def _build_graph(self):
# build input
self.context_word = tf.placeholder(tf.int32, [None, None])
self.context_len = tf.placeholder(tf.int32, [None])
self.question_word = tf.placeholder(tf.int32, [None, None])
self.question_len = tf.placeholder(tf.int32, [None])
self.answer_start = tf.placeholder(tf.int32, [None])
self.answer_end = tf.placeholder(tf.int32, [None])
self.training = tf.placeholder(tf.bool, [])
self.match_lemma = tf.placeholder(tf.int32, [None, None])
self.match_lower = tf.placeholder(tf.int32, [None, None])
self.pos_feature = tf.placeholder(tf.int32, [None, None])
self.ner_feature = tf.placeholder(tf.int32, [None, None])
self.normalized_tf = tf.placeholder(tf.int32, [None, None])
# 1. Word encoding
word_embedding = PartiallyTrainableEmbedding(
trainable_num=self.finetune_word_size,
pretrained_embedding=self.pretrained_word_embedding,
embedding_shape=(len(self.vocab.get_word_vocab()) + 1,
self.word_embedding_size))
# 1.1 Embedding
context_word_repr = word_embedding(self.context_word)
question_word_repr = word_embedding(self.question_word)
# pos embedding
if len(self.features) > 0 and 'pos' in self.features:
pos_embedding = Embedding(
pretrained_embedding=None,
embedding_shape=(len(self.feature_vocab['pos']) + 1,
self.pos_embedding_size))
context_pos_feature = pos_embedding(self.pos_feature)
if len(self.features) > 0 and 'ner' in self.features:
ner_embedding = Embedding(
pretrained_embedding=None,
embedding_shape=(len(self.feature_vocab['ner']) + 1,
self.ner_embedding_size))
context_ner_feature = ner_embedding(self.ner_feature)
# add dropout
dropout = VariationalDropout(self.dropout_keep_prob)
context_word_repr = dropout(context_word_repr, self.training)
question_word_repr = dropout(question_word_repr, self.training)
glove_word_repr = context_word_repr
glove_question_repr = question_word_repr
if self.use_cove_emb and self.cove_path is not None:
cove_embedding = CoveEmbedding(
cove_path=self.cove_path,
pretrained_word_embedding=self.pretrained_word_embedding,
vocab=self.vocab)
cove_context_repr = cove_embedding(self.context_word,
self.context_len)
cove_question_repr = cove_embedding(self.question_word,
self.question_len)
cove_context_repr = dropout(cove_context_repr, self.training)
cove_question_repr = dropout(cove_question_repr, self.training)
# exact_match feature
context_expanded = tf.tile(
tf.expand_dims(self.context_word, axis=-1),
[1, 1, tf.shape(self.question_word)[1]])
query_expanded = tf.tile(tf.expand_dims(self.question_word, axis=1),
[1, tf.shape(self.context_word)[1], 1])
exact_match_feature = tf.cast(
tf.reduce_any(tf.equal(context_expanded, query_expanded), axis=-1),
tf.float32)
exact_match_feature = tf.expand_dims(exact_match_feature, axis=-1)
if len(self.features) > 0 and 'match_lower' in self.features:
match_lower_feature = tf.expand_dims(tf.cast(
self.match_lower, tf.float32),
axis=-1)
exact_match_feature = tf.concat(
[exact_match_feature, match_lower_feature], axis=-1)
if len(self.features) > 0 and 'match_lemma' in self.features:
exact_match_feature = tf.concat([
exact_match_feature,
tf.expand_dims(tf.cast(self.match_lemma, tf.float32), axis=-1)
],
axis=-1)
# features
if len(self.features) > 0 and 'pos' in self.features:
context_word_repr = tf.concat(
[context_word_repr, context_pos_feature], axis=-1)
if len(self.features) > 0 and 'ner' in self.features:
context_word_repr = tf.concat(
[context_word_repr, context_ner_feature], axis=-1)
if len(self.features) > 0 and 'context_tf' in self.features:
context_word_repr = tf.concat([
context_word_repr,
tf.cast(tf.expand_dims(self.normalized_tf, axis=-1),
tf.float32)
],
axis=-1)
if self.use_cove_emb:
context_word_repr = tf.concat(
[cove_context_repr, context_word_repr], axis=-1)
question_word_repr = tf.concat(
[cove_question_repr, question_word_repr], axis=-1)
# 1.2word fusion
sim_function = ProjectedDotProduct(self.rnn_hidden_size,
activation=tf.nn.relu,
reuse_weight=True)
word_fusion = UniAttention(sim_function)
context2question_fusion = word_fusion(glove_word_repr,
glove_question_repr,
self.question_len)
enhanced_context_repr = tf.concat(
[context_word_repr, exact_match_feature, context2question_fusion],
axis=-1)
enhanced_context_repr = dropout(enhanced_context_repr, self.training)
# 1.3.1 context encoder
context_encoder = [
CudnnBiLSTM(self.rnn_hidden_size),
CudnnBiLSTM(self.rnn_hidden_size)
]
context_low_repr, _ = context_encoder[0](enhanced_context_repr,
self.context_len)
context_low_repr = dropout(context_low_repr, self.training)
context_high_repr, _ = context_encoder[1](context_low_repr,
