• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1# Copyright 2021 Huawei Technologies Co., Ltd
2#
3# Licensed under the Apache License, Version 2.0 (the "License");
4# you may not use this file except in compliance with the License.
5# You may obtain a copy of the License at
6#
7# http://www.apache.org/licenses/LICENSE-2.0
8#
9# Unless required by applicable law or agreed to in writing, software
10# distributed under the License is distributed on an "AS IS" BASIS,
11# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12# See the License for the specific language governing permissions and
13# limitations under the License.
14# ============================================================================
15"""test jvp in pynative mode"""
16import numpy as np
17import pytest
18import mindspore.nn as nn
19import mindspore.context as context
20from mindspore import Tensor
21from mindspore.nn.grad import Vjp
22
23context.set_context(mode=context.PYNATIVE_MODE)
24
25
26class SingleInputNet(nn.Cell):
27    def construct(self, x):
28        return x**3
29
30
31class MultipleInputsOutputNet(nn.Cell):
32    def construct(self, x, y):
33        return 2*x, y**3
34
35
36@pytest.mark.level0
37@pytest.mark.platform_x86_cpu
38@pytest.mark.env_onecard
39def test_vjp_single_input_graph():
40    x = Tensor(np.array([[1, 2], [3, 4]]).astype(np.float32))
41    v = Tensor(np.array([[1, 1], [1, 1]]).astype(np.float32))
42    net = SingleInputNet()
43    expect_primal = Tensor(np.array([[1, 8], [27, 64]]).astype(np.float32))
44    expect_grad = Tensor(np.array([[3, 12], [27, 48]]).astype(np.float32))
45    primal, grad = Vjp(net)(x, v)
46    assert np.allclose(primal.asnumpy(), expect_primal.asnumpy())
47    assert np.allclose(grad.asnumpy(), expect_grad.asnumpy())
48
49
50
51@pytest.mark.level0
52@pytest.mark.platform_x86_cpu
53@pytest.mark.env_onecard
54def test_vjp_multiple_inputs_default_v_graph():
55    x = Tensor(np.array([[1, 2], [3, 4]]).astype(np.float32))
56    y = Tensor(np.array([[1, 2], [3, 4]]).astype(np.float32))
57    v = Tensor(np.array([[1, 1], [1, 1]]).astype(np.float32))
58    net = MultipleInputsOutputNet()
59    expect_primal_0 = Tensor(np.array([[2, 4], [6, 8]]).astype(np.float32))
60    expect_primal_1 = Tensor(np.array([[1, 8], [27, 64]]).astype(np.float32))
61    expect_grad_0 = Tensor(np.array([[2, 2], [2, 2]]).astype(np.float32))
62    expect_grad_1 = Tensor(np.array([[3, 12], [27, 48]]).astype(np.float32))
63    primal, grad = Vjp(net)(x, y, (v, v))
64    assert isinstance(primal, tuple)
65    assert len(primal) == 2
66    assert np.allclose(primal[0].asnumpy(), expect_primal_0.asnumpy())
67    assert np.allclose(primal[1].asnumpy(), expect_primal_1.asnumpy())
68    assert isinstance(grad, tuple)
69    assert len(grad) == 2
70    assert np.allclose(grad[0].asnumpy(), expect_grad_0.asnumpy())
71    assert np.allclose(grad[1].asnumpy(), expect_grad_1.asnumpy())
72