Support for the TSV format dataset about pytorch-nli HOT 2 CLOSED

Currently, it is reading from the JSON format dataset. But it can be modified very easily to read from TSV format. I am using torchtext dataset splits method to prepare the SNLI and MultiNLI datasets.

By default splits methods reads from JSON files which are stored into .data folder. You can give your .tsv file name here and put your TSV file in .data folder. It should work right away. If you want to give own custom file path I would recommend you to use the tabular dataset from the torchtext and specify the path of TSV files.

from pytorch-nli.

Thanks for your suggestions. :-)

from pytorch-nli.