Secure your code as it's written. Use Snyk Code to scan source code in minutes - no build needed - and fix issues immediately.
def __init__(self,
hidden_size,
num_intents,
num_slots,
dropout,
**kwargs):
TrainableNM.__init__(self, **kwargs)
self.hidden_size = hidden_size
self.num_intents = num_intents
self.num_slots = num_slots
self.dropout = nn.Dropout(dropout)
self.intent_dense = nn.Linear(self.hidden_size, self.hidden_size)
self.intent_classifier = nn.Linear(self.hidden_size, self.num_intents)
self.slot_dense = nn.Linear(self.hidden_size, self.hidden_size)
self.slot_classifier = nn.Linear(self.hidden_size, self.num_slots)
self.apply(
lambda module: transformer_weights_init(module, xavier=False))
self.to(self._device)
def __init__(self, *, from_dim, to_dim, dropout=0.0, **kwargs):
TrainableNM.__init__(self, **kwargs)
self.from_dim = from_dim
self.to_dim = to_dim
self.dropout = dropout
self.projection = nn.Linear(self.from_dim, self.to_dim, bias=False)
if self.dropout != 0.0:
self.embedding_dropout = nn.Dropout(self.dropout)
def __init__(self, *, d_model, num_classes, **kwargs):
TrainableNM.__init__(self, **kwargs)
self.log_softmax = ClassificationLogSoftmax(
hidden_size=d_model,
num_classes=num_classes
)
self.log_softmax.apply(transformer_weights_init)
self.log_softmax.to(self._device)
def __init__(self, *,
pretrained_model_name=None,
config_filename=None,
vocab_size=None,
hidden_size=768,
num_hidden_layers=12,
num_attention_heads=12,
intermediate_size=3072,
hidden_act="gelu",
max_position_embeddings=512,
**kwargs):
TrainableNM.__init__(self, **kwargs)
# Check that only one of pretrained_model_name, config_filename, and
# vocab_size was passed in
total = 0
if pretrained_model_name is not None:
total += 1
if config_filename is not None:
total += 1
if vocab_size is not None:
total += 1
if total != 1:
raise ValueError("Only one of pretrained_model_name, vocab_size, "
+ "or config_filename should be passed into the "
+ "BERT constructor.")
def __init__(self, *, voc_size, hidden_size, dropout=0.0, **kwargs):
TrainableNM.__init__(self, **kwargs)
self.voc_size = voc_size
self.hidden_size = hidden_size
self.dropout = dropout
self.embedding = nn.Embedding(self.voc_size, self.hidden_size)
if self.dropout != 0.0:
self.embedding_dropout = nn.Dropout(self.dropout)
def __init__(self, **kwargs):
TrainableNM.__init__(self, **kwargs)
def __init__(self, *, vocab_size, d_model, **kwargs):
TrainableNM.__init__(self, **kwargs)
self.log_softmax = TransformerLogSoftmax(
vocab_size=vocab_size,
hidden_size=d_model)
self.log_softmax.apply(transformer_weights_init)
self.log_softmax.to(self._device)
def __init__(self, **kwargs):
TrainableNM.__init__(self, **kwargs)
def __init__(self,
vocab_size,
d_model,
d_inner,
num_layers,
max_seq_length,
num_attn_heads,
ffn_dropout=0.0,
embedding_dropout=0.0,
attn_score_dropout=0.0,
attn_layer_dropout=0.0,
learn_positional_encodings=False,
hidden_act='relu',
**kwargs):
TrainableNM.__init__(self, **kwargs)
self.embedding_layer = TransformerEmbedding(
vocab_size=vocab_size,
hidden_size=d_model,
max_sequence_length=max_seq_length,
embedding_dropout=embedding_dropout,
learn_positional_encodings=learn_positional_encodings
)
self.decoder = TransformerDecoder(
num_layers=num_layers,
hidden_size=d_model,
num_attention_heads=num_attn_heads,
inner_size=d_inner,
ffn_dropout=ffn_dropout,
hidden_act=hidden_act,
attn_score_dropout=attn_score_dropout,