wehub-resource-sync
235 lines
8.5 KiB
Python
235 lines
8.5 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.
|
|
|
|
# Unit test for paddle.nn.functional.pooling helper functions
|
|
# Target: cover uncovered helper functions and error paths in pooling.py
|
|
|
|
import unittest
|
|
|
|
import paddle
|
|
import paddle.nn.functional as F
|
|
from paddle.nn.functional.pooling import (
|
|
_channel_last,
|
|
_check_input,
|
|
_check_instance,
|
|
_check_value_limitation,
|
|
_update_padding_nd,
|
|
)
|
|
|
|
|
|
class TestCheckInput(unittest.TestCase):
|
|
"""Test _check_input dimension validation."""
|
|
|
|
def test_check_input_1d_ok(self):
|
|
"""1D tensor should pass for dimension=1."""
|
|
x = paddle.to_tensor([1.0, 2.0, 3.0])
|
|
_check_input(x, 1) # should not raise
|
|
|
|
def test_check_input_2d_ok(self):
|
|
"""2D tensor should pass for dimension=2."""
|
|
x = paddle.to_tensor([[1.0, 2.0]])
|
|
_check_input(x, 2)
|
|
|
|
def test_check_input_wrong_dimension(self):
|
|
"""Wrong dimension should raise ValueError."""
|
|
x = paddle.to_tensor([1.0, 2.0, 3.0])
|
|
with self.assertRaises(ValueError):
|
|
_check_input(x, 3)
|
|
|
|
|
|
class TestCheckInstance(unittest.TestCase):
|
|
"""Test _check_instance type validation."""
|
|
|
|
def test_check_instance_int_ok(self):
|
|
"""Integer value should pass."""
|
|
_check_instance(3, 'kernel_size', (int,))
|
|
|
|
def test_check_instance_float_ok(self):
|
|
"""Float value should pass."""
|
|
_check_instance(1.5, 'padding', (int, float))
|
|
|
|
def test_check_instance_wrong_type(self):
|
|
"""Wrong type should raise ValueError."""
|
|
with self.assertRaises(ValueError):
|
|
_check_instance("bad", 'kernel_size', (int, float))
|
|
|
|
|
|
class TestCheckValueLimitation(unittest.TestCase):
|
|
"""Test _check_value_limitation min value check.
|
|
The function only raises for int values where x < min_limit.
|
|
"""
|
|
|
|
def test_check_value_above_min(self):
|
|
"""Values above min_limit should pass."""
|
|
_check_value_limitation([2, 3], 'kernel_size', min_limit=1)
|
|
|
|
def test_check_value_below_min(self):
|
|
"""Int value below min_limit should raise ValueError."""
|
|
with self.assertRaises(ValueError):
|
|
_check_value_limitation([0], 'kernel_size', min_limit=1)
|
|
|
|
def test_check_value_equal_min(self):
|
|
"""Int value equal to min_limit should NOT raise (only x < min_limit raises)."""
|
|
_check_value_limitation([1], 'kernel_size', min_limit=1)
|
|
|
|
def test_check_value_float_not_checked(self):
|
|
"""Float values are not checked by _check_value_limitation (only ints)."""
|
|
_check_value_limitation([0.001], 'kernel_size', min_limit=1)
|
|
|
|
|
|
class TestChannelLast(unittest.TestCase):
|
|
"""Test _channel_last for different data_format and num_dims."""
|
|
|
|
def test_1d_nlc(self):
|
|
"""1D NLC format should be channel_last."""
|
|
self.assertTrue(_channel_last('NLC', 1))
|
|
|
|
def test_1d_ncl(self):
|
|
"""1D NCL format should not be channel_last."""
|
|
self.assertFalse(_channel_last('NCL', 1))
|
|
|
|
def test_2d_nhwc(self):
|
|
"""2D NHWC format should be channel_last."""
|
|
self.assertTrue(_channel_last('NHWC', 2))
|
|
|
|
def test_2d_nchw(self):
|
|
"""2D NCHW format should not be channel_last."""
|
|
self.assertFalse(_channel_last('NCHW', 2))
|
|
|
|
def test_3d_ndhwc(self):
|
|
"""3D NDHWC format should be channel_last."""
|
|
self.assertTrue(_channel_last('NDHWC', 3))
|
|
|
|
def test_3d_ncdhw(self):
|
|
"""3D NCDHW format should not be channel_last."""
|
|
self.assertFalse(_channel_last('NCDHW', 3))
|
|
|
|
def test_invalid_1d_format(self):
|
|
"""Invalid 1D format should raise ValueError."""
|
|
with self.assertRaises(ValueError):
|
|
_channel_last('NCHW', 1)
|
|
|
|
def test_invalid_2d_format(self):
|
|
"""Invalid 2D format should raise ValueError."""
|
|
with self.assertRaises(ValueError):
|
|
_channel_last('NCL', 2)
|
|
|
|
def test_invalid_3d_format(self):
|
|
"""Invalid 3D format should raise ValueError."""
|
|
with self.assertRaises(ValueError):
|
|
_channel_last('NCHW', 3)
|
|
|
|
|
|
class TestUpdatePaddingNd(unittest.TestCase):
|
|
"""Test _update_padding_nd logic.
