wehub-resource-sync
229 lines
7.2 KiB
Python
229 lines
7.2 KiB
Python
# 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.layer.rnn (RNN, LSTM, GRU)
|
|
# 自动生成的单测，覆盖 paddle.nn.layer.rnn 模块中未覆盖的代码路径
|
|
# Target: cover uncovered lines in paddle/python/paddle/nn/layer/rnn.py
|
|
# 目标：覆盖 RNN、LSTM、GRU 的各种初始化参数和前向传播路径
|
|
|
|
"""
|
|
This test covers the following modules and code paths:
|
|
这个测试覆盖以下模块和代码路径：
|
|
|
|
1. SimpleRNN - 初始化和前向传播 (各种参数组合)
|
|
2. LSTM - 初始化和前向传播 (bidirectional, layers)
|
|
3. GRU - 初始化和前向传播
|
|
4. RNNCell, LSTMCell, GRUCell - 基本功能
|
|
5. time_major 参数 / dropout 参数
|
|
"""
|
|
|
|
import unittest
|
|
|
|
import paddle
|
|
from paddle import nn
|
|
|
|
|
|
class TestSimpleRNN(unittest.TestCase):
|
|
"""Test SimpleRNN.
|
|
测试 SimpleRNN。
|
|
"""
|
|
|
|
def setUp(self):
|
|
paddle.disable_static()
|
|
|
|
def test_rnn_basic(self):
|
|
"""Basic SimpleRNN."""
|
|
rnn = nn.SimpleRNN(input_size=16, hidden_size=32)
|
|
x = paddle.randn([4, 10, 16])
|
|
out, h = rnn(x)
|
|
self.assertEqual(out.shape, [4, 10, 32])
|
|
self.assertEqual(h.shape, [1, 4, 32])
|
|
|
|
def test_rnn_multi_layer(self):
|
|
"""SimpleRNN with multiple layers."""
|
|
rnn = nn.SimpleRNN(input_size=16, hidden_size=32, num_layers=3)
|
|
x = paddle.randn([4, 10, 16])
|
|
out, h = rnn(x)
|
|
self.assertEqual(out.shape, [4, 10, 32])
|
|
self.assertEqual(h.shape, [3, 4, 32])
|
|
|
|
def test_rnn_bidirectional(self):
|
|
"""Bidirectional SimpleRNN."""
|
|
rnn = nn.SimpleRNN(
|
|
input_size=16, hidden_size=32, direction='bidirectional'
|
|
)
|
|
x = paddle.randn([4, 10, 16])
|
|
out, h = rnn(x)
|
|
self.assertEqual(out.shape, [4, 10, 64])
|
|
self.assertEqual(h.shape, [2, 4, 32])
|
|
|
|
def test_rnn_dropout(self):
|
|
"""SimpleRNN with dropout."""
|
|
rnn = nn.SimpleRNN(
|
|
input_size=16,
|
|
hidden_size=32,
|
|
num_layers=2,
|
|
dropout=0.1,
|
|
)
|
|
rnn.train()
|
|
x = paddle.randn([4, 10, 16])
|
|
out, h = rnn(x)
|
|
self.assertEqual(out.shape, [4, 10, 32])
|
|
|
|
def test_rnn_time_major(self):
|
|
"""SimpleRNN with time_major=True."""
|
|
rnn = nn.SimpleRNN(input_size=16, hidden_size=32, time_major=True)
|
|
x = paddle.randn([10, 4, 16])
|
|
out, h = rnn(x)
|
|
self.assertEqual(out.shape, [10, 4, 32])
|
|
|
|
def test_rnn_with_initial_state(self):
|
|
"""SimpleRNN with initial state."""
|
|
rnn = nn.SimpleRNN(input_size=16, hidden_size=32)
|
|
x = paddle.randn([4, 10, 16])
|
|
h0 = paddle.zeros([1, 4, 32])
|
|
out, h = rnn(x, h0)
|
|
self.assertEqual(out.shape, [4, 10, 32])
|
|
|
|
|
|
class TestLSTM(unittest.TestCase):
|
|
"""Test LSTM.
|
|
测试 LSTM。
|
|
"""
|
|
|
|
def setUp(self):
|
|
paddle.disable_static()
|
|
|
|
def test_lstm_basic(self):
|
|
"""Basic LSTM."""
|
|
lstm = nn.LSTM(input_size=16, hidden_size=32)
|
|
x = paddle.randn([4, 10, 16])
|
|
out, (h, c) = lstm(x)
|
|
self.assertEqual(out.shape, [4, 10, 32])
|
|
self.assertEqual(h.shape, [1, 4, 32])
|
|
self.assertEqual(c.shape, [1, 4, 32])
|
|
|
|
def test_lstm_multi_layer(self):
|
|
"""LSTM with multiple layers."""
