Training with DDP could fail at startup due to FileExistsError

[alumae]

In multi-GPU mode with DDP, starting the training can fail with the following error:

Traceback (most recent call last):
  File "train.py", line 109, in <module>
    main(hyperparams)
  File "train.py", line 62, in main
    trainer.fit(model)
  File "/home/tanel/miniconda3/lib/python3.7/site-packages/pytorch_lightning/trainer/trainer.py", line 343, in fit
    mp.spawn(self.ddp_train, nprocs=self.num_gpus, args=(model,))
  File "/home/tanel/miniconda3/lib/python3.7/site-packages/torch/multiprocessing/spawn.py", line 171, in spawn
    while not spawn_context.join():
  File "/home/tanel/miniconda3/lib/python3.7/site-packages/torch/multiprocessing/spawn.py", line 118, in join
    raise Exception(msg)
Exception: 

-- Process 1 terminated with the following error:
Traceback (most recent call last):
  File "/home/tanel/miniconda3/lib/python3.7/site-packages/torch/multiprocessing/spawn.py", line 19, in _wrap
    fn(i, *args)
  File "/home/tanel/miniconda3/lib/python3.7/site-packages/pytorch_lightning/trainer/ddp_mixin.py", line 181, in ddp_train
    self.run_pretrain_routine(model)
  File "/home/tanel/miniconda3/lib/python3.7/site-packages/pytorch_lightning/trainer/trainer.py", line 415, in run_pretrain_routine
    self.configure_checkpoint_callback()
  File "/home/tanel/miniconda3/lib/python3.7/site-packages/pytorch_lightning/trainer/callback_config_mixin.py", line 28, in configure_checkpoint_callback
    filepath=ckpt_path
  File "/home/tanel/miniconda3/lib/python3.7/site-packages/pytorch_lightning/callbacks/pt_callbacks.py", line 202, in __init__
    os.makedirs(filepath)
  File "/home/tanel/miniconda3/lib/python3.7/os.py", line 221, in makedirs
    mkdir(name, mode)
FileExistsError: [Errno 17] File exists: '/home/tanel/devel/torch-xvectors/lightning_logs/version_151/checkpoints'

It happens quite rarely.

Desktop (please complete the following information):

• OS: Linux
• Version 0.5.3.2 (git as of 2019-11-20)

Fix is easy, I'll create a PR.