wehub-resource-sync
160 lines
5.2 KiB
Python
160 lines
5.2 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.
|
|
|
|
"""
|
|
Embedding层高级测试 / Advanced Embedding Layer Tests
|
|
|
|
测试目标 / Test Target:
|
|
paddle.nn.Embedding 高级用法
|
|
|
|
覆盖的模块 / Covered Modules:
|
|
- paddle.nn.Embedding: 基本嵌入
|
|
- 稀疏更新嵌入
|
|
- 带padding的嵌入
|
|
- 嵌入层梯度
|
|
|
|
作用 / Purpose:
|
|
补充Embedding层API的测试，提升覆盖率。
|
|
"""
|
|
|
|
import unittest
|
|
|
|
import numpy as np
|
|
|
|
import paddle
|
|
from paddle import nn
|
|
|
|
paddle.disable_static()
|
|
|
|
|
|
class TestEmbeddingBasic(unittest.TestCase):
|
|
"""测试基本Embedding / Test basic Embedding"""
|
|
|
|
def test_embedding_basic(self):
|
|
"""测试基本嵌入 / Test basic embedding"""
|
|
emb = nn.Embedding(100, 16)
|
|
x = paddle.to_tensor([0, 1, 2, 3])
|
|
result = emb(x)
|
|
self.assertEqual(result.shape, [4, 16])
|
|
|
|
def test_embedding_2d_input(self):
|
|
"""测试2D输入嵌入 / Test embedding with 2D input"""
|
|
emb = nn.Embedding(100, 16)
|
|
x = paddle.to_tensor([[0, 1, 2], [3, 4, 5]])
|
|
result = emb(x)
|
|
self.assertEqual(result.shape, [2, 3, 16])
|
|
|
|
def test_embedding_padding_idx(self):
|
|
"""测试padding_idx嵌入 / Test embedding with padding_idx"""
|
|
emb = nn.Embedding(100, 16, padding_idx=0)
|
|
x = paddle.to_tensor([0, 1, 2])
|
|
result = emb(x)
|
|
self.assertEqual(result.shape, [3, 16])
|
|
# Padding index should produce zero vector
|
|
np.testing.assert_allclose(result[0].numpy(), np.zeros(16), atol=1e-7)
|
|
|
|
def test_embedding_sparse(self):
|
|
"""测试稀疏嵌入 / Test sparse embedding"""
|
|
emb = nn.Embedding(100, 16, sparse=True)
|
|
x = paddle.to_tensor([5, 10, 15])
|
|
result = emb(x)
|
|
self.assertEqual(result.shape, [3, 16])
|
|
|
|
def test_embedding_dtype(self):
|
|
"""测试嵌入数据类型 / Test embedding dtype"""
|
|
emb = nn.Embedding(
|
|
100,
|
|
16,
|
|
weight_attr=paddle.ParamAttr(
|
|
initializer=paddle.nn.initializer.Normal()
|
|
),
|
|
)
|
|
x = paddle.to_tensor([1, 2, 3])
|
|
result = emb(x)
|
|
self.assertEqual(result.dtype, paddle.float32)
|
|
|
|
|
|
class TestEmbeddingGradient(unittest.TestCase):
|
|
"""测试嵌入梯度 / Test embedding gradient"""
|
|
|
|
def test_embedding_gradient(self):
|
|
"""测试嵌入梯度更新 / Test embedding gradient update"""
|
|
emb = nn.Embedding(10, 4)
|
|
x = paddle.to_tensor([0, 2, 4])
|
|
output = emb(x)
|
|
loss = output.sum()
|
|
loss.backward()
|
|
self.assertIsNotNone(emb.weight.grad)
|
|
|
|
def test_embedding_with_linear(self):
|
|
"""测试嵌入与线性层组合 / Test embedding combined with linear layer"""
|
|
emb = nn.Embedding(100, 16)
|
|
linear = nn.Linear(16, 8)
|
|
x = paddle.to_tensor([1, 2, 3, 4])
|
|
embedded = emb(x)
|
|
output = linear(embedded)
|
|
self.assertEqual(output.shape, [4, 8])
|
|
|
|
|
|
class TestMultipleEmbeddings(unittest.TestCase):
|
|
"""测试多嵌入组合 / Test multiple embeddings combination"""
|
|
|
|
def test_category_embeddings(self):
|
|
"""测试类别嵌入组合 / Test category embeddings combination"""
|
|
# Common in recommendation systems
|
|
item_emb = nn.Embedding(1000, 32)
|
|
user_emb = nn.Embedding(500, 32)
|
|
items = paddle.to_tensor([1, 5, 10])
|
|
users = paddle.to_tensor([2, 3, 4])
|
|
item_vecs = item_emb(items)
|
|
user_vecs = user_emb(users)
|
|
# Dot product similarity
|
|
scores = (item_vecs * user_vecs).sum(axis=1)
|
|
self.assertEqual(scores.shape, [3])
|
|
|
|
def test_position_embedding(self):
|
|
"""测试位置嵌入 / Test position embedding"""
|
|
max_seq_len = 128
|
|
d_model = 64
|
|
pos_emb = nn.Embedding(max_seq_len, d_model)
|
|
positions = paddle.arange(10)
|
|
pos_vecs = pos_emb(positions)
|
|
self.assertEqual(pos_vecs.shape, [10, d_model])
|
|
|
|
|
|
class TestEmbeddingWeight(unittest.TestCase):
|
|
"""测试嵌入权重操作 / Test embedding weight operations"""
|
|
|
|
def test_embedding_weight_init(self):
|
|
"""测试嵌入权重初始化 / Test embedding weight initialization"""
|
|
# Initialize with specific weights
|
|
weight = paddle.randn([100, 16])
|
|
emb = nn.Embedding(100, 16)
|
|
emb.weight.set_value(weight)
|
|
x = paddle.to_tensor([0, 1, 2])
|
|
result = emb(x)
|
|
np.testing.assert_allclose(
|
|
result.numpy(), weight[:3].numpy(), rtol=1e-5
|
|
)
|
|
|
|
def test_embedding_num_embeddings(self):
|
|
"""测试嵌入数量属性 / Test embedding num_embeddings attribute"""
|
|
emb = nn.Embedding(200, 32)
|
|
self.assertEqual(emb._num_embeddings, 200)
|
|
self.assertEqual(emb._embedding_dim, 32)
|
|
|
|
|
|
if __name__ == '__main__':
|
|
unittest.main()
|