def build_gnn_parsing_model2(
flags,
data_reader: DatasetReader,
vocab: Vocabulary,
is_test: bool = False,
source_namespace: str = 'source_tokens',
target_namespace: str = 'target_tokens',
segment_namespace: str = 'segment_tokens',
) -> Model:
metric = SequenceAccuracy()
model = GNNCopyTransformer2(
vocab=vocab,
source_namespace=source_namespace,
target_namespace=target_namespace,
segment_namespace=segment_namespace,
max_decoding_step=flags.max_decode_length,
token_based_metric=metric,
source_embedding_dim=flags.source_embedding_dim,
target_embedding_dim=flags.target_embedding_dim,
encoder_d_model=flags.transformer_encoder_hidden_dim,
decoder_d_model=flags.transformer_decoder_hidden_dim,
encoder_nhead=flags.transformer_encoder_nhead,
decoder_nhead=flags.transformer_decoder_nhead,
num_decoder_layers=flags.transformer_num_decoder_layers,
num_encoder_layers=flags.transformer_num_encoder_layers,
encoder_dim_feedforward=flags.transformer_encoder_feedforward_dim,
decoder_dim_feedforward=flags.transformer_decoder_feedforward_dim,
dropout=flags.dropout,
beam_size=1,
nlabels=flags.gnn_transformer_num_edge_labels,
max_decode_clip_range=flags.gnn_max_decode_clip_range,
encode_edge_label_with_matrix=flags.gnn_encode_edge_label_with_matrix,
is_test=is_test)
return model
def build_parsing_seq2seq_model(
flags,
data_reader,
vocab: Vocabulary,
source_namespace: str = 'source_tokens',
target_namespace: str = 'target_tokens') -> Model:
source_embedding = Embedding(
vocab.get_vocab_size(namespace=source_namespace),
embedding_dim=flags.source_embedding_dim)
source_embedder = BasicTextFieldEmbedder({'tokens': source_embedding})
lstm_encoder = PytorchSeq2SeqWrapper(
torch.nn.LSTM(flags.source_embedding_dim,
flags.encoder_hidden_dim,
batch_first=True,
bidirectional=flags.encoder_bidirectional))
attention = DotProductAttention()
metric = SequenceAccuracy()
model = Seq2SeqModel(vocab,
source_embedder,
lstm_encoder,
flags.max_decode_length,
target_embedding_dim=flags.decoder_hidden_dim,
target_namespace=target_namespace,
attention=attention,
beam_size=flags.beam_size,
use_bleu=False,
seq_metrics=metric)
return model
def build_parsing_recombination_seq2seq_copy_model(
flags,
data_reader,
vocab: Vocabulary,
source_namespace: str = 'source_tokens',
target_namespace: str = 'target_tokens') -> Model:
source_embedding = Embedding(
vocab.get_vocab_size(namespace=source_namespace),
embedding_dim=flags.source_embedding_dim)
lstm = PytorchSeq2SeqWrapper(
torch.nn.LSTM(flags.source_embedding_dim,
flags.encoder_hidden_dim,
batch_first=True,
bidirectional=flags.encoder_bidirectional))
attention = BilinearAttention(flags.attention_hidden_dim,
flags.attention_hidden_dim,
normalize=False)
source_embedder = BasicTextFieldEmbedder({'tokens': source_embedding})
initializer = InitializerApplicator.from_params([
(".*bias", Params({
"type": "constant",
"val": 0
})), ('.*', Params({
"type": "uniform",
"a": -0.1,
"b": 0.1
}))
])
metric = SequenceAccuracy()
model = RecombinationSeq2SeqWithCopy(
vocab,
source_embedder,
lstm,
flags.max_decode_length,
seq_metrics=metric,
source_namespace=source_namespace,
target_namespace=target_namespace,
target_embedding_dim=flags.target_embedding_dim,
attention=attention,
beam_size=flags.beam_size,
use_bleu=False,
encoder_input_dropout=flags.encoder_input_dropout,
encoder_output_dropout=flags.encoder_output_dropout,
dropout=flags.dropout,
feed_output_attention_to_decoder=True,
keep_decoder_output_dim_same_as_encoder=True,
initializer=initializer)
return model
def build_grammar_copy_model(
flags,
data_reader: DatasetReader,
vocab: Vocabulary,
grammar: Grammar,
source_namespace: str = 'source_tokens',
) -> Model:
source_embedding = Embedding(
vocab.get_vocab_size(namespace=source_namespace),
embedding_dim=flags.source_embedding_dim)
source_embedder = BasicTextFieldEmbedder({'tokens': source_embedding})
lstm_encoder = PytorchSeq2SeqWrapper(
torch.nn.LSTM(flags.source_embedding_dim,
flags.encoder_hidden_dim,
batch_first=True,
bidirectional=flags.encoder_bidirectional))
decoder = LSTMGrammarCopyDecoder(
grammar,
AST,
lstm_hidden_dim=flags.decoder_hidden_dim,
num_lstm_layers=flags.decoder_num_layers,
rule_pad_index=data_reader.rule_pad_index,
rule_embedding_dim=flags.rule_embedding_dim,
nonterminal_pad_index=data_reader.nonterminal_pad_index,
nonterminal_end_index=data_reader.nonterminal_end_index,
nonterminal_embedding_dim=flags.nonterminal_embedding_dim,
source_encoding_dim=flags.encoder_hidden_dim * 2,
dropout=flags.dropout,
max_target_length=flags.max_decode_length)
metric = SequenceAccuracy()
model = GrammarModel(vocab,
source_embedder,
lstm_encoder,
decoder,
metric,
flags,
regularizer=None)
return model