Refactor hub interface for batched inference (#1539) (#1539)

Summary:
# Before submitting

- [x] Was this discussed/approved via a Github issue? (no need for typos, doc improvements)
- [x] Did you read the [contributor guideline](https://github.com/pytorch/fairseq/blob/master/CONTRIBUTING.md)?
- [x] Did you make sure to update the docs?
- [] Did you write any new necessary tests?

## What does this PR do?
Fixes facebookresearch/fairseq#1508.
Pull Request resolved: facebookresearch/fairseq#1539

Pulled By: myleott

Differential Revision: D19216104

fbshipit-source-id: 14917c1459b8794eeb74c09a16b9899c366242d2

examples/language_model/README.md

@@ -26,6 +26,10 @@ torch.hub.list('pytorch/fairseq')  # [..., 'transformer_lm.wmt19.en', ...]
 
 # Load an English LM trained on WMT'19 News Crawl data
 en_lm = torch.hub.load('pytorch/fairseq', 'transformer_lm.wmt19.en', tokenizer='moses', bpe='fastbpe')
+en_lm.eval()  # disable dropout
+
+# Move model to GPU
+en_lm.cuda()
 
 # Sample from the language model
 en_lm.sample('Barack Obama', beam=1, sampling=True, sampling_topk=10, temperature=0.8)

examples/translation/README.md

@@ -31,13 +31,21 @@ torch.hub.list('pytorch/fairseq')  # [..., 'transformer.wmt16.en-de', ... ]
 
 # Load a transformer trained on WMT'16 En-De
 en2de = torch.hub.load('pytorch/fairseq', 'transformer.wmt16.en-de', tokenizer='moses', bpe='subword_nmt')
+en2de.eval()  # disable dropout
 
 # The underlying model is available under the *models* attribute
 assert isinstance(en2de.models[0], fairseq.models.transformer.TransformerModel)
 
+# Move model to GPU for faster translation
+en2de.cuda()
+
 # Translate a sentence
 en2de.translate('Hello world!')
 # 'Hallo Welt!'
+
+# Batched translation
+en2de.translate(['Hello world!', 'The cat sat on the mat.'])
+# ['Hallo Welt!', 'Die Katze saß auf der Matte.']
 ```
 
 Loading custom models:

fairseq/hub_utils.py

@@ -11,6 +11,7 @@
 from typing import List, Dict, Iterator, Tuple, Any
 
 import os
+from typing import List, Dict, Iterator, Tuple, Any
 
 import torch
 from torch import nn