self.context_len)
context_high_repr = dropout(context_high_repr, self.training)
# 1.3.2 question encoder
question_encoder = [
CudnnBiLSTM(self.rnn_hidden_size),
CudnnBiLSTM(self.rnn_hidden_size)
]
question_low_repr, _ = question_encoder[0](question_word_repr,
self.question_len)
question_low_repr = dropout(question_low_repr, self.training)
question_high_repr, _ = question_encoder[1](question_low_repr,
self.question_len)
question_high_repr = dropout(question_high_repr, self.training)
# 1.4 question understanding
question_understanding_encoder = CudnnBiLSTM(self.rnn_hidden_size)
question_understanding, _ = question_understanding_encoder(
tf.concat([question_low_repr, question_high_repr], axis=-1),
self.question_len)
question_understanding = dropout(question_understanding, self.training)
# history of context
context_history = tf.concat(
[glove_word_repr, context_low_repr, context_high_repr], axis=-1)
# histor of question
question_history = tf.concat(
[glove_question_repr, question_low_repr, question_high_repr],
axis=-1)
# concat cove emb
if self.use_cove_emb:
context_history = tf.concat([cove_context_repr, context_history],
axis=-1)
question_history = tf.concat(
[cove_question_repr, question_history], axis=-1)
# 1.5.1 low level fusion
low_level_attn = UniAttention(SymmetricProject(
self.attention_hidden_size),
name='low_level_fusion')
low_level_fusion = low_level_attn(context_history, question_history,
self.question_len, question_low_repr)
low_level_fusion = dropout(low_level_fusion, self.training)
# 1.5.2 high level fusion
high_level_attn = UniAttention(SymmetricProject(
self.attention_hidden_size, name='high_sim_func'),
name='high_level_fusion')
high_level_fusion = high_level_attn(context_history, question_history,
self.question_len,
question_high_repr)
high_level_fusion = dropout(high_level_fusion, self.training)
# 1.5.3 understanding level fusion
understanding_attn = UniAttention(SymmetricProject(
self.attention_hidden_size, name='understanding_sim_func'),
name='understanding_level_fusion')
understanding_fusion = understanding_attn(context_history,
question_history,
self.question_len,
question_understanding)
understanding_fusion = dropout(understanding_fusion, self.training)
# merge context attention
fully_aware_encoder = CudnnBiLSTM(self.rnn_hidden_size)
full_fusion_context = tf.concat([
context_low_repr, context_high_repr, low_level_fusion,
high_level_fusion, understanding_fusion
],
axis=-1)
full_fusion_context_repr, _ = fully_aware_encoder(
full_fusion_context, self.context_len)
# history of context
context_history = tf.concat(
[glove_word_repr, full_fusion_context, full_fusion_context_repr],
axis=-1)
if self.use_cove_emb:
context_history = tf.concat([cove_context_repr, context_history],
axis=-1)
# 1.6 self boosted fusion
self_boosted_attn = UniAttention(SymmetricProject(
self.attention_hidden_size, name='self_boosted_attn'),
name='boosted_fusion')
boosted_fusion = self_boosted_attn(context_history, context_history,
self.context_len,
full_fusion_context_repr)
boosted_fusion = dropout(boosted_fusion, self.training)
# 1.7 context vectors
context_final_encoder = CudnnBiLSTM(self.rnn_hidden_size)
context_repr, _ = context_final_encoder(
tf.concat([full_fusion_context_repr, boosted_fusion], axis=-1),
self.context_len)
context_repr = dropout(context_repr, self.training)
self_attn = SelfAttn()
U_Q = self_attn(question_understanding, self.question_len)
# start project
start_project = tf.keras.layers.Dense(self.rnn_hidden_size * 2,
use_bias=False)
context_repr = dropout(context_repr, self.training)
start_logits = tf.squeeze(tf.matmul(start_project(context_repr),
tf.expand_dims(U_Q, axis=-1)),
axis=-1)
start_logits_masked = mask_logits(start_logits, self.context_len)
self.start_prob = tf.nn.softmax(start_logits_masked)
gru_input = tf.reduce_sum(tf.expand_dims(self.start_prob, axis=-1) *
context_repr,
axis=1)
GRUCell = tf.contrib.rnn.GRUCell(self.rnn_hidden_size * 2)
V_Q, _ = GRUCell(gru_input, U_Q)
# end project
end_project = tf.keras.layers.Dense(self.rnn_hidden_size * 2,
use_bias=False)
end_logits = tf.squeeze(tf.matmul(end_project(context_repr),
tf.expand_dims(V_Q, axis=-1)),
axis=-1)
end_logits_masked = mask_logits(end_logits, self.context_len)
self.end_prob = tf.nn.softmax(end_logits_masked)
# 7. Loss and input/output dict
self.start_loss = tf.reduce_mean(
tf.nn.sparse_softmax_cross_entropy_with_logits(
logits=mask_logits(start_logits, self.context_len),
labels=self.answer_start))
self.end_loss = tf.reduce_mean(
