Is the `.data` directory supposed to be auto-created when using the experimental datasets?

[thvasilo]

I just tried running the following on a fresh instance:

import torch
import torchtext

# import datasets
from torchtext.experimental.datasets import IMDB

SEED = 11

torch.manual_seed(SEED)
torch.backends.cudnn.deterministic = True

# use the default tokenizer
train_dataset, test_dataset = IMDB()
vocab = train_dataset.get_vocab()

and got the error RuntimeError: Download directory .data does not exist. Did you create it?

I expected that the library would create the directory on its own, like it does for the .vector_cache directory. Is that not the case?

[zhangguanheng66]

Fixed by #797