1# Copyright 2019 Huawei Technologies Co., Ltd 2# 3# Licensed under the Apache License, Version 2.0 (the "License"); 4# you may not use this file except in compliance with the License. 5# You may obtain a copy of the License at 6# 7# http://www.apache.org/licenses/LICENSE-2.0 8# 9# Unless required by applicable law or agreed to in writing, software 10# distributed under the License is distributed on an "AS IS" BASIS, 11# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 12# See the License for the specific language governing permissions and 13# limitations under the License. 14 15import numpy as np 16 17import mindspore.context as context 18import mindspore.nn as nn 19from mindspore import Tensor, Parameter 20from mindspore.common.api import _cell_graph_executor 21from mindspore.communication.management import init 22from mindspore.nn import Dense 23from mindspore.nn import Momentum 24from mindspore.nn import TrainOneStepCell, WithLossCell 25from mindspore.ops import operations as P 26from mindspore.context import ParallelMode 27from mindspore.communication._comm_helper import GlobalComm 28 29class Net(nn.Cell): 30 def __init__(self, input_channel, out_channel): 31 super(Net, self).__init__() 32 weight_init1 = np.ones([64, 128]).astype(np.float32) 33 weight_init2 = np.ones([32, 64]).astype(np.float32) 34 self.weight1 = Parameter(Tensor(weight_init1), "loss_weight1", layerwise_parallel=True) 35 self.weight2 = Parameter(Tensor(weight_init2), "loss_weight2", layerwise_parallel=True) 36 self.fc = P.MatMul(transpose_b=True) 37 self.dense = Dense(input_channel, out_channel) 38 39 def construct(self, x): 40 x = self.dense(x) 41 x = self.fc(x, self.weight1) 42 x = self.fc(x, self.weight2) 43 return x 44 45 46def test_dense_gen_graph(): 47 context.set_context(mode=context.GRAPH_MODE) 48 context.reset_auto_parallel_context() 49 context.set_auto_parallel_context(parallel_mode=ParallelMode.HYBRID_PARALLEL, gradients_mean=True, device_num=8) 50 GlobalComm.CHECK_ENVS = False 51 init() 52 GlobalComm.CHECK_ENVS = True 53 network = Net(512, 128) 54 55 loss_fn = nn.SoftmaxCrossEntropyWithLogits() 56 optimizer = Momentum(filter(lambda x: x.requires_grad, network.get_parameters()), 57 learning_rate=0.1, 58 momentum=0.9) 59 network = WithLossCell(network, loss_fn) 60 network = TrainOneStepCell(network, optimizer) 61 62 predict = Tensor(np.ones([64, 512]).astype(np.float32) * 0.01) 63 label = Tensor(np.zeros([64, 32]).astype(np.float32)) 64 network.set_auto_parallel() 65 _cell_graph_executor.compile(network, predict, label) 66