tf.nn.sparse_softmax_cross_entropy_with_logits(
logits=mask_logits(end_logits, self.context_len),
labels=self.answer_end))
self.loss = self.start_loss + self.end_loss
global_step = tf.train.get_or_create_global_step()
input_dict = {
"context_word": self.context_word,
"context_len": self.context_len,
"question_word": self.question_word,
"question_len": self.question_len,
"answer_start": self.answer_start,
"answer_end": self.answer_end,
"training": self.training,
}
if len(self.features) > 0 and 'match_lower' in self.features:
input_dict['match_lower'] = self.match_lower
if len(self.features) > 0 and 'match_lemma' in self.features:
input_dict['match_lemma'] = self.match_lemma
if self.use_outer_embedding:
input_dict['context'] = self.context_string,
input_dict['question'] = self.question_string
if len(self.features) > 0 and 'pos' in self.features:
input_dict['pos'] = self.pos_feature
if len(self.features) > 0 and 'ner' in self.features:
input_dict['ner'] = self.ner_feature
if len(self.features) > 0 and 'context_tf' in self.features:
input_dict['context_tf'] = self.normalized_tf
self.input_placeholder_dict = OrderedDict(input_dict)
self.output_variable_dict = OrderedDict({
"start_prob": self.start_prob,
"end_prob": self.end_prob
})
# 8. Metrics and summary
with tf.variable_scope("train_metrics"):
self.train_metrics = {'loss': tf.metrics.mean(self.loss)}
self.train_update_metrics = tf.group(
*[op for _, op in self.train_metrics.values()])
metric_variables = tf.get_collection(tf.GraphKeys.LOCAL_VARIABLES,
scope="train_metrics")
self.train_metric_init_op = tf.variables_initializer(metric_variables)
with tf.variable_scope("eval_metrics"):
self.eval_metrics = {'loss': tf.metrics.mean(self.loss)}
self.eval_update_metrics = tf.group(
*[op for _, op in self.eval_metrics.values()])
metric_variables = tf.get_collection(tf.GraphKeys.LOCAL_VARIABLES,
scope="eval_metrics")
self.eval_metric_init_op = tf.variables_initializer(metric_variables)
tf.summary.scalar('loss', self.loss)
self.summary_op = tf.summary.merge_all()
def _build_graph(self):
self.context_word = tf.placeholder(tf.int32, [None, None])
self.context_char = tf.placeholder(tf.int32, [None, None, None])
self.context_len = tf.placeholder(tf.int32, [None])
self.question_word = tf.placeholder(tf.int32, [None, None])
self.question_char = tf.placeholder(tf.int32, [None, None, None])
self.question_len = tf.placeholder(tf.int32, [None])
self.answer_start = tf.placeholder(tf.int32, [None])
self.answer_end = tf.placeholder(tf.int32, [None])
self.training = tf.placeholder(tf.bool, [])
self.question_tokens = tf.placeholder(tf.string, [None, None])
self.context_tokens = tf.placeholder(tf.string,[None,None])
if self.enable_na_answer:
self.na = tf.placeholder(tf.int32, [None])
# 1. Word encoding
word_embedding = Embedding(pretrained_embedding=self.pretrained_word_embedding,
embedding_shape=(len(self.vocab.get_word_vocab()) + 1, self.word_embedding_size),
trainable=self.word_embedding_trainable)
char_embedding = Embedding(embedding_shape=(len(self.vocab.get_char_vocab()) + 1, self.char_embedding_size),
trainable=True, init_scale=0.2)
# 1.1 Embedding
context_word_repr = word_embedding(self.context_word)
context_char_repr = char_embedding(self.context_char)
question_word_repr = word_embedding(self.question_word)
question_char_repr = char_embedding(self.question_char)
# 1.2 Char convolution
dropout = Dropout(self.keep_prob)
conv1d = Conv1DAndMaxPooling(self.char_conv_filters, self.char_conv_kernel_size)
context_char_repr = dropout(conv1d(context_char_repr), self.training)
question_char_repr = dropout(conv1d(question_char_repr), self.training)
#elmo embedding
if self.use_elmo:
elmo_emb = ElmoEmbedding(local_path=self.elmo_local_path)
context_elmo_repr = elmo_emb(self.context_tokens,self.context_len)
context_elmo_repr = dropout(context_elmo_repr,self.training)
question_elmo_repr = elmo_emb(self.question_tokens,self.question_len)
question_elmo_repr = dropout(question_elmo_repr,self.training)
#concat word and char
context_repr = tf.concat([context_word_repr, context_char_repr],axis=-1)
question_repr = tf.concat([question_word_repr,question_char_repr],axis=-1)
if self.use_elmo:
context_repr= tf.concat([context_repr,context_elmo_repr],axis=-1)
question_repr = tf.concat([question_repr,question_elmo_repr],axis=-1)
# 1.3 Highway network
highway1 = Highway()
highway2 = Highway()
context_repr = highway2(highway1(context_repr))
question_repr = highway2(highway1(question_repr))
# 2. Phrase encoding
phrase_lstm = CudnnBiLSTM(self.rnn_hidden_size)
