mindspore/tests/ut/python/communication/test_data_parallel_resnet.py

# Copyright 2020 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.

"""
resnet50 example
"""
import numpy as np

import mindspore.context as context
import mindspore.nn as nn
from mindspore import Tensor, Model, ParallelMode
from mindspore.nn.optim import Momentum
from mindspore.ops.operations import TensorAdd
from ....dataset_mock import MindData


def conv3x3(in_channels, out_channels, stride=1, padding=1, pad_mode='pad'):
    """3x3 convolution """
    return nn.Conv2d(in_channels, out_channels,
                     kernel_size=3, stride=stride, padding=padding, pad_mode=pad_mode)


def conv1x1(in_channels, out_channels, stride=1, padding=0, pad_mode='pad'):
    """1x1 convolution"""
    return nn.Conv2d(in_channels, out_channels,
                     kernel_size=1, stride=stride, padding=padding, pad_mode=pad_mode)


class ResidualBlock(nn.Cell):
    """
    residual Block
    """
    expansion = 4

    def __init__(self,
                 in_channels,
                 out_channels,
                 stride=1,
                 down_sample=False):
        super(ResidualBlock, self).__init__()

        out_chls = out_channels // self.expansion
        self.conv1 = conv1x1(in_channels, out_chls, stride=1, padding=0)
        self.bn1 = nn.BatchNorm2d(out_chls)

        self.conv2 = conv3x3(out_chls, out_chls, stride=stride, padding=1)
        self.bn2 = nn.BatchNorm2d(out_chls)

        self.conv3 = conv1x1(out_chls, out_channels, stride=1, padding=0)
        self.bn3 = nn.BatchNorm2d(out_channels)

        self.relu = nn.ReLU()
        self.downsample = down_sample

        self.conv_down_sample = conv1x1(in_channels, out_channels,
                                        stride=stride, padding=0)
        self.bn_down_sample = nn.BatchNorm2d(out_channels)
        self.add = TensorAdd()

    def construct(self, x):
        """
        :param x:
        :return:
        """
        identity = x

        out = self.conv1(x)
        out = self.bn1(out)
        out = self.relu(out)

        out = self.conv2(out)
        out = self.bn2(out)
        out = self.relu(out)

        out = self.conv3(out)
        out = self.bn3(out)

        if self.downsample:
            identity = self.conv_down_sample(identity)
            identity = self.bn_down_sample(identity)

        out = self.add(out, identity)
        out = self.relu(out)

        return out


class ResNet18(nn.Cell):
    """
    resnet nn.Cell
    """

    def __init__(self, block, num_classes=100):
        super(ResNet18, self).__init__()

        self.conv1 = nn.Conv2d(3, 64, kernel_size=7, stride=2, padding=3, pad_mode='pad')
        self.bn1 = nn.BatchNorm2d(64)
        self.relu = nn.ReLU()
        self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, pad_mode='same')

        self.layer1 = self.MakeLayer(
            block, 2, in_channels=64, out_channels=256, stride=1)
        self.layer2 = self.MakeLayer(
            block, 2, in_channels=256, out_channels=512, stride=2)
        self.layer3 = self.MakeLayer(
            block, 2, in_channels=512, out_channels=1024, stride=2)
        self.layer4 = self.MakeLayer(
            block, 2, in_channels=1024, out_channels=2048, stride=2)

        self.avgpool = nn.AvgPool2d(7, 1)
        self.flatten = nn.Flatten()
        self.fc = nn.Dense(512 * block.expansion, num_classes)

    def MakeLayer(self, block, layer_num, in_channels, out_channels, stride):
        """
        make block layer
        :param block:
        :param layer_num:
        :param in_channels:
        :param out_channels:
        :param stride:
        :return:
        """
        layers = []
        resblk = block(in_channels, out_channels,
                       stride=stride, down_sample=True)
        layers.append(resblk)

        for _ in range(1, layer_num):
            resblk = block(out_channels, out_channels, stride=1)
            layers.append(resblk)

        return nn.SequentialCell(layers)

    def construct(self, x):
        """
        :param x:
        :return:
        """
        x = self.conv1(x)
        x = self.bn1(x)
        x = self.relu(x)
        x = self.maxpool(x)

        x = self.layer1(x)
        x = self.layer2(x)
        x = self.layer3(x)
        x = self.layer4(x)

        x = self.avgpool(x)
        x = self.flatten(x)
        x = self.fc(x)

        return x


class ResNet9(nn.Cell):
    """
    resnet nn.Cell
    """

