wehub-resource-sync
191 lines
7.7 KiB
Python
191 lines
7.7 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.
|
|
|
|
# [AUTO-GENERATED] Test file for paddle.nn.functional.input
|
|
# 覆盖模块: paddle/nn/functional/input.py
|
|
# 未覆盖行: 118,119,121,122,124,125,127,128,129,130,137,315,316,318,324,326,332
|
|
# Covered module: paddle/nn/functional/input.py
|
|
# Uncovered lines: 118,119,121,122,124,125,127,128,129,130,137,315,316,318,324,326,332
|
|
|
|
import unittest
|
|
|
|
import numpy as np
|
|
|
|
import paddle
|
|
|
|
|
|
class TestOneHot(unittest.TestCase):
|
|
"""测试 one_hot 函数
|
|
Test one_hot function"""
|
|
|
|
def test_one_hot_basic(self):
|
|
"""测试基本的 one_hot
|
|
Test basic one_hot"""
|
|
label = paddle.to_tensor([1, 1, 3, 0], dtype='int64')
|
|
result = paddle.nn.functional.one_hot(label, num_classes=4)
|
|
self.assertEqual(result.shape, [4, 4])
|
|
self.assertEqual(result.dtype, paddle.float32)
|
|
expected = np.array(
|
|
[
|
|
[0.0, 1.0, 0.0, 0.0],
|
|
[0.0, 1.0, 0.0, 0.0],
|
|
[0.0, 0.0, 0.0, 1.0],
|
|
[1.0, 0.0, 0.0, 0.0],
|
|
]
|
|
)
|
|
np.testing.assert_array_equal(result.numpy(), expected)
|
|
|
|
def test_one_hot_int32(self):
|
|
"""测试 int32 输入的 one_hot
|
|
Test one_hot with int32 input"""
|
|
label = paddle.to_tensor([0, 1, 2], dtype='int32')
|
|
result = paddle.nn.functional.one_hot(label, num_classes=3)
|
|
self.assertEqual(result.shape, [3, 3])
|
|
|
|
def test_one_hot_auto_num_classes(self):
|
|
"""测试自动推断 num_classes 的 one_hot (num_classes=-1)
|
|
Test one_hot with auto num_classes (num_classes=-1)"""
|
|
label = paddle.to_tensor([1, 0, 2], dtype='int64')
|
|
result = paddle.nn.functional.one_hot(label, num_classes=-1)
|
|
# max(label) + 1 = 3
|
|
self.assertEqual(result.shape, [3, 3])
|
|
|
|
def test_one_hot_2d_input(self):
|
|
"""测试2D输入的 one_hot
|
|
Test one_hot with 2D input"""
|
|
label = paddle.to_tensor([[0, 1], [2, 0]], dtype='int64')
|
|
result = paddle.nn.functional.one_hot(label, num_classes=3)
|
|
self.assertEqual(result.shape, [2, 2, 3])
|
|
|
|
def test_one_hot_alias_input(self):
|
|
"""测试使用 input 别名的 one_hot
|
|
Test one_hot with 'input' alias parameter"""
|
|
label = paddle.to_tensor([1, 0], dtype='int64')
|
|
result = paddle.nn.functional.one_hot(input=label, num_classes=2)
|
|
self.assertEqual(result.shape, [2, 2])
|
|
|
|
def test_one_hot_3d_input(self):
|
|
"""测试3D输入的 one_hot
|
|
Test one_hot with 3D input"""
|
|
label = paddle.to_tensor([[[0, 1]]], dtype='int64')
|
|
result = paddle.nn.functional.one_hot(label, num_classes=2)
|
|
self.assertEqual(result.shape, [1, 1, 2, 2])
|
|
|
|
|
|
class TestEmbedding(unittest.TestCase):
|
|
"""测试 embedding 函数
|
|
Test embedding function"""
|
|
|
|
def test_embedding_basic(self):
|
|
"""测试基本的 embedding
|
|
Test basic embedding"""
|
|
x = paddle.to_tensor([0, 1, 2], dtype='int64')
|
|
weight = paddle.randn([10, 4])
|
|
result = paddle.nn.functional.embedding(x, weight)
|
|
self.assertEqual(result.shape, [3, 4])
|
|
|
|
def test_embedding_2d_indices(self):
|
|
"""测试2D索引的 embedding
|
|
Test embedding with 2D indices"""
|
|
x = paddle.arange(3, 6).reshape((3, 1)).astype(paddle.int64)
|
|
weight = paddle.full(shape=(10, 3), fill_value=2.0).astype(
|
|
paddle.float32
|
|
)
|
|
result = paddle.nn.functional.embedding(x, weight, sparse=True)
|
|
self.assertEqual(result.shape, [3, 1, 3])
|
|
np.testing.assert_allclose(result.numpy(), np.full((3, 1, 3), 2.0))
