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
01950ceb
P
Paddle
项目概览
PaddlePaddle
/
Paddle
12 个月 前同步成功
通知
2292
Star
20931
Fork
5422
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
体验新版 GitCode,发现更多精彩内容 >>
未验证
提交
01950ceb
编写于
12月 25, 2020
作者:
L
lilong12
提交者:
GitHub
12月 25, 2020
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
fix the bug in pipeline data parallelism (#29731)
* update, test=develop
上级
0f4b2186
变更
5
隐藏空白更改
内联
并排
Showing
5 changed file
with
265 addition
and
94 deletion
+265
-94
python/paddle/distributed/fleet/meta_optimizers/pipeline_optimizer.py
...e/distributed/fleet/meta_optimizers/pipeline_optimizer.py
+48
-58
python/paddle/fluid/optimizer.py
python/paddle/fluid/optimizer.py
+70
-35
python/paddle/fluid/tests/unittests/pipeline_mnist_one_device.py
...paddle/fluid/tests/unittests/pipeline_mnist_one_device.py
+138
-0
python/paddle/fluid/tests/unittests/test_fleet_pipeline_meta_optimizer.py
...uid/tests/unittests/test_fleet_pipeline_meta_optimizer.py
+1
-1
python/paddle/fluid/tests/unittests/test_pipeline.py
python/paddle/fluid/tests/unittests/test_pipeline.py
+8
-0
未找到文件。
python/paddle/distributed/fleet/meta_optimizers/pipeline_optimizer.py
浏览文件 @
01950ceb
...
@@ -41,35 +41,36 @@ class PipelineHelper(object):
...
@@ -41,35 +41,36 @@ class PipelineHelper(object):
inner_parallelism
=
None
):
inner_parallelism
=
None
):
self
.
startup_program
=
startup_program
self
.
startup_program
=
startup_program
endpoints
=
self
.
role_maker
.
_get_trainer_endpoints
()
current_endpoint
=
endpoints
[
self
.
role_maker
.
_worker_index
()]
node_num
=
_get_node_num
(
endpoints
)
assert
len
(
endpoints
)
%
node_num
==
0
nranks
=
self
.
role_maker
.
_worker_num
()
nranks
=
self
.
role_maker
.
_worker_num
()
rank
=
self
.
role_maker
.
_worker_index
()
rank
=
self
.
role_maker
.
_worker_index
()
endpoints
=
self
.
role_maker
.
_get_trainer_endpoints
()
# Create ring 0 for all gpus in a pipeline
current_endpoint
=
endpoints
[
rank
]
pipeline_endpoints
=
[]
node_num
=
_get_node_num
(
endpoints
)
pipeline_rank
=
rank
%
inner_parallelism
assert
nranks
%
node_num
==
0
pipeline_id
=
rank
//
inner_parallelism
for
idx
,
ep
in
enumerate
(
endpoints
):
# Create ring 0 for all gpus in the same pipeline
if
idx
//
inner_parallelism
==
pipeline_id
:
if
inner_parallelism
>
1
:
pipeline_endpoints
.
append
(
ep
)
pipeline_rank
=
rank
%
inner_parallelism
self
.
_init_communicator
(
self
.
startup_program
,
current_endpoint
,
pipeline_id
=
rank
//
inner_parallelism
pipeline_endpoints
,
pipeline_rank
,
0
,
start_index
=
pipeline_id
*
inner_parallelism
self
.
wait_port
)
pipeline_endpoints
=
endpoints
[
start_index
:
start_index
+
inner_parallelism
]
self
.
_init_communicator
(
self
.
startup_program
,
current_endpoint
,
pipeline_endpoints
,
pipeline_rank
,
0
,
self
.
wait_port
)
pipeline_num
=
len
(
endpoints
)
//
inner_parallelism
pipeline_num
=
len
(
endpoints
)
//
inner_parallelism
if
pipeline_num
==
1
:
return
if
pipeline_num
==
1
:
return
# Create rings for gpus with the same
gpu id
# Create rings for gpus with the same
pipeline id for data parallel
eps
=
[]
eps
=
[]
local_rank
=
self
.
role_maker
.
_worker_index
()
%
inner_parallelism
pipeline_rank
=
rank
%
inner_parallelism
ring_id
=
local
_rank
+
1
ring_id
=
pipeline
_rank
+
1
for
i
in
range
(
pipeline_num
):
for
i
in
range
(
pipeline_num
):
eps
.
append
(
endpoints
[
i
*
inner_parallelism
+
local_rank
])
eps
.
append
(
endpoints
[
i
*
inner_parallelism
+
pipeline_rank
])
temp_rank
=
self
.
role_maker
.
_worker_index
()
//
inner_parallelism
# rank in a ring of gpus with the same pipeline id for data parallel
dp_rank
=
rank
//
inner_parallelism
self
.
_init_communicator
(
self
.
startup_program
,
current_endpoint
,
eps
,
self
.
_init_communicator
(
self
.
startup_program
,
current_endpoint
,
eps
,
tem
p_rank
,
ring_id
,
self
.
wait_port
)
d
p_rank
,
ring_id
,
self
.
wait_port
)
self
.
_broadcast_params
(
ring_id
)
self
.
