UNPKG

@tensorflow/tfjs-core

Version:

Hardware-accelerated JavaScript library for machine intelligence

197 lines 8.28 kB
/** * @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. * ============================================================================= */ import * as tf from '../index'; import { ALL_ENVS, describeWithFlags } from '../jasmine_util'; import { expectArraysClose } from '../test_util'; describeWithFlags('slice2d', ALL_ENVS, () => { it('slicing a 1x1 from a 1x1 returns a 1x1', () => { const a = tf.tensor2d([0], [1, 1]); const b = tf.slice2d(a, [0, 0], [1, 1]); expect(b.shape).toEqual([1, 1]); }); it('returns a tensor of slice size', () => { const a = tf.zeros([100, 100]); const b = tf.slice2d(a, [0, 0], [12, 34]); expect(b.shape).toEqual([12, 34]); }); it('returns the upper-left submatrix when begin is [0, 0]', async () => { const a = tf.randomUniform([10, 10], -1, 1); const b = tf.slice2d(a, [0, 0], [2, 2]); const aValues = await a.data(); expectArraysClose(await b.data(), [aValues[0], aValues[1], aValues[10], aValues[11]]); }); it('returns the rectangle specified', async () => { const a = tf.tensor2d([1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12], [4, 3]); const b = tf.slice2d(a, [1, 1], [3, 2]); expectArraysClose(await b.data(), [5, 6, 8, 9, 11, 12]); }); it('throws when requesting out of bounds slice', () => { const a = tf.tensor2d([1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12], [4, 3]); expect(() => tf.slice2d(a, [1, 1], [10, 10])).toThrowError(); }); it('grad', async () => { const a = tf.tensor2d([[1, 2, 3], [4, 5, 6]]); const dy = tf.tensor2d([[20], [50]]); const da = tf.grad((x) => tf.slice2d(a, [0, 1], [2, 1]))(a, dy); expect(da.shape).toEqual([2, 3]); expectArraysClose(await da.data(), [0, 20, 0, 0, 50, 0]); }); it('accepts a tensor-like object', () => { const a = [[0]]; // 1x1 const b = tf.slice2d(a, [0, 0], [1, 1]); expect(b.shape).toEqual([1, 1]); }); it('slice an already sliced tensor, first was not continous', async () => { const a = [ [1, 2, 3, 4], [5, 6, 7, 8], [9, 10, 11, 12], ]; // 3x4. const b = tf.slice(a, [0, 1]); const c = tf.slice(b, [1, 1], [1, 1]); expect(c.shape).toEqual([1, 1]); expectArraysClose(await c.data(), [7]); }); it('slice an already sliced tensor, first was continous', async () => { const a = [ [1, 2, 3, 4], [5, 6, 7, 8], [9, 10, 11, 12], ]; // 3x4. const b = tf.slice(a, [1, 0]); const c = tf.slice(b, [1, 0]); expect(c.shape).toEqual([1, 4]); expectArraysClose(await c.data(), [9, 10, 11, 12]); }); it('slice an already sliced tensor and do async read', async () => { const a = [ [1, 2, 3, 4], [5, 6, 7, 8], [9, 10, 11, 12], ]; // 3x4. const b = tf.slice(a, [0, 1]); const c = tf.slice(b, [1, 1], [1, 1]); expect(c.shape).toEqual([1, 1]); expectArraysClose(await c.data(), new Float32Array([7])); }); it('square a sliced texture, followed by non-sliced texture of same shape', async () => { // Make a 2x3 tensor, upload to gpu and reshape to 3x2. const input = tf.tensor([[1, 2, 3], [4, 5, 6]]).abs().as2D(3, 2); const slicedInput = tf.slice(input, [0, 0], [3, 2]); // First square program takes the sliced input. const a = slicedInput.square(); expectArraysClose(await a.data(), [1, 4, 9, 16, 25, 36]); // Second square program takes the non-sliced