|
|
_update_padding_nd returns (padding, padding_algorithm) tuple,
|
|
where padding is first and padding_algorithm is second.
|
|
"""
|
|
|
|
def test_same_padding(self):
|
|
"""SAME padding should return padding_algorithm='SAME'."""
|
|
result = _update_padding_nd('SAME', num_dims=2)
|
|
self.assertEqual(result[1], 'SAME')
|
|
|
|
def test_valid_padding(self):
|
|
"""VALID padding should return padding_algorithm='VALID'."""
|
|
result = _update_padding_nd('VALID', num_dims=2)
|
|
self.assertEqual(result[1], 'VALID')
|
|
|
|
def test_valid_padding_with_ceil_mode(self):
|
|
"""VALID padding with ceil_mode=True should raise ValueError."""
|
|
with self.assertRaises(ValueError):
|
|
_update_padding_nd('VALID', num_dims=2, ceil_mode=True)
|
|
|
|
def test_unknown_padding_string(self):
|
|
"""Unknown padding string should raise ValueError."""
|
|
with self.assertRaises(ValueError):
|
|
_update_padding_nd('UNKNOWN', num_dims=2)
|
|
|
|
def test_explicit_padding_with_batch_channel_channel_last(self):
|
|
"""Explicit padding including batch and channel dims (channel_last=True)."""
|
|
padding = [[0, 0], [1, 1], [0, 0]]
|
|
result = _update_padding_nd(padding, num_dims=1, channel_last=True)
|
|
self.assertEqual(result[1], 'EXPLICIT')
|
|
|
|
def test_non_zero_batch_padding_raises(self):
|
|
"""Non-zero batch/channel padding should raise ValueError."""
|
|
padding = [[1, 0], [1, 1], [0, 0]]
|
|
with self.assertRaises(ValueError):
|
|
_update_padding_nd(padding, num_dims=1, channel_last=True)
|
|
|
|
def test_explicit_padding_no_batch_channel(self):
|
|
"""Explicit padding without batch/channel dims."""
|
|
result = _update_padding_nd([1, 2], num_dims=2)
|
|
self.assertEqual(result[1], 'EXPLICIT')
|
|
self.assertEqual(result[0], [1, 2])
|
|
|
|
def test_same_padding_produces_zero_padding(self):
|
|
"""SAME padding should produce zero padding list."""
|
|
result = _update_padding_nd('SAME', num_dims=2)
|
|
self.assertEqual(result[0], [0, 0])
|
|
|
|
|
|
class TestPool2DErrorPaths(unittest.TestCase):
|
|
"""Test avg_pool2d / max_pool2d / avg_pool3d / max_pool3d operations."""
|
|
|
|
def setUp(self):
|
|
paddle.disable_static()
|
|
|
|
def test_avg_pool2d_basic(self):
|
|
"""Basic avg_pool2d should work correctly."""
|
|
x = paddle.randn([2, 3, 8, 8])
|
|
out = F.avg_pool2d(x, kernel_size=2, stride=2)
|
|
self.assertEqual(out.shape, [2, 3, 4, 4])
|
|
|
|
def test_avg_pool2d_adaptive(self):
|
|
"""Adaptive avg_pool2d should work."""
|
|
x = paddle.randn([2, 3, 8, 8])
|
|
out = F.adaptive_avg_pool2d(x, output_size=4)
|
|
self.assertEqual(out.shape, [2, 3, 4, 4])
|
|
|
|
def test_max_pool2d_basic(self):
|
|
"""Basic max_pool2d should work."""
|
|
x = paddle.randn([2, 3, 8, 8])
|
|
out = F.max_pool2d(x, kernel_size=2, stride=2)
|
|
self.assertEqual(out.shape, [2, 3, 4, 4])
|
|
|
|
def test_avg_pool3d_basic(self):
|
|
"""Basic avg_pool3d should work."""
|
|
x = paddle.randn([2, 3, 4, 4, 4])
|
|
out = F.avg_pool3d(x, kernel_size=2, stride=2)
|
|
self.assertEqual(out.shape, [2, 3, 2, 2, 2])
|
|
|
|
def test_max_pool3d_basic(self):
|
|
"""Basic max_pool3d should work."""
|
|
x = paddle.randn([2, 3, 4, 4, 4])
|
|
out = F.max_pool3d(x, kernel_size=2, stride=2)
|
|
self.assertEqual(out.shape, [2, 3, 2, 2, 2])
|
|
|
|
def test_avg_pool2d_adaptive_global(self):
|
|
"""Adaptive avg_pool2d with output_size=1 acts as global pooling."""
|
|
x = paddle.randn([2, 3, 8, 8])
|
|
out = F.adaptive_avg_pool2d(x, output_size=1)
|
|
self.assertEqual(out.shape, [2, 3, 1, 1])
|
|
|
|
def test_avg_pool2d_with_padding(self):
|
|
"""avg_pool2d with explicit padding."""
|
|
x = paddle.randn([2, 3, 8, 8])
|
|
out = F.avg_pool2d(x, kernel_size=3, stride=1, padding=1)
|
|
self.assertEqual(out.shape, [2, 3, 8, 8])
|
|
|
|
|
|
if __name__ == '__main__':
|
|
unittest.main()
|