_broadcast_params
(
ring_id
)
def
_init_communicator
(
self
,
program
,
current_endpoint
,
endpoints
,
rank
,
def
_init_communicator
(
self
,
program
,
current_endpoint
,
endpoints
,
rank
,
...
@@ -108,8 +109,10 @@ class PipelineHelper(object):
...
@@ -108,8 +109,10 @@ class PipelineHelper(object):
def
_broadcast_params
(
self
,
ring_id
):
def
_broadcast_params
(
self
,
ring_id
):
block
=
self
.
startup_program
.
global_block
()
block
=
self
.
startup_program
.
global_block
()
for
param
in
block
.
iter_parameters
():
for
var_name
in
block
.
vars
:
if
param
.
is_distributed
:
if
"nccl_id"
in
var_name
:
continue
param
=
block
.
var
(
var_name
)
if
not
param
.
persistable
:
continue
continue
block
.
append_op
(
block
.
append_op
(
...
@@ -136,7 +139,7 @@ class PipelineOptimizer(MetaOptimizerBase):
...
@@ -136,7 +139,7 @@ class PipelineOptimizer(MetaOptimizerBase):
self
.
inner_opt
=
optimizer
self
.
inner_opt
=
optimizer
# we do not allow meta optimizer to be inner optimizer currently
# we do not allow meta optimizer to be inner optimizer currently
self
.
meta_optimizers_white_list
=
[]
self
.
meta_optimizers_white_list
=
[]
self
.
meta_optimizers_black_list
=
[]
self
.
meta_optimizers_black_list
=
[
"GraphExecutionOptimizer"
,
]
def
_set_basic_info
(
self
,
loss
,
role_maker
,
user_defined_optimizer
,
def
_set_basic_info
(
self
,
loss
,
role_maker
,
user_defined_optimizer
,
user_defined_strategy
):
user_defined_strategy
):
...
@@ -161,14 +164,6 @@ class PipelineOptimizer(MetaOptimizerBase):
...
@@ -161,14 +164,6 @@ class PipelineOptimizer(MetaOptimizerBase):
dist_strategy
.
pipeline
=
True
dist_strategy
.
pipeline
=
True
dist_strategy
.
pipeline_configs
=
{
"micro_batch"
:
1
,
}
dist_strategy
.
pipeline_configs
=
{
"micro_batch"
:
1
,
}
def
_get_local_rank
(
self
,
current_endpoint
,
endpoints
):
cur_node_endpoints
=
[]
cur_ip
=
current_endpoint
.
split
(
':'
)[
0
].
strip
()
for
ep
in
endpoints
:
if
cur_ip
==
ep
.
split
(
':'
)[
0
].
strip
():
cur_node_endpoints
.
append
(
ep
)
return
cur_node_endpoints
.
index
(
current_endpoint
)
def
minimize_impl
(
self
,
def
minimize_impl
(
self
,
loss
,
loss
,
startup_program
=
None
,
startup_program
=
None
,
...
@@ -176,56 +171,51 @@ class PipelineOptimizer(MetaOptimizerBase):
...
@@ -176,56 +171,51 @@ class PipelineOptimizer(MetaOptimizerBase):
no_grad_set
=
None
):
no_grad_set
=
None
):
endpoints
=
self
.
role_maker
.
_get_trainer_endpoints
()
endpoints
=
self
.
role_maker
.
_get_trainer_endpoints
()
current_endpoint
=
endpoints
[
self
.
role_maker
.
_worker_index
()]
current_endpoint
=
endpoints
[
self
.
role_maker
.
_worker_index
()]
self
.
local_rank
=
self
.
_get_local_rank
(
current_endpoint
,
endpoints
)
self
.
wrapped_opt
=
PO
(
self
.
inner_opt
,
self
.
wrapped_opt
=
PO
(
self
.
inner_opt
,
num_microbatches
=
self
.
num_microbatches
,
num_microbatches
=
self
.
num_microbatches
)
start_cpu_core_id
=
self
.
local_rank
)
node_num
=
_get_node_num
(
endpoints
)
node_num
=
_get_node_num
(
endpoints
)
gpus_per_node
=
len
(
endpoints
)
//
node_num
gpus_per_node
=
len
(
endpoints
)
//
node_num
self
.
startup_program
=
startup_program
self
.
startup_program
=
startup_program
self
.
local_rank
=
self
.
_get_local_rank
(
current_endpoint
,
endpoints
)
if
startup_program
is
None
:
if
startup_program
is
None
:
self
.
startup_program
=
fluid
.
default_startup_program
()
self
.
startup_program
=
fluid
.
default_startup_program
()
loss
.
block
.
program
.
_pipeline_opt
=
dict
()
self
.
rank
=
self
.
role_maker
.
_worker_index
()
loss
.
block
.
program
.
_pipeline_opt
[
'local_rank'
]
=
self
.
local_rank
self
.
nranks
=
self
.
role_maker
.
_worker_num
()
optimize_ops
,
params_grads
,
prog_list
=
\
assert
self
.
nranks
%
node_num
==
0
self
.
wrapped_opt
.
minimize
(
loss
,
startup_program
,
parameter_list
,
no_grad_set
)
loss
.
block
.
program
.
_pipeline_opt
=
dict
()
loss
.
block
.
program
.
