Fix lint after updating pylint

Author: leezu

src/gluonnlp/data/batchify/embedding.py

@@ -206,21 +206,21 @@ def _context_generator(sentence_boundaries, window, batch_size,
                     # In SkipGram mode, there may be some leftover contexts
                     # form the last batch
                     continue
-                elif i < num_rows:
-                    num_context_skip = 0
-                    context_row.append(i)
-                    context_col.append(context)
-                    if cbow:
-                        context_data.append(1.0 / len(contexts))
-                    else:
-                        center_batch.append(center)
-                        context_data.append(1)
-                        i += 1
-                else:
+                if i >= num_rows:
                     num_context_skip = j
                     assert not cbow
                     break
 
+                num_context_skip = 0
+                context_row.append(i)
+                context_col.append(context)
+                if cbow:
+                    context_data.append(1.0 / len(contexts))
+                else:
+                    center_batch.append(center)
+                    context_data.append(1)
+                    i += 1
+
             if cbow:
                 center_batch.append(center)
                 i += 1

src/gluonnlp/data/stream.py

@@ -270,8 +270,7 @@ def run(self):
                 c = self._controlq.get(False)
                 if c is None:
                     break
-                else:
-                    raise RuntimeError('Got unexpected control code {}'.format(repr(c)))
+                raise RuntimeError('Got unexpected control code {}'.format(repr(c)))
             except queue.Empty:
                 pass
             except RuntimeError as e:

src/gluonnlp/data/transforms.py

@@ -178,7 +178,7 @@ class SacreMosesTokenizer:
     """
 
     def __init__(self):
-        from sacremoses import MosesTokenizer
+        from sacremoses import MosesTokenizer  # pylint: disable=import-outside-toplevel
         self._tokenizer = MosesTokenizer()
 
     def __call__(self, sample: str, return_str: bool = False):
@@ -228,8 +228,8 @@ class SpacyTokenizer:
 
     def __init__(self, lang='en_core_web_sm'):
         try:
-            import spacy
-            from pkg_resources import parse_version
+            import spacy  # pylint: disable=import-outside-toplevel
+            from pkg_resources import parse_version  # pylint: disable=import-outside-toplevel
             assert parse_version(spacy.__version__) >= parse_version('2.0.0'),\
                 'We only support spacy>=2.0.0'
         except ImportError:
@@ -292,7 +292,7 @@ class SacreMosesDetokenizer:
 
     def __init__(self, return_str=True):
         self._return_str = return_str
-        from sacremoses import MosesDetokenizer
+        from sacremoses import MosesDetokenizer  # pylint: disable=import-outside-toplevel
         self._detokenizer = MosesDetokenizer()
 
     def __call__(self, sample: List[str], return_str: Optional[bool] = None):
@@ -338,7 +338,7 @@ class JiebaTokenizer:
 
     def __init__(self):
         try:
-            import jieba
+            import jieba  # pylint: disable=import-outside-toplevel
         except ImportError:
             raise ImportError(
                 'jieba is not installed. You must install jieba in order to use the '
@@ -404,7 +404,7 @@ def __init__(self,
         assert is_java_exist == 0, 'Java is not installed. You must install Java 8.0' \
                                    'in order to use the NLTKStanfordSegmenter'
         try:
-            from nltk.tokenize import StanfordSegmenter
+            from nltk.tokenize import StanfordSegmenter  # pylint: disable=import-outside-toplevel
         except ImportError:
             raise ImportError(
                 'NLTK or relevant packages are not installed. You must install NLTK '
@@ -474,13 +474,13 @@ def __call__(self, sample):
         ret : list of strs
             List of tokens
         """
-        return [tok for tok in self._tokenizer.segment(sample).strip().split()]
+        return self._tokenizer.segment(sample).strip().split()
 
