1# Copyright 2020 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 as ms 18from mindspore.common import dtype as mstype 19from mindspore import context, Tensor, Parameter 20from mindspore.nn import Cell, Momentum 21from mindspore.ops import operations as P 22from mindspore.train import Model 23from tests.dataset_mock import MindData 24 25 26class Dataset(MindData): 27 def __init__(self, predict, label, length=3): 28 super(Dataset, self).__init__(size=length) 29 self.predict = predict 30 self.label = label 31 self.index = 0 32 self.length = length 33 34 def __iter__(self): 35 return self 36 37 def __next__(self): 38 if self.index >= self.length: 39 raise StopIteration 40 self.index += 1 41 return self.predict, self.label 42 43 def reset(self): 44 self.index = 0 45 46 47class Net(Cell): 48 def __init__(self, weight, start, limit, delta, strategy1=None, strategy2=None, strategy3=None): 49 super().__init__() 50 self.mul = P.Mul().shard(strategy1) 51 if isinstance(start, float): 52 self.type = mstype.float32 53 else: 54 self.type = mstype.int32 55 self.start = Tensor(start, self.type) 56 self.limit = Tensor(limit, self.type) 57 self.delta = Tensor(delta, self.type) 58 self.range = P.Range() 59 self.range.shard(strategy2) 60 self.mul2 = P.Mul().shard(strategy3) 61 self.weight = Parameter(weight, "w") 62 63 def construct(self, x, b): 64 r_out = self.range(self.start, self.limit, self.delta) 65 out = self.mul(x, self.weight) 66 out = self.mul2(out, r_out) 67 return out 68 69 70dev_num = 4 71_x = Tensor(np.ones([64 // dev_num, 8]), dtype=ms.float32) 72_b = Tensor(np.ones([8]), dtype=ms.float32) 73_w1 = Tensor(np.ones([64, 8]), dtype=ms.float32) 74 75 76def compile_net(net): 77 learning_rate = 0.1 78 momentum = 0.9 79 epoch_size = 2 80 dataset = Dataset(_x, _b) 81 opt = Momentum(net.trainable_params(), learning_rate, momentum) 82 model = Model(net, optimizer=opt) 83 model.train(epoch_size, dataset, dataset_sink_mode=False) 84 context.reset_auto_parallel_context() 85 86 87def test_range(): 88 context.set_auto_parallel_context(parallel_mode="semi_auto_parallel", device_num=dev_num, global_rank=2) 89 strategy1 = ((2, 2), (2, 2)) 90 strategy2 = ((2,),) 91 strategy3 = ((2, 2), (2,)) 92 net = Net(_w1, 0, 8, 1, strategy1, strategy2, strategy3) 93 compile_net(net) 94 95 96def test_range2(): 97 context.set_auto_parallel_context(parallel_mode="semi_auto_parallel", device_num=dev_num, global_rank=0) 98 strategy1 = ((4, 1), (4, 1)) 99 strategy2 = ((1,),) 100 strategy3 = ((4, 1), (1,)) 101 net = Net(_w1, 0.0, 4.0, 0.5, strategy1, strategy2, strategy3) 102 compile_net(net) 103 104 105def test_range3(): 106 context.set_auto_parallel_context(parallel_mode="auto_parallel", device_num=dev_num, global_rank=2) 107 net = Net(_w1, 0.0, 4.0, 0.5) 108 compile_net(net) 109