def step(state_tup, inp):
"""Single step of the improved Neural GPU."""
state, _ = state_tup
x = state
for layer in range(hparams.num_hidden_layers):
x, new_loss = common_layers.diagonal_conv_gru(
x, (hparams.kernel_height, hparams.kernel_width),
hparams.hidden_size,
dropout=hparams.dropout,
name="dcgru_%d" % layer)
# Padding input is zeroed-out in the modality, we check this by summing.
padding_inp = tf.less(tf.reduce_sum(tf.abs(inp), axis=[1, 2]), 0.00001)
new_state = tf.where(padding_inp, state, x) # No-op where inp is padding.
return new_state, new_loss
def step(state_tup, inp):
"""Single step of the improved Neural GPU."""
state, _ = state_tup
x = state
for layer in xrange(hparams.num_hidden_layers):
x, new_loss = common_layers.diagonal_conv_gru(
x, (hparams.kernel_height, hparams.kernel_width),
hparams.hidden_size,
dropout=hparams.dropout,
name="dcgru_%d" % layer)
# Padding input is zeroed-out in the modality, we check this by summing.
padding_inp = tf.less(tf.reduce_sum(tf.abs(inp), axis=[1, 2]), 0.00001)
new_state = tf.where(padding_inp, state, x) # No-op where inp is padding.
return new_state, new_loss