166 lines
6.0 KiB
Python
166 lines
6.0 KiB
Python
# Copyright (c) 2026 PaddlePaddle Authors. 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.
|
|
|
|
"""
|
|
随机操作高级测试 / Advanced Random Operations Tests
|
|
|
|
测试目标 / Test Target:
|
|
paddle.tensor.random 随机张量操作
|
|
|
|
覆盖的模块 / Covered Modules:
|
|
- paddle.seed/get_cuda_rng_state/set_cuda_rng_state: 随机状态
|
|
- paddle.randperm: 随机排列
|
|
- paddle.bernoulli: 伯努利采样
|
|
- paddle.multinomial: 多项式采样
|
|
|
|
作用 / Purpose:
|
|
补充随机操作API的测试,提升覆盖率。
|
|
"""
|
|
|
|
import unittest
|
|
|
|
import numpy as np
|
|
|
|
import paddle
|
|
|
|
paddle.disable_static()
|
|
|
|
|
|
class TestRandomSeed(unittest.TestCase):
|
|
"""测试随机种子 / Test random seed"""
|
|
|
|
def test_seed_reproducibility(self):
|
|
"""测试随机种子可重复性 / Test seed reproducibility"""
|
|
paddle.seed(42)
|
|
x1 = paddle.randn([5])
|
|
paddle.seed(42)
|
|
x2 = paddle.randn([5])
|
|
np.testing.assert_allclose(x1.numpy(), x2.numpy())
|
|
|
|
def test_different_seeds(self):
|
|
"""测试不同种子 / Test different seeds"""
|
|
paddle.seed(42)
|
|
x1 = paddle.randn([10])
|
|
paddle.seed(99)
|
|
x2 = paddle.randn([10])
|
|
# Different seeds should produce different results (very likely)
|
|
self.assertFalse(np.allclose(x1.numpy(), x2.numpy()))
|
|
|
|
|
|
class TestRandperm(unittest.TestCase):
|
|
"""测试随机排列 / Test random permutation"""
|
|
|
|
def test_randperm_basic(self):
|
|
"""测试基本随机排列 / Test basic randperm"""
|
|
result = paddle.randperm(10)
|
|
self.assertEqual(result.shape[0], 10)
|
|
# All values 0-9 should be present
|
|
sorted_result = np.sort(result.numpy())
|
|
np.testing.assert_array_equal(sorted_result, np.arange(10))
|
|
|
|
def test_randperm_dtype(self):
|
|
"""测试随机排列数据类型 / Test randperm dtype"""
|
|
result = paddle.randperm(5, dtype='int64')
|
|
self.assertEqual(result.dtype, paddle.int64)
|
|
|
|
def test_randperm_shuffle(self):
|
|
"""测试随机排列用于数据打乱 / Test randperm for shuffling"""
|
|
data = paddle.to_tensor([10.0, 20.0, 30.0, 40.0, 50.0])
|
|
perm = paddle.randperm(5)
|
|
shuffled = data[perm]
|
|
self.assertEqual(shuffled.shape, [5])
|
|
# All values should still be present
|
|
np.testing.assert_array_equal(
|
|
np.sort(shuffled.numpy()), [10.0, 20.0, 30.0, 40.0, 50.0]
|
|
)
|
|
|
|
|
|
class TestBernoulli(unittest.TestCase):
|
|
"""测试伯努利采样 / Test Bernoulli sampling"""
|
|
|
|
def test_bernoulli_basic(self):
|
|
"""测试基本伯努利采样 / Test basic Bernoulli sampling"""
|
|
probs = paddle.to_tensor([0.5, 0.5, 0.5, 0.5, 0.5])
|
|
result = paddle.bernoulli(probs)
|
|
# All values should be 0 or 1
|
|
self.assertTrue(bool(((result == 0) | (result == 1)).all().numpy()))
|
|
|
|
def test_bernoulli_all_ones(self):
|
|
"""测试全1伯努利采样 / Test Bernoulli with all ones"""
|
|
probs = paddle.ones([5])
|
|
result = paddle.bernoulli(probs)
|
|
np.testing.assert_array_equal(result.numpy(), np.ones(5))
|
|
|
|
def test_bernoulli_all_zeros(self):
|
|
"""测试全0伯努利采样 / Test Bernoulli with all zeros"""
|
|
probs = paddle.zeros([5])
|
|
result = paddle.bernoulli(probs)
|
|
np.testing.assert_array_equal(result.numpy(), np.zeros(5))
|
|
|
|
|
|
class TestMultinomial(unittest.TestCase):
|
|
"""测试多项式采样 / Test multinomial sampling"""
|
|
|
|
def test_multinomial_basic(self):
|
|
"""测试基本多项式采样 / Test basic multinomial sampling"""
|
|
weights = paddle.to_tensor([1.0, 2.0, 3.0, 4.0])
|
|
result = paddle.multinomial(weights, num_samples=100, replacement=True)
|
|
self.assertEqual(result.shape, [100])
|
|
# All indices should be valid
|
|
self.assertTrue(bool((result >= 0).all().numpy()))
|
|
self.assertTrue(bool((result < 4).all().numpy()))
|
|
|
|
def test_multinomial_without_replacement(self):
|
|
"""测试无重复多项式采样 / Test multinomial without replacement"""
|
|
weights = paddle.to_tensor([1.0, 1.0, 1.0, 1.0, 1.0])
|
|
result = paddle.multinomial(weights, num_samples=3, replacement=False)
|
|
self.assertEqual(result.shape, [3])
|
|
# All values should be unique
|
|
self.assertEqual(len(np.unique(result.numpy())), 3)
|
|
|
|
def test_multinomial_2d(self):
|
|
"""测试2D多项式采样 / Test 2D multinomial sampling"""
|
|
weights = paddle.to_tensor([[1.0, 2.0, 3.0], [3.0, 2.0, 1.0]])
|
|
result = paddle.multinomial(weights, num_samples=5, replacement=True)
|
|
self.assertEqual(result.shape, [2, 5])
|
|
|
|
|
|
class TestRandomDropout(unittest.TestCase):
|
|
"""测试随机丢弃 / Test random dropout"""
|
|
|
|
def test_dropout_training(self):
|
|
"""测试训练模式dropout / Test dropout in training mode"""
|
|
dropout = paddle.nn.Dropout(p=0.5)
|
|
dropout.train()
|
|
x = paddle.ones([100, 100])
|
|
result = dropout(x)
|
|
# In training mode, some values should be 0
|
|
zero_fraction = float((result == 0).sum().numpy()) / result.numel()
|
|
# Should be approximately 0.5 (with tolerance)
|
|
self.assertGreater(zero_fraction, 0.3)
|
|
self.assertLess(zero_fraction, 0.7)
|
|
|
|
def test_dropout_eval(self):
|
|
"""测试评估模式dropout / Test dropout in eval mode"""
|
|
dropout = paddle.nn.Dropout(p=0.5)
|
|
dropout.eval()
|
|
x = paddle.ones([10, 10])
|
|
result = dropout(x)
|
|
# In eval mode, no dropout
|
|
np.testing.assert_allclose(result.numpy(), x.numpy())
|
|
|
|
|
|
if __name__ == '__main__':
|
|
unittest.main()
|