Skip to content
项目
群组
代码片段
帮助
当前项目
正在载入...
登录 / 注册
切换导航面板
P
pytensor
项目
项目
详情
活动
周期分析
仓库
仓库
文件
提交
分支
标签
贡献者
图表
比较
统计图
议题
0
议题
0
列表
看板
标记
里程碑
合并请求
0
合并请求
0
CI / CD
CI / CD
流水线
作业
日程
统计图
Wiki
Wiki
代码片段
代码片段
成员
成员
折叠边栏
关闭边栏
活动
图像
聊天
创建新问题
作业
提交
问题看板
Open sidebar
testgroup
pytensor
Commits
bdd1daaf
提交
bdd1daaf
authored
11月 08, 2016
作者:
Alexander Matyasko
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
Fix bug in max pooling grad grad
Correct way is to first compute window end and only after that clip window start.
上级
7a777b2c
显示空白字符变更
内嵌
并排
正在显示
1 个修改的文件
包含
6 行增加
和
6 行删除
+6
-6
pool_grad_grad.c
theano/gpuarray/pool_grad_grad.c
+6
-6
没有找到文件。
theano/gpuarray/pool_grad_grad.c
浏览文件 @
bdd1daaf
...
@@ -18,11 +18,11 @@ KERNEL void max_pool2d_grad_grad_kernel(const ga_size nthreads,
...
@@ -18,11 +18,11 @@ KERNEL void max_pool2d_grad_grad_kernel(const ga_size nthreads,
const
ga_size
c
=
(
index
/
pooled_width
/
pooled_height
)
%
channels
;
const
ga_size
c
=
(
index
/
pooled_width
/
pooled_height
)
%
channels
;
const
ga_size
n
=
(
index
/
pooled_width
/
pooled_height
/
channels
);
const
ga_size
n
=
(
index
/
pooled_width
/
pooled_height
/
channels
);
ga_int
hstart
=
static_cast
<
ga_int
>
(
ph
*
stride_h
)
-
static_cast
<
ga_int
>
(
pad_h
);
ga_int
hstart
=
static_cast
<
ga_int
>
(
ph
*
stride_h
)
-
static_cast
<
ga_int
>
(
pad_h
);
hstart
=
max
(
hstart
,
0
);
const
ga_size
hend
=
min
(
hstart
+
kernel_h
,
height
);
const
ga_size
hend
=
min
(
hstart
+
kernel_h
,
height
);
ga_int
wstart
=
static_cast
<
ga_int
>
(
pw
*
stride_w
)
-
static_cast
<
ga_int
>
(
pad_w
);
ga_int
wstart
=
static_cast
<
ga_int
>
(
pw
*
stride_w
)
-
static_cast
<
ga_int
>
(
pad_w
);
wstart
=
max
(
wstart
,
0
);
const
ga_size
wend
=
min
(
wstart
+
kernel_w
,
width
);
const
ga_size
wend
=
min
(
wstart
+
kernel_w
,
width
);
hstart
=
max
(
hstart
,
0
);
wstart
=
max
(
wstart
,
0
);
const
ga_size
offset
=
(
n
*
channels
+
c
)
*
height
*
width
;
const
ga_size
offset
=
(
n
*
channels
+
c
)
*
height
*
width
;
...
@@ -63,14 +63,14 @@ KERNEL void max_pool3d_grad_grad_kernel(const ga_size nthreads,
...
@@ -63,14 +63,14 @@ KERNEL void max_pool3d_grad_grad_kernel(const ga_size nthreads,
const
ga_size
c
=
(
index
/
pooled_width
/
pooled_height
/
pooled_depth
)
%
channels
;
const
ga_size
c
=
(
index
/
pooled_width
/
pooled_height
/
pooled_depth
)
%
channels
;
const
ga_size
n
=
(
index
/
pooled_width
/
pooled_height
/
pooled_depth
/
channels
);
const
ga_size
n
=
(
index
/
pooled_width
/
pooled_height
/
pooled_depth
/
channels
);
ga_int
dstart
=
static_cast
<
ga_int
>
(
pd
*
stride_d
)
-
static_cast
<
ga_int
>
(
pad_d
);
ga_int
dstart
=
static_cast
<
ga_int
>
(
pd
*
stride_d
)
-
static_cast
<
ga_int
>
(
pad_d
);
dstart
=
max
(
dstart
,
0
);
const
ga_size
dend
=
min
(
dstart
+
kernel_d
,
depth
);
const
ga_size
dend
=
min
(
dstart
+
kernel_d
,
depth
);
ga_int
hstart
=
static_cast
<
ga_int
>
(
ph
*
stride_h
)
-
static_cast
<
ga_int
>
(
pad_h
);
ga_int
hstart
=
static_cast
<
ga_int
>
(
ph
*
stride_h
)
-
static_cast
<
ga_int
>
(
pad_h
);
hstart
=
max
(
hstart
,
0
);
const
ga_size
hend
=
min
(
hstart
+
kernel_h
,
height
);
const
ga_size
hend
=
min
(
hstart
+
kernel_h
,
height
);
ga_int
wstart
=
static_cast
<
ga_int
>
(
pw
*
stride_w
)
-
static_cast
<
ga_int
>
(
pad_w
);
ga_int
wstart
=
static_cast
<
ga_int
>
(
pw
*
stride_w
)
-
static_cast
<
ga_int
>
(
pad_w
);
wstart
=
max
(
wstart
,
0
);
const
ga_size
wend
=
min
(
wstart
+
kernel_w
,
width
);
const
ga_size
wend
=
min
(
wstart
+
kernel_w
,
width
);
dstart
=
max
(
dstart
,
0
);
hstart
=
max
(
hstart
,
0
);
wstart
=
max
(
wstart
,
0
);
const
ga_size
offset
=
(
n
*
channels
+
c
)
*
depth
*
height
*
width
;
const
ga_size
offset
=
(
n
*
channels
+
c
)
*
depth
*
height
*
width
;
...
@@ -142,7 +142,7 @@ int APPLY_SPECIFIC(pool_grad_grad)(PyGpuArrayObject *x,
...
@@ -142,7 +142,7 @@ int APPLY_SPECIFIC(pool_grad_grad)(PyGpuArrayObject *x,
const
size_t
*
z_dims
=
PyGpuArray_DIMS
(
z
);
const
size_t
*
z_dims
=
PyGpuArray_DIMS
(
z
);
const
size_t
*
x_dims
=
PyGpuArray_DIMS
(
x
);
const
size_t
*
x_dims
=
PyGpuArray_DIMS
(
x
);
//
G
et the max threads per blocks
//
g
et the max threads per blocks
err
=
gpucontext_property
(
ctx
->
ctx
,
GA_CTX_PROP_MAXLSIZE0
,
&
max_threads_dim
);
err
=
gpucontext_property
(
ctx
->
ctx
,
GA_CTX_PROP_MAXLSIZE0
,
&
max_threads_dim
);
if
(
err
!=
GA_NO_ERROR
){
if
(
err
!=
GA_NO_ERROR
){
PyErr_SetString
(
PyExc_RuntimeError
,
"Could not fetch max_threads_dims"
);
PyErr_SetString
(
PyExc_RuntimeError
,
"Could not fetch max_threads_dims"
);
...
...
编写
预览
Markdown
格式
0%
重试
或
添加新文件
添加附件
取消
您添加了
0
人
到此讨论。请谨慎行事。
请先完成此评论的编辑!
取消
请
注册
或者
登录
后发表评论