base.py

import os
from collections import OrderedDict

import torch
import torch.nn as nn
import torch.nn.functional as F
from torch import optim
from torch.utils.data import DataLoader
from torch.utils.data.distributed import DistributedSampler
from torchvision import transforms
from torchvision.datasets import MNIST

try:
    from test_tube import HyperOptArgumentParser
except ImportError:
    # TODO: this should be discussed and moved out of this package
    raise ImportError('Missing test-tube package.')

from pytorch_lightning.core.decorators import data_loader
from pytorch_lightning.core.lightning import LightningModule


class TestingMNIST(MNIST):

    def __init__(self, root, train=True, transform=None, target_transform=None,
                 download=False, num_samples=8000):
        super(TestingMNIST, self).__init__(
            root,
            train=train,
            transform=transform,
            target_transform=target_transform,
            download=download
        )
        # take just a subset of MNIST dataset
        self.data = self.data[:num_samples]
        self.targets = self.targets[:num_samples]


class LightningTestModelBase(LightningModule):
    """
    Base LightningModule for testing. Implements only the required
    interface
    """

    def __init__(self, hparams, force_remove_distributed_sampler=False):
        """
        Pass in parsed HyperOptArgumentParser to the model
        :param hparams:
        """
        # init superclass
        super(LightningTestModelBase, self).__init__()
        self.hparams = hparams

        self.batch_size = hparams.batch_size

        # if you specify an example input, the summary will show input/output for each layer
        self.example_input_array = torch.rand(5, 28 * 28)

        # remove to test warning for dist sampler
        self.force_remove_distributed_sampler = force_remove_distributed_sampler

        # build model
        self.__build_model()

    # ---------------------
    # MODEL SETUP
    # ---------------------
    def __build_model(self):
        """
        Layout model
        :return:
        """
        self.c_d1 = nn.Linear(in_features=self.hparams.in_features,
                              out_features=self.hparams.hidden_dim)
        self.c_d1_bn = nn.BatchNorm1d(self.hparams.hidden_dim)
        self.c_d1_drop = nn.Dropout(self.hparams.drop_prob)

        self.c_d2 = nn.Linear(in_features=self.hparams.hidden_dim,
                              out_features=self.hparams.out_features)

    # ---------------------
    # TRAINING
    # ---------------------
    def forward(self, x):
        """
        No special modification required for lightning, define as you normally would
        :param x:
        :return:
        """

        x = self.c_d1(x)
        x = torch.tanh(x)
        x = self.c_d1_bn(x)
        x = self.c_d1_drop(x)

        x = self.c_d2(x)
        logits = F.log_softmax(x, dim=1)

        return logits

    def loss(self, labels, logits):
        nll = F.nll_loss(logits, labels)
        return nll

    def training_step(self, batch, batch_idx):
        """
        Lightning calls this inside the training loop
        :param batch:
        :return:
        """
        # forward pass
        x, y = batch
        x = x.view(x.size(0), -1)

        y_hat = self.forward(x)

        # calculate loss
        loss_val = self.loss(y, y_hat)

        # in DP mode (default) make sure if result is scalar, there's another dim in the beginning
        if self.trainer.use_dp:
            loss_val = loss_val.unsqueeze(0)

        # alternate possible outputs to test
        if self.trainer.batch_idx % 1 == 0:
            output = OrderedDict({
                'loss': loss_val,
                'progress_bar': {'some_val': loss_val * loss_val},
                'log': {'train_some_val': loss_val * loss_val},
            })

            return output
        if self.trainer.batch_idx % 2 == 0:
            return loss_val

    # ---------------------
    # TRAINING SETUP
    # ---------------------
    def configure_optimizers(self):
        """
        return whatever optimizers we want here.
        :return: list of optimizers
        """
        # try no scheduler for this model (testing purposes)
        if self.hparams.optimizer_name == 'lbfgs':
            optimizer = optim.LBFGS(self.parameters(), lr=self.hparams.learning_rate)
        else:
            optimizer = optim.Adam(self.parameters(), lr=self.hparams.learning_rate)

        # test returning only 1 list instead of 2
        return optimizer

    def _dataloader(self, train):
        # init data generators
        transform = transforms.Compose([transforms.ToTensor(),
                                        transforms.Normalize((0.5,), (1.0,))])
        dataset = TestingMNIST(root=self.hparams.data_root, train=train,
                               transform=transform, download=True, num_samples=2000)

        # when using multi-node we need to add the datasampler
        train_sampler = None
        batch_size = self.hparams.batch_size

        try:
            if self.use_ddp and not self.force_remove_distributed_sampler:
                train_sampler = DistributedSampler(dataset, rank=self.trainer.proc_rank)
                batch_size = batch_size // self.trainer.world_size  # scale batch size
        except Exception:
            pass

        should_shuffle = train_sampler is None
        loader = DataLoader(
            dataset=dataset,
            batch_size=batch_size,
            shuffle=should_shuffle,
            sampler=train_sampler
        )

        return loader

    @data_loader
    def train_dataloader(self):
        return self._dataloader(train=True)

    @staticmethod
    def add_model_specific_args(parent_parser, root_dir):  # pragma: no cover
        """
        Parameters you define here will be available to your model through self.hparams
        :param parent_parser:
        :param root_dir:
        :return:
        """
        parser = HyperOptArgumentParser(strategy=parent_parser.strategy, parents=[parent_parser])

        # param overwrites
        # parser.set_defaults(gradient_clip_val=5.0)

        # network params
        parser.opt_list('--drop_prob', default=0.2, options=[0.2, 0.5], type=float, tunable=False)
        parser.add_argument('--in_features', default=28 * 28, type=int)
        parser.add_argument('--out_features', default=10, type=int)
        # use 500 for CPU, 50000 for GPU to see speed difference
        parser.add_argument('--hidden_dim', default=50000, type=int)

        # data
        parser.add_argument('--data_root', default=os.path.join(root_dir, 'mnist'), type=str)

        # training params (opt)
        parser.opt_list('--learning_rate', default=0.001 * 8, type=float,
                        options=[0.0001, 0.0005, 0.001, 0.005],
                        tunable=False)
        parser.opt_list('--optimizer_name', default='adam', type=str,
                        options=['adam'], tunable=False)

        # if using 2 nodes with 4 gpus each the batch size here
        #  (256) will be 256 / (2*8) = 16 per gpu
        parser.opt_list('--batch_size', default=256 * 8, type=int,
                        options=[32, 64, 128, 256], tunable=False,
                        help='batch size will be divided over all gpus being used across all nodes')
        return parser