_pipeline_opt
[
'local_rank'
]
=
self
.
rank
optimize_ops
,
params_grads
,
prog_list
=
self
.
wrapped_opt
.
minimize
(
loss
,
startup_program
,
parameter_list
,
no_grad_set
)
assert
prog_list
assert
prog_list
self
.
main_program_list
=
prog_list
self
.
main_program_list
=
prog_list
self
.
main_program
=
loss
.
block
.
program
self
.
main_program
=
loss
.
block
.
program
self
.
inner_parallelism
=
loss
.
block
.
program
.
_pipeline_opt
[
self
.
inner_parallelism
=
loss
.
block
.
program
.
_pipeline_opt
[
'inner_parallelism'
]
'inner_parallelism'
]
nranks
=
len
(
endpoints
)
assert
self
.
nranks
%
self
.
inner_parallelism
==
0
self
.
nranks
=
nranks
self
.
nrings
=
len
(
self
.
main_program_list
)
self
.
rank
=
self
.
role_maker
.
_worker_index
()
self
.
endpoints
=
endpoints
self
.
current_endpoint
=
current_endpoint
pipeline_helper
=
PipelineHelper
(
self
.
role_maker
)
pipeline_helper
=
PipelineHelper
(
self
.
role_maker
)
pipeline_helper
.
update_startup_program
(
pipeline_helper
.
update_startup_program
(
self
.
startup_program
.
_pipeline_opt
[
"startup_program"
],
self
.
startup_program
.
_pipeline_opt
[
"startup_program"
],
self
.
inner_parallelism
)
self
.
inner_parallelism
)
self
.
_transpile_main_program
(
loss
,
node_num
,
gpus_per_node
)
pipeline_num
=
self
.
nranks
//
self
.
inner_parallelism
self
.
_transpile_main_program
(
loss
,
pipeline_num
,
self
.
inner_parallelism
)
return
optimize_ops
,
params_grads
return
optimize_ops
,
params_grads
def
_transpile_main_program
(
self
,
loss
,
node_num
,
gpus_per_node
):
def
_transpile_main_program
(
self
,
loss
,
pipeline_num
,
inner_parallelism
):
self
.
_insert_loss_grad_ops
(
loss
,
gpus_per_node
,
node_num
)
if
pipeline_num
<=
1
:
return
for
ring_id
in
range
(
1
,
gpus_per_node
+
1
):
self
.
_insert_loss_grad_ops
(
loss
,
pipeline_num
)
for
ring_id
in
range
(
1
,
inner_parallelism
+
1
):
self
.
_insert_allreduce_ops
(
ring_id
)
self
.
_insert_allreduce_ops
(
ring_id
)
def
_insert_loss_grad_ops
(
self
,
loss
,
gpus_per_node
,
nod
e_num
):
def
_insert_loss_grad_ops
(
self
,
loss
,
pipelin
e_num
):
"""
"""
In order to keep the learning rate consistent in different numbers of
In order to keep the learning rate consistent in different numbers of
training workers, we scale the loss grad by the number of workers
training workers, we scale the loss grad by the number of workers
"""
"""
block
=
self
.
main_program_list
[
gpus_per_node
-
1
][
block
=
self
.
main_program_list
[
-
1
][
'program'
].
global_block
()
'program'
].
global_block
()
for
idx
,
op
in
reversed
(
list
(
enumerate
(
block
.
ops
))):
for
idx
,
op
in
reversed
(
list
(
enumerate
(
block
.
ops
))):
if
is_loss_grad_op
(
op
):
if
is_loss_grad_op
(
op
):
loss_grad_var
=
block
.
vars
[
op
.
output_arg_names
[
0
]]
loss_grad_var
=
block
.
vars
[
op
.
output_arg_names
[
0
]]
...
@@ -235,7 +225,7 @@ class PipelineOptimizer(MetaOptimizerBase):
...
@@ -235,7 +225,7 @@ class PipelineOptimizer(MetaOptimizerBase):
inputs
=
{
'X'
:
loss_grad_var
},
inputs
=
{
'X'
:
loss_grad_var
},
outputs
=
{
'Out'
:
loss_grad_var
},
outputs
=
{
'Out'
:
loss_grad_var
},
attrs
=
{
attrs
=
{
'scale'
:
1.0
/
nod
e_num
,
'scale'
:
1.0
/
pipelin
e_num
,
OP_ROLE_KEY
:
OpRole
.
Backward
OP_ROLE_KEY
:
OpRole
.
Backward
})
})
...
@@ -269,7 +259,7 @@ class PipelineOptimizer(MetaOptimizerBase):
...
@@ -269,7 +259,7 @@ class PipelineOptimizer(MetaOptimizerBase):
block
.
_insert_op
(
block
.
_insert_op
(
offset
,
offset
,
type
=
'c_
sync_calc_strea
m'
,
type
=
'c_
allreduce_su
m'
,
inputs
=
{
'X'
:
grad
},
inputs
=
{
'X'
:
grad
},
outputs
=
{
'Out'
:
grad
},
outputs
=
{
'Out'
:
grad
},
attrs
=
{
attrs
=
{
...
@@ -283,7 +273,7 @@ class PipelineOptimizer(MetaOptimizerBase):
...
@@ -283,7 +273,7 @@ class PipelineOptimizer(MetaOptimizerBase):
for
idx
,
op
in
enumerate
(
block
.
ops
):
for
idx
,
op
in
enumerate
(
block
.
ops
):
if
is_optimizer_op
(
op
):
if
is_optimizer_op
(
op
):
block
.
