end, strides, beginMask, endMask, ellipsisMask);
expect(output.shape).toEqual([3, 2, 1]);
expectArraysClose(await output.data(), [1, 2, 3, 4, 5, 6]);
});
it('with ellipsisMask=2', async () => {
const t = tf.tensor3d([
[[1, 1, 1], [2, 2, 2]], [[3, 3, 3], [4, 4, 4]], [[5, 5, 5], [6, 6, 6]]
]);
const begin = [1, 0, 0];
const end = [2, 1, 3];
const strides = [1, 1, 1];
const beginMask = 0;
const endMask = 0;
const ellipsisMask = 2;
const output = tf.stridedSlice(t, begin, end, strides, beginMask, endMask, ellipsisMask);
expect(output.shape).toEqual([1, 2, 3]);
expectArraysClose(await output.data(), [3, 3, 3, 4, 4, 4]);
});
it('with ellipsisMask=2 and start / end normalization', async () => {
const t = tf.tensor4d([
[[[1, 1], [1, 1], [1, 1]], [[2, 2], [2, 2], [2, 2]]],
[[[3, 3], [3, 3], [3, 3]], [[4, 4], [4, 4], [4, 4]]],
[[[5, 5], [5, 5], [5, 5]], [[6, 6], [6, 6], [6, 6]]]
]);
const begin = [1, 0, 0];
const end = [2, 1, 1];
const strides = [1, 1, 1];
const beginMask = 0;
const endMask = 0;
const ellipsisMask = 2;
const output = tf.stridedSlice(t, begin, end, strides, beginMask, endMask, ellipsisMask);
expect(output.shape).toEqual([1, 2, 3, 1]);
expectArraysClose(await output.data(), [3, 3, 3, 4, 4, 4]);
});
it('stridedSlice should fail if ellipsis mask is set and newAxisMask or ' +
'shrinkAxisMask are also set', async () => {
const tensor = tf.tensor1d([0, 1, 2, 3]);
expect(() => tf.stridedSlice(tensor, [0], [3], [2], 0, 0, 1, 1))
.toThrow();
expect(() => tf.stridedSlice(tensor, [0], [3], [2], 0, 0, 1, 0, 1))
.toThrow();
});
it('stridedSlice with first axis being new', async () => {
// Python slice code: t[tf.newaxis,0:3]
const t = tf.tensor1d([0, 1, 2, 3]);
const begin = [0, 0];
const end = [1, 3];
const strides = [1, 2];
const beginMask = 0;
const endMask = 0;
const ellipsisMask = 0;
const newAxisMask = 1;
const output = tf.stridedSlice(t, begin, end, strides, beginMask, endMask, ellipsisMask, newAxisMask);
expect(output.shape).toEqual([1, 2]);
expectArraysClose(await output.data(), [0, 2]);
});
it('strided slice with several new axes', async () => {
// Python slice code: t[1:2,tf.newaxis,0:3,tf.newaxis,2:5]
const t = tf.zeros([2, 3, 4, 5]);
const begin = [1, 0, 0, 0, 2];
const end = [2, 1, 3, 1, 5];
const strides = null;
const beginMask = 0;
const endMask = 0;
const ellipsisMask = 0;
const newAxisMask = 0b1010;
const output = tf.stridedSlice(t, begin, end, strides, beginMask, endMask, ellipsisMask, newAxisMask);
expect(output.shape).toEqual([1, 1, 3, 1, 2, 5]);
expectArraysClose(await output.data(), new Array(30).fill(0));
});
it('strided slice with new axes and shrink axes', () => {
// Python slice code: t[1:2,tf.newaxis,1,tf.newaxis,2,2:5]
const t = tf.zeros([2, 3, 4, 5]);
const begin = [1, 0, 1, 0, 2, 2];
const end = [2, 1, 2, 1, 3, 5];
const strides = null;
const beginMask = 0;
const endMask = 0;
const ellipsisMask = 0;
const newAxisMask = 0b1010;
const shrinkAxisMask = 0b10100;
const output = tf.stridedSlice(t, begin, end, strides, beginMask, endMask, ellipsisMask, newAxisMask, shrinkAxisMask);
expect(output.shape).toEqual([1, 1, 1, 3]);
});
it('stridedSlice should support 1d tensor', async () => {
const tensor = tf.tensor1d([0, 1, 2, 3]);
const output = tf.stridedSlice(tensor, [0], [3], [2]);
expect(output.shape).toEqual([2]);
expectArraysClose(await output.data(), [0, 2]);
});
it('stridedSlice should support 1d tensor', async () => {
const tensor = tf.tensor1d([0, 1, 2, 3]);
const output = tf.stridedSlice(tensor, [0], [3], [2]);
expect(output.shape).toEqual([2]);
expectArraysClose(await output.data(), [0, 2]);
});
it('stridedSlice with 1d tensor should be used by tensor directly', async () => {