    def __init__(self, block, num_classes=100):
        super(ResNet9, self).__init__()

        self.conv1 = nn.Conv2d(3, 64, kernel_size=7, stride=2, padding=3, pad_mode='pad')
        self.bn1 = nn.BatchNorm2d(64)
        self.relu = nn.ReLU()
        self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, pad_mode='same')

        self.layer1 = self.MakeLayer(
            block, 1, in_channels=64, out_channels=256, stride=1)
        self.layer2 = self.MakeLayer(
            block, 1, in_channels=256, out_channels=512, stride=2)
        self.layer3 = self.MakeLayer(
            block, 1, in_channels=512, out_channels=1024, stride=2)
        self.layer4 = self.MakeLayer(
            block, 1, in_channels=1024, out_channels=2048, stride=2)

        self.avgpool = nn.AvgPool2d(7, 1)
        self.flatten = nn.Flatten()
        self.fc = nn.Dense(512 * block.expansion, num_classes)

    def MakeLayer(self, block, layer_num, in_channels, out_channels, stride):
        """
        make block layer
        :param block:
        :param layer_num:
        :param in_channels:
        :param out_channels:
        :param stride:
        :return:
        """
        layers = []
        resblk = block(in_channels, out_channels,
                       stride=stride, down_sample=True)
        layers.append(resblk)

        for _ in range(1, layer_num):
            resblk = block(out_channels, out_channels, stride=1)
            layers.append(resblk)

        return nn.SequentialCell(layers)

    def construct(self, x):
        """
        :param x:
        :return:
        """
        x = self.conv1(x)
        x = self.bn1(x)
        x = self.relu(x)
        x = self.maxpool(x)

        x = self.layer1(x)
        x = self.layer2(x)
        x = self.layer3(x)
        x = self.layer4(x)

        x = self.avgpool(x)
        x = self.flatten(x)
        x = self.fc(x)

        return x


def resnet9(classnum):
    return ResNet9(ResidualBlock, classnum)


class DatasetLenet(MindData):
    """DatasetLenet definition"""

    def __init__(self, predict, label, length=3, size=None, batch_size=None,
                 np_types=None, output_shapes=None, input_indexs=()):
        super(DatasetLenet, self).__init__(size=size, batch_size=batch_size,
                                           np_types=np_types, output_shapes=output_shapes,
                                           input_indexs=input_indexs)
        self.predict = predict
        self.label = label
        self.index = 0
        self.length = length

    def __iter__(self):
        return self

    def __next__(self):
        if self.index >= self.length:
            raise StopIteration
        self.index += 1
        return self.predict, self.label

    def reset(self):
        self.index = 0


def test_resnet_train_tensor():
    """test_resnet_train_tensor"""
    batch_size = 1
    size = 2
    context.set_context(mode=context.GRAPH_MODE)
    context.reset_auto_parallel_context()
    context.set_auto_parallel_context(parallel_mode=ParallelMode.DATA_PARALLEL, device_num=size,
                                      parameter_broadcast=True)
    one_hot_len = 10
    dataset_types = (np.float32, np.float32)
    dataset_shapes = [[batch_size, 3, 224, 224], [batch_size, one_hot_len]]
    predict = Tensor(np.ones([batch_size, 3, 224, 224]).astype(np.float32) * 0.01)
    label = Tensor(np.zeros([batch_size, one_hot_len]).astype(np.float32))
    dataset = DatasetLenet(predict, label, 2,
                           size=2, batch_size=2,
                           np_types=dataset_types,
                           output_shapes=dataset_shapes,
                           input_indexs=(0, 1))
    dataset.reset()
    network = resnet9(one_hot_len)
    network.set_train()
    loss_fn = nn.SoftmaxCrossEntropyWithLogits()
    optimizer = Momentum(filter(lambda x: x.requires_grad, network.get_parameters()), learning_rate=0.1, momentum=0.9)
    model = Model(network=network, loss_fn=loss_fn, optimizer=optimizer)
    model.train(epoch=2, train_dataset=dataset, dataset_sink_mode=False)
    context.set_context(mode=context.GRAPH_MODE)
    context.reset_auto_parallel_context()


class_num = 10


def get_dataset():
    dataset_types = (np.float32, np.float32)
    dataset_shapes = ((32, 3, 224, 224), (32, class_num))

    dataset = MindData(size=2, batch_size=1,
                       np_types=dataset_types,
                       output_shapes=dataset_shapes,
                       input_indexs=(0, 1))
    return dataset