_insert_op
(
block
.
_insert_op
(
idx
+
ring_id
,
idx
,
type
=
'c_sync_comm_stream'
,
type
=
'c_sync_comm_stream'
,
inputs
=
{
'X'
:
grad
},
inputs
=
{
'X'
:
grad
},
outputs
=
{
'Out'
:
grad
},
outputs
=
{
'Out'
:
grad
},
...
...
python/paddle/fluid/optimizer.py
浏览文件 @
01950ceb
...
@@ -16,6 +16,7 @@ from __future__ import print_function
...
@@ -16,6 +16,7 @@ from __future__ import print_function
import
numpy
as
np
import
numpy
as
np
import
six
import
six
import
os
import
logging
import
logging
from
collections
import
defaultdict
from
collections
import
defaultdict
...
@@ -39,6 +40,7 @@ from .dygraph.learning_rate_scheduler import LearningRateDecay, _LearningRateEpo
...
@@ -39,6 +40,7 @@ from .dygraph.learning_rate_scheduler import LearningRateDecay, _LearningRateEpo
from
paddle.fluid
import
core
from
paddle.fluid
import
core
from
paddle.fluid.layers
import
tensor
from
paddle.fluid.layers
import
tensor
from
functools
import
reduce
from
functools
import
reduce
from
functools
import
cmp_to_key
from
.wrapped_decorator
import
signature_safe_contextmanager
from
.wrapped_decorator
import
signature_safe_contextmanager
from
..
import
compat
as
cpt
from
..
import
compat
as
cpt
...
@@ -3773,7 +3775,7 @@ class PipelineOptimizer(object):
...
@@ -3773,7 +3775,7 @@ class PipelineOptimizer(object):
self
.
_op_device_key
=
op_maker
.
kOpDeviceAttrName
()
self
.
_op_device_key
=
op_maker
.
kOpDeviceAttrName
()
self
.
_param_device_map
=
None
self
.
_param_device_map
=
None
def
_create_vars
(
self
,
block
,
main_program
):
def
_create_vars
(
self
,
block
,
ori_block
):
# Create vars for block, copied from main_program's global block
# Create vars for block, copied from main_program's global block
used_var_set
=
set
()
used_var_set
=
set
()
for
op_idx
in
range
(
block
.
desc
.
op_size
()):
for
op_idx
in
range
(
block
.
desc
.
op_size
()):
...
@@ -3785,7 +3787,8 @@ class PipelineOptimizer(object):
...
@@ -3785,7 +3787,8 @@ class PipelineOptimizer(object):
if
var
in
used_var_set
or
"_blocking_queue"
in
var
:
if
var
in
used_var_set
or
"_blocking_queue"
in
var
:
continue
continue
used_var_set
.
add
(
var
)
used_var_set
.
add
(
var
)
source_var
=
main_program
.
block
(
0
).
var
(
str
(
var
))
if
block
.
_find_var_recursive
(
str
(
var
)):
continue
source_var
=
ori_block
.
_var_recursive
(
str
(
var
))
if
source_var
.
type
==
core
.
VarDesc
.
VarType
.
READER
:
if
source_var
.
type
==
core
.
VarDesc
.
VarType
.
READER
:
block
.
create_var
(
block
.
create_var
(
name
=
var
,
name
=
var
,
...
@@ -3840,45 +3843,65 @@ class PipelineOptimizer(object):
...
@@ -3840,45 +3843,65 @@ class PipelineOptimizer(object):
op_desc
=
op
.
desc
op_desc
=
op
.
desc
ap_op
=
program
[
"program"
].
block
(
0
).
desc
.
append_op
()
ap_op
=
program
[
"program"
].
block
(
0
).
desc
.
append_op
()
ap_op
.
copy_from
(
op_desc
)
ap_op
.
copy_from
(
op_desc
)
ap_op
.
_set_attr
(
self
.
_op_device_key
,
""
)
#
ap_op._set_attr(self._op_device_key, "")
elif
op
.
type
==
"create_py_reader"
or
op
.
type
==
"read"
:
elif
op
.
type
==
"create_py_reader"
or
op
.
type
==
"read"
or
op
.
type
==
"create_double_buffer_reader"
:
# Copy read related ops to all section to make them exit after each epoch.
# Copy read related ops to all section to make them exit after each epoch.
for
device
in
device_program_map
.
keys
():
for
device
in
device_program_map
.
keys
():
program
=
device_program_map
[
device
]
program
=
device_program_map
[
device
]
op_desc
=
op
.
desc
op_desc
=
op
.
desc
ap_op
=
program
[
"program"
].
block
(
0
).
desc
.
append_op
()
ap_op
=
program
[
"program"
].
block
(
0
).
desc
.
append_op
()
ap_op
.
copy_from
(
op_desc
)
ap_op
.
copy_from
(
op_desc
)
ap_op
.
_set_attr
(
self
.
_op_device_key
,
""
)
else
:
else
:
program
=
device_program_map
[
device
]
program
=
device_program_map
[
device
]
op_desc
=
op
.
desc
op_desc
=
op
.
desc
ap_op
=
program
[
"program"
].
block
(
0
).
desc
.
append_op
()
ap_op
=
program
[
"program"
].
block
(
0
).
desc
.
append_op
()
ap_op
.
copy_from
(
op_desc
)
ap_op
.
copy_from
(
op_desc
)
ap_op
.