context_repr, _ = phrase_lstm(dropout(context_repr, self.training), self.context_len)
question_repr, _ = phrase_lstm(dropout(question_repr, self.training), self.question_len)
# 3. Bi-Attention
bi_attention = BiAttention(TriLinear())
c2q, q2c = bi_attention(context_repr, question_repr, self.context_len, self.question_len)
# 4. Modeling layer
final_merged_context = tf.concat([context_repr, c2q, context_repr * c2q, context_repr * q2c], axis=-1)
modeling_lstm1 = CudnnBiLSTM(self.rnn_hidden_size)
modeling_lstm2 = CudnnBiLSTM(self.rnn_hidden_size)
modeled_context1, _ = modeling_lstm1(dropout(final_merged_context, self.training), self.context_len)
modeled_context2, _ = modeling_lstm2(dropout(modeled_context1, self.training), self.context_len)
modeled_context = modeled_context1 + modeled_context2
# 5. Start prediction
start_pred_layer = tf.keras.layers.Dense(1, use_bias=False)
start_logits = start_pred_layer(
dropout(tf.concat([final_merged_context, modeled_context], axis=-1), self.training))
start_logits = tf.squeeze(start_logits, axis=-1)
self.start_prob = masked_softmax(start_logits, self.context_len)
# 6. End prediction
start_repr = weighted_sum(modeled_context, self.start_prob)
tiled_start_repr = tf.tile(tf.expand_dims(start_repr, axis=1), [1, tf.shape(modeled_context)[1], 1])
end_lstm = CudnnBiLSTM(self.rnn_hidden_size)
encoded_end_repr, _ = end_lstm(dropout(tf.concat(
[final_merged_context, modeled_context, tiled_start_repr, modeled_context * tiled_start_repr], axis=-1),
self.training),
self.context_len)
end_pred_layer = tf.keras.layers.Dense(1, use_bias=False)
end_logits = end_pred_layer(dropout(tf.concat(
[final_merged_context, encoded_end_repr], axis=-1), self.training))
end_logits = tf.squeeze(end_logits, axis=-1)
self.end_prob = masked_softmax(end_logits, self.context_len)
# 7. Loss and input/output dict
if self.enable_na_answer:
self.na_bias = tf.get_variable("na_bias", shape=[1], dtype='float')
self.na_bias_tiled = tf.tile(tf.reshape(self.na_bias, [1, 1]), [tf.shape(self.context_word)[0], 1])
self.concat_start_na_logits = tf.concat([self.na_bias_tiled, start_logits], axis=-1)
concat_start_na_prob = masked_softmax(self.concat_start_na_logits, self.context_len + 1)
self.na_prob = tf.squeeze(tf.slice(concat_start_na_prob, [0, 0], [-1, 1]), axis=1)
self.start_prob = tf.slice(concat_start_na_prob, [0, 1], [-1, -1])
self.concat_end_na_logits = tf.concat([self.na_bias_tiled,end_logits],axis=-1)
concat_end_na_prob = masked_softmax(self.concat_end_na_logits,self.context_len+1)
self.na_prob2 =tf.squeeze(tf.slice(concat_end_na_prob,[0,0],[-1,1]),axis=1)
self.end_prob = tf.slice(concat_end_na_prob,[0,1],[-1,-1])
max_len =tf.reduce_max(self.context_len)
start_label = tf.cast(tf.one_hot(self.answer_start,max_len),tf.float32)
start_label =(1.0-tf.cast(tf.expand_dims(self.na,axis=-1),tf.float32))*start_label
na = tf.cast(tf.expand_dims(self.na,axis=-1),tf.float32)
start_na_label = tf.concat([na,start_label],axis=-1)
self.start_loss = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits_v2(
logits=mask_logits(self.concat_start_na_logits,self.context_len+1),
labels=start_na_label))
end_label = tf.cast(tf.one_hot(self.answer_end,max_len),tf.float32)
end_label = (1.0-tf.cast(tf.expand_dims(self.na,axis=-1),tf.float32))*end_label
end_na_label = tf.concat([na,end_label],axis=-1)
self.end_loss = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits_v2(
logits=mask_logits(self.concat_end_na_logits,self.context_len+1),
labels=end_na_label))
else:
self.start_loss = tf.reduce_mean(
tf.nn.sparse_softmax_cross_entropy_with_logits(logits=mask_logits(start_logits, self.context_len),
labels=self.answer_start))
self.end_loss = tf.reduce_mean(
tf.nn.sparse_softmax_cross_entropy_with_logits(logits=mask_logits(end_logits, self.context_len),
labels=self.answer_end))
self.loss = self.start_loss + self.end_loss
global_step = tf.train.get_or_create_global_step()
input_dict = {
"context_word": self.context_word,
"context_char": self.context_char,
"context_len": self.context_len,
"question_word": self.question_word,
"question_char": self.question_char,
"question_len": self.question_len,
"answer_start": self.answer_start,
"answer_end": self.answer_end,
"training": self.training
}
if self.use_elmo:
input_dict['context_tokens'] = self.context_tokens
input_dict['question_tokens'] = self.question_tokens
if self.enable_na_answer:
input_dict["is_impossible"] = self.na
self.input_placeholder_dict = OrderedDict(input_dict)
output_dict = {
"start_prob": self.start_prob,
"end_prob": self.end_prob
}
