wehub-resource-sync
227 lines
8.2 KiB
Python
227 lines
8.2 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.
|
|
|
|
# Unit test for paddle.nn.layer.norm (InstanceNorm, GroupNorm, BatchNorm layers)
|
|
# Target: cover _InstanceNormBase._check_input_dim, GroupNorm invalid data_format,
|
|
# _BatchNormBase extra_repr, _check_input_dim, _check_data_format
|
|
|
|
import unittest
|
|
|
|
import paddle
|
|
from paddle import nn
|
|
from paddle.nn.layer.norm import _BatchNormBase, _InstanceNormBase
|
|
|
|
|
|
class TestInstanceNormBaseCheckInputDim(unittest.TestCase):
|
|
"""Test that _InstanceNormBase._check_input_dim raises NotImplementedError.
|
|
The subclass InstanceNorm1D overrides this, so we must call the base class method.
|
|
"""
|
|
|
|
def test_base_check_input_dim_raises(self):
|
|
"""Base class _check_input_dim should raise NotImplementedError."""
|
|
layer = _InstanceNormBase(3)
|
|
with self.assertRaises(NotImplementedError):
|
|
layer._check_input_dim(paddle.randn([2, 3, 10]))
|
|
|
|
|
|
class TestInstanceNormInputDimCheck(unittest.TestCase):
|
|
"""Test InstanceNorm1D checks input dimension via forward."""
|
|
|
|
def test_instance_norm1d_wrong_dim_raises(self):
|
|
"""InstanceNorm1D should reject non-2D/3D input."""
|
|
layer = nn.InstanceNorm1D(3)
|
|
# 4D input should be rejected by InstanceNorm1D._check_input_dim
|
|
x = paddle.randn([2, 3, 4, 4])
|
|
with self.assertRaises(ValueError):
|
|
layer(x)
|
|
|
|
|
|
class TestGroupNormErrorPaths(unittest.TestCase):
|
|
"""Test GroupNorm error paths."""
|
|
|
|
def test_invalid_data_format(self):
|
|
"""Invalid data_format should raise ValueError."""
|
|
with self.assertRaises(ValueError):
|
|
nn.GroupNorm(2, 4, data_format='INVALID')
|
|
|
|
def test_group_norm_nchw(self):
|
|
"""GroupNorm with NCHW format."""
|
|
layer = nn.GroupNorm(2, 4, data_format='NCHW')
|
|
x = paddle.randn([2, 4, 8, 8])
|
|
out = layer(x)
|
|
self.assertEqual(out.shape, [2, 4, 8, 8])
|
|
|
|
def test_group_norm_nhwc(self):
|
|
"""GroupNorm with NHWC format."""
|
|
layer = nn.GroupNorm(2, 4, data_format='NHWC')
|
|
x = paddle.randn([2, 8, 8, 4])
|
|
out = layer(x)
|
|
self.assertEqual(out.shape, [2, 8, 8, 4])
|
|
|
|
def test_group_norm_ncl(self):
|
|
"""GroupNorm with NCL format (1D)."""
|
|
layer = nn.GroupNorm(2, 4, data_format='NCL')
|
|
x = paddle.randn([2, 4, 10])
|
|
out = layer(x)
|
|
self.assertEqual(out.shape, [2, 4, 10])
|
|
|
|
def test_group_norm_no_affine(self):
|
|
"""GroupNorm without affine transformation."""
|
|
layer = nn.GroupNorm(2, 4, affine=False)
|
|
self.assertIsNone(layer.weight)
|
|
self.assertIsNone(layer.bias)
|
|
|
|
def test_group_norm_with_affine(self):
|
|
"""GroupNorm with affine transformation."""
|
|
layer = nn.GroupNorm(2, 4, affine=True)
|
|
self.assertIsNotNone(layer.weight)
|
|
self.assertIsNotNone(layer.bias)
|
|
|
|
|
|
class TestBatchNormBaseErrorPaths(unittest.TestCase):
|
|
"""Test _BatchNormBase error paths and extra_repr."""
|
|
|
|
def test_batch_norm_base_check_input_dim_raises(self):
|
|
"""_BatchNormBase._check_input_dim should raise NotImplementedError."""
|
|
layer = _BatchNormBase(3)
|
|
with self.assertRaises(NotImplementedError):
|
|
layer._check_input_dim(paddle.randn([2, 3, 4, 4]))
|
|
|
|
def test_batch_norm_base_check_data_format_raises(self):
|
|
"""_BatchNormBase._check_data_format should raise NotImplementedError."""
|
|
layer = _BatchNormBase(3)
|
|
with self.assertRaises(NotImplementedError):
|
|
layer._check_data_format('NCHW')
|
|
|
|
def test_batch_norm_base_extra_repr(self):
|
|
"""_BatchNormBase extra_repr should contain num_features, momentum, epsilon."""
|
|
layer = _BatchNormBase(3, momentum=0.9, epsilon=1e-5)
|
|
repr_str = layer.extra_repr()
|
|
self.assertIn('num_features=3', repr_str)
|
|
self.assertIn('momentum=0.9', repr_str)
|
|
self.assertIn('epsilon=1e-05', repr_str)
|
|
|
|
def test_batch_norm_base_extra_repr_no_name(self):
|
|
"""_BatchNormBase extra_repr without name should not include name."""
