def __init__(self, vocab_size, tag_to_ix, embedding_dim, hidden_dim):
super(BilstmCrf, self).__init__()
self.embedding_dim = embedding_dim
self.hidden_dim = hidden_dim
self.vocab_size = vocab_size
self.tag_to_ix = tag_to_ix
self.tagset_size = len(tag_to_ix)
self.word_embeds = nn.Embedding(vocab_size, embedding_dim)
self.lstm = nn.LSTM(embedding_dim,
hidden_dim // 2,
num_layers=1,
bidirectional=True)
# Maps the output of the LSTM into tag space.
self.hidden2tag = nn.Linear(hidden_dim, self.tagset_size)
# Matrix of transition parameters. Entry i,j is the score of
# transitioning *to* i *from* j.
# 转移矩阵
param = torch.randn(self.tagset_size, self.tagset_size)
self.transitions = Parameter.__new__(param)
# These two statements enforce the constraint that we never transfer
# to the start tag and we never transfer from the stop tag
self.transitions.data[tag_to_ix[START_TAG], :] = -10000
self.transitions.data[:, tag_to_ix[STOP_TAG]] = -10000
self.hidden = self.init_hidden()
def __new__(cls: Type['AnalogContext'],
analog_tile: 'BaseTile',
parameter: Optional[Parameter] = None) -> 'AnalogContext':
# pylint: disable=signature-differs
if parameter is None:
return Parameter.__new__(cls,
data=ones((), device=analog_tile.device),
requires_grad=True)
parameter.__class__ = cls
return parameter
def __new__(cls: Type['AnalogContext'],
analog_tile: 'BaseTile') -> 'AnalogContext':
# pylint: disable=signature-differs
return Parameter.__new__(cls,
data=ones((), device=analog_tile.device),
requires_grad=True)
def __new__(cls, data, requires_grad, manifold, c):
return Parameter.__new__(cls, data, requires_grad)
def __new__(cls, data=None, requires_grad=True, name=None):
"""Wrap __new__ of torch Parameter."""
return Torch_Parameter.__new__(cls, data, requires_grad)