if self.enable_na_answer:
output_dict['na_prob'] = self.na_prob*self.na_prob2
self.output_variable_dict = OrderedDict(output_dict)
# 8. Metrics and summary
with tf.variable_scope("train_metrics"):
self.train_metrics = {
'loss': tf.metrics.mean(self.loss)
}
self.train_update_metrics = tf.group(*[op for _, op in self.train_metrics.values()])
metric_variables = tf.get_collection(tf.GraphKeys.LOCAL_VARIABLES, scope="train_metrics")
self.train_metric_init_op = tf.variables_initializer(metric_variables)
with tf.variable_scope("eval_metrics"):
self.eval_metrics = {
'loss': tf.metrics.mean(self.loss)
}
self.eval_update_metrics = tf.group(*[op for _, op in self.eval_metrics.values()])
metric_variables = tf.get_collection(tf.GraphKeys.LOCAL_VARIABLES, scope="eval_metrics")
self.eval_metric_init_op = tf.variables_initializer(metric_variables)
tf.summary.scalar('loss', self.loss)
self.summary_op = tf.summary.merge_all()
def _build_graph(self):
self.context_word = tf.placeholder(tf.int32, [None, None])
self.context_len = tf.placeholder(tf.int32, [None])
self.context_char = tf.placeholder(tf.int32, [None, None, None])
self.context_word_len = tf.placeholder(tf.int32, [None, None])
self.question_word = tf.placeholder(tf.int32, [None, None])
self.question_len = tf.placeholder(tf.int32, [None])
self.question_char = tf.placeholder(tf.int32, [None, None, None])
self.question_word_len = tf.placeholder(tf.int32, [None, None])
self.answer_start = tf.placeholder(tf.int32, [None])
self.answer_end = tf.placeholder(tf.int32, [None])
self.abstractive_answer_mask = tf.placeholder(tf.int32, [None, self.abstractive_answer_num])
self.training = tf.placeholder(tf.bool, [])
self.question_tokens = tf.placeholder(tf.string, [None, None])
self.context_tokens = tf.placeholder(tf.string,[None,None])
# 1. Word encoding
word_embedding = Embedding(pretrained_embedding=self.pretrained_word_embedding,
embedding_shape=(len(self.vocab.get_word_vocab()) + 1, self.word_embedding_size),
trainable=self.word_embedding_trainable)
char_embedding = Embedding(embedding_shape=(len(self.vocab.get_char_vocab()) + 1, self.char_embedding_size), trainable=True, init_scale=0.05)
# 1.1 Embedding
dropout = Dropout(self.keep_prob)
context_word_repr = word_embedding(self.context_word)
context_char_repr = char_embedding(self.context_char)
question_word_repr = word_embedding(self.question_word)
question_char_repr = char_embedding(self.question_char)
if self.use_elmo:
elmo_emb = ElmoEmbedding(local_path=self.elmo_local_path)
context_elmo_repr = elmo_emb(self.context_tokens,self.context_len)
context_elmo_repr = dropout(context_elmo_repr,self.training)
question_elmo_repr = elmo_emb(self.question_tokens,self.question_len)
question_elmo_repr = dropout(question_elmo_repr,self.training)
# 1.2 Char convolution
conv1d = Conv1DAndMaxPooling(self.char_conv_filters, self.char_conv_kernel_size)
if self.max_pooling_mask:
question_char_repr = conv1d(dropout(question_char_repr, self.training), self.question_word_len)
context_char_repr = conv1d(dropout(context_char_repr, self.training), self.context_word_len)
else:
question_char_repr = conv1d(dropout(question_char_repr, self.training))
context_char_repr = conv1d(dropout(context_char_repr, self.training))
# 2. Phrase encoding
context_embs = [context_word_repr, context_char_repr]
question_embs = [question_word_repr, question_char_repr]
if self.use_elmo:
context_embs.append(context_elmo_repr)
question_embs.append(question_elmo_repr)
context_repr = tf.concat(context_embs, axis=-1)
question_repr = tf.concat(question_embs, axis=-1)
variational_dropout = VariationalDropout(self.keep_prob)
emb_enc_gru = CudnnBiGRU(self.rnn_hidden_size)
context_repr = variational_dropout(context_repr, self.training)
context_repr, _ = emb_enc_gru(context_repr, self.context_len)
context_repr = variational_dropout(context_repr, self.training)
question_repr = variational_dropout(question_repr, self.training)
question_repr, _ = emb_enc_gru(question_repr, self.question_len)
question_repr = variational_dropout(question_repr, self.training)
# 3. Bi-Attention
bi_attention = BiAttention(TriLinear(bias=True, name="bi_attention_tri_linear"))
c2q, q2c = bi_attention(context_repr, question_repr, self.context_len, self.question_len)
context_repr = tf.concat([context_repr, c2q, context_repr * c2q, context_repr * q2c], axis=-1)
# 4. Self-Attention layer
dense1 = tf.keras.layers.Dense(self.rnn_hidden_size*2, use_bias=True, activation=tf.nn.relu)
gru = CudnnBiGRU(self.rnn_hidden_size)
dense2 = tf.keras.layers.Dense(self.rnn_hidden_size*2, use_bias=True, activation=tf.nn.relu)
