提交 8b7516ed 编写于 作者: Q qijun

test memopt fit a line demo with nccl

上级 abb94057
...@@ -26,8 +26,16 @@ fluid.default_startup_program().random_seed = 111 ...@@ -26,8 +26,16 @@ fluid.default_startup_program().random_seed = 111
x = fluid.layers.data(name='x', shape=[13], dtype='float32') x = fluid.layers.data(name='x', shape=[13], dtype='float32')
y = fluid.layers.data(name='y', shape=[1], dtype='float32') y = fluid.layers.data(name='y', shape=[1], dtype='float32')
places = fluid.layers.get_places(device_count=2, device_type='CPU') device_type = 'CPU'
pd = fluid.layers.ParallelDo(places) use_nccl = False
place = fluid.CPUPlace()
if fluid.core.is_compiled_with_cuda():
device_type = 'CUDA'
use_nccl = True
place = fluid.CUDAPlace(0)
places = fluid.layers.get_places(device_count=2, device_type=device_type)
pd = fluid.layers.ParallelDo(places, use_nccl=use_nccl)
with pd.do(): with pd.do():
x_ = pd.read_input(x) x_ = pd.read_input(x)
y_ = pd.read_input(y) y_ = pd.read_input(y)
...@@ -54,7 +62,6 @@ train_reader = paddle.batch( ...@@ -54,7 +62,6 @@ train_reader = paddle.batch(
# paddle.dataset.uci_housing.train(), buf_size=500), # paddle.dataset.uci_housing.train(), buf_size=500),
# batch_size=BATCH_SIZE) # batch_size=BATCH_SIZE)
place = fluid.CPUPlace()
feeder = fluid.DataFeeder(place=place, feed_list=[x, y]) feeder = fluid.DataFeeder(place=place, feed_list=[x, y])
exe = fluid.Executor(place) exe = fluid.Executor(place)
......
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