const t = tf.tensor1d([0, 1, 2, 3]);
const output = t.stridedSlice([0], [3], [2]);
expect(output.shape).toEqual([2]);
expectArraysClose(await output.data(), [0, 2]);
});
it('stridedSlice should support 1d tensor empty result', async () => {
const tensor = tf.tensor1d([0, 1, 2, 3]);
const output = tf.stridedSlice(tensor, [10], [3], [2]);
expect(output.shape).toEqual([0]);
expectArraysClose(await output.data(), []);
});
it('stridedSlice should support 1d tensor negative begin', async () => {
const tensor = tf.tensor1d([0, 1, 2, 3]);
const output = tf.stridedSlice(tensor, [-3], [3], [1]);
expect(output.shape).toEqual([2]);
expectArraysClose(await output.data(), [1, 2]);
});
it('stridedSlice should support 1d tensor out of range begin', async () => {
const tensor = tf.tensor1d([0, 1, 2, 3]);
const output = tf.stridedSlice(tensor, [-5], [3], [1]);
expect(output.shape).toEqual([3]);
expectArraysClose(await output.data(), [0, 1, 2]);
});
it('stridedSlice should support 1d tensor negative end', async () => {
const tensor = tf.tensor1d([0, 1, 2, 3]);
const output = tf.stridedSlice(tensor, [1], [-2], [1]);
expect(output.shape).toEqual([1]);
expectArraysClose(await output.data(), [1]);
});
it('stridedSlice should support 1d tensor out of range end', async () => {
const tensor = tf.tensor1d([0, 1, 2, 3]);
const output = tf.stridedSlice(tensor, [-3], [5], [1]);
expect(output.shape).toEqual([3]);
expectArraysClose(await output.data(), [1, 2, 3]);
});
it('stridedSlice should support 1d tensor begin mask', async () => {
const tensor = tf.tensor1d([0, 1, 2, 3]);
const output = tf.stridedSlice(tensor, [1], [3], [1], 1);
expect(output.shape).toEqual([3]);
expectArraysClose(await output.data(), [0, 1, 2]);
});
it('stridedSlice should support 1d tensor nagtive begin and stride', async () => {
const tensor = tf.tensor1d([0, 1, 2, 3]);
const output = tf.stridedSlice(tensor, [-2], [-3], [-1]);
expect(output.shape).toEqual([1]);
expectArraysClose(await output.data(), [2]);
});
it('stridedSlice should support 1d tensor' +
' out of range begin and negative stride', async () => {
const tensor = tf.tensor1d([0, 1, 2, 3]);
const output = tf.stridedSlice(tensor, [5], [-2], [-1]);
expect(output.shape).toEqual([1]);
expectArraysClose(await output.data(), [3]);
});
it('stridedSlice should support 1d tensor nagtive end and stride', async () => {
const tensor = tf.tensor1d([0, 1, 2, 3]);
const output = tf.stridedSlice(tensor, [2], [-4], [-1]);
expect(output.shape).toEqual([2]);
expectArraysClose(await output.data(), [2, 1]);
});
it('stridedSlice should support 1d tensor' +
' out of range end and negative stride', async () => {
const tensor = tf.tensor1d([0, 1, 2, 3]);
const output = tf.stridedSlice(tensor, [-3], [-5], [-1]);
expect(output.shape).toEqual([2]);
expectArraysClose(await output.data(), [1, 0]);
});
it('stridedSlice should support 1d tensor end mask', async () => {
const tensor = tf.tensor1d([0, 1, 2, 3]);
const output = tf.stridedSlice(tensor, [1], [3], [1], 0, 1);
expect(output.shape).toEqual([3]);
expectArraysClose(await output.data(), [1, 2, 3]);
});
it('stridedSlice should support 1d tensor shrink axis mask', async () => {
const tensor = tf.tensor1d([0, 1, 2, 3]);
const output = tf.stridedSlice(tensor, [1], [3], [1], 0, 0, 0, 0, 1);
expect(output.shape).toEqual([]);
expectArraysClose(await output.data(), [1]);
});
it('stridedSlice should support 1d tensor negative stride', async () => {
const tensor = tf.tensor1d([0, 1, 2, 3]);
const output = tf.stridedSlice(tensor, [-1], [-4], [-1]);
expect(output.shape).toEqual([3]);
expectArraysClose(await output.data(), [3, 2, 1]);
});