self_attention = SelfAttention(TriLinear(bias=True, name="self_attention_tri_linear"))
inputs = dense1(context_repr)
outputs = variational_dropout(inputs, self.training)
outputs, _ = gru(outputs, self.context_len)
outputs = variational_dropout(outputs, self.training)
c2c = self_attention(outputs, self.context_len)
outputs = tf.concat([c2c, outputs, c2c * outputs], axis=len(c2c.shape)-1)
outputs = dense2(outputs)
context_repr = inputs + outputs
context_repr = variational_dropout(context_repr, self.training)
# 5. Modeling layer
sum_max_encoding = SumMaxEncoder()
context_modeling_gru1 = CudnnBiGRU(self.rnn_hidden_size)
context_modeling_gru2 = CudnnBiGRU(self.rnn_hidden_size)
question_modeling_gru1 = CudnnBiGRU(self.rnn_hidden_size)
question_modeling_gru2 = CudnnBiGRU(self.rnn_hidden_size)
self.max_context_len = tf.reduce_max(self.context_len)
self.max_question_len = tf.reduce_max(self.question_len)
modeled_context1, _ = context_modeling_gru1(context_repr, self.context_len)
modeled_context2, _ = context_modeling_gru2(tf.concat([context_repr, modeled_context1], axis=2), self.context_len)
encoded_context = sum_max_encoding(modeled_context1, self.context_len, self.max_context_len)
modeled_question1, _ = question_modeling_gru1(question_repr, self.question_len)
modeled_question2, _ = question_modeling_gru2(tf.concat([question_repr, modeled_question1], axis=2), self.question_len)
encoded_question = sum_max_encoding(modeled_question2, self.question_len, self.max_question_len)
# 6. Predictions
start_dense = tf.keras.layers.Dense(1, activation=None)
start_logits = tf.squeeze(start_dense(modeled_context1), squeeze_dims=[2])
start_logits = mask_logits(start_logits, self.context_len)
end_dense = tf.keras.layers.Dense(1, activation=None)
end_logits = tf.squeeze(end_dense(modeled_context2), squeeze_dims=[2])
end_logits = mask_logits(end_logits, self.context_len)
abstractive_answer_logits = None
if self.abstractive_answer_num != 0:
abstractive_answer_logits = []
for i in range(self.abstractive_answer_num):
tri_linear = TriLinear(name="cls"+str(i))
abstractive_answer_logits.append(tf.squeeze(tri_linear(encoded_context, encoded_question), squeeze_dims=[2]))
abstractive_answer_logits = tf.concat(abstractive_answer_logits, axis=-1)
# 7. Loss and input/output dict
seq_length = tf.shape(start_logits)[1]
start_mask = tf.one_hot(self.answer_start, depth=seq_length, dtype=tf.float32)
end_mask = tf.one_hot(self.answer_end, depth=seq_length, dtype=tf.float32)
if self.abstractive_answer_num != 0:
abstractive_answer_mask = tf.cast(self.abstractive_answer_mask, dtype=tf.float32)
extractive_mask = 1. - tf.reduce_max(abstractive_answer_mask, axis=-1, keepdims=True)
start_mask = extractive_mask * start_mask
end_mask = extractive_mask * end_mask
concated_start_masks = tf.concat([start_mask, abstractive_answer_mask], axis=1)
concated_end_masks = tf.concat([end_mask, abstractive_answer_mask], axis=1)
concated_start_logits = tf.concat([start_logits, abstractive_answer_logits], axis=1)
concated_end_logits = tf.concat([end_logits, abstractive_answer_logits], axis=1)
else:
concated_start_masks = start_mask
concated_end_masks = end_mask
concated_start_logits = start_logits
concated_end_logits = end_logits
start_log_norm = tf.reduce_logsumexp(concated_start_logits, axis=1)
start_log_score = tf.reduce_logsumexp(concated_start_logits + VERY_NEGATIVE_NUMBER * (1- tf.cast(concated_start_masks, tf.float32)), axis=1)
self.start_loss = tf.reduce_mean(-(start_log_score - start_log_norm))
end_log_norm = tf.reduce_logsumexp(concated_end_logits, axis=1)
end_log_score = tf.reduce_logsumexp(concated_end_logits + VERY_NEGATIVE_NUMBER * (1- tf.cast(concated_end_masks, tf.float32)), axis=1)
self.end_loss = tf.reduce_mean(-(end_log_score - end_log_norm))
self.loss = self.start_loss + self.end_loss
global_step = tf.train.get_or_create_global_step()
self.input_placeholder_dict = OrderedDict({
"context_word": self.context_word,
"question_word": self.question_word,
"context_char": self.context_char,
"question_char": self.question_char,
"context_len": self.context_len,
"question_len": self.question_len,
"answer_start": self.answer_start,
"answer_end": self.answer_end,
"training": self.training
})
if self.max_pooling_mask:
self.input_placeholder_dict['context_word_len'] = self.context_word_len
self.input_placeholder_dict['question_word_len'] = self.question_word_len
if self.use_elmo:
self.input_placeholder_dict['context_tokens'] = self.context_tokens
