Skip to content
体验新版
项目
组织
正在加载...
登录
切换导航
打开侧边栏
wmsofts
Paddle
提交
64e2f10c
P
Paddle
项目概览
wmsofts
/
Paddle
与 Fork 源项目一致
Fork自
PaddlePaddle / Paddle
通知
1
Star
0
Fork
0
代码
文件
提交
分支
Tags
贡献者
分支图
Diff
Issue
0
列表
看板
标记
里程碑
合并请求
0
Wiki
0
Wiki
分析
仓库
DevOps
项目成员
Pages
P
Paddle
项目概览
项目概览
详情
发布
仓库
仓库
文件
提交
分支
标签
贡献者
分支图
比较
Issue
0
Issue
0
列表
看板
标记
里程碑
合并请求
0
合并请求
0
Pages
分析
分析
仓库分析
DevOps
Wiki
0
Wiki
成员
成员
收起侧边栏
关闭侧边栏
动态
分支图
创建新Issue
提交
Issue看板
体验新版 GitCode,发现更多精彩内容 >>
未验证
提交
64e2f10c
编写于
6月 15, 2022
作者:
Z
zhangkaihuo
提交者:
GitHub
6月 15, 2022
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
Conv3 support bias (#43458)
上级
81abaaf5
变更
2
隐藏空白更改
内联
并排
Showing
2 changed file
with
44 addition
and
3 deletion
+44
-3
python/paddle/fluid/tests/unittests/test_sparse_conv_op.py
python/paddle/fluid/tests/unittests/test_sparse_conv_op.py
+41
-0
python/paddle/incubate/sparse/nn/layer/conv.py
python/paddle/incubate/sparse/nn/layer/conv.py
+3
-3
未找到文件。
python/paddle/fluid/tests/unittests/test_sparse_conv_op.py
浏览文件 @
64e2f10c
...
...
@@ -117,3 +117,44 @@ class TestSparseConv(unittest.TestCase):
#Currently, only support data_format='NDHWC'
conv3d
=
paddle
.
incubate
.
sparse
.
nn
.
SubmConv3D
(
1
,
1
,
(
1
,
3
,
3
),
data_format
=
'NCDHW'
)
def
test_Conv3D_bias
(
self
):
with
_test_eager_guard
():
paddle
.
seed
(
0
)
shape
=
[
1
,
4
,
4
,
4
,
3
]
x
=
paddle
.
randn
(
shape
)
sp_x
=
x
.
to_sparse_coo
(
4
)
conv3d
=
paddle
.
nn
.
Conv3D
(
3
,
2
,
3
,
data_format
=
'NDHWC'
)
sp_conv3d
=
paddle
.
incubate
.
sparse
.
nn
.
Conv3D
(
3
,
2
,
3
,
data_format
=
'NDHWC'
)
sp_conv3d
.
weight
.
set_value
(
paddle
.
to_tensor
(
conv3d
.
weight
.
numpy
().
transpose
(
2
,
3
,
4
,
1
,
0
)))
sp_conv3d
.
bias
.
set_value
(
paddle
.
to_tensor
(
conv3d
.
bias
.
numpy
()))
x
.
stop_gradient
=
False
out
=
conv3d
(
x
)
loss
=
out
.
mean
()
loss
.
backward
()
sp_x
.
stop_gradient
=
False
sp_out
=
sp_conv3d
(
sp_x
)
dense_out
=
sp_out
.
to_dense
()
sp_loss
=
dense_out
.
mean
()
sp_loss
.
backward
()
assert
np
.
allclose
(
out
.
numpy
(),
dense_out
.
numpy
(),
atol
=
1e-3
,
rtol
=
1e-3
)
assert
np
.
allclose
(
conv3d
.
weight
.
grad
.
numpy
().
transpose
(
2
,
3
,
4
,
1
,
0
),
sp_conv3d
.
weight
.
grad
.
numpy
(),
atol
=
1e-3
,
rtol
=
1e-3
)
assert
np
.
allclose
(
conv3d
.
bias
.
grad
.
numpy
(),
sp_conv3d
.
bias
.
grad
.
numpy
(),
atol
=
1e-5
,
rtol
=
1e-5
)
python/paddle/incubate/sparse/nn/layer/conv.py
浏览文件 @
64e2f10c
...
...
@@ -82,9 +82,9 @@ class _Conv3D(Layer):
shape
=
filter_shape
,
attr
=
self
.
_param_attr
,
default_initializer
=
_get_default_param_initializer
())
#self.bias = self.create_parameter(
# attr=self._bias_attr, shape=[self._out_channels], is_bias=True)
self
.
bias
=
None
self
.
bias
=
self
.
create_parameter
(
attr
=
self
.
_bias_attr
,
shape
=
[
self
.
_out_channels
],
is_bias
=
True
)
def
forward
(
self
,
x
):
out
=
F
.
conv
.
_conv3d
(
x
,
...
...
编辑
预览
Markdown
is supported
0%
请重试
或
添加新附件
.
添加附件
取消
You are about to add
0
people
to the discussion. Proceed with caution.
先完成此消息的编辑!
取消
想要评论请
注册
或
登录