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
PaddlePaddle
Paddle
提交
ba51e6ea
P
Paddle
项目概览
PaddlePaddle
/
Paddle
1 年多 前同步成功
通知
2305
Star
20932
Fork
5423
代码
文件
提交
分支
Tags
贡献者
分支图
Diff
Issue
1423
列表
看板
标记
里程碑
合并请求
543
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
1,423
Issue
1,423
列表
看板
标记
里程碑
合并请求
543
合并请求
543
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
提交
ba51e6ea
编写于
7年前
作者:
Q
qiaolongfei
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
fix style problem
上级
d6e8d5cd
3 合并请求
!11636
[IMPORTANT] MKLDNN layout: Support for sum operator
,
!2081
Release/0.10.0
,
!1761
support Beam search in v2 api
变更
5
隐藏空白更改
内联
并排
Showing
5 changed file
with
16 addition
and
15 deletion
+16
-15
demo/seqToseq/seqToseq_net.py
demo/seqToseq/seqToseq_net.py
+3
-2
python/paddle/v2/config_base.py
python/paddle/v2/config_base.py
+2
-1
python/paddle/v2/layer.py
python/paddle/v2/layer.py
+3
-4
python/paddle/v2/layers/__init__.py
python/paddle/v2/layers/__init__.py
+1
-1
python/paddle/v2/layers/beam_search.py
python/paddle/v2/layers/beam_search.py
+7
-7
未找到文件。
demo/seqToseq/seqToseq_net.py
浏览文件 @
ba51e6ea
...
...
@@ -131,8 +131,9 @@ def gru_encoder_decoder(data_conf,
decoder_group_name
=
"decoder_group"
group_inputs
=
[
StaticInput
(
input
=
encoded_vector
,
is_seq
=
True
),
StaticInput
(
input
=
encoded_proj
,
is_seq
=
True
)
StaticInput
(
input
=
encoded_vector
,
is_seq
=
True
),
StaticInput
(
input
=
encoded_proj
,
is_seq
=
True
)
]
if
not
is_generating
:
...
...
This diff is collapsed.
Click to expand it.
python/paddle/v2/config_base.py
浏览文件 @
ba51e6ea
...
...
@@ -114,7 +114,8 @@ class Layer(object):
# 4. parse myself and add myself into context.
ret_val
=
self
.
to_proto_impl
(
context
=
context
,
**
kwargs
)
if
self
.
context_name
()
is
not
None
and
self
.
context_name
()
not
in
context
:
if
self
.
context_name
()
is
not
None
and
self
.
context_name
(
)
not
in
context
:
context
[
self
.
context_name
()]
=
ret_val
# 5. parse children that should be pased after this layer.
...
...
This diff is collapsed.
Click to expand it.
python/paddle/v2/layer.py
浏览文件 @
ba51e6ea
...
...
@@ -292,7 +292,8 @@ class RecurrentLayerInput(Layer):
else
:
self
.
__parents__
=
parent_layers
.
values
()[
0
]
self
.
__recurrent_name__
=
recurrent_name
name
=
self
.
__parents__
[
index
].
name
if
index
>=
0
else
self
.
context_name
()
name
=
self
.
__parents__
[
index
].
name
if
index
>=
0
else
self
.
context_name
()
super
(
RecurrentLayerInput
,
self
).
__init__
(
name
=
name
,
parent_layers
=
parent_layers
)
...
...
@@ -402,9 +403,7 @@ def recurrent_group(step, input, name=None):
extra_input
=
None
if
len
(
non_static_inputs
)
==
0
:
extra_input
=
RecurrentLayerInput
(
recurrent_name
=
name
,
index
=-
1
,
parent_layers
=
{})
recurrent_name
=
name
,
index
=-
1
,
parent_layers
=
{})
def
__real_step__
(
*
args
):
rnn_input
=
list
(
args
)
...
...
This diff is collapsed.
Click to expand it.
python/paddle/v2/layers/__init__.py
浏览文件 @
ba51e6ea
import
beam_search
\ No newline at end of file
import
beam_search
This diff is collapsed.
Click to expand it.
python/paddle/v2/layers/beam_search.py
浏览文件 @
ba51e6ea
...
...
@@ -63,6 +63,7 @@ class RecurrentLayerGroupSetGeneratorV2(Layer):
def
use_context_name
(
self
):
return
True
@
wrap_name_default
()
def
beam_search
(
step
,
input
,
...
...
@@ -75,9 +76,10 @@ def beam_search(step,
if
num_results_per_sample
is
None
:
num_results_per_sample
=
beam_size
assert
num_results_per_sample
<=
beam_size
# logger.warning("num_results_per_sample should be less than beam_size")
# logger.warning("num_results_per_sample should be less than beam_size")
if
isinstance
(
input
,
paddle
.
layer
.
StaticInputV2
)
or
isinstance
(
input
,
BaseGeneratedInputV2
):
if
isinstance
(
input
,
paddle
.
layer
.
StaticInputV2
)
or
isinstance
(
input
,
BaseGeneratedInputV2
):
input
=
[
input
]
generated_input_index
=
-
1
...
...
@@ -107,8 +109,8 @@ def beam_search(step,
args
=
list
(
args
)
before_step_layer
=
gipt
.
before_real_step
()
before_step_layer
.
append_child
(
layer
=
generator
,
parent_names
=
[
before_step_layer
.
name
])
before_step_layer
.
append_child
(
layer
=
generator
,
parent_names
=
[
before_step_layer
.
name
])
args
.
insert
(
generated_input_index
,
before_step_layer
)
predict
=
gipt
.
after_real_step
(
step
(
*
args
))
...
...
@@ -125,8 +127,6 @@ def beam_search(step,
# name=name,
# is_generating=True)
tmp
=
paddle
.
layer
.
recurrent_group
(
step
=
__real_step__
,
input
=
real_input
,
name
=
name
)
step
=
__real_step__
,
input
=
real_input
,
name
=
name
)
return
tmp
This diff is collapsed.
Click to expand it.
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
反馈
建议
客服
返回
顶部