self.input_placeholder_dict['question_tokens'] = self.question_tokens
if self.abstractive_answer_num != 0:
self.input_placeholder_dict["abstractive_answer_mask"] = self.abstractive_answer_mask
self.output_variable_dict = OrderedDict({
"start_logits": start_logits,
"end_logits": end_logits,
})
if self.abstractive_answer_num != 0:
self.output_variable_dict["abstractive_answer_logits"] = abstractive_answer_logits
# 8. Metrics and summary
with tf.variable_scope("train_metrics"):
self.train_metrics = {
'loss': tf.metrics.mean(self.loss)
}
self.train_update_metrics = tf.group(*[op for _, op in self.train_metrics.values()])
metric_variables = tf.get_collection(tf.GraphKeys.LOCAL_VARIABLES, scope="train_metrics")
self.train_metric_init_op = tf.variables_initializer(metric_variables)
with tf.variable_scope("eval_metrics"):
self.eval_metrics = {
'loss': tf.metrics.mean(self.loss)
}
self.eval_update_metrics = tf.group(*[op for _, op in self.eval_metrics.values()])
metric_variables = tf.get_collection(tf.GraphKeys.LOCAL_VARIABLES, scope="eval_metrics")
self.eval_metric_init_op = tf.variables_initializer(metric_variables)
tf.summary.scalar('loss', self.loss)
self.summary_op = tf.summary.merge_all()
def _build_graph(self):
# build input
self.training = tf.placeholder(tf.bool, [])
self.context_word = tf.placeholder(tf.int32, [None, None])
self.slice_context_len = tf.where(self.training, 400, 1000)
self.slice_question_len = tf.where(self.training, 50, 100)
self.context_char = tf.placeholder(tf.int32, [None, None, None])
self.context_char = tf.placeholder(tf.int32, [None, None, None])
batch_size, original_context_len, max_context_word_len = tf.shape(self.context_char)[0],tf.shape(self.context_char)[1],tf.shape(self.context_char)[2]
slice_context_word = tf.slice(self.context_word,[0,0],[batch_size,tf.minimum(self.slice_context_len,original_context_len)])
slice_context_char = tf.slice(self.context_char, [0, 0, 0], [batch_size,tf.minimum(self.slice_context_len,original_context_len),max_context_word_len])
self.context_len = tf.placeholder(tf.int32, [None])
self.question_word = tf.placeholder(tf.int32, [None, None])
self.question_char = tf.placeholder(tf.int32, [None, None, None])
original_question_len,max_question_word_len = tf.shape(self.question_char)[1],tf.shape(self.question_char)[2]
slice_question_word = tf.slice(self.question_word,[0,0],[batch_size,tf.minimum(self.slice_question_len,original_question_len)])
slice_question_char = tf.slice(self.question_char, [0, 0, 0],[batch_size, tf.minimum(self.slice_question_len, original_question_len),max_question_word_len])
self.question_len = tf.placeholder(tf.int32, [None])
self.answer_start = tf.placeholder(tf.int32, [None])
self.answer_end = tf.placeholder(tf.int32, [None])
max_context_len = tf.shape(slice_context_word)[1]
max_question_len = tf.shape(slice_question_word)[1]
slice_context_len = tf.clip_by_value(self.context_len,0,max_context_len)
slice_question_len = tf.clip_by_value(self.question_len,0,max_question_len)
context_mask = (tf.sequence_mask(slice_context_len, max_context_len, dtype=tf.float32)-1)*100
question_mask = (tf.sequence_mask(slice_question_len, max_question_len, dtype=tf.float32)-1) * 100
divisors = tf.pow(tf.constant([10000.0] * (self.filters // 2), dtype=tf.float32),tf.range(0, self.filters, 2, dtype=tf.float32) / self.filters)
quotients = tf.cast(tf.expand_dims(tf.range(0, max_context_len), -1),tf.float32) / tf.expand_dims(divisors, 0)
position_repr = tf.concat([tf.sin(quotients), tf.cos(quotients)], -1) # CL*F
# 1. Word encoding
word_embedding = Embedding(pretrained_embedding=self.pretrained_word_embedding,
embedding_shape=(len(self.vocab.get_word_vocab()) + 1, self.word_embedding_size),
trainable=self.word_embedding_trainable)
char_embedding = Embedding(embedding_shape=(len(self.vocab.get_char_vocab()) + 1, self.char_embedding_size),
trainable=True, init_scale=0.2)
dropout = Dropout(self.keep_prob)
# 1.1 Embedding
context_word_embedding = Dropout(0.9)(word_embedding(slice_context_word),self.training) # B*CL*WD
context_char_embedding = Dropout(0.95)(char_embedding(slice_context_char),self.training) # B*CL*WL*CD
question_word_embedding = Dropout(0.9)(word_embedding(slice_question_word),self.training) # B*QL*WD
question_char_embedding = Dropout(0.95)(char_embedding(slice_question_char),self.training) # B*QL*WL*CD
char_cnn = Conv1DAndMaxPooling(self.char_filters,self.kernel_size1,padding='same',activation=tf.nn.relu)
embedding_dense = tf.keras.layers.Dense(self.filters)