 
 class _SentencepieceProcessor:
     def __init__(self, path):
         try:
-            import sentencepiece
+            import sentencepiece  # pylint: disable=import-outside-toplevel
         except ImportError:
             raise ImportError(
                 'sentencepiece is not installed. You must install sentencepiece '
@@ -1235,7 +1235,8 @@ class GPT2BPETokenizer(_GPT2BPE):
                               '1a770728fd102bc9dc332f322e6bfb294767a685')
     def __init__(self, root=os.path.join(get_home_dir(), 'models')):
         try:
-            import regex as re
+            import regex  # pylint: disable=import-outside-toplevel
+            self._regex = regex
         except ImportError:
             raise ImportError(
                 'GPT2BPETokenizer requires regex. '
@@ -1285,7 +1286,7 @@ def __init__(self, root=os.path.join(get_home_dir(), 'models')):
                 raise ValueError('Downloaded file has different hash. Please try again.')
         self._read_bpe_ranks(file_path)
         self._cache = {}
-        self._token_pattern = re.compile(
+        self._token_pattern = self._regex.compile(
             r'\'s|\'t|\'re|\'ve|\'m|\'ll|\'d| ?\p{L}+'
             r'| ?\p{N}+| ?[^\s\p{L}\p{N}]+|\s+(?!\S)|\s+')
 
@@ -1347,9 +1348,8 @@ def __call__(self, sample):
         -------
         ret : list(str)
         """
-        import regex as re
         ret = []
-        for word_token in re.findall(self._token_pattern, sample):
+        for word_token in self._regex.findall(self._token_pattern, sample):
             word_token = bytearray(word_token.encode('utf-8'))
             word_token = ''.join(self._byte_encoder[code] for code in word_token)
             ret.extend(self.get_bpe_subword(word_token))

src/gluonnlp/data/utils.py

@@ -348,6 +348,7 @@ def _load_pretrained_vocab(name, root, cls=None):
 def _load_vocab_file(file_path, cls):
     with open(file_path, 'r') as f:
         if cls is None:
+            # pylint: disable=import-outside-toplevel
             from ..vocab import Vocab
             cls = Vocab
 

src/gluonnlp/data/word_embedding_evaluation.py

@@ -330,8 +330,7 @@ def __init__(self, root=os.path.join(get_home_dir(), 'datasets',
     def _get_data(self):
         datafilepath = os.path.join(self.root, self._archive_file[0])
 
-        dataset = CorpusDataset(datafilepath, tokenizer=lambda x: x.split(','))
-        return [row for row in dataset]
+        return list(CorpusDataset(datafilepath, tokenizer=lambda x: x.split(',')))
 
 
 @register

src/gluonnlp/initializer/initializer.py

@@ -80,7 +80,7 @@ class TruncNorm(Initializer):
     def __init__(self, mean=0, stdev=0.01, **kwargs):
         super(TruncNorm, self).__init__(**kwargs)
         try:
-            from scipy.stats import truncnorm
+            from scipy.stats import truncnorm  # pylint: disable=import-outside-toplevel
         except ImportError:
             raise ImportError('SciPy is not installed. '
                               'You must install SciPy >= 1.0.0 in order to use the '

src/gluonnlp/model/attention_cell.py

@@ -23,6 +23,7 @@
 import mxnet as mx
 from mxnet.gluon.block import HybridBlock
 from mxnet.gluon import nn
+from mxnet.contrib.amp import amp
 from .block import L2Normalization
 
 
@@ -46,13 +47,9 @@ def _apply_mask(F, att_score, mask, dtype):
     if np.dtype(dtype) == np.float16:
         neg = -1e4
     else:
-        try:
-            # if AMP (automatic mixed precision) is enabled, -1e18 will cause NaN.
-            from mxnet.contrib import amp
-            if amp.amp._amp_initialized:
-                neg = -1e4
-        except ImportError:
-            pass
+        # if AMP (automatic mixed precision) is enabled, -1e18 will cause NaN.
+        if amp._amp_initialized:
+            neg = -1e4
     att_score = F.where(mask, att_score, neg * F.ones_like(att_score))
     return att_score
 

src/gluonnlp/model/bert.py

@@ -32,6 +32,7 @@
 from .seq2seq_encoder_decoder import Seq2SeqEncoder
 from .transformer import TransformerEncoderCell
 from .utils import _load_pretrained_params, _load_vocab
+from ..vocab import Vocab, BERTVocab
 
