Skip to content

hierarchical encoder decoder #8

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
junbeomlee wants to merge 2 commits into
base: master
Choose a base branch
from
Open

hierarchical encoder decoder #8

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
e794929
hierarchical encoder decoder added
junbeomlee Dec 7, 2017
a020a56
Merge pull request #1 from junbeomlee/hier_en_de
junbeomlee Dec 7, 2017
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions config/check_tiny.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21,6 +21,7 @@ model:
dropout: 0.2
encoder_type: bi
attention_mechanism: bahdanau
session_level_encoder: False

train:
batch_size: 2
Expand Down
42 changes: 41 additions & 1 deletion model.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -59,6 +59,8 @@ def _init_placeholder(self, features, labels):
def build_graph(self):
self._build_embed()
self._build_encoder()
if Config.model.session_level_encoder:
self._build_session_level_encoder()
self._build_decoder()

if self.mode != tf.estimator.ModeKeys.PREDICT:
Expand Down Expand Up @@ -106,6 +108,44 @@ def _build_encoder(self):
self.encoder_outputs = tf.contrib.seq2seq.tile_batch(self.encoder_outputs, beam_width)
self.encoder_input_lengths = tf.contrib.seq2seq.tile_batch(self.encoder_input_lengths, beam_width)

def _build_session_level_encoder(self):

if Config.model.num_layers > 1:
if Config.model.cell_type == "LSTM":
if Config.model.encoder_type == "bi":
session_level_encoder_input = tf.stack(self.encoder_final_state)
else:
session_level_encoder_input = tf.stack(self.encoder_final_state[-1])
else:
if Config.model.encoder_type == "bi":
session_level_encoder_input = tf.stack([self.encoder_final_state])
else:
session_level_encoder_input = tf.stack([self.encoder_final_state[-1]])
else:
if Config.model.cell_type == "LSTM":
if Config.model.encoder_type == "bi":
session_level_encoder_input = tf.stack(self.encoder_final_state)
else:
session_level_encoder_input = tf.stack(self.encoder_final_state[0])
else:
if Config.model.encoder_type == "bi":
session_level_encoder_input = tf.stack([self.encoder_final_state])
else:
session_level_encoder_input = tf.stack(self.encoder_final_state)

with tf.variable_scope("session_level_encoder"):

if Config.model.cell_type == "LSTM":
session_level_encoder_cell = self._single_cell("LSTM", Config.model.dropout,Config.model.num_units)
else:
session_level_encoder_cell = self._single_cell("GRU", Config.model.dropout,Config.model.num_units)

session_level_encoder_outputs, session_level_encoder_final_state = tf.nn.dynamic_rnn(
session_level_encoder_cell, session_level_encoder_input,
dtype=tf.float32)

self.encoder_final_state = tf.unstack(session_level_encoder_outputs)[0]

def _build_unidirectional_rnn(self):
cells = self._build_rnn_cells(Config.model.num_units)
return tf.nn.dynamic_rnn(
Expand Down Expand Up @@ -202,7 +242,7 @@ def decode(helper=None, scope="decode"):
attention_mechanism,
attention_layer_size=attention_layer_size,
alignment_history=alignment_history,
name="attention")
name="attention")

out_cell = tf.contrib.rnn.OutputProjectionWrapper(
attn_cell, Config.data.vocab_size)
Expand Down