|
|
lstm = nn.LSTM(input_size=16, hidden_size=32, num_layers=3)
|
|
x = paddle.randn([4, 10, 16])
|
|
out, (h, c) = lstm(x)
|
|
self.assertEqual(h.shape, [3, 4, 32])
|
|
|
|
def test_lstm_bidirectional(self):
|
|
"""Bidirectional LSTM."""
|
|
lstm = nn.LSTM(input_size=16, hidden_size=32, direction='bidirectional')
|
|
x = paddle.randn([4, 10, 16])
|
|
out, (h, c) = lstm(x)
|
|
self.assertEqual(out.shape, [4, 10, 64])
|
|
self.assertEqual(h.shape, [2, 4, 32])
|
|
|
|
def test_lstm_dropout(self):
|
|
"""LSTM with dropout."""
|
|
lstm = nn.LSTM(
|
|
input_size=16,
|
|
hidden_size=32,
|
|
num_layers=3,
|
|
dropout=0.1,
|
|
)
|
|
lstm.train()
|
|
x = paddle.randn([4, 10, 16])
|
|
out, _ = lstm(x)
|
|
self.assertEqual(out.shape, [4, 10, 32])
|
|
|
|
def test_lstm_time_major(self):
|
|
"""LSTM with time_major=True."""
|
|
lstm = nn.LSTM(input_size=16, hidden_size=32, time_major=True)
|
|
x = paddle.randn([10, 4, 16])
|
|
out, _ = lstm(x)
|
|
self.assertEqual(out.shape, [10, 4, 32])
|
|
|
|
|
|
class TestGRU(unittest.TestCase):
|
|
"""Test GRU.
|
|
测试 GRU。
|
|
"""
|
|
|
|
def setUp(self):
|
|
paddle.disable_static()
|
|
|
|
def test_gru_basic(self):
|
|
"""Basic GRU."""
|
|
gru = nn.GRU(input_size=16, hidden_size=32)
|
|
x = paddle.randn([4, 10, 16])
|
|
out, h = gru(x)
|
|
self.assertEqual(out.shape, [4, 10, 32])
|
|
|
|
def test_gru_bidirectional(self):
|
|
"""Bidirectional GRU."""
|
|
gru = nn.GRU(input_size=16, hidden_size=32, direction='bidirectional')
|
|
x = paddle.randn([4, 10, 16])
|
|
out, h = gru(x)
|
|
self.assertEqual(out.shape, [4, 10, 64])
|
|
|
|
def test_gru_multi_layer(self):
|
|
"""GRU with multiple layers."""
|
|
gru = nn.GRU(input_size=16, hidden_size=32, num_layers=3)
|
|
x = paddle.randn([4, 10, 16])
|
|
out, h = gru(x)
|
|
self.assertEqual(h.shape, [3, 4, 32])
|
|
|
|
|
|
class TestRNNCells(unittest.TestCase):
|
|
"""Test RNN cells.
|
|
测试 RNN 单元。
|
|
"""
|
|
|
|
def setUp(self):
|
|
paddle.disable_static()
|
|
|
|
def test_rnn_cell(self):
|
|
"""SimpleRNNCell - returns output and hidden state."""
|
|
cell = nn.SimpleRNNCell(input_size=16, hidden_size=32)
|
|
x = paddle.randn([4, 16])
|
|
h = paddle.zeros([4, 32])
|
|
out, _ = cell(x, h)
|
|
self.assertEqual(out.shape, [4, 32])
|
|
|
|
def test_lstm_cell(self):
|
|
"""LSTMCell - states is a list/tuple, forward returns (output, (h, c))."""
|
|
cell = nn.LSTMCell(input_size=16, hidden_size=32)
|
|
x = paddle.randn([4, 16])
|
|
h = paddle.zeros([4, 32])
|
|
c = paddle.zeros([4, 32])
|
|
out, (h_new, c_new) = cell(x, states=[h, c])
|
|
self.assertEqual(h_new.shape, [4, 32])
|
|
self.assertEqual(c_new.shape, [4, 32])
|
|
|
|
def test_gru_cell(self):
|
|
"""GRUCell - forward returns (h_new, new_states)."""
|
|
cell = nn.GRUCell(input_size=16, hidden_size=32)
|
|
x = paddle.randn([4, 16])
|
|
h = paddle.zeros([4, 32])
|
|
h_new, _ = cell(x, h)
|
|
self.assertEqual(h_new.shape, [4, 32])
|
|
|
|
def test_rnn_cell_sequence(self):
|
|
"""Process sequence with SimpleRNNCell."""
|
|
cell = nn.SimpleRNNCell(input_size=16, hidden_size=32)
|
|
x = paddle.randn([10, 4, 16])
|
|
h = paddle.zeros([4, 32])
|
|
for t in range(10):
|
|
h, _ = cell(x[t], h)
|
|
self.assertEqual(h.shape, [4, 32])
|
|
|
|
|
|
if __name__ == '__main__':
|
|
unittest.main()
|