def AttentiveCNN_match(context,
query,
context_mask,
query_mask,
causality=False,
scope='AttentiveCNN_Block',
reuse=None,
**kwargs):
with tf.variable_scope(scope, reuse=reuse):
direction = 'forward' if causality else 'none'
cnn_wo_att = CNN_encode(context,
filter_size=3,
direction=direction,
act_fn=None)
att_context, _ = Attentive_match(context,
query,
context_mask,
query_mask,
causality=causality)
cnn_att = CNN_encode(att_context,
filter_size=1,
direction=direction,
act_fn=None)
output = tf.nn.tanh(cnn_wo_att + cnn_att)
return dropout_res_layernorm(context, output, **kwargs)
def Res_DualCNN_encode(seqs, use_spatial_dropout=True, scope='res_biconv_block', reuse=None, **kwargs):
with tf.variable_scope(scope, reuse=reuse):
out1 = CNN_encode(seqs, scope='first_conv1d', **kwargs)
if use_spatial_dropout:
out1 = spatial_dropout(out1)
out2 = CNN_encode(out1, scope='second_conv1d', **kwargs)
if use_spatial_dropout:
out2 = CNN_encode(out2)
return dropout_res_layernorm(seqs, out2, **kwargs)