# Copyright (c) 2024 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] Unit test for paddle.nn.functional.input # 自动生成的单测，覆盖 paddle.nn.functional.input 模块中未覆盖的代码 # Target: cover uncovered lines 118-137, 315-332 in paddle/python/paddle/nn/functional/input.py """ 测试模块：paddle.nn.functional.input (embedding, one_hot) Test Module: paddle.nn.functional.input (embedding, one_hot) 本测试覆盖以下功能： This test covers the following functions: 1. one_hot - 独热编码 / One-hot encoding - 静态图路径 / Static graph path (lines 118-137) - 自动推断num_classes / Auto-infer num_classes 2. embedding - 嵌入查找 / Embedding lookup - max_norm 参数：嵌入向量范数裁剪 / max_norm parameter: embedding norm clipping - embedding_renorm_ 函数 / embedding_renorm_ function - scale_grad_by_freq 参数 / scale_grad_by_freq parameter (lines 315-332) - padding_idx 负索引 / negative padding_idx 覆盖的未覆盖行：118-137（one_hot静态图），315-332（embedding scale_grad_by_freq） """ import unittest import numpy as np import paddle class TestOneHotDynamic(unittest.TestCase): """测试动态图模式下的one_hot Test one_hot in dynamic graph mode""" def setUp(self): paddle.disable_static() def test_basic_one_hot(self): """测试基本的one_hot编码 Test basic one_hot encoding""" x = paddle.to_tensor([0, 1, 2, 3], dtype='int64') out = paddle.nn.functional.one_hot(x, num_classes=4) expected = np.array( [ [1, 0, 0, 0], [0, 1, 0, 0], [0, 0, 1, 0], [0, 0, 0, 1], ], dtype='float32', ) np.testing.assert_array_equal(out.numpy(), expected) def test_one_hot_auto_num_classes(self): """测试one_hot自动推断num_classes（num_classes=-1） Test one_hot with automatic num_classes inference (num_classes=-1)""" x = paddle.to_tensor([0, 1, 2], dtype='int64') out = paddle.nn.functional.one_hot(x) # num_classes defaults to -1 self.assertEqual(list(out.shape), [3, 3]) def test_one_hot_2d_input(self): """测试2D输入的one_hot Test one_hot with 2D input""" x = paddle.to_tensor([[0, 1], [2, 0]], dtype='int64') out = paddle.nn.functional.one_hot(x, num_classes=3) self.assertEqual(list(out.shape), [2, 2, 3]) class TestOneHotStatic(unittest.TestCase): """测试静态图模式下的one_hot，覆盖未覆盖行118-137 Test one_hot in static graph mode to cover uncovered lines 118-137""" def test_static_graph_one_hot(self): """测试静态图模式下的one_hot基本功能 Test basic one_hot in static graph mode""" paddle.enable_static() try: main_prog = paddle.static.Program() startup_prog = paddle.static.Program() with paddle.static.program_guard(main_prog, startup_prog): x = paddle.static.data(name='x', shape=[4], dtype='int64') out = paddle.nn.functional.one_hot(x, num_classes=5) exe = paddle.static.Executor(paddle.CPUPlace()) exe.run(startup_prog) x_np = np.array([0, 1, 3, 4], dtype='int64') result = exe.run(main_prog, feed={'x': x_np}, fetch_list=[out]) self.assertEqual(list(result[0].shape), [4, 5]) # 验证第一个向量 / Verify first vector np.testing.assert_array_equal(result[0][0], [1, 0, 0, 0, 0]) finally: paddle.disable_static() class TestEmbeddingMaxNorm(unittest.TestCase): """测试embedding的max_norm功能 Test embedding max_norm feature""" def setUp(self): paddle.disable_static() def test_embedding_with_max_norm(self): """测试embedding的max_norm范数裁剪，覆盖embedding_renorm_函数 Test embedding max_norm norm clipping, covers embedding_renorm_ function""" x = paddle.to_tensor([0, 1, 2], dtype='int64') # 创建一个权重矩阵，其中某些行的范数大于max_norm # Create a weight matrix where some rows have norm > max_norm weight = paddle.to_tensor( [ [10.0, 10.0, 10.0], # norm = sqrt(300) ≈ 17.3 [1.0, 0.0, 0.0], # norm = 1.0 [0.0, 2.0, 0.0], # norm = 2.0 ], dtype='float32', ) weight.stop_gradient = False out = paddle.nn.functional.embedding( x, weight, max_norm=5.0, norm_type=2.0 ) self.assertEqual(list(out.shape), [3, 3]) # 第一行的范数应被裁剪到不超过5.0 # First row norm should be clipped to <= 5.0 row0_norm = float(paddle.norm(out[0], p=2).item()) self.assertLessEqual(row0_norm, 5.0 + 1e-3) def test_embedding_with_negative_padding_idx(self): """测试embedding的负padding_idx Test embedding with negative padding_idx""" x = paddle.to_tensor([0, 1, 4], dtype='int64') weight = paddle.full(shape=(5, 3), fill_value=2.0, dtype='float32') out = paddle.nn.functional.embedding(x, weight, padding_idx=-1) self.assertEqual(list(out.shape), [3, 3]) # padding_idx=-1 → 实际index=4, 输出应全0 # padding_idx=-1 → actual index=4, output should be all zeros np.testing.assert_array_equal(out[2].numpy(), [0.0, 0.0, 0.0]) def test_embedding_with_scale_grad_by_freq(self): """测试embedding的scale_grad_by_freq参数 Test embedding scale_grad_by_freq parameter""" x = paddle.to_tensor([0, 0, 1, 2], dtype='int64') weight = paddle.randn([5, 3]) weight.stop_gradient = False out = paddle.nn.functional.embedding( x, weight, scale_grad_by_freq=True, sparse=False ) self.assertEqual(list(out.shape), [4, 3]) # 验证可以正常前向计算 / Verify forward pass works loss = out.sum() loss.backward() def test_embedding_padding_idx_out_of_range(self): """测试embedding的padding_idx超出范围应报错 Test embedding with padding_idx out of range should raise""" x = paddle.to_tensor([0, 1], dtype='int64') weight = paddle.randn([5, 3]) with self.assertRaises(ValueError): paddle.nn.functional.embedding(x, weight, padding_idx=10) if __name__ == '__main__': unittest.main()