m53297601
/
Paddle
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

update the api en doc of BuildStrategy (#20445)

Browse Source 
* update the api en doc of BuildStrategy and its setting, test=develop, test=document_fix

* update api.spec, test=develop, test=document_fix

* update the en doc of fuse_relu_depthwise_conv, test=develop, test=document_fix
revert-20712-fix_depthwise_conv
liu zhengxi 6 years ago committed by GitHub
parent a010d883b4
commit f855a86c93
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
2 changed files with 49 additions and 26 deletions
Show Stats Download Patch File Download Diff File

2
paddle/fluid/API.spec
Unescape View File

@ -63,7 +63,7 @@ paddle.fluid.CompiledProgram.__init__ (ArgSpec(args=['self', 'program_or_graph',
paddle.fluid.CompiledProgram.with_data_parallel (ArgSpec(args=['self', 'loss_name', 'build_strategy', 'exec_strategy', 'share_vars_from', 'places'], varargs=None, keywords=None, defaults=(None, None, None, None, None)), ('document', '3b61147fc4f54e1724aa9ead8a1d5f26'))
paddle.fluid.ExecutionStrategy ('paddle.fluid.core_avx.ExecutionStrategy', ('document', '535ce28c4671176386e3cd283a764084'))
paddle.fluid.ExecutionStrategy.__init__ __init__(self: paddle.fluid.core_avx.ParallelExecutor.ExecutionStrategy) -> None
paddle.fluid.BuildStrategy ('paddle.fluid.core_avx.BuildStrategy', ('document', 'eec64b9b7cba58b0a63687b4c34ffe56'))
paddle.fluid.BuildStrategy ('paddle.fluid.core_avx.BuildStrategy', ('document', '9c7ee090a0ab6896f5de996d59a2f645'))
paddle.fluid.BuildStrategy.GradientScaleStrategy ('paddle.fluid.core_avx.GradientScaleStrategy', ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
paddle.fluid.BuildStrategy.GradientScaleStrategy.__init__ __init__(self: paddle.fluid.core_avx.ParallelExecutor.BuildStrategy.GradientScaleStrategy, arg0: int) -> None
paddle.fluid.BuildStrategy.ReduceStrategy ('paddle.fluid.core_avx.ReduceStrategy', ('document', '6adf97f83acf6453d4a6a4b1070f3754'))

