# Copyright 2021 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 os import numpy as np from mindspore.communication.management import get_rank from mindspore import Tensor from mindspore import Parameter from mindspore import context from mindspore.ops import operations as P import mindspore.nn as nn from mindspore.train import Model from mindspore.context import ParallelMode from mindspore.communication.management import init from mindspore.communication.management import get_group_size class FakeDataInitMode: RandomInit = 0 OnesInit = 1 UniqueInit = 2 ZerosInit = 3 class FakeData: def __init__(self, size=1024, batch_size=32, image_size=(3, 224, 224), num_class=10, random_offset=0, use_parallel=False, fakedata_mode=FakeDataInitMode.RandomInit): self.size = size self.rank_batch_size = batch_size self.total_batch_size = self.rank_batch_size self.random_offset = random_offset self.image_size = image_size self.num_class = num_class self.rank_size = 1 self.rank_id = 0 self.batch_index = 0 self.image_data_type = np.float32 self.label_data_type = np.float32 self.is_onehot = True self.fakedata_mode = fakedata_mode if use_parallel: if 'CONTEXT_DEVICE_TARGET' in os.environ and os.environ['CONTEXT_DEVICE_TARGET'] == 'GPU': init(backend_name='nccl') else: init(backend_name='hccl') self.rank_size = get_group_size() self.rank_id = get_rank() self.total_batch_size = self.rank_batch_size * self.rank_size assert self.size % self.total_batch_size == 0 self.total_batch_data_size = (self.rank_size, self.rank_batch_size) + image_size def get_dataset_size(self): return int(self.size / self.total_batch_size) def get_reeat_count(self): return 1 def set_image_data_type(self, data_type): self.image_data_type = data_type def set_label_data_type(self, data_type): self.label_data_type = data_type def set_label_onehot(self, is_onehot=True): self.is_onehot = is_onehot def create_tuple_iterator(self, num_epochs=-1, do_copy=False): return self def __getitem__(self, batch_index): if batch_index * self.total_batch_size >= len(self): raise IndexError("{} index out of range".format(self.__class__.__name__)) rng_state = np.random.get_state() np.random.seed(batch_index + self.random_offset) if self.fakedata_mode == FakeDataInitMode.OnesInit: img = np.ones(self.total_batch_data_size) elif self.fakedata_mode == FakeDataInitMode.ZerosInit: img = np.zeros(self.total_batch_data_size) elif self.fakedata_mode == FakeDataInitMode.UniqueInit: total_size = 1 for i in self.total_batch_data_size: total_size = total_size* i img = np.reshape(np.arange(total_size)*0.0001, self.total_batch_data_size) else: img = np.random.randn(*self.total_batch_data_size) target = np.random.randint(0, self.num_class, size=(self.rank_size, self.rank_batch_size)) np.random.set_state(rng_state) img = img[self.rank_id] target = target[self.rank_id] img_ret = img.astype(self.image_data_type) target_ret = target.astype(self.label_data_type) if self.is_onehot: target_onehot = np.zeros(shape=(self.rank_batch_size, self.num_class)) target_onehot[np.arange(self.rank_batch_size), target] = 1 target_ret = target_onehot.astype(self.label_data_type) return Tensor(img_ret), Tensor(target_ret) def __len__(self): return self.size def __iter__(self): self.batch_index = 0 return self def reset(self): self.batch_index = 0 def __next__(self): if self.batch_index * self.total_batch_size < len(self): data = self[self.batch_index] self.batch_index += 1 return data raise StopIteration class NetWithSparseGatherV2(nn.Cell): def __init__(self, strategy=None, sparse=True): super(NetWithSparseGatherV2, self).__init__() self.axis = 0 self.sparse = sparse if sparse: self.weight = Parameter(Tensor(np.ones([8, 8]).astype(np.float32)), name="weight") self.gather = P.SparseGatherV2() else: self.weight = Parameter(Tensor(np.ones([8, 8]).astype(np.float32)), name="weight") self.gather = P.Gather() if strategy is not None: self.gather.shard(strategy) def construct(self, indices): x = self.gather(self.weight, indices, self.axis) return x def train_mindspore_impl(self, indices, epoch, batch_size, use_parallel=True): ds = FakeData(size=8, batch_size=batch_size, num_class=8, image_size=(), use_parallel=use_parallel) ds.set_image_data_type(np.int32) net = self net.set_train() loss = nn.SoftmaxCrossEntropyWithLogits() optimizer = nn.Adam(net.trainable_params()) optimizer.target = "CPU" model = Model(net, loss, optimizer) for _ in range(epoch): model.train(1, ds, dataset_sink_mode=False) output = net(indices) return output def test_allreduce_sparsegatherv2_adam_auto_parallel(): context.set_context(mode=context.GRAPH_MODE, device_target='Ascend') init(backend_name='hccl') context.set_auto_parallel_context(parallel_mode=ParallelMode.AUTO_PARALLEL, device_num=8, gradients_mean=True) indices = Tensor(np.array([0, 1, 2, 3, 4, 5, 6, 7]).astype(np.int32)) epoch = 3 batch_size = 1 context.set_context(enable_sparse=True) net = NetWithSparseGatherV2(sparse=True) output_sparse = net.train_mindspore_impl(indices, epoch, batch_size) net = NetWithSparseGatherV2(sparse=False) output = net.train_mindspore_impl(indices, epoch, batch_size) assert np.allclose(output.asnumpy(), output_sparse.asnumpy(), 0.001, 0.001)