|
|
layer = _BatchNormBase(3)
|
|
repr_str = layer.extra_repr()
|
|
self.assertNotIn('name=', repr_str)
|
|
|
|
def test_batch_norm_base_extra_repr_nhwc(self):
|
|
"""_BatchNormBase extra_repr with NHWC data_format."""
|
|
layer = _BatchNormBase(3, data_format='NHWC')
|
|
repr_str = layer.extra_repr()
|
|
self.assertIn('NHWC', repr_str)
|
|
|
|
def test_batch_norm_base_extra_repr_with_name(self):
|
|
"""_BatchNormBase extra_repr with name parameter should include name."""
|
|
layer = _BatchNormBase(3, name='my_bn')
|
|
repr_str = layer.extra_repr()
|
|
self.assertIn('name=my_bn', repr_str)
|
|
|
|
def test_batch_norm_base_no_weight_bias(self):
|
|
"""_BatchNormBase with weight_attr=False and bias_attr=False."""
|
|
layer = _BatchNormBase(3, weight_attr=False, bias_attr=False)
|
|
self.assertIsNone(layer.weight)
|
|
self.assertIsNone(layer.bias)
|
|
|
|
|
|
class TestBatchNormSubclassForward(unittest.TestCase):
|
|
"""Test BatchNorm1D/2D/3D forward passes."""
|
|
|
|
def test_batch_norm2d_basic(self):
|
|
"""BatchNorm2D basic forward."""
|
|
layer = nn.BatchNorm2D(3)
|
|
x = paddle.randn([2, 3, 4, 4])
|
|
out = layer(x)
|
|
self.assertEqual(out.shape, [2, 3, 4, 4])
|
|
|
|
def test_batch_norm1d_basic(self):
|
|
"""BatchNorm1D basic forward."""
|
|
layer = nn.BatchNorm1D(3)
|
|
x = paddle.randn([2, 3, 10])
|
|
out = layer(x)
|
|
self.assertEqual(out.shape, [2, 3, 10])
|
|
|
|
def test_batch_norm3d_basic(self):
|
|
"""BatchNorm3D basic forward."""
|
|
layer = nn.BatchNorm3D(3)
|
|
x = paddle.randn([2, 3, 4, 4, 4])
|
|
out = layer(x)
|
|
self.assertEqual(out.shape, [2, 3, 4, 4, 4])
|
|
|
|
def test_batch_norm2d_eval_mode(self):
|
|
"""BatchNorm2D in eval mode."""
|
|
layer = nn.BatchNorm2D(3)
|
|
layer.eval()
|
|
x = paddle.randn([2, 3, 4, 4])
|
|
out = layer(x)
|
|
self.assertEqual(out.shape, [2, 3, 4, 4])
|
|
|
|
def test_batch_norm2d_momentum(self):
|
|
"""BatchNorm2D with custom momentum."""
|
|
layer = nn.BatchNorm2D(3, momentum=0.1)
|
|
x = paddle.randn([2, 3, 4, 4])
|
|
out = layer(x)
|
|
self.assertEqual(out.shape, [2, 3, 4, 4])
|
|
|
|
def test_batch_norm2d_no_weight_bias(self):
|
|
"""BatchNorm2D without weight and bias."""
|
|
layer = nn.BatchNorm2D(3, weight_attr=False, bias_attr=False)
|
|
self.assertIsNone(layer.weight)
|
|
self.assertIsNone(layer.bias)
|
|
|
|
|
|
class TestInstanceNormLayers(unittest.TestCase):
|
|
"""Test InstanceNorm layers."""
|
|
|
|
def test_instance_norm1d_basic(self):
|
|
"""InstanceNorm1D basic forward."""
|
|
layer = nn.InstanceNorm1D(3)
|
|
x = paddle.randn([2, 3, 10])
|
|
out = layer(x)
|
|
self.assertEqual(out.shape, [2, 3, 10])
|
|
|
|
def test_instance_norm2d_basic(self):
|
|
"""InstanceNorm2D basic forward."""
|
|
layer = nn.InstanceNorm2D(3)
|
|
x = paddle.randn([2, 3, 4, 4])
|
|
out = layer(x)
|
|
self.assertEqual(out.shape, [2, 3, 4, 4])
|
|
|
|
def test_instance_norm3d_basic(self):
|
|
"""InstanceNorm3D basic forward."""
|
|
layer = nn.InstanceNorm3D(3)
|
|
x = paddle.randn([2, 3, 4, 4, 4])
|
|
out = layer(x)
|
|
self.assertEqual(out.shape, [2, 3, 4, 4, 4])
|
|
|
|
def test_instance_norm_no_affine(self):
|
|
"""InstanceNorm without affine transformation."""
|
|
layer = nn.InstanceNorm1D(3, weight_attr=False, bias_attr=False)
|
|
self.assertIsNone(layer.scale)
|
|
self.assertIsNone(layer.bias)
|
|
|
|
def test_instance_norm_eval(self):
|
|
"""InstanceNorm in eval mode."""
|
|
layer = nn.InstanceNorm1D(3)
|
|
layer.eval()
|
|
x = paddle.randn([2, 3, 10])
|
|
out = layer(x)
|
|
self.assertEqual(out.shape, [2, 3, 10])
|
|
|
|
|
|
if __name__ == '__main__':
|
|
unittest.main()
|