_set_attr
(
self
.
_op_device_key
,
""
)
for
key
in
sorted
(
device_program_map
.
keys
())
:
for
key
in
devices
:
program
=
device_program_map
[
key
]
program
=
device_program_map
[
key
]
program
[
'program'
].
_sync_with_cpp
()
program
[
'program'
].
_sync_with_cpp
()
programs
.
append
(
program
)
programs
.
append
(
program
)
return
programs
return
programs
def
_get_op_device_for_startup_program
(
self
,
var_name
):
"""
For adam optimizer, it will add accumulators and initialize them
with fill_constant, and force the op device to cpu. Hence, we should
get the real op_device attribute of the fill_constant as the device
where the corresponding parameters on.
"""
assert
"beta1_pow_acc"
in
var_name
or
"beta2_pow_acc"
in
var_name
param_name
=
var_name
[
0
:
var_name
.
index
(
'_beta'
)]
device
=
self
.
_param_device_map
[
param_name
]
return
device
def
_split_startup_program
(
self
,
startup_program
,
local_rank
):
def
_split_startup_program
(
self
,
startup_program
,
local_rank
):
block
=
startup_program
.
block
(
0
)
block
=
startup_program
.
block
(
0
)
new_startup_program
=
Program
()
new_startup_program
=
Program
()
for
op
in
block
.
ops
:
for
op
in
block
.
ops
:
device
=
op
.
attr
(
self
.
_op_device_key
)
device
=
op
.
attr
(
self
.
_op_device_key
)
if
device
==
"cpu"
:
assert
op
.
type
==
"fill_constant"
,
(
"For ops in startup "
"program that with the op_device attribute of cpu, "
"they must be fill_constant."
)
output_var
=
op
.
output_arg_names
[
0
]
device
=
self
.
_get_op_device_for_startup_program
(
output_var
)
if
device
:
if
device
:
device_index
=
int
(
device
.
split
(
":"
)[
1
])
device_index
=
int
(
device
.
split
(
':'
)[
1
])
else
:
else
:
device_index
=
None
# LR related ops
if
device_index
is
not
None
and
device_index
!=
local_rank
:
continue
device
=
None
if
device
and
device_index
!=
local_rank
:
continue
op_desc
=
op
.
desc
op_desc
=
op
.
desc
ap_op
=
new_startup_program
.
block
(
0
).
desc
.
append_op
()
ap_op
=
new_startup_program
.
block
(
0
).
desc
.
append_op
()
ap_op
.
copy_from
(
op_desc
)
ap_op
.
copy_from
(
op_desc
)
ap_op
.
_set_attr
(
self
.
_op_device_key
,
""
)
ap_op
.
_set_attr
(
self
.
_op_device_key
,
""
)
new_startup_program
.
_sync_with_cpp
()
new_startup_program
.
_sync_with_cpp
()
self
.
_create_vars
(
new_startup_program
.
block
(
0
),
startup_program
)
self
.
_create_vars
(
new_startup_program
.
block
(
0
),
startup_program
.
global_block
())
return
new_startup_program
return
new_startup_program
def
_find_post_op
(
self
,
ops
,
cur_op
,
var_name
):
def
_find_post_op
(
self
,
ops
,
cur_op
,
var_name
):
...
@@ -4093,6 +4116,8 @@ class PipelineOptimizer(object):
...
@@ -4093,6 +4116,8 @@ class PipelineOptimizer(object):
first_device
=
op
.
attr
(
self
.
_op_device_key
)
first_device
=
op
.
attr
(
self
.
_op_device_key
)
break
break
assert
first_device
assert
first_device
first_device_type
=
first_device
.
split
(
":"
)[
0
]
assert
first_device_type
==
"gpu"
# set op_device attr for lr-related ops
# set op_device attr for lr-related ops
lrsched_role
=
int
(
self
.
_op_role
.
LRSched
)
lrsched_role
=
int
(
self
.
_op_role
.
LRSched
)
...
@@ -4136,10 +4161,11 @@ class PipelineOptimizer(object):
...
@@ -4136,10 +4161,11 @@ class PipelineOptimizer(object):
dev_spec
=
op
.
attr
(
self
.
_op_device_key
)
dev_spec
=
op
.
attr
(
self
.
_op_device_key
)
assert
dev_spec
,
(
"op_device attribute for op "
assert
dev_spec
,
(
"op_device attribute for op "
"{} has not been set."
.
format
(
op
.
type
))
"{} has not been set."
.
format
(
op
.
type
))
dev_type
=
dev_spec
.
split
(
':'
)[
0
]
assert
dev_type
==
"gpu"
,
(
"Now only gpu devices are supported "
"for pipeline parallelism."
)
if
not
dev_spec
in
device_specs
:
if
not
dev_spec
in
device_specs
:
device_specs
.
append
(
dev_spec
)
device_specs
.
append
(
dev_spec
)
sorted_device_specs
=
sorted
(
device_specs
)
assert
sorted_device_specs
==
device_specs
return
device_specs
return
device_specs
def
_insert_sendrecv_ops_for_boundaries
(
self
,
block
):
def
_insert_sendrecv_ops_for_boundaries
(
self
,
block
):
...
