229 lines
7.7 KiB
Python
229 lines
7.7 KiB
Python
# Copyright (c) 2022 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.
|
|
import copy
|
|
import unittest
|
|
|
|
import numpy as np
|
|
from op_test import is_custom_device
|
|
from utils import compare_legacy_with_pt
|
|
|
|
import paddle
|
|
from paddle import base, sparse
|
|
from paddle.sparse import nn
|
|
|
|
|
|
class TestSparseBatchNorm(unittest.TestCase):
|
|
def test(self):
|
|
paddle.seed(0)
|
|
channels = 4
|
|
shape = [2, 3, 6, 6, channels]
|
|
# there is no zero in dense_x
|
|
dense_x = paddle.randn(shape)
|
|
dense_x.stop_gradient = False
|
|
|
|
batch_norm = paddle.nn.BatchNorm3D(channels, data_format="NDHWC")
|
|
dense_y = batch_norm(dense_x)
|
|
dense_y.backward(dense_y)
|
|
|
|
sparse_dim = 4
|
|
dense_x2 = copy.deepcopy(dense_x)
|
|
dense_x2.stop_gradient = False
|
|
sparse_x = dense_x2.to_sparse_coo(sparse_dim)
|
|
sparse_x.retain_grads()
|
|
sparse_batch_norm = paddle.sparse.nn.BatchNorm(channels)
|
|
# set same params
|
|
sparse_batch_norm._mean.set_value(batch_norm._mean)
|
|
sparse_batch_norm._variance.set_value(batch_norm._variance)
|
|
sparse_batch_norm.weight.set_value(batch_norm.weight)
|
|
|
|
sparse_y = sparse_batch_norm(sparse_x)
|
|
# compare the result with dense batch_norm
|
|
np.testing.assert_allclose(
|
|
dense_y.flatten().numpy(),
|
|
sparse_y.values().flatten().numpy(),
|
|
atol=1e-5,
|
|
rtol=1e-5,
|
|
)
|
|
|
|
# test backward
|
|
sparse_y.backward(sparse_y)
|
|
np.testing.assert_allclose(
|
|
dense_x.grad.flatten().numpy(),
|
|
sparse_x.grad.values().flatten().numpy(),
|
|
atol=1e-5,
|
|
rtol=1e-5,
|
|
)
|
|
|
|
def test_error_layout(self):
|
|
with self.assertRaises(ValueError):
|
|
shape = [2, 3, 6, 6, 3]
|
|
x = paddle.randn(shape)
|
|
sparse_x = x.to_sparse_coo(4)
|
|
sparse_batch_norm = paddle.sparse.nn.BatchNorm(
|
|
3, data_format='NCDHW'
|
|
)
|
|
sparse_batch_norm(sparse_x)
|
|
|
|
def test2(self):
|
|
paddle.seed(123)
|
|
channels = 3
|
|
x_data = paddle.randn((1, 6, 6, 6, channels)).astype('float32')
|
|
dense_x = paddle.to_tensor(x_data)
|
|
sparse_x = dense_x.to_sparse_coo(4)
|
|
batch_norm = paddle.sparse.nn.BatchNorm(channels)
|
|
batch_norm_out = batch_norm(sparse_x)
|
|
dense_bn = paddle.nn.BatchNorm1D(channels)
|
|
dense_x = dense_x.reshape((-1, dense_x.shape[-1]))
|
|
dense_out = dense_bn(dense_x)
|
|
np.testing.assert_allclose(
|
|
dense_out.numpy(), batch_norm_out.values().numpy()
|
|
)
|
|
# [1, 6, 6, 6, 3]
|
|
|
|
def check(self, shape):
|
|
np.random.seed(0)
|
|
data = np.random.uniform(-0.01, 0.01, shape).astype("float32")
|
|
x = paddle.to_tensor(data)
|
|
x.stop_gradient = False
|
|
dim = len(shape)
|
|
data_format = "NHWC" if dim == 4 else "NDHWC"
|
|
if dim == 4:
|
|
bn = paddle.nn.BatchNorm2D(shape[-1], data_format=data_format)
|
|
else:
|
|
bn = paddle.nn.BatchNorm3D(shape[-1], data_format=data_format)
|
|
y = bn(x)
|
|
np.random.seed(5)
|
|
loss_data = np.random.uniform(-0.01, 0.01, y.shape).astype("float32")
|
|
loss = paddle.to_tensor(loss_data)
|
|
y.backward(loss)
|
|
|
|
sp_x = paddle.to_tensor(data).to_sparse_coo(dim - 1)
|
|
sp_x.stop_gradient = False
|
|
