提交 1e2acd97 编写于 作者: L Luo Tao

refine ParallelDo doc

上级 21b98d29
...@@ -234,8 +234,56 @@ class BlockGuard(object): ...@@ -234,8 +234,56 @@ class BlockGuard(object):
class ParallelDo(object): class ParallelDo(object):
""" """
ParallelDo class is used to create a ParallelDo. ParallelDo is used to represent multi-thread data parallel processing.
It will be soon deprecated, please use ParallelExecutor instead.
Its vanilla implementation can be shown as the following (:math:`|` means
single thread and :math:`||||` means multiple threads)
.. code-block:: text
In the forward pass
| Split input onto different devices
| Copy parameter onto different devices
|||| Compute forward pass in parallel
| Merge output from different devices
In the backward pass
| Split output@grad onto different devices
|||| Compute backward pass in parallel
| accumulate param@grad from different devices to the first device
| Merge input@grad from different devices
 | Copy param@grad to the place of parallel_do_op
Examples:
.. code-block:: python
images = fluid.layers.data(name='pixel', shape=[1, 28, 28], dtype=DTYPE)
label = fluid.layers.data(name='label', shape=[1], dtype='int64')
# ParallelDo version & Single-thread version
if thread_num > 1:
places = fluid.layers.get_places(thread_num)
pd = fluid.layers.ParallelDo(places)
with pd.do():
images = pd.read_input(images)
label = pd.read_input(label)
predict = cnn_model(images)
cost = fluid.layers.cross_entropy(input=predict, label=label)
avg_cost = fluid.layers.mean(x=cost)
pd.write_output(avg_cost)
avg_cost = pd()
avg_cost = fluid.layers.mean(avg_cost)
else:
predict = cnn_model(images)
cost = fluid.layers.cross_entropy(input=predict, label=label)
avg_cost = fluid.layers.mean(x=cost)
.. warning::
It will be soon deprecated, please use ParallelExecutor instead.
""" """
def __init__(self, places, use_nccl=False, name=None): def __init__(self, places, use_nccl=False, name=None):
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册