Support PyTorch 1.3

[gpengzhi]

resolve #247
resolve #234
resolve #231

[codecov]

Codecov Report

Merging #249 into master will decrease coverage by 0.05%.
The diff coverage is 86.89%.

@@            Coverage Diff             @@
##           master     #249      +/-   ##
==========================================
- Coverage   83.09%   83.04%   -0.06%     
==========================================
  Files         192      195       +3     
  Lines       15196    15290      +94     
==========================================
+ Hits        12627    12697      +70     
- Misses       2569     2593      +24

Impacted Files	Coverage Δ
texar/torch/data/data/__init__.py	100% <100%> (ø)	⬆️
texar/torch/data/data/data_iterators.py	81.88% <100%> (-2.77%)	⬇️
texar/torch/data/data/data_iterators_test.py	97.6% <100%> (-0.33%)	⬇️
texar/torch/data/data/data_iterators_utils.py	72.72% <72.72%> (ø)
texar/torch/data/data/sampler.py	90.44% <90.44%> (ø)
texar/torch/data/data/sampler_test.py	97.87% <97.87%> (ø)
texar/torch/core/layers.py	87.7% <0%> (-0.54%)	⬇️
... and 1 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update f2f4f95...4c6702e. Read the comment docs.

[gpengzhi]

Tested on seq2seq_attn, gpt-2, and sentence_classifier examples.

[huzecong]

LGTM.

[AvinashBukkittu]

Thanks for doing this! With new releases, Pytorch is making changes to a few attribute names in its core module, hence we end up making some updates to our data modules. LGTM.