def __init__(self,
args=None,
vocab=None,
pretrain=None,
model_file=None,
use_cuda=False):
self.use_cuda = use_cuda
if model_file is not None:
# load everything from file
self.load(pretrain, model_file)
else:
assert all(var is not None for var in [args, vocab, pretrain])
# build model from scratch
self.args = args
self.vocab = vocab
self.model = Tagger(args,
vocab,
emb_matrix=pretrain.emb,
share_hid=args['share_hid'])
self.parameters = [
p for p in self.model.parameters() if p.requires_grad
]
if self.use_cuda:
self.model.cuda()
else:
self.model.cpu()
self.optimizer = utils.get_optimizer(self.args['optim'],
self.parameters,
self.args['lr'],
betas=(0.9, self.args['beta2']),
eps=1e-6)
def __init__(self, args=None, vocab=None, emb_matrix=None, model_file=None, use_cuda=False):
self.use_cuda = use_cuda
if model_file is not None:
# load everything from file
self.load(model_file, use_cuda)
else:
# build model from scratch
self.args = args
self.model = None if args['dict_only'] else Seq2SeqModel(args, emb_matrix=emb_matrix, use_cuda=use_cuda)
self.vocab = vocab
# dict-based components
self.fallback_dict = dict()
self.composite_dict = dict()
if not self.args['dict_only']:
if self.args.get('edit', False):
self.crit = loss.MixLoss(self.vocab['char'].size, self.args['alpha'])
print("[Running seq2seq lemmatizer with edit classifier]")
else:
self.crit = loss.SequenceLoss(self.vocab['char'].size)
self.parameters = [p for p in self.model.parameters() if p.requires_grad]
if use_cuda:
self.model.cuda()
self.crit.cuda()
else:
self.model.cpu()
self.crit.cpu()
self.optimizer = utils.get_optimizer(self.args['optim'], self.parameters, self.args['lr'])
def __init__(self,
args=None,
vocab=None,
emb_matrix=None,
model_file=None,
use_cuda=False):
self.use_cuda = use_cuda
if model_file is not None:
# load from file
self.load(model_file, use_cuda)
else:
self.args = args
self.model = None if args['dict_only'] else Seq2SeqModel(
args, emb_matrix=emb_matrix)
self.vocab = vocab
self.expansion_dict = dict()
if not self.args['dict_only']:
self.crit = loss.SequenceLoss(self.vocab.size)
self.parameters = [
p for p in self.model.parameters() if p.requires_grad
]
if use_cuda:
self.model.cuda()
self.crit.cuda()
else:
self.model.cpu()
self.crit.cpu()
self.optimizer = utils.get_optimizer(self.args['optim'],
self.parameters,
self.args['lr'])
def __init__(self,
args=None,
vocab=None,
pretrain=None,
model_file=None,
use_cuda=False):
self.use_cuda = use_cuda
if model_file is not None:
# load everything from file
self.load(model_file, args)
else:
assert all(var is not None for var in [args, vocab, pretrain])
# build model from scratch
self.args = args
self.vocab = vocab
self.model = NERTagger(args, vocab, emb_matrix=pretrain.emb)
self.parameters = [
p for p in self.model.parameters() if p.requires_grad
]
if self.use_cuda:
self.model.cuda()
else:
self.model.cpu()
self.optimizer = utils.get_optimizer(self.args['optim'],
self.parameters,
self.args['lr'],
momentum=self.args['momentum'])
def __init__(self,
args=None,
vocab=None,
pretrain=None,
model_file=None,
use_cuda=False):
self.use_cuda = use_cuda
if model_file is not None:
# load everything from file
self.load(model_file, pretrain)
else:
# build model from scratch
self.args = args
self.vocab = vocab
self.model = Tagger(
args,
vocab,
emb_matrix=pretrain.emb if pretrain is not None else None,
share_hid=args['share_hid'])
self.constrain_via_lexicon = args[
'constrain_via_lexicon'] if args is not None and 'constrain_via_lexicon' in args else None
self.inflectional_lexicon = None
if self.constrain_via_lexicon:
inflectional_lexicon = LemmaTrainer(
model_file=self.constrain_via_lexicon).composite_dict
args['shorthand'] = args[
'shorthand'] if 'shorthand' in args else self.args['shorthand']
self.inflectional_lexicon = InflectionalLexicon(
inflectional_lexicon, args['shorthand'], self.vocab, pretrain)
self.parameters = [
p for p in self.model.parameters() if p.requires_grad
]
if self.use_cuda:
self.model.cuda()
else:
self.model.cpu()
self.optimizer = utils.get_optimizer(self.args['optim'],
self.parameters,
self.args['lr'],
betas=(0.9, self.args['beta2']),
eps=1e-6)