@@ -4216,6 +4242,7 @@ class PipelineOptimizer(object):
...
@@ -4216,6 +4242,7 @@ class PipelineOptimizer(object):
device
=
self
.
_param_device_map
[
param_name
]
device
=
self
.
_param_device_map
[
param_name
]
if
device
!=
dev_spec
:
continue
if
device
!=
dev_spec
:
continue
grad_name
=
self
.
_append_grad_suffix
(
param_name
)
grad_name
=
self
.
_append_grad_suffix
(
param_name
)
if
not
main_block
.
has_var
(
grad_name
):
continue
grad_var
=
main_block
.
vars
[
grad_name
]
grad_var
=
main_block
.
vars
[
grad_name
]
main_block
.
_insert_op
(
main_block
.
_insert_op
(
index
=
0
,
index
=
0
,
...
@@ -4297,6 +4324,7 @@ class PipelineOptimizer(object):
...
@@ -4297,6 +4324,7 @@ class PipelineOptimizer(object):
ap_op
=
new_sub_block
.
desc
.
append_op
()
ap_op
=
new_sub_block
.
desc
.
append_op
()
ap_op
.
copy_from
(
op_desc
)
ap_op
.
copy_from
(
op_desc
)
new_sub_block
.
_sync_with_cpp
()
new_sub_block
.
_sync_with_cpp
()
self
.
_create_vars
(
new_sub_block
,
origin_sub_block
)
op
.
_set_attr
(
'sub_block:'
,
new_sub_block
)
op
.
_set_attr
(
'sub_block:'
,
new_sub_block
)
def
_get_device_info
(
self
,
block
):
def
_get_device_info
(
self
,
block
):
...
@@ -4318,6 +4346,7 @@ class PipelineOptimizer(object):
...
@@ -4318,6 +4346,7 @@ class PipelineOptimizer(object):
prog
=
prog_info
[
'program'
]
prog
=
prog_info
[
'program'
]
block
=
prog
.
block
(
0
)
block
=
prog
.
block
(
0
)
for
var_name
in
block
.
vars
:
for
var_name
in
block
.
vars
:
if
var_name
==
"double_buffer_0"
:
continue
var
=
block
.
var
(
var_name
)
var
=
block
.
var
(
var_name
)
if
not
var
.
persistable
:
continue
if
not
var
.
persistable
:
continue
if
not
var_name
in
var_info
:
if
not
var_name
in
var_info
:
...
@@ -4413,30 +4442,33 @@ class PipelineOptimizer(object):
...
@@ -4413,30 +4442,33 @@ class PipelineOptimizer(object):
self
.
_add_default_opdevice_attr
(
main_block
)
self
.
_add_default_opdevice_attr
(
main_block
)
device_specs
=
self
.
_check_validation
(
main_block
)
device_specs
=
self
.
_check_validation
(
main_block
)
assert
len
(
device_specs
)
>
1
def
device_cmp
(
device1
,
device2
):
dev1_id
=
int
(
device1
.
split
(
':'
)[
1
])
dev2_id
=
int
(
device2
.
split
(
':'
)[
1
])
if
dev1_id
<
dev2_id
:
return
-
1
elif
dev1_id
>
dev2_id
:
return
1
else
:
return
0
sorted_device_spec
=
sorted
(
device_specs
,
key
=
cmp_to_key
(
device_cmp
))
assert
sorted_device_spec
==
device_specs
,
(
"With pipeline "
"parallelism, you must use gpu devices one after another "
"in the order of their ids."
)
# Step3: add send and recv ops between section boundaries
# Step3: add send and recv ops between section boundaries
self
.
_insert_sendrecv_ops_for_boundaries
(
main_block
)
self
.
_insert_sendrecv_ops_for_boundaries
(
main_block
)
place_list
=
[]
place_id_list
=
[]
for
dev_spec
in
device_specs
:
if
dev_spec
==
"cpu"
:
place_list
.
append
(
core
.
CPUPlace
())
place_id_list
.
append
(
-
1
)
elif
"gpu"
in
dev_spec
and
":"
in
dev_spec
:
dev_index
=
dev_spec
.
split
(
":"
)[
1
]
place_list
.
append
(
core
.
CUDAPlace
(
int
(
dev_index
)))
place_id_list
.
append
(
int
(
dev_index
))
else
:
raise
ValueError
(
"Unknown device type: %s"
,
dev_spec
)
# Step4: split program into sections and add pairs of
# Step4: split program into sections and add pairs of
# send and recv ops for data var.
# send and recv ops for data var.
main_program
=
main_block
.
program
main_program
=
main_block
.
program
program_list
=
self
.
_split_program
(
main_program
,
device_specs
)
program_list
=
self
.
_split_program
(
main_program
,
device_specs
)
for
p
in
program_list
:
for
p
in
program_list
:
self
.
_create_vars
(
p
[
"program"
].
block
(
0
),
main_program
)
self
.
_create_vars
(
p
[
"program"
].
block
(
0
),
main_program
.
global_block
())
self
.
_insert_sendrecv_for_data_var
(
main_block
,
program_list
,
self
.
_insert_sendrecv_for_data_var
(
main_block
,
program_list
,
startup_program
,
device_specs
)
startup_program
,
device_specs
)
...
@@ -4452,7 +4484,13 @@ class PipelineOptimizer(object):
...