|
|
|
|
def test_embedding_padding_idx(self):
|
|
"""测试带 padding_idx 的 embedding
|
|
Test embedding with padding_idx"""
|
|
x = paddle.to_tensor([0, 1, 2, 3], dtype='int64')
|
|
weight = paddle.randn([5, 4])
|
|
result = paddle.nn.functional.embedding(x, weight, padding_idx=3)
|
|
self.assertEqual(result.shape, [4, 4])
|
|
# padding_idx=3 的位置应该全为0
|
|
np.testing.assert_array_equal(result[3].numpy(), np.zeros(4))
|
|
|
|
def test_embedding_negative_padding_idx(self):
|
|
"""测试负 padding_idx 的 embedding
|
|
Test embedding with negative padding_idx"""
|
|
x = paddle.to_tensor([0, 1, 5], dtype='int64')
|
|
weight = paddle.randn([6, 4])
|
|
result = paddle.nn.functional.embedding(x, weight, padding_idx=-1)
|
|
self.assertEqual(result.shape, [3, 4])
|
|
# padding_idx=-1 means last row (index 5)
|
|
np.testing.assert_array_equal(result[2].numpy(), np.zeros(4))
|
|
|
|
def test_embedding_invalid_padding_idx(self):
|
|
"""测试无效 padding_idx 的报错
|
|
Test embedding raises error for invalid padding_idx"""
|
|
x = paddle.to_tensor([0, 1], dtype='int64')
|
|
weight = paddle.randn([5, 4])
|
|
with self.assertRaises(ValueError):
|
|
paddle.nn.functional.embedding(x, weight, padding_idx=10)
|
|
|
|
def test_embedding_sparse(self):
|
|
"""测试 sparse 模式的 embedding
|
|
Test embedding in sparse mode"""
|
|
x = paddle.to_tensor([0, 1, 2], dtype='int64')
|
|
weight = paddle.randn([10, 4])
|
|
result = paddle.nn.functional.embedding(x, weight, sparse=True)
|
|
self.assertEqual(result.shape, [3, 4])
|
|
|
|
def test_embedding_max_norm(self):
|
|
"""测试带 max_norm 的 embedding (会 renorm weight)
|
|
Test embedding with max_norm (renorms weight)"""
|
|
x = paddle.to_tensor([0, 1], dtype='int64')
|
|
weight = paddle.randn([5, 4]) * 10 # large values
|
|
result = paddle.nn.functional.embedding(x, weight, max_norm=1.0)
|
|
self.assertEqual(result.shape, [2, 4])
|
|
|
|
def test_embedding_alias_input(self):
|
|
"""测试使用 input 别名的 embedding
|
|
Test embedding with 'input' alias parameter"""
|
|
x = paddle.to_tensor([0, 1], dtype='int64')
|
|
weight = paddle.randn([5, 4])
|
|
result = paddle.nn.functional.embedding(input=x, weight=weight)
|
|
self.assertEqual(result.shape, [2, 4])
|
|
|
|
|
|
class TestEmbeddingRenorm(unittest.TestCase):
|
|
"""测试 embedding_renorm_ 函数
|
|
Test embedding_renorm_ function"""
|
|
|
|
def test_embedding_renorm_basic(self):
|
|
"""测试基本的 embedding_renorm_
|
|
Test basic embedding_renorm_"""
|
|
x = paddle.to_tensor([0, 1, 2], dtype='int64')
|
|
weight = paddle.randn([5, 4]) * 10
|
|
original_norms = paddle.norm(weight[:3], p=2, axis=1)
|
|
result = paddle.nn.functional.embedding_renorm_(x, weight, max_norm=1.0)
|
|
new_norms = paddle.norm(result[:3], p=2, axis=1)
|
|
# All norms should be <= max_norm
|
|
self.assertTrue(paddle.all(new_norms <= 1.0 + 1e-5).item())
|
|
|
|
def test_embedding_renorm_no_change(self):
|
|
"""测试 norm 已经小于 max_norm 时不会改变 weight
|
|
Test embedding_renorm_ doesn't change weight when norms are already small"""
|
|
x = paddle.to_tensor([0, 1], dtype='int64')
|
|
weight = paddle.randn([5, 4]) * 0.01 # small values
|
|
result = paddle.nn.functional.embedding_renorm_(
|
|
x, weight, max_norm=100.0
|
|
)
|
|
# Should not change since norms are already small
|
|
np.testing.assert_allclose(result.numpy(), weight.numpy(), rtol=1e-5)
|
|
|
|
|
|
if __name__ == '__main__':
|
|
unittest.main()
|