Minor modifications to run locally + comments for explanation

[jjc10]

Minor changes to have this run locally (for instance removed num_workers) on dataloaders which don't work for us (or at least seem to slow things down a lot since we're running this on a single machine).

To run the code:

python train_cifar100.py --data-root data/cifar10 --dataset cifar10 --result_dir results/msdnet_ge_cifar10 --arch msdnet_ge --ensemble_reweight 0.5 --batch-size 64 --nBlocks 10 --stepmode even --step 2 --base 4 --nChannels 16 --workers 1 --lr_f 0.1 --lr_milestones 150,225 --epochs 200 --weight-decay 1e-4

Also when running this we will likely run into an issue with tensorboard pytorch/pytorch#69894.

The file to fix is: site-packages/torch/utils/tensorboard/__init__.py
To get the exact location, activation your environment, start a python shell, import torch and print(torch.file)

The fix is to use proper import syntax for LooseVersion
The file should look like this:

import tensorboard
from distutils.version import LooseVersion


if not hasattr(tensorboard, '__version__') or LooseVersion(tensorboard.__version__) < LooseVersion('1.15'):
    raise ImportError('TensorBoard logging requires TensorBoard version 1.15 or above')


del LooseVersion
del tensorboard

from .writer import FileWriter, SummaryWriter  # noqa: F401
from tensorboard.summary.writer.record_writer import RecordWriter  # noqa: F401