 ###############################################################################
 #                              COMPONENTS                                     #
@@ -1161,8 +1162,6 @@ def get_roberta_model(model_name=None, dataset_name=None, vocab=None, pretrained
                           use_residual=predefined_args['use_residual'],
                           activation=predefined_args.get('activation', 'gelu'),
                           layer_norm_eps=predefined_args.get('layer_norm_eps', 1e-5))
-
-    from ..vocab import Vocab
     bert_vocab = _load_vocab(dataset_name, vocab, root, cls=Vocab)
     # BERT
     net = RoBERTaModel(encoder, len(bert_vocab),
@@ -1269,8 +1268,6 @@ def get_bert_model(model_name=None, dataset_name=None, vocab=None, pretrained=Tr
                           use_residual=predefined_args['use_residual'],
                           activation=predefined_args.get('activation', 'gelu'),
                           layer_norm_eps=predefined_args.get('layer_norm_eps', 1e-12))
-
-    from ..vocab import BERTVocab
     # bert_vocab
     bert_vocab = _load_vocab(dataset_name, vocab, root, cls=BERTVocab)
     # BERT

src/gluonnlp/optimizer/bert_adam.py

@@ -20,6 +20,7 @@
 import numpy
 from mxnet.optimizer import Optimizer, register
 from mxnet.ndarray import zeros, NDArray, full
+from mxnet.ndarray.contrib import mp_adamw_update, adamw_update
 
 __all__ = ['BERTAdam']
 
@@ -93,14 +94,6 @@ def update_multi_precision(self, index, weight, grad, state):
 
     def _update_impl(self, indices, weight, grad, state, multi_precision=False):
         """update function"""
-        try:
-            from mxnet.ndarray.contrib import adamw_update
-        except ImportError:
-            raise ImportError('Failed to import nd.contrib.adamw_update from MXNet. '
-                              'BERTAdam optimizer requires mxnet>=1.5.0b20190220. '
-                              'Please upgrade your MXNet version. For example: '
-                              'pip install mxnet-cu90 --pre. Otherwise, please consider '
-                              'Adam optimizer with different hyper-parameters.')
         self._update_count(indices)
         lr = self._get_lr(indices)
         wd = self._get_wd(indices)
@@ -120,15 +113,6 @@ def _update_impl(self, indices, weight, grad, state, multi_precision=False):
             adamw_update(weight, grad, mean, var, out=weight,
                          lr=1, wd=wd, eta=lr, **kwargs)
         else:
-            try:
-                from mxnet.ndarray.contrib import mp_adamw_update
-            except ImportError:
-                raise ImportError('Failed to import '
-                                  'nd.contrib.mp_adamw_update from MXNet. '
-                                  'BERTAdam optimizer requires mxnet>=1.5.0b20190220. '
-                                  'Please upgrade your MXNet version. For example: '
-                                  'pip install mxnet-cu90 --pre. Otherwise, please consider '
-                                  'Adam optimizer with different hyper-parameters.')
             mean, var = state[0]
             mp_adamw_update(weight, grad, mean, var, state[1], out=weight,
                             lr=1, wd=wd, eta=lr, **kwargs)

src/gluonnlp/utils/files.py

@@ -85,7 +85,7 @@ def __exit__(self, exec_type, exec_value, traceback):
 def _transfer_file_s3(filename, s3_filename, upload=True):
     """Transfer a file between S3 and local file system."""
     try:
-        import boto3
+        import boto3  # pylint: disable=import-outside-toplevel
     except ImportError:
         raise ImportError('boto3 is required to support s3 URI. Please install'
                           'boto3 via `pip install boto3`')