73
paddle/fluid/pybind/pybind.cc
Unescape View File

@ -1618,9 +1618,26 @@ All parameter, weight, gradient are variables in Paddle.
Examples:
.. code-block:: python
import os
import numpy as np
import paddle.fluid as fluid
os.environ["CPU_NUM"] = '2'
places = fluid.cpu_places()
data = fluid.layers.data(name="x", shape=[1], dtype="float32")
hidden = fluid.layers.fc(input=data, size=10)
loss = fluid.layers.mean(hidden)
fluid.optimizer.SGD(learning_rate=0.01).minimize(loss)
build_strategy = fluid.BuildStrategy()
build_strategy.enable_inplace = True
build_strategy.memory_optimize = True
build_strategy.reduce_strategy = fluid.BuildStrategy.ReduceStrategy.Reduce
program = fluid.compiler.CompiledProgram(fluid.default_main_program())
program = program.with_data_parallel(loss_name=loss.name,
build_strategy=build_strategy,
places=places)
)DOC");
py::enum_<BuildStrategy::ReduceStrategy>(build_strategy, "ReduceStrategy")
@ -1642,13 +1659,13 @@ All parameter, weight, gradient are variables in Paddle.
"BuildStrategy is finlaized.");
self.reduce_ = strategy;
},
R"DOC(The type is fluid.BuildStrategy.ReduceStrategy, there are two reduce
R"DOC((fluid.BuildStrategy.ReduceStrategy, optional): there are two reduce
strategies in ParallelExecutor, AllReduce and Reduce. If you want
that all the parameters' optimization are done on all devices independently,
you should choose AllReduce; if you choose Reduce, all the parameters'
you should choose AllReduce; otherwise, if you choose Reduce, all the parameters'
optimization will be evenly distributed to different devices, and then
broadcast the optimized parameter to other devices.
Default 'AllReduce'.
Default is 'AllReduce'.
Examples:
.. code-block:: python
@ -1666,11 +1683,11 @@ All parameter, weight, gradient are variables in Paddle.
"BuildStrategy is finalized.");
self.gradient_scale_ = strategy;
},
R"DOC(The type is fluid.BuildStrategy.GradientScaleStrategy, there are three
ways of defining :math:`loss@grad` in ParallelExecutor, CoeffNumDevice,
R"DOC((fluid.BuildStrategy.GradientScaleStrategy, optional): there are three
ways of defining :math:`loss@grad` in ParallelExecutor, that is, CoeffNumDevice,
One and Customized. By default, ParallelExecutor sets the :math:`loss@grad`
according to the number of devices. If you want to customize :math:`loss@grad`,
you can choose Customized. Default 'CoeffNumDevice'.
you can choose Customized. Default is 'CoeffNumDevice'.
Examples:
.. code-block:: python
@ -1728,9 +1745,9 @@ All parameter, weight, gradient are variables in Paddle.
"BuildStrategy is finlaized.");
self.debug_graphviz_path_ = path;
},
R"DOC(The type is STR, debug_graphviz_path indicates the path that
R"DOC((str, optional): debug_graphviz_path indicates the path that
writing the SSA Graph to file in the form of graphviz.
It is useful for debugging. Default ""
It is useful for debugging. Default is empty string, that is, ""
Examples:
.. code-block:: python
@ -1750,8 +1767,8 @@ All parameter, weight, gradient are variables in Paddle.
"BuildStrategy is finlaized.");
self.enable_sequential_execution_ = b;
},
R"DOC(The type is BOOL. If set True, the execution order of ops would
be the same as what is in the program. Default False.
R"DOC((bool, optional): If set True, the execution order of ops would
be the same as what is in the program. Default is False.
Examples:
.. code-block:: python
@ -1770,8 +1787,8 @@ All parameter, weight, gradient are variables in Paddle.
"BuildStrategy is finlaized.");
self.remove_unnecessary_lock_ = b;
},
R"DOC(The type is BOOL. If set True, some locks in GPU ops would be
released and ParallelExecutor would run faster. Default True.
R"DOC((bool, optional): If set True, some locks in GPU ops would be
released and ParallelExecutor would run faster. Default is True.
Examples:
.. code-block:: python
@ -1832,9 +1849,9 @@ All parameter, weight, gradient are variables in Paddle.
"BuildStrategy is finlaized.");
self.fuse_elewise_add_act_ops_ = b;
},
R"DOC(The type is BOOL, fuse_elewise_add_act_ops indicate whether
R"DOC((bool, optional): fuse_elewise_add_act_ops indicate whether
to fuse elementwise_add_op and activation_op,
it may make the execution faster. Default False
it may make the execution faster. Default is False.
Examples:
.. code-block:: python
@ -1853,11 +1870,11 @@ All parameter, weight, gradient are variables in Paddle.
"BuildStrategy is finlaized.");
self.fuse_relu_depthwise_conv_ = b;
},
R"DOC(The type is BOOL, fuse_relu_depthwise_conv indicate whether
R"DOC((bool, optional): fuse_relu_depthwise_conv indicate whether
to fuse relu and depthwise_conv2d,
it will save GPU memory and may make the execution faster.
This options is only available in GPU devices.
Default False.
Default is False.
Examples:
.. code-block:: python
@ -1876,12 +1893,20 @@ All parameter, weight, gradient are variables in Paddle.
"BuildStrategy is finlaized.");
self.fuse_broadcast_ops_ = b;
},
R"DOC(The type is BOOL, fuse_broadcast_op indicates whether
R"DOC((bool, optional): fuse_broadcast_op indicates whether
to fuse the broadcast ops. Note that, in Reduce mode,
fusing broadcast ops may make the program faster. Because
fusing broadcast OP equals delaying the execution of all
broadcast Ops, in this case, all nccl streams are used only
for NCCLReduce operations for a period of time. Default False.)DOC")
for NCCLReduce operations for a period of time. Default False.
Examples:
.. code-block:: python
import paddle.fluid as fluid
build_strategy = fluid.BuildStrategy()
build_strategy.fuse_broadcast_ops = True
)DOC")
.def_property("fuse_all_optimizer_ops",
[](const BuildStrategy &self) {
return self.fuse_all_optimizer_ops_ == true ||
@ -1900,14 +1925,12 @@ All parameter, weight, gradient are variables in Paddle.
"BuildStrategy is finlaized.");
self.sync_batch_norm_ = b;
},
R"DOC(The type is BOOL, sync_batch_norm indicates whether to use
R"DOC((bool, optional): sync_batch_norm indicates whether to use
synchronous batch normalization which synchronizes the mean
and variance through multi-devices in training phase.
Current implementation doesn't support FP16 training and CPU.
And only synchronous on one machine, not all machines.
Default False
And only synchronous on one machine, not all machines.
Default is False.
Examples:
.. code-block:: python
@ -1937,13 +1960,13 @@ All parameter, weight, gradient are variables in Paddle.
"True");
}
},
R"DOC(The type is BOOL or None, memory opitimize aims to save total memory
R"DOC((bool, optional): memory opitimize aims to save total memory
consumption, set to True to enable it.
Default None. None means framework would choose to use or not use
this strategy automatically. Currently, None means that it is
enabled when GC is disabled, and disabled when GC is enabled.
True means enabling and False means disabling. Default None.)DOC")
True means enabling and False means disabling. Default is None.)DOC")
.def_property(
"is_distribution",
[](const BuildStrategy &self) { return self.is_distribution_; },

Write Preview
Loading…
Cancel
Save