def build_encoder(opt, embeddings):
"""
Various encoder dispatcher function.
Args:
opt: the option in current environment.
embeddings (Embeddings): vocab embeddings for this encoder.
"""
if opt.encoder_type == "transformer":
return TransformerEncoder(opt.enc_layers, opt.enc_rnn_size,
opt.heads, opt.transformer_ff,
opt.dropout, embeddings)
elif opt.encoder_type == "cnn":
return CNNEncoder(opt.enc_layers, opt.enc_rnn_size,
opt.cnn_kernel_width,
opt.dropout, embeddings)
elif opt.encoder_type == "mean":
return MeanEncoder(opt.enc_layers, embeddings)
elif opt.encoder_type == "dnc":
#import pdb; pdb.set_trace()
return MEMEncoder(opt.rnn_type, opt.brnn, opt.enc_layers,
opt.enc_rnn_size, opt.dropout, embeddings)
else:
# "rnn" or "brnn"
return RNNEncoder(opt.rnn_type, opt.brnn, opt.enc_layers,
opt.enc_rnn_size, opt.dropout, embeddings,
opt.bridge)
def build_encoder(opt, embeddings, embeddings_latt = False, feat_vec_size = 512):
"""
Various encoder dispatcher function.
Args:
opt: the option in current environment.
embeddings (Embeddings): vocab embeddings for this encoder.
embeddings_latt: embeddings of senses if lattice is used # latt!!!
feat_vec_size: for adaptable feat_vec_size # latt!!!
"""
#latt
if opt.encoder_type == "transformer":
if embeddings_latt != False: #latt
return TransformerEncoder(opt.enc_layers, opt.rnn_size,
opt.heads, opt.transformer_ff,
opt.dropout, embeddings, embeddings_latt, feat_vec_size) #latt
else:
return TransformerEncoder(opt.enc_layers, opt.rnn_size,
opt.heads, opt.transformer_ff,
opt.dropout, embeddings, embeddings_latt, feat_vec_size) #latt
#latt
elif opt.encoder_type == "cnn":
return CNNEncoder(opt.enc_layers, opt.rnn_size,
opt.cnn_kernel_width,
opt.dropout, embeddings)
elif opt.encoder_type == "mean":
return MeanEncoder(opt.enc_layers, embeddings)
else:
# "rnn" or "brnn"
return RNNEncoder(opt.rnn_type, opt.brnn, opt.enc_layers,
opt.rnn_size, opt.dropout, embeddings,
opt.bridge)
def build_encoder(opt, embeddings):
"""
Various encoder dispatcher function.
Args:
opt: the option in current environment.
embeddings (Embeddings): vocab embeddings for this encoder.
"""
if opt.encoder_type == "transformer":
return TransformerEncoder(opt.enc_layers, opt.rnn_size,
opt.heads, opt.transformer_ff,
opt.dropout, embeddings), None
elif opt.encoder_type == "cnn":
return CNNEncoder(opt.enc_layers, opt.rnn_size,
opt.cnn_kernel_width,
opt.dropout, embeddings), None
elif opt.encoder_type == "mean":
return MeanEncoder(opt.enc_layers, embeddings), None
else:
# "rnn" or "brnn"
word_encoder = RNNEncoder(opt.rnn_type, opt.brnn, opt.enc_layers,
opt.rnn_size, opt.dropout, embeddings, None, opt.bridge)
if opt.rnn_type == "LSTM":
emb_size = opt.enc_layers * opt.rnn_size * 2
else:
emb_size = opt.enc_layers * opt.rnn_size
sen_encoder = RNNEncoder(opt.rnn_type, opt.brnn, opt.sen_enc_layers,
opt.sen_rnn_size, opt.dropout, None, emb_size, opt.bridge)
return word_encoder, sen_encoder
def build_encoder(opt, embeddings, main_encoder=None):
"""
Various encoder dispatcher function.