it('stridedSlice should support 1d tensor even length stride', async () => {
const tensor = tf.tensor1d([0, 1, 2, 3]);
const output = tf.stridedSlice(tensor, [0], [2], [2]);
expect(output.shape).toEqual([1]);
expectArraysClose(await output.data(), [0]);
});
it('stridedSlice should support 1d tensor odd length stride', async () => {
const tensor = tf.tensor1d([0, 1, 2, 3]);
const output = tf.stridedSlice(tensor, [0], [3], [2]);
expect(output.shape).toEqual([2]);
expectArraysClose(await output.data(), [0, 2]);
});
it('stridedSlice should support 2d tensor identity', async () => {
const tensor = tf.tensor2d([1, 2, 3, 4, 5, 6], [2, 3]);
const output = tf.stridedSlice(tensor, [0, 0], [2, 3], [1, 1]);
expect(output.shape).toEqual([2, 3]);
expectArraysClose(await output.data(), [1, 2, 3, 4, 5, 6]);
});
it('stridedSlice should support 2d tensor', async () => {
const tensor = tf.tensor2d([1, 2, 3, 4, 5, 6], [2, 3]);
const output = tf.stridedSlice(tensor, [1, 0], [2, 2], [1, 1]);
expect(output.shape).toEqual([1, 2]);
expectArraysClose(await output.data(), [4, 5]);
});
it('stridedSlice should support 2d tensor strides', async () => {
const tensor = tf.tensor2d([1, 2, 3, 4, 5, 6], [2, 3]);
const output = tf.stridedSlice(tensor, [0, 0], [2, 3], [2, 2]);
expect(output.shape).toEqual([1, 2]);
expectArraysClose(await output.data(), [1, 3]);
});
it('stridedSlice with 2d tensor should be used by tensor directly', async () => {
const t = tf.tensor2d([1, 2, 3, 4, 5, 6], [2, 3]);
const output = t.stridedSlice([1, 0], [2, 2], [1, 1]);
expect(output.shape).toEqual([1, 2]);
expectArraysClose(await output.data(), [4, 5]);
});
it('stridedSlice should support 2d tensor negative strides', async () => {
const tensor = tf.tensor2d([1, 2, 3, 4, 5, 6], [2, 3]);
const output = tf.stridedSlice(tensor, [1, -1], [2, -4], [2, -1]);
expect(output.shape).toEqual([1, 3]);
expectArraysClose(await output.data(), [6, 5, 4]);
});
it('stridedSlice should support 2d tensor begin mask', async () => {
const tensor = tf.tensor2d([1, 2, 3, 4, 5, 6], [2, 3]);
const output = tf.stridedSlice(tensor, [1, 0], [2, 2], [1, 1], 1);
expect(output.shape).toEqual([2, 2]);
expectArraysClose(await output.data(), [1, 2, 4, 5]);
});
it('stridedSlice should support 2d tensor shrink mask', async () => {
const tensor = tf.tensor2d([1, 2, 3, 4, 5, 6], [2, 3]);
const output = tf.stridedSlice(tensor, [1, 0], [2, 2], [1, 1], 0, 0, 0, 0, 1);
expect(output.shape).toEqual([2]);
expectArraysClose(await output.data(), [4, 5]);
});
it('stridedSlice should support 2d tensor end mask', async () => {
const tensor = tf.tensor2d([1, 2, 3, 4, 5, 6], [2, 3]);
const output = tf.stridedSlice(tensor, [1, 0], [2, 2], [1, 1], 0, 2);
expect(output.shape).toEqual([1, 3]);
expectArraysClose(await output.data(), [4, 5, 6]);
});
it('stridedSlice should support 2d tensor' +
' negative strides and begin mask', async () => {
const tensor = tf.tensor2d([1, 2, 3, 4, 5, 6], [2, 3]);
const output = tf.stridedSlice(tensor, [1, -2], [2, -4], [1, -1], 2);
expect(output.shape).toEqual([1, 3]);
expectArraysClose(await output.data(), [6, 5, 4]);
});
it('stridedSlice should support 2d tensor' +
' negative strides and end mask', async () => {
const tensor = tf.tensor2d([1, 2, 3, 4, 5, 6], [2, 3]);
const output = tf.stridedSlice(tensor, [1, -2], [2, -3], [1, -1], 0, 2);
expect(output.shape).toEqual([1, 2]);
expectArraysClose(await output.data(), [5, 4]);
});
it('stridedSlice should support 3d tensor identity', async () => {
const tensor = tf.tensor3d([1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12], [2, 3, 2]);
const output = tf.stridedSlice(tensor, [0, 0, 0], [2, 3, 2], [1, 1, 1]);
expect(output.shape).toEqual([2, 3, 2]);