input. const b = tf.square(input); expectArraysClose(await b.data(), [1, 4, 9, 16, 25, 36]); }); it('square a non-sliced texture, followed by a sliced texture of same shape', async () => { // Make a 2x3 tensor, upload to gpu and reshape to 3x2. const input = tf.tensor([[1, 2, 3], [4, 5, 6]]).abs().as2D(3, 2); // Make a sliced version of the same tensor with the same shape. const slicedInput = tf.slice(input, [0, 0], [3, 2]); // First square program takes the non-sliced input. const a = input.square(); expectArraysClose(await a.data(), [1, 4, 9, 16, 25, 36]); // Second square program takes the sliced input. const b = tf.square(slicedInput); expectArraysClose(await b.data(), [1, 4, 9, 16, 25, 36]); }); it('slice a tensor and do async read', async () => { const a = [ [1, 2, 3, 4], [5, 6, 7, 8], [9, 10, 11, 12], ]; // 3x4. const b = tf.slice(a, [0, 1], [3, 2]); expect(b.shape).toEqual([3, 2]); const vals = await b.data(); expectArraysClose(vals, new Float32Array([2, 3, 6, 7, 10, 11])); }); it('flatten a sliced tensor that was continuous in memory', async () => { const a = [ [1, 2, 3, 4], [5, 6, 7, 8], [9, 10, 11, 12], ]; // 3x4. const b = tf.slice(a, [1, 0]).flatten(); expect(b.shape).toEqual([8]); expectArraysClose(await b.data(), [5, 6, 7, 8, 9, 10, 11, 12]); }); it('slice a tensor that was not continuous in memory', async () => { const a = [ [1, 2, 3, 4], [5, 6, 7, 8], [9, 10, 11, 12], ]; // 3x4. const b = tf.slice(a, [0, 1]); expect(b.shape).toEqual([3, 3]); expectArraysClose(await b.data(), [2, 3, 4, 6, 7, 8, 10, 11, 12]); }); it('flatten a sliced tensor that was not continuous in memory', async () => { const a = [ [1, 2, 3, 4], [5, 6, 7, 8], [9, 10, 11, 12], ]; // 3x4. const b = tf.slice(a, [0, 1]).flatten(); expect(b.shape).toEqual([9]); expectArraysClose(await b.data(), [2, 3, 4, 6, 7, 8, 10, 11, 12]); }); it('flatten a sliced tensor not continuous in memory and run program', async () => { const a = [ [1, 2, 3, 4], [5, 6, 7, 8], [9, 10, 11, 12], ]; // 3x4. const b = tf.slice(a, [0, 1]).flatten(); const c = tf.square(b); expectArraysClose(await c.data(), [4, 9, 16, 36, 49, 64, 100, 121, 144]); }); it('reshape a sliced 1d into a 2d tensor', async () => { const a = [1, 2, 3, 4, 5]; const b = tf.slice(a, 1).as2D(2, 2); expect(b.shape).toEqual([2, 2]); expectArraysClose(await b.data(), [2, 3, 4, 5]); }); it('reshape a sliced 1d into a 2d tensor and run program', async () => { const a = [1, 2, 3, 4, 5]; const b = tf.slice(a, 1).as2D(2, 2).square(); expect(b.shape).toEqual([2, 2]); expectArraysClose(await b.data(), [4, 9, 16, 25]); }); it('broadcast the original with the sliced tensor', async () => { const a = [[1, 2], [3, 4]]; const b = tf.slice(a, [0, 1]); const c = tf.add(a, b); expect(c.shape).toEqual([2, 2]); expectArraysClose(await c.data(), [3, 4, 7, 8]); }); it('zero-sized slice out of a non-zero sized tensor', async () => { const a = tf.zeros([4, 2]); const res = tf.slice(a, [0, 0], [0, 2]); expect(res.shape).toEqual([0, 2]); expectArraysClose(await res.data(), []); }); it('zero-sized slice out of a zero-sized tensor', async () => { const a = tf.zeros([0, 4]); const res = tf.slice(a, [0, 1], [0, 3]); expect(res.shape).toEqual([0, 3]); expectArraysClose(await res.data(), []); }); }); //# sourceMappingURL=slice2d_test.js.map