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
飞不过沧海
Paddle
提交
31ccaf09
P
Paddle
项目概览
飞不过沧海
/
Paddle
与 Fork 源项目一致
Fork自
PaddlePaddle / Paddle
通知
1
Star
1
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 >>
提交
31ccaf09
编写于
3月 11, 2019
作者:
L
luotao1
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
add all_kernels_must_compute_runtime_shape example for speedup infershape
test=develop
上级
ad80bde8
变更
4
隐藏空白更改
内联
并排
Showing
4 changed file
with
33 addition
and
11 deletion
+33
-11
paddle/fluid/framework/operator.cc
paddle/fluid/framework/operator.cc
+9
-2
paddle/fluid/operators/fused/fused_embedding_seq_pool_op.cc
paddle/fluid/operators/fused/fused_embedding_seq_pool_op.cc
+8
-3
paddle/fluid/operators/hash_op.cc
paddle/fluid/operators/hash_op.cc
+8
-3
paddle/fluid/operators/sequence_ops/sequence_enumerate_op.cc
paddle/fluid/operators/sequence_ops/sequence_enumerate_op.cc
+8
-3
未找到文件。
paddle/fluid/framework/operator.cc
浏览文件 @
31ccaf09
...
...
@@ -926,8 +926,15 @@ void OperatorWithKernel::RunImpl(const Scope& scope,
dev_ctx
=
pool
.
Get
(
expected_kernel_key
.
place_
);
}
RuntimeInferShapeContext
infer_shape_ctx
(
*
this
,
exec_scope
,
ctx
);
this
->
InferShape
(
&
infer_shape_ctx
);
// If Op has attribute all_kernels_must_compute_runtime_shape,
// all the kernels of this Op would compute runtime shape,
// and skip infershape in runtime for speedup.
// TODO(luotao): Note that it is a temporal attribute, after all ops
// implement computing runtime shape, this attribute would be deleted.
if
(
!
HasAttr
(
"all_kernels_must_compute_runtime_shape"
))
{
RuntimeInferShapeContext
infer_shape_ctx
(
*
this
,
exec_scope
,
ctx
);
this
->
InferShape
(
&
infer_shape_ctx
);
}
// TODO(panyx0718): ExecutionContext should only depend on RuntimeContext
// not Scope. Imperative mode only pass inputs and get outputs.
kernel_iter
->
second
(
...
...
paddle/fluid/operators/fused/fused_embedding_seq_pool_op.cc
浏览文件 @
31ccaf09
...
...
@@ -23,9 +23,6 @@ class FusedEmbeddingSeqPoolOp : public framework::OperatorWithKernel {
using
framework
::
OperatorWithKernel
::
OperatorWithKernel
;
void
InferShape
(
framework
::
InferShapeContext
*
ctx
)
const
override
{
if
(
ctx
->
IsRuntime
())
{
return
;
}
PADDLE_ENFORCE
(
ctx
->
HasInput
(
"W"
),
"Input W of FusedEmbeddingSeqPoolOp should not be null."
);
PADDLE_ENFORCE
(
ctx
->
HasInput
(
"Ids"
),
...
...
@@ -91,6 +88,14 @@ class FusedEmbeddingSeqPoolOpMaker : public framework::OpProtoAndCheckerMaker {
"(boolean, default false) "
"Sparse update."
)
.
SetDefault
(
false
);
AddAttr
<
bool
>
(
"all_kernels_must_compute_runtime_shape"
,
"(boolean, default true) "
"An attribute to speed up OperatorWithKernel::RunImpl."
"If true, all the kernels of this Op would compute runtime "
"shape, but skip infershape in runtime. Note that it is a temporal "
"attribute, please do DOT set it in python layer."
)
.
SetDefault
(
true
);
AddComment
(
R"DOC(
FusedEmbeddingSeqPool Operator.
...
...
paddle/fluid/operators/hash_op.cc
浏览文件 @
31ccaf09
...
...
@@ -26,9 +26,6 @@ class HashOp : public framework::OperatorWithKernel {
:
OperatorWithKernel
(
type
,
inputs
,
outputs
,
attrs
)
{}
void
InferShape
(
framework
::
InferShapeContext
*
ctx
)
const
override
{
if
(
ctx
->
IsRuntime
())
{
return
;
}
PADDLE_ENFORCE
(
ctx
->
HasInput
(
"X"
),
"Input(X) of HashOp should not be null."
);
PADDLE_ENFORCE
(
ctx
->
HasOutput
(
"Out"
),
...
...
@@ -57,6 +54,14 @@ $$Out = scale * X$$
)DOC"
);
AddAttr
<
int
>
(
"num_hash"
,
""
).
SetDefault
(
1
);
AddAttr
<
int
>
(
"mod_by"
,
""
).
SetDefault
(
100000
);
AddAttr
<
bool
>
(
"all_kernels_must_compute_runtime_shape"
,
"(boolean, default true) "
"An attribute to speed up OperatorWithKernel::RunImpl."
"If true, all the kernels of this Op would compute runtime "
"shape, but skip infershape in runtime. Note that it is a temporal "
"attribute, please do DOT set it in python layer."
)
.
SetDefault
(
true
);
}
};
...
...
paddle/fluid/operators/sequence_ops/sequence_enumerate_op.cc
浏览文件 @
31ccaf09
...
...
@@ -22,9 +22,6 @@ class SequenceEnumerateOp : public framework::OperatorWithKernel {
using
framework
::
OperatorWithKernel
::
OperatorWithKernel
;
void
InferShape
(
framework
::
InferShapeContext
*
ctx
)
const
override
{
if
(
ctx
->
IsRuntime
())
{
return
;
}
PADDLE_ENFORCE
(
ctx
->
HasInput
(
"X"
),
"Input(X) of SequecceEnumerate operator should not be null."
);
...
...
@@ -62,6 +59,14 @@ class SequenceEnumerateOpMaker : public framework::OpProtoAndCheckerMaker {
});
AddAttr
<
int
>
(
"pad_value"
,
"(int) The enumerate sequence padding value."
)
.
SetDefault
(
0
);
AddAttr
<
bool
>
(
"all_kernels_must_compute_runtime_shape"
,
"(boolean, default true) "
"An attribute to speed up OperatorWithKernel::RunImpl."
"If true, all the kernels of this Op would compute runtime "
"shape, but skip infershape in runtime. Note that it is a temporal "
"attribute, please do DOT set it in python layer."
)
.
SetDefault
(
true
);
AddComment
(
R"DOC(
Sequence Enumerate Operator.
...
...
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