PyTorch-NLP 0.5.0

[PetrochukM]

These changes come after using the library for another couple months at https://wellsaidlabs.com/

Major Changes

• The README has been changed to have more a story and vision. The original README didn't concretely layout a focus apart from "Supporting Rapid Prototyping".
• Added torchnlp.random enabling control over numpy, torch, and random states.
• Reworked samplers to be more flexible and extensible.
• Added support for Python 3.5.

Minor Changes

• Better naming with regards to stoi and itos
• stack_and_pad_tensors returns a namedtuple
• Deprecated torchnlp.datasets.Dataset in favor of third party libraries like pandas
• Deprecated torchnlp.utils.datasets_iterator in favor of Pythons itertools.chain.
• Deprecated torchnlp.utils.shuffle in favor of torchnlp.random
• Fixed examples/snli
• Added torch.utils.get_total_parameters
• Depcrecated torch.utils.resplit_datasets in favor of torch.utils.split_list and torchnlp.random.
• Remove pandas and requests as dependencies, slimming the package down
• Fix issue Allow StaticTokenizerEncoder to take any iterable #85

[codecov-io]

Codecov Report

Merging #84 into master will increase coverage by 0.24%.
The diff coverage is 97.68%.

@@            Coverage Diff             @@
##           master      #84      +/-   ##
==========================================
+ Coverage   94.17%   94.41%   +0.24%     
==========================================
  Files          57       63       +6     
  Lines        1545     1594      +49     
==========================================
+ Hits         1455     1505      +50     
+ Misses         90       89       -1

Impacted Files	Coverage Δ
torchnlp/_third_party/lazy_loader.py	81.81% <ø> (ø)
torchnlp/samplers/bptt_sampler.py	100% <ø> (ø)	⬆️
torchnlp/datasets/penn_treebank.py	95.23% <ø> (ø)	⬆️
torchnlp/datasets/wikitext_2.py	95.23% <ø> (ø)	⬆️
torchnlp/datasets/__init__.py	100% <ø> (ø)	⬆️
torchnlp/samplers/oom_batch_sampler.py	100% <100%> (ø)
torchnlp/encoders/label_encoder.py	94.11% <100%> (ø)	⬆️
torchnlp/datasets/trec.py	100% <100%> (ø)	⬆️
torchnlp/datasets/iwslt.py	95.91% <100%> (-0.09%)	⬇️
torchnlp/samplers/sorted_sampler.py	100% <100%> (ø)	⬆️
... and 42 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 49bb1d7...7f82397. Read the comment docs.