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
Greenplum
DeepSpeed
提交
3354cff3
D
DeepSpeed
项目概览
Greenplum
/
DeepSpeed
上一次同步 大约 1 年
通知
10
Star
0
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
DevOps
流水线
流水线任务
计划
Wiki
0
Wiki
分析
仓库
DevOps
项目成员
Pages
D
DeepSpeed
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
DevOps
DevOps
流水线
流水线任务
计划
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
流水线任务
提交
Issue看板
体验新版 GitCode,发现更多精彩内容 >>
提交
3354cff3
编写于
1月 31, 2020
作者:
J
Jeff Rasley
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
add deepspeed exec and dataloader
上级
c53cabb0
变更
3
隐藏空白更改
内联
并排
Showing
3 changed file
with
99 addition
and
0 deletion
+99
-0
bin/deepspeed
bin/deepspeed
+1
-0
bin/ds
bin/ds
+6
-0
deepspeed/pt/deepspeed_dataloader.py
deepspeed/pt/deepspeed_dataloader.py
+92
-0
未找到文件。
bin/deepspeed
0 → 120000
浏览文件 @
3354cff3
ds
\ No newline at end of file
bin/ds
0 → 100755
浏览文件 @
3354cff3
#!/usr/bin/env python
from
deepspeed.pt.deepspeed_run
import
main
if
__name__
==
'__main__'
:
main
()
deepspeed/pt/deepspeed_dataloader.py
0 → 100755
浏览文件 @
3354cff3
'''
Copyright 2019 The Microsoft DeepSpeed Team
'''
import
torch
import
logging
from
torch.utils.data
import
DataLoader
,
RandomSampler
from
torch.utils.data.distributed
import
DistributedSampler
from
tqdm
import
tqdm
class
DeepSpeedDataSource
(
object
):
def
__init__
(
self
,
filenames
):
all_lines
=
[]
for
filename
in
filenames
:
logging
.
info
(
"Start reading file %s"
%
filename
)
with
open
(
filename
,
"r"
)
as
f
:
for
i
,
line
in
enumerate
(
tqdm
(
f
)):
all_lines
.
append
(
line
.
strip
())
self
.
all_lines
=
all_lines
self
.
len
=
len
(
self
.
all_lines
)
def
__len__
(
self
):
return
self
.
len
class
DeepSpeedDataLoader
(
object
):
def
__init__
(
self
,
dataset
,
batch_size
,
pin_memory
,
local_rank
,
tput_timer
,
collate_fn
=
None
,
num_local_io_workers
=
None
,
data_sampler
=
None
):
self
.
tput_timer
=
tput_timer
self
.
batch_size
=
batch_size
if
local_rank
>=
0
:
if
data_sampler
is
None
:
data_sampler
=
DistributedSampler
(
dataset
)
device_count
=
1
else
:
if
data_sampler
is
None
:
data_sampler
=
RandomSampler
(
dataset
)
device_count
=
torch
.
cuda
.
device_count
()
batch_size
*=
device_count
if
num_local_io_workers
is
None
:
num_local_io_workers
=
2
*
device_count
self
.
num_local_io_workers
=
num_local_io_workers
self
.
data_sampler
=
data_sampler
self
.
dataset
=
dataset
self
.
collate_fn
=
collate_fn
self
.
device_count
=
device_count
self
.
batch_size
=
batch_size
self
.
pin_memory
=
pin_memory
self
.
len
=
len
(
self
.
data_sampler
)
self
.
data
=
None
def
__iter__
(
self
):
self
.
_create_dataloader
()
return
self
def
__len__
(
self
):
return
self
.
len
def
__next__
(
self
):
if
self
.
tput_timer
:
self
.
tput_timer
.
start
()
return
next
(
self
.
data
)
def
_create_dataloader
(
self
):
if
self
.
collate_fn
is
None
:
self
.
dataloader
=
DataLoader
(
self
.
dataset
,
batch_size
=
self
.
batch_size
,
pin_memory
=
self
.
pin_memory
,
sampler
=
self
.
data_sampler
,
num_workers
=
self
.
num_local_io_workers
)
else
:
self
.
dataloader
=
DataLoader
(
self
.
dataset
,
batch_size
=
self
.
batch_size
,
pin_memory
=
self
.
pin_memory
,
sampler
=
self
.
data_sampler
,
collate_fn
=
self
.
collate_fn
,
num_workers
=
self
.
num_local_io_workers
)
self
.
data
=
(
x
for
x
in
self
.
dataloader
)
return
self
.
dataloader
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