@@ -4452,7 +4484,13 @@ class PipelineOptimizer(object):
isinstance
(
main_program
.
_pipeline_opt
,
dict
)
and
isinstance
(
main_program
.
_pipeline_opt
,
dict
)
and
'local_rank'
in
main_program
.
_pipeline_opt
),
\
'local_rank'
in
main_program
.
_pipeline_opt
),
\
"You must use pipeline with fleet"
"You must use pipeline with fleet"
local_rank
=
main_program
.
_pipeline_opt
[
'local_rank'
]
local_rank
=
main_program
.
_pipeline_opt
[
'local_rank'
]
%
len
(
device_specs
)
place_list
=
[]
for
dev_spec
in
device_specs
:
dev_index
=
dev_spec
.
split
(
":"
)[
1
]
place_list
.
append
(
core
.
CUDAPlace
(
local_rank
))
# Step7: Split startup program
# Step7: Split startup program
new_startup_program
=
self
.
_split_startup_program
(
startup_program
,
new_startup_program
=
self
.
_split_startup_program
(
startup_program
,
...
@@ -4466,21 +4504,18 @@ class PipelineOptimizer(object):
...
@@ -4466,21 +4504,18 @@ class PipelineOptimizer(object):
self
.
_accumulate_gradients
(
program_list
[
local_rank
][
'program'
]
self
.
_accumulate_gradients
(
program_list
[
local_rank
][
'program'
]
.
global_block
())
.
global_block
())
with
open
(
"startup_prog_%d"
%
local_rank
,
'w'
)
as
f
:
f
.
writelines
(
str
(
new_startup_program
))
with
open
(
"main_prog_%d"
%
local_rank
,
'w'
)
as
f
:
f
.
writelines
(
str
(
program_list
[
local_rank
][
'program'
]))
startup_program
.
_pipeline_opt
=
{
startup_program
.
_pipeline_opt
=
{
"startup_program"
:
new_startup_program
,
"startup_program"
:
new_startup_program
,
}
}
place_id
=
int
(
os
.
getenv
(
"FLAGS_selected_gpus"
,
"0"
))
main_program
.
_pipeline_opt
=
{
main_program
.
_pipeline_opt
=
{
"trainer"
:
"PipelineTrainer"
,
"trainer"
:
"PipelineTrainer"
,
"device_worker"
:
"Section"
,
"device_worker"
:
"Section"
,
"inner_parallelism"
:
len
(
device_specs
),
"inner_parallelism"
:
len
(
device_specs
),
"section_program"
:
program_list
[
local_rank
],
"section_program"
:
program_list
[
local_rank
],
"place"
:
place_list
[
local_rank
],
"place"
:
place_list
[
local_rank
],
"place_id"
:
place_id
_list
[
local_rank
]
,
"place_id"
:
place_id
,
"sync_steps"
:
-
1
,
"sync_steps"
:
-
1
,
"num_microbatches"
:
self
.
_num_microbatches
,
"num_microbatches"
:
self
.
_num_microbatches
,
"start_cpu_core_id"
:
self
.
_start_cpu_core_id
,
"start_cpu_core_id"
:
self
.
_start_cpu_core_id
,
...
...
python/paddle/fluid/tests/unittests/pipeline_mnist_one_device.py
0 → 100644
浏览文件 @
01950ceb
# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from
__future__
import
print_function
import
numpy
as
np
import
argparse
import
time
import
math
import
paddle
import
paddle.fluid
as
fluid
import
paddle.fluid.profiler
as
profiler
from
paddle.fluid
import
core
import
unittest
from
multiprocessing
import
Process
import
os
import
signal
from
functools
import
reduce
from
test_dist_base
import
TestDistRunnerBase
,
runtime_main
import
paddle.distributed.fleet
as
fleet
paddle
.
enable_static
()
DTYPE
=
"float32"
paddle
.
dataset
.
mnist
.
fetch
()
# Fix seed for test
fluid
.
default_startup_program
().
random_seed
=
1
fluid
.
default_main_program
().
random_seed
=
1
def
cnn_model
(
data
):
conv_pool_1
=
fluid
.
nets
.
simple_img_conv_pool
(
input
=
data
,
filter_size
=
5
,
num_filters
=
20
,
pool_size
=
2
,
pool_stride
=
2
,
act
=
"relu"
,
param_attr
=
fluid
.
ParamAttr
(
initializer
=
fluid
.
initializer
.
Constant
(
value
=
0.01
)))
conv_pool_2
=
fluid
.
nets
.
simple_img_conv_pool
(
input
=
conv_pool_1
,
filter_size
=
5
,
num_filters
=
50
,
pool_size
=
2
,
pool_stride
=
2
,
act
=
"relu"
,
param_attr
=
fluid
.
ParamAttr
(
initializer
=
fluid
.
initializer
.
Constant
(
value
=
0.01
)))
SIZE
=
10
input_shape
=
conv_pool_2
.
shape
param_shape
=
[
reduce
(
lambda
a
,
b
:
a
*
b
,
input_shape
[
1
:],
1
)]
+
[
SIZE
]
scale
=
(
2.0
/
(
param_shape
[
0
]
**
2
*
SIZE
))
**
0.5
predict
=
fluid
.
layers
.
fc
(
input
=
conv_pool_2
,
size
=
SIZE
,
act
=
"softmax"
,
param_attr
=
fluid
.
param_attr
.