Args:
opt: the option in current environment.
embeddings (Embeddings): vocab embeddings for this encoder.
"""
if opt.encoder_type == "transformer":
encoder = TransformerEncoder(opt.enc_layers,
opt.enc_rnn_size,
opt.heads,
opt.transformer_ff,
opt.dropout,
embeddings,
main_encoder=main_encoder,
mtl_opt=opt)
elif opt.encoder_type == "cnn":
encoder = CNNEncoder(opt.enc_layers, opt.enc_rnn_size,
opt.cnn_kernel_width, opt.dropout, embeddings)
elif opt.encoder_type == "mean":
encoder = MeanEncoder(opt.enc_layers, embeddings)
else:
encoder = RNNEncoder(opt.rnn_type,
opt.brnn,
opt.enc_layers,
opt.enc_rnn_size,
opt.dropout,
embeddings,
opt.bridge,
main_encoder=main_encoder,
mtl_opt=opt)
return encoder
def __init__(self,
rnn_type,
encoder_type,
enc_layers,
hidden_size,
dropout=0.0,
embeddings=None):
super(QueryEncoder, self).__init__()
assert embeddings is not None
self.embeddings = embeddings
self.rnn_type = rnn_type
bidirectional = True if encoder_type == 'brnn' else False
self.src_encoder = RNNEncoder(rnn_type, bidirectional, enc_layers,
hidden_size, dropout,
embeddings.embedding_size)
self.answer_encoder = MeanEncoder()
def build_encoder(opt, embeddings):
"""
Various encoder dispatcher function.
Args:
opt: the option in current environment.
embeddings (Embeddings): vocab embeddings for this encoder.
"""
if opt.encoder_type == "transformer":
encoder = TransformerEncoder(opt.enc_layers, opt.enc_rnn_size,
opt.heads, opt.transformer_ff,
opt.dropout, embeddings)
elif opt.encoder_type == "cnn":
encoder = CNNEncoder(opt.enc_layers, opt.enc_rnn_size,
opt.cnn_kernel_width, opt.dropout, embeddings)
elif opt.encoder_type == "mean":
encoder = MeanEncoder(opt.enc_layers, embeddings)
elif opt.encoder_type == "hr_brnn":
bi_enc = True
encoder = HREncoder(opt.rnn_type, bi_enc, opt.enc_layers,
opt.enc_rnn_size, opt.dropout, embeddings,
opt.bridge)
elif opt.encoder_type == "seq_hr_brnn":
bi_enc = True
encoder = SeqHREncoder(opt.rnn_type, bi_enc, opt.enc_layers,
opt.enc_rnn_size, opt.dropout, embeddings,
opt.bridge)
elif opt.encoder_type == "tg_brnn":
bi_enc = True
encoder = TGEncoder(opt.rnn_type, bi_enc, opt.enc_layers,
opt.enc_rnn_size, opt.dropout, embeddings)
else:
bi_enc = 'brnn' in opt.encoder_type
encoder = RNNEncoder(opt.rnn_type,
bi_enc,
opt.enc_layers,
opt.enc_rnn_size,
opt.dropout,
embeddings,
opt.bridge,
use_catSeq_dp=opt.use_catSeq_dp)
return encoder
def build_encoder(opt, embeddings):
"""
Various encoder dispatcher function.
Args:
opt: the option in current environment.
embeddings (Embeddings): vocab embeddings for this encoder.
"""
if opt.encoder_type == "transformer":
return TransformerEncoder(
opt.enc_layers,
opt.enc_rnn_size,
opt.heads,
opt.transformer_ff,
opt.dropout,
embeddings,
opt.self_attention_function,
opt.self_attention_alpha,
opt.self_attention_bisect_iter,
)
elif opt.encoder_type == "cnn":
return CNNEncoder(
opt.enc_layers,
opt.enc_rnn_size,
opt.cnn_kernel_width,
opt.dropout,
embeddings,
)
elif opt.encoder_type == "mean":
return MeanEncoder(opt.enc_layers, embeddings)
else:
# "rnn" or "brnn"
return RNNEncoder(
opt.rnn_type,
opt.brnn,
opt.enc_layers,
opt.enc_rnn_size,
opt.dropout,
embeddings,
opt.bridge,
)
def build_encoder(opt, embeddings):
"""
Various encoder dispatcher function.
