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
机器未来
Paddle
提交
a6482258
P
Paddle
项目概览
机器未来
/
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
1
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
1
Issue
1
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
未验证
提交
a6482258
编写于
1月 04, 2021
作者:
G
GaoWei8
提交者:
GitHub
1月 04, 2021
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
add REGISTER_OP_VERSION for LSTM (#30038)
上级
6e93fb92
变更
1
隐藏空白更改
内联
并排
Showing
1 changed file
with
24 addition
and
2 deletion
+24
-2
paddle/fluid/operators/cudnn_lstm_op.cc
paddle/fluid/operators/cudnn_lstm_op.cc
+24
-2
未找到文件。
paddle/fluid/operators/cudnn_lstm_op.cc
浏览文件 @
a6482258
...
@@ -309,7 +309,9 @@ REGISTER_OP_CPU_KERNEL(cudnn_lstm_grad, ops::NotImpleKernel<float>);
...
@@ -309,7 +309,9 @@ REGISTER_OP_CPU_KERNEL(cudnn_lstm_grad, ops::NotImpleKernel<float>);
REGISTER_OP_VERSION
(
cudnn_lstm
)
REGISTER_OP_VERSION
(
cudnn_lstm
)
.
AddCheckpoint
(
.
AddCheckpoint
(
R"ROC(
R"ROC(
Upgrade cudnn_lstm add a new input [WeightList] and modify input [W] to dispensable.)ROC"
,
Upgrade cudnn_lstm add new inputs [WeightList, SequenceLength], modify the input [W] to dispensable, delete the input [Cache].
Upgrade cudnn_lstm add new outputs [StateOut, Reserve, LastC, LastH], delete output [last_c, last_h].
Upgrade cudnn_lstm modify the attr [seed] default value to 0, delete the attr [max_len].)ROC"
,
paddle
::
framework
::
compatible
::
OpVersionDesc
()
paddle
::
framework
::
compatible
::
OpVersionDesc
()
.
NewInput
(
.
NewInput
(
"WeightList"
,
"WeightList"
,
...
@@ -318,6 +320,26 @@ REGISTER_OP_VERSION(cudnn_lstm)
...
@@ -318,6 +320,26 @@ REGISTER_OP_VERSION(cudnn_lstm)
.
NewInput
(
"SequenceLength"
,
.
NewInput
(
"SequenceLength"
,
"When the input data is padding, set this parameter. "
"When the input data is padding, set this parameter. "
"SequenceLength is dispensable."
)
"SequenceLength is dispensable."
)
.
ModifyInput
(
"W"
,
"The new LSTM use WeightList instead of W. The W "
"concatenate all the weight to one Tensor."
)
.
DeleteInput
(
"Cache"
,
"The new LSTM use the Reserve Output to store the "
"data of dropout."
)
.
NewOutput
(
"StateOut"
,
"Store the global drop state when training"
)
.
NewOutput
(
"StateOut"
,
"Store the global drop state when training"
)
.
NewOutput
(
"Reserve"
,
.
NewOutput
(
"Reserve"
,
"A temporary output Tensor to store the reserve_data"
));
"A temporary output Tensor to store the reserve_data"
)
.
DeleteOutput
(
"last_c"
,
"Modify the name of the output from 'last_c' to 'LastC'."
)
.
NewOutput
(
"LastC"
,
"The cell state of the last step."
)
.
DeleteOutput
(
"last_h"
,
"Modify the name of the output from 'last_h' to 'LastH'."
)
.
NewOutput
(
"LastH"
,
"The hidden state of the last step."
)
.
ModifyAttr
(
"seed"
,
"Set the default value of seed from '-1' to '0'."
,
0
)
.
DeleteAttr
(
"max_len"
,
"The length of Inputs is achieved form the input data "
"which is difficult to know the information in "
"advance."
));
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