expectArraysClose(await output.data(), [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12]);
});
it('stridedSlice should support 3d tensor negative stride', async () => {
const tensor = tf.tensor3d([1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12], [2, 3, 2]);
const output = tf.stridedSlice(tensor, [-1, -1, -1], [-3, -4, -3], [-1, -1, -1]);
expect(output.shape).toEqual([2, 3, 2]);
expectArraysClose(await output.data(), [12, 11, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1]);
});
it('stridedSlice should support 3d tensor strided 2', async () => {
const tensor = tf.tensor3d([1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12], [2, 3, 2]);
const output = tf.stridedSlice(tensor, [0, 0, 0], [2, 3, 2], [2, 2, 2]);
expect(output.shape).toEqual([1, 2, 1]);
expectArraysClose(await output.data(), [1, 5]);
});
it('stridedSlice should support 3d tensor shrink mask', async () => {
const tensor = tf.tensor3d([1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12], [2, 3, 2]);
const output = tf.stridedSlice(tensor, [0, 0, 0], [2, 3, 2], [1, 1, 1], 0, 0, 0, 0, 1);
expect(output.shape).toEqual([3, 2]);
expectArraysClose(await output.data(), [1, 2, 3, 4, 5, 6]);
});
it('stridedSlice should support 3d with smaller length of begin array', async () => {
const tensor = tf.tensor4d([1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12], [2, 3, 1, 2]);
const output = tf.stridedSlice(tensor, [1, 0], [2, 3, 1, 2], [1, 1, 1, 1], 0, 0, 0, 0, 0);
expect(output.shape).toEqual([1, 3, 1, 2]);
expectArraysClose(await output.data(), [7, 8, 9, 10, 11, 12]);
});
it('stridedSlice should support 3d with smaller length of end array', async () => {
const tensor = tf.tensor4d([1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12], [2, 3, 1, 2]);
const output = tf.stridedSlice(tensor, [1, 0, 0, 0], [2, 3], [1, 1, 1, 1], 0, 0, 0, 0, 0);
expect(output.shape).toEqual([1, 3, 1, 2]);
expectArraysClose(await output.data(), [7, 8, 9, 10, 11, 12]);
});
it('stridedSlice should support 3d with smaller length of stride array', async () => {
const tensor = tf.tensor4d([1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12], [2, 3, 1, 2]);
const output = tf.stridedSlice(tensor, [1, 0, 0, 0], [2, 3, 1, 2], [1, 1], 0, 0, 0, 0, 0);
expect(output.shape).toEqual([1, 3, 1, 2]);
expectArraysClose(await output.data(), [7, 8, 9, 10, 11, 12]);
});
it('stridedSlice should throw when passed a non-tensor', () => {
expect(() => tf.stridedSlice({}, [0], [0], [1]))
.toThrowError(/Argument 'x' passed to 'stridedSlice' must be a Tensor/);
});
it('stridedSlice should handle negative end with ellipsisMask', () => {
const a = tf.ones([1, 240, 1, 10]);
const output = tf.stridedSlice(a, [0, 0, 0], [0, -1, 0], [1, 1, 1], 3, 1, 4);
expect(output.shape).toEqual([1, 239, 1, 10]);
});
it('accepts a tensor-like object', async () => {
const tensor = [0, 1, 2, 3];
const output = tf.stridedSlice(tensor, [0], [3], [2]);
expect(output.shape).toEqual([2]);
expectArraysClose(await output.data(), [0, 2]);
});
it('ensure no memory leak', async () => {
const numTensorsBefore = tf.memory().numTensors;
const numDataIdBefore = tf.engine().backend.numDataIds();
const tensor = tf.tensor1d([0, 1, 2, 3]);
const output = tf.stridedSlice(tensor, [0], [3], [2]);
expect(output.shape).toEqual([2]);
expectArraysClose(await output.data(), [0, 2]);
tensor.dispose();
output.dispose();
const numTensorsAfter = tf.memory().numTensors;
const numDataIdAfter = tf.engine().backend.numDataIds();
expect(numTensorsAfter).toBe(numTensorsBefore);
expect(numDataIdAfter).toBe(numDataIdBefore);
});
});
//# sourceMappingURL=strided_slice_test.js.map