ParamAttr
(
initializer
=
fluid
.
initializer
.
Constant
(
value
=
0.01
)))
return
predict
class
TestDistMnist2x2
(
TestDistRunnerBase
):
def
get_model
(
self
,
batch_size
=
2
,
use_dgc
=
False
,
dist_strategy
=
None
):
# Input data
device_id
=
0
if
dist_strategy
:
fleet
.
init
(
is_collective
=
True
)
with
fluid
.
device_guard
(
"gpu:0"
):
images
=
fluid
.
layers
.
data
(
name
=
'pixel'
,
shape
=
[
1
,
28
,
28
],
dtype
=
DTYPE
)
label
=
fluid
.
layers
.
data
(
name
=
'label'
,
shape
=
[
1
],
dtype
=
'int64'
)
if
dist_strategy
:
data_loader
=
fluid
.
io
.
DataLoader
.
from_generator
(
feed_list
=
[
images
,
label
],
capacity
=
64
,
use_double_buffer
=
False
,
iterable
=
False
)
# Train program
predict
=
cnn_model
(
images
)
with
fluid
.
device_guard
(
"gpu:0"
):
cost
=
fluid
.
layers
.
cross_entropy
(
input
=
predict
,
label
=
label
)
avg_cost
=
fluid
.
layers
.
mean
(
x
=
cost
)
# Evaluator
with
fluid
.
device_guard
(
"gpu:0"
):
batch_size_tensor
=
fluid
.
layers
.
create_tensor
(
dtype
=
'int64'
)
batch_acc
=
fluid
.
layers
.
accuracy
(
input
=
predict
,
label
=
label
,
total
=
batch_size_tensor
)
inference_program
=
fluid
.
default_main_program
().
clone
()
base_lr
=
self
.
lr
passes
=
[
30
,
60
,
80
,
90
]
steps_per_pass
=
10
bd
=
[
steps_per_pass
*
p
for
p
in
passes
]
lr
=
[
base_lr
*
(
0.1
**
i
)
for
i
in
range
(
len
(
bd
)
+
1
)]
lr_val
=
fluid
.
layers
.
piecewise_decay
(
boundaries
=
bd
,
values
=
lr
)
opt
=
fluid
.
optimizer
.
Momentum
(
learning_rate
=
lr_val
,
momentum
=
0.9
)
# Reader
train_reader
=
paddle
.
batch
(
paddle
.
dataset
.
mnist
.
test
(),
batch_size
=
batch_size
)
test_reader
=
paddle
.
batch
(
paddle
.
dataset
.
mnist
.
test
(),
batch_size
=
batch_size
)
if
dist_strategy
:
strategy
=
fleet
.
DistributedStrategy
()
strategy
.
pipeline
=
True
dist_opt
=
fleet
.
distributed_optimizer
(
optimizer
=
opt
,
strategy
=
strategy
)
dist_opt
.
minimize
(
avg_cost
)
else
:
opt
.
minimize
(
avg_cost
)
if
dist_strategy
:
return
inference_program
,
avg_cost
,
train_reader
,
test_reader
,
batch_acc
,
predict
,
data_loader
else
:
return
inference_program
,
avg_cost
,
train_reader
,
test_reader
,
batch_acc
,
predict
if
__name__
==
"__main__"
:
runtime_main
(
TestDistMnist2x2
)
python/paddle/fluid/tests/unittests/test_fleet_pipeline_meta_optimizer.py
浏览文件 @
01950ceb
...
@@ -50,7 +50,7 @@ class TestFleetMetaOptimizer(unittest.TestCase):
...
@@ -50,7 +50,7 @@ class TestFleetMetaOptimizer(unittest.TestCase):
strategy
.
pipeline
=
True
strategy
.
pipeline
=
True
strategy
.
pipeline_configs
=
{
'micro_batch'
:
2
}
strategy
.
pipeline_configs
=
{
'micro_batch'
:
2
}
optimizer
=
paddle
.
fluid
.
optimizer
.
SGD
(
learning_rate
=
0.01
)
optimizer
=
paddle
.
fluid
.
optimizer
.
Adam
(
0.01
)
optimizer
=
fleet
.
distributed_optimizer
(
optimizer
,
strategy
=
strategy
)
optimizer
=
fleet
.
distributed_optimizer
(
optimizer
,
strategy
=
strategy
)
optimizer
.
minimize
(
avg_cost
)
optimizer
.
minimize
(
avg_cost
)
...
...
python/paddle/fluid/tests/unittests/test_pipeline.py
浏览文件 @
01950ceb
...
@@ -40,6 +40,14 @@ class TestPipeline(TestDistBase):
...
@@ -40,6 +40,14 @@ class TestPipeline(TestDistBase):
check_error_log
=
True
,
check_error_log
=
True
,
log_name
=
flag_name
)
log_name
=
flag_name
)
def
test_dist_train_one_device
(
self
):
import
paddle.fluid
as
fluid
if
fluid
.
core
.
is_compiled_with_cuda
():
self
.
check_with_place
(
"pipeline_mnist_one_device.py"
,
check_error_log
=
True
,
log_name
=
flag_name
)
if
__name__
==
'__main__'
:
if
__name__
==
'__main__'
:
unittest
.
main
()
unittest
.
main
()
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