highway = Highway(gate_activation=tf.nn.relu,trans_activation=tf.nn.tanh,hidden_units=self.filters,keep_prob=self.keep_prob)
context_char_repr = tf.reshape(char_cnn(context_char_embedding),[-1,max_context_len,self.char_filters]) # B*CL*CF
context_repr = highway(dropout(embedding_dense(tf.concat([context_word_embedding,context_char_repr],-1)), self.training),self.training) # B*CL*F
question_char_repr = tf.reshape(char_cnn(question_char_embedding),[-1,max_question_len,self.char_filters]) # B*QL*CF
question_repr = highway(dropout(embedding_dense(tf.concat([question_word_embedding,question_char_repr],-1)), self.training),self.training) # B*QL*CF
# 1.2 Embedding Encoder
embedding_encoder = EncoderBlock(self.kernel_size1,self.filters,self.conv_layers1,self.heads,self.keep_prob)
embedding_context = tf.contrib.layers.layer_norm(embedding_encoder(context_repr + position_repr,self.training,context_mask), begin_norm_axis=-1) # B*CL*F
embedding_question = tf.contrib.layers.layer_norm(embedding_encoder(question_repr + position_repr[:max_question_len], self.training,question_mask), begin_norm_axis=-1) # B*QL*F
# 1.3 co-attention
co_attention_context = tf.keras.layers.Dense(1)(embedding_context) # B*CL*1
co_attention_question = tf.keras.layers.Dense(1)(embedding_question) # B*QL*1
cq = tf.matmul(tf.expand_dims(tf.transpose(embedding_context, [0, 2, 1]), -1),tf.expand_dims(tf.transpose(embedding_question, [0, 2, 1]), -2)) # B*F*CL*QL
cq_score = tf.keras.layers.Dense(1)(tf.transpose(cq, [0, 2, 3, 1]))[:, :, :, 0] + co_attention_context + tf.transpose(co_attention_question, [0, 2, 1]) # B*CL*QL
question_similarity = tf.nn.softmax(cq_score + tf.expand_dims(question_mask, -2),2) # B*CL*QL
context_similarity = tf.nn.softmax(cq_score + tf.expand_dims(context_mask, -1), 1) # B*CL*QL
cqa = tf.matmul(question_similarity, embedding_question) # B*CL*F
qca = tf.matmul(question_similarity, tf.matmul(context_similarity, embedding_context, transpose_a=True)) # B*CL*F
co_attention_output = dropout(tf.keras.layers.Dense(self.filters)(tf.concat([embedding_context, cqa, embedding_context * cqa, embedding_context * qca], -1)), self.training) # B*CL*F
# 1.4 Model Encoder
model_encoder_blocks=[EncoderBlock(self.kernel_size2,self.filters,self.conv_layers2,self.heads,self.keep_prob) for _ in range(self.model_encoder_layers)]
m0 = co_attention_output
for model_encoder_block in model_encoder_blocks:
m0 = model_encoder_block(m0 + position_repr,self.training,context_mask)
m1 = m0
for model_encoder_block in model_encoder_blocks:
m1 = model_encoder_block(m1 + position_repr,self.training,context_mask)
m2 = m1
for model_encoder_block in model_encoder_blocks:
m2 = model_encoder_block(m2 + position_repr,self.training,context_mask)
norm_m0 = tf.contrib.layers.layer_norm(m0, begin_norm_axis=-1)
norm_m1 = tf.contrib.layers.layer_norm(m1, begin_norm_axis=-1)
norm_m2 = tf.contrib.layers.layer_norm(m2, begin_norm_axis=-1)
# predict start
start_logits = tf.keras.layers.Dense(1)(tf.concat([norm_m0, norm_m1], -1))[:, :, 0] + context_mask
self.start_prob = tf.nn.softmax(start_logits,-1)
self.start_loss = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=start_logits,labels=tf.one_hot(self.answer_start,max_context_len)))
# predict end
end_logits = tf.keras.layers.Dense(1)(tf.concat([norm_m0, norm_m2], -1))[:, :, 0] + context_mask
self.end_prob = tf.nn.softmax(end_logits,-1)
self.end_loss = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=end_logits, labels=tf.one_hot(self.answer_end,max_context_len)))
self.loss = self.start_loss + self.end_loss
self.global_step = tf.train.get_or_create_global_step()
input_dict = {
"context_word": self.context_word,
"context_char": self.context_char,
"context_len": self.context_len,
"question_word": self.question_word,
"question_char": self.question_char,
"question_len": self.question_len,
"answer_start": self.answer_start,
"answer_end": self.answer_end,
"training": self.training
}
self.input_placeholder_dict = OrderedDict(input_dict)
print(self.input_placeholder_dict)# = OrderedDict(input_dict)
self.output_variable_dict = OrderedDict({
"start_prob": self.start_prob,
"end_prob": self.end_prob
})
# 8. Metrics and summary
with tf.variable_scope("train_metrics"):
self.train_metrics = {
'loss': tf.metrics.mean(self.loss)
}
self.train_update_metrics = tf.group(*[op for _, op in self.train_metrics.values()])
metric_variables = tf.get_collection(tf.GraphKeys.LOCAL_VARIABLES, scope="train_metrics")
self.train_metric_init_op = tf.variables_initializer(metric_variables)
with tf.variable_scope("eval_metrics"):
self.eval_metrics = {
'