Args:
opt: the option in current environment.
embeddings (Embeddings): vocab embeddings for this encoder.
"""
if opt.encoder_type == "transformer":
return TransformerEncoder(opt.enc_layers, opt.enc_rnn_size,
opt.heads, opt.transformer_ff,
opt.dropout, embeddings)
elif opt.encoder_type == "cnn":
return CNNEncoder(opt.enc_layers, opt.enc_rnn_size,
opt.cnn_kernel_width,
opt.dropout, embeddings)
elif opt.encoder_type == "mean":
return MeanEncoder(opt.enc_layers, embeddings)
elif opt.encoder_type == "rnntreelstm" or opt.encoder_type == "treelstm":
opt.brnn = True if opt.encoder_type == "rnntreelstm" else False
return TreeLSTMEncoder(opt.rnn_type, opt.brnn, opt.enc_layers,
opt.rnn_size, opt.dropout, embeddings,
opt.bridge, False)
elif opt.encoder_type == "rnnbitreelstm" or opt.encoder_type == "bitreelstm":
opt.brnn = True if opt.encoder_type == "rnnbitreelstm" else False
return TreeLSTMEncoder(opt.rnn_type, opt.brnn, opt.enc_layers,
opt.rnn_size, opt.dropout, embeddings,
opt.bridge, True)
elif opt.encoder_type == "rnngcn" or opt.encoder_type == "gcn":
opt.brnn = True if opt.encoder_type == "rnngcn" else False
return GCNEncoder(opt.rnn_type, opt.brnn, opt.enc_layers,
opt.rnn_size, opt.dropout, embeddings,
opt.bridge, opt.gcn_dropout,
opt.gcn_edge_dropout, opt.n_gcn_layers,
opt.activation, opt.highway)
else:
# "rnn" or "brnn"
return RNNEncoder(opt.rnn_type, opt.brnn, opt.enc_layers,
opt.enc_rnn_size, opt.dropout, embeddings,
opt.bridge)
def build_encoder(opt, embeddings, fields=None):
"""
Various encoder dispatcher function.
Args:
opt: the option in current environment.
embeddings (Embeddings): vocab embeddings for this encoder.
"""
if opt.encoder_type == 'simple_context_0':
# bottom n-1 layers are shared
return SimpleContextTransformerEncoder(
opt.enc_layers - 1, opt.rnn_size,
opt.heads, opt.transformer_ff,
opt.dropout, embeddings,
selected_ctx=0)
elif opt.encoder_type == 'simple_context_1':
# bottom n-1 layers are shared
return SimpleContextTransformerEncoder(
opt.enc_layers - 1, opt.rnn_size,
opt.heads, opt.transformer_ff,
opt.dropout, embeddings,
selected_ctx=1)
elif opt.encoder_type == "transformer":
return TransformerEncoder(opt.enc_layers, opt.rnn_size,
opt.heads, opt.transformer_ff,
opt.dropout, embeddings)
elif opt.encoder_type == "cnn":
return CNNEncoder(opt.enc_layers, opt.rnn_size,
opt.cnn_kernel_width,
opt.dropout, embeddings)
elif opt.encoder_type == "mean":
return MeanEncoder(opt.enc_layers, embeddings)
else:
# "rnn" or "brnn"
return RNNEncoder(opt.rnn_type, opt.brnn, opt.enc_layers,
opt.rnn_size, opt.dropout, embeddings,
opt.bridge)