mindspore/tests/ut/python/parallel/test_step_parallel.py

80 lines
2.7 KiB
Python

# Copyright 2019 Huawei Technologies Co., Ltd
#
# 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.
import numpy as np
import mindspore as ms
import mindspore.nn as nn
from mindspore import Tensor
from mindspore import context
from mindspore.common.api import _executor
from mindspore.ops import composite as C
from mindspore.ops import operations as P
from tests.ut.python.ops.test_math_ops import VirtualLoss
grad_all = C.GradOperation(get_all=True)
class NetWithLoss(nn.Cell):
def __init__(self, network):
super(NetWithLoss, self).__init__()
self.loss = VirtualLoss()
self.network = network
def construct(self, x, y, b, a):
predict = self.network(x, y, b, a)
return self.loss(predict)
class GradWrap(nn.Cell):
def __init__(self, network):
super(GradWrap, self).__init__()
self.network = network
def construct(self, x, y, b, a):
return grad_all(self.network)(x, y, b, a)
def test_two_matmul():
class Net(nn.Cell):
def __init__(self, strategy1, strategy2, strategy3, strategy4):
super().__init__()
self.matmul1 = P.MatMul().set_strategy(strategy1)
self.matmul2 = P.MatMul().set_strategy(strategy2)
self.matmul3 = P.MatMul().set_strategy(strategy3)
self.matmul4 = P.MatMul().set_strategy(strategy4)
def construct(self, x, y, b, a):
out = self.matmul1(x, y)
out1 = self.matmul2(out, b)
out2 = self.matmul3(out, a)
out3 = self.matmul4(out1, out2)
return out3
context.set_auto_parallel_context(device_num=8, global_rank=0)
strategy1 = ((2, 2), (2, 2))
strategy2 = ((1, 8), (8, 1))
strategy3 = ((4, 1), (1, 2))
strategy4 = ((4, 2), (2, 1))
net = GradWrap(NetWithLoss(Net(strategy1, strategy2, strategy3, strategy4)))
context.set_auto_parallel_context(parallel_mode="semi_auto_parallel")
x = Tensor(np.ones([128, 32]), dtype=ms.float32)
y = Tensor(np.ones([32, 128]), dtype=ms.float32)
b = Tensor(np.ones([128, 128]), dtype=ms.float32)
a = Tensor(np.ones([128, 128]), dtype=ms.float32)
net.set_auto_parallel()
_executor.compile(net, x, y, b, a)