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
夏有凉风,冬有雪
Paddle
提交
3712822e
P
Paddle
项目概览
夏有凉风,冬有雪
/
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
体验新版 GitCode,发现更多精彩内容 >>
提交
3712822e
编写于
5月 25, 2017
作者:
Y
Yibing Liu
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
modify seq2seq demo to show gradient/error clipping
上级
b4b967a7
变更
1
隐藏空白更改
内联
并排
Showing
1 changed file
with
27 addition
and
5 deletion
+27
-5
demo/seqToseq/api_train_v2.py
demo/seqToseq/api_train_v2.py
+27
-5
未找到文件。
demo/seqToseq/api_train_v2.py
浏览文件 @
3712822e
...
...
@@ -21,9 +21,12 @@ def seqToseq_net(source_dict_dim, target_dict_dim, is_generating=False):
size
=
word_vector_dim
,
param_attr
=
paddle
.
attr
.
ParamAttr
(
name
=
'_source_language_embedding'
))
src_forward
=
paddle
.
networks
.
simple_gru
(
input
=
src_embedding
,
size
=
encoder_size
)
name
=
'src_forward_gru'
,
input
=
src_embedding
,
size
=
encoder_size
)
src_backward
=
paddle
.
networks
.
simple_gru
(
input
=
src_embedding
,
size
=
encoder_size
,
reverse
=
True
)
name
=
'src_backward_gru'
,
input
=
src_embedding
,
size
=
encoder_size
,
reverse
=
True
)
encoded_vector
=
paddle
.
layer
.
concat
(
input
=
[
src_forward
,
src_backward
])
#### Decoder
...
...
@@ -34,7 +37,9 @@ def seqToseq_net(source_dict_dim, target_dict_dim, is_generating=False):
backward_first
=
paddle
.
layer
.
first_seq
(
input
=
src_backward
)
with
paddle
.
layer
.
mixed
(
size
=
decoder_size
,
act
=
paddle
.
activation
.
Tanh
())
as
decoder_boot
:
name
=
"decoder_boot_mixed"
,
size
=
decoder_size
,
act
=
paddle
.
activation
.
Tanh
())
as
decoder_boot
:
decoder_boot
+=
paddle
.
layer
.
full_matrix_projection
(
input
=
backward_first
)
...
...
@@ -44,11 +49,17 @@ def seqToseq_net(source_dict_dim, target_dict_dim, is_generating=False):
name
=
'gru_decoder'
,
size
=
decoder_size
,
boot_layer
=
decoder_boot
)
context
=
paddle
.
networks
.
simple_attention
(
name
=
"simple_attention"
,
encoded_sequence
=
enc_vec
,
encoded_proj
=
enc_proj
,
decoder_state
=
decoder_mem
)
with
paddle
.
layer
.
mixed
(
size
=
decoder_size
*
3
)
as
decoder_inputs
:
with
paddle
.
layer
.
mixed
(
name
=
"input_recurrent"
,
size
=
decoder_size
*
3
,
# enable error clipping
layer_attr
=
paddle
.
attr
.
ExtraAttr
(
error_clipping_threshold
=
100.0
))
as
decoder_inputs
:
decoder_inputs
+=
paddle
.
layer
.
full_matrix_projection
(
input
=
context
)
decoder_inputs
+=
paddle
.
layer
.
full_matrix_projection
(
input
=
current_word
)
...
...
@@ -57,9 +68,12 @@ def seqToseq_net(source_dict_dim, target_dict_dim, is_generating=False):
name
=
'gru_decoder'
,
input
=
decoder_inputs
,
output_mem
=
decoder_mem
,
# uncomment to enable local threshold for gradient clipping
# param_attr=paddle.attr.ParamAttr(gradient_clipping_threshold=9.9),
size
=
decoder_size
)
with
paddle
.
layer
.
mixed
(
name
=
"gru_step_output"
,
size
=
target_dict_dim
,
bias_attr
=
True
,
act
=
paddle
.
activation
.
Softmax
())
as
out
:
...
...
@@ -125,7 +139,13 @@ def seqToseq_net(source_dict_dim, target_dict_dim, is_generating=False):
def
main
():
paddle
.
init
(
use_gpu
=
False
,
trainer_count
=
1
)
paddle
.
init
(
use_gpu
=
False
,
trainer_count
=
1
,
# log gradient clipping info
log_clipping
=
True
,
# log error clipping info
log_error_clipping
=
True
)
is_generating
=
False
# source and target dict dim.
...
...
@@ -140,6 +160,8 @@ def main():
# define optimize method and trainer
optimizer
=
paddle
.
optimizer
.
Adam
(
learning_rate
=
5e-5
,
# uncomment to enable global threshold for gradient clipping
# gradient_clipping_threshold=10.0,
regularization
=
paddle
.
optimizer
.
L2Regularization
(
rate
=
8e-4
))
trainer
=
paddle
.
trainer
.
SGD
(
cost
=
cost
,
parameters
=
parameters
,
...
...
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