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
BaiXuePrincess
Paddle
提交
118aee6f
P
Paddle
项目概览
BaiXuePrincess
/
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
未验证
提交
118aee6f
编写于
1月 31, 2023
作者:
Z
zhangbo9674
提交者:
GitHub
1月 31, 2023
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
not use shm cache default (#50089)
上级
c64296bf
变更
2
显示空白变更内容
内联
并排
Showing
2 changed file
with
17 addition
and
8 deletion
+17
-8
paddle/phi/core/flags.cc
paddle/phi/core/flags.cc
+2
-2
python/paddle/fluid/dataloader/dataloader_iter.py
python/paddle/fluid/dataloader/dataloader_iter.py
+15
-6
未找到文件。
paddle/phi/core/flags.cc
浏览文件 @
118aee6f
...
@@ -1198,11 +1198,11 @@ PADDLE_DEFINE_EXPORTED_bool(trt_ibuilder_cache,
...
@@ -1198,11 +1198,11 @@ PADDLE_DEFINE_EXPORTED_bool(trt_ibuilder_cache,
* mmap_allocator related FLAG
* mmap_allocator related FLAG
* Name: use_shm_cache
* Name: use_shm_cache
* Since Version: 2.5.0
* Since Version: 2.5.0
* Value Range: bool, default=
tru
e
* Value Range: bool, default=
fals
e
* Example:
* Example:
* Note: . If True, mmap_allocator will cache shm file to decrease munmap
* Note: . If True, mmap_allocator will cache shm file to decrease munmap
* operation.
* operation.
*/
*/
PADDLE_DEFINE_EXPORTED_bool
(
use_shm_cache
,
PADDLE_DEFINE_EXPORTED_bool
(
use_shm_cache
,
tru
e
,
fals
e
,
"Use shm cache in mmap_allocator."
);
"Use shm cache in mmap_allocator."
);
python/paddle/fluid/dataloader/dataloader_iter.py
浏览文件 @
118aee6f
...
@@ -410,6 +410,13 @@ class _DataLoaderIterMultiProcess(_DataLoaderIterBase):
...
@@ -410,6 +410,13 @@ class _DataLoaderIterMultiProcess(_DataLoaderIterBase):
# Note(zhangbo): shm_buffer_size is used for MemoryMapAllocationPool.
# Note(zhangbo): shm_buffer_size is used for MemoryMapAllocationPool.
# MemoryMapAllocationPool is used to cache and reuse shm, thus reducing munmap in dataloader.
# MemoryMapAllocationPool is used to cache and reuse shm, thus reducing munmap in dataloader.
# For more details, please see: paddle/fluid/memory/allocation/mmap_allocator.h
# For more details, please see: paddle/fluid/memory/allocation/mmap_allocator.h
if
os
.
environ
.
get
(
'FLAGS_use_shm_cache'
,
False
)
in
[
1
,
'1'
,
True
,
'True'
,
'true'
,
]:
try
:
try
:
self
.
_worker_shm_buffer_size
=
(
2
+
1
)
*
len
(
self
.
_dataset
[
0
])
self
.
_worker_shm_buffer_size
=
(
2
+
1
)
*
len
(
self
.
_dataset
[
0
])
except
:
except
:
...
@@ -417,6 +424,8 @@ class _DataLoaderIterMultiProcess(_DataLoaderIterBase):
...
@@ -417,6 +424,8 @@ class _DataLoaderIterMultiProcess(_DataLoaderIterBase):
warnings
.
warn
(
warnings
.
warn
(
"Setting the shm cache buffer size to 0, equivalent to not using the shm cache policy."
"Setting the shm cache buffer size to 0, equivalent to not using the shm cache policy."
)
)
else
:
self
.
_worker_shm_buffer_size
=
0
self
.
_main_thread_shm_buffer_size
=
(
self
.
_main_thread_shm_buffer_size
=
(
(
self
.
_worker_shm_buffer_size
)
*
2
*
self
.
_num_workers
(
self
.
_worker_shm_buffer_size
)
*
2
*
self
.
_num_workers
)
)
...
...
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