Make lstm tagger not depend on pytorch

examples/lstm_tagger.py

@@ -20,18 +20,9 @@
 from thinc.neural.optimizers import SGD
 from thinc.neural.util import get_array_module
 
-import torch
-import torch.nn
-import torch.autograd
-
 from thinc.extra.datasets import ancora_pos_tags
 
 
-def PyTorchBiLSTM(nO, nI, depth):
-    model = torch.nn.LSTM(nI, nO//2, depth, bidirectional=True)
-    return with_square_sequences(PyTorchWrapperRNN(model))
-
-
 def FeatureExtracter(lang, attrs=[LOWER, SHAPE, PREFIX, SUFFIX], tokenized=True):
     nlp = spacy.blank(lang)
     nlp.vocab.lex_attr_getters[PREFIX] = lambda string: string[:3]
@@ -102,14 +93,12 @@ def debug(X, drop=0.):
     L2=("L2 regularization penalty", "option", "L", float),
 )
 def main(width=32, depth=1, vector_length=32,
-         min_batch_size=1, max_batch_size=4, learn_rate=0.001,
+         min_batch_size=16, max_batch_size=16, learn_rate=0.001,
          momentum=0.9, dropout=0.5, dropout_decay=1e-4,
          nb_epoch=20, L2=1e-6):
     cfg = dict(locals())
     print(cfg)
     train_data, check_data, nr_tag = ancora_pos_tags()
-    train_data = train_data[:100]
-    check_data = check_data[:100]
 
     extracter = FeatureExtracter('es', attrs=[LOWER, SHAPE, PREFIX, SUFFIX])
     with Model.define_operators({'**': clone, '>>': chain, '+': add,