sp_bn = paddle.sparse.nn.BatchNorm(shape[-1], data_format=data_format)
|
|
sp_y = sp_bn(sp_x)
|
|
sp_loss = loss.to_sparse_coo(dim - 1)
|
|
sp_y.backward(sp_loss)
|
|
|
|
np.testing.assert_allclose(
|
|
sp_y.to_dense().numpy(), y.numpy(), rtol=1e-5
|
|
)
|
|
np.testing.assert_allclose(
|
|
sp_x.grad.to_dense().numpy(), x.grad.numpy(), rtol=1e-5
|
|
)
|
|
|
|
def test_nd(self):
|
|
# 2D
|
|
self.check([2, 8, 8, 3])
|
|
# 3D
|
|
self.check([2, 8, 8, 3, 4])
|
|
|
|
|
|
class TestSyncBatchNorm(unittest.TestCase):
|
|
def test_sync_batch_norm(self):
|
|
x = np.array(
|
|
[[[[0.3, 0.4], [0.3, 0.07]], [[0.83, 0.37], [0.18, 0.93]]]]
|
|
).astype('float32')
|
|
x = paddle.to_tensor(x)
|
|
sparse_x = x.to_sparse_coo(len(x.shape) - 1)
|
|
|
|
if paddle.is_compiled_with_cuda() or is_custom_device():
|
|
sparse_sync_bn = nn.SyncBatchNorm(2)
|
|
sparse_hidden = sparse_sync_bn(sparse_x)
|
|
|
|
dense_sync_bn = paddle.nn.SyncBatchNorm(2)
|
|
x = x.reshape((-1, x.shape[-1]))
|
|
dense_hidden = dense_sync_bn(x)
|
|
np.testing.assert_allclose(
|
|
sparse_hidden.values().numpy(), dense_hidden.numpy()
|
|
)
|
|
|
|
def test_convert(self):
|
|
base_model = paddle.nn.Sequential(
|
|
nn.Conv3D(3, 5, 3), nn.BatchNorm(5), nn.BatchNorm(5)
|
|
)
|
|
|
|
model = paddle.nn.Sequential(
|
|
nn.Conv3D(3, 5, 3),
|
|
nn.BatchNorm(5),
|
|
nn.BatchNorm(
|
|
5,
|
|
weight_attr=base.ParamAttr(name='bn.scale'),
|
|
bias_attr=base.ParamAttr(name='bn.bias'),
|
|
),
|
|
)
|
|
model = nn.SyncBatchNorm.convert_sync_batchnorm(model)
|
|
for idx, sublayer in enumerate(base_model.sublayers()):
|
|
if isinstance(sublayer, nn.BatchNorm):
|
|
self.assertEqual(isinstance(model[idx], nn.SyncBatchNorm), True)
|
|
|
|
|
|
class TestStatic(unittest.TestCase):
|
|
@compare_legacy_with_pt
|
|
def test(self):
|
|
paddle.enable_static()
|
|
main_program = paddle.base.Program()
|
|
startup_program = paddle.base.Program()
|
|
with paddle.base.program_guard(main_program, startup_program):
|
|
indices = paddle.static.data(
|
|
name='indices', shape=[4, 4], dtype='int32'
|
|
)
|
|
values = paddle.static.data(
|
|
name='values', shape=[4, 1], dtype='float32'
|
|
)
|
|
channels = 1
|
|
dense_shape = [1, 1, 3, 4, channels]
|
|
sp_x = sparse.sparse_coo_tensor(indices, values, dense_shape)
|
|
|
|
sparse_batch_norm = paddle.sparse.nn.BatchNorm(channels)
|
|
sp_y = sparse_batch_norm(sp_x)
|
|
out = sp_y.to_dense()
|
|
|
|
exe = paddle.static.Executor()
|
|
exe.run(startup_program)
|
|
indices_data = [
|
|
[0, 0, 0, 0],
|
|
[0, 0, 0, 0],
|
|
[0, 0, 1, 2],
|
|
[1, 3, 2, 3],
|
|
]
|
|
values_data = np.array([[1.0], [2.0], [3.0], [4.0]]).astype(
|
|
'float32'
|
|
)
|
|
fetch = exe.run(
|
|
main_program,
|
|
feed={
|
|
'indices': indices_data,
|
|
'values': values_data,
|
|
},
|
|
fetch_list=[out],
|
|
return_numpy=True,
|
|
)
|
|
correct_out = np.array(
|
|
[
|
|
[
|
|
[
|
|
[[0.0], [-1.3416353], [0.0], [-0.44721174]],
|
|
[[0.0], [0.0], [0.44721198], [0.0]],
|
|
[[0.0], [0.0], [0.0], [1.3416355]],
|
|
]
|
|
]
|
|
]
|
|
).astype('float32')
|
|
np.testing.assert_allclose(correct_out, fetch[0], rtol=1e-5)
|
|
paddle.disable_static()
|
|
|
|
|
|
if __name__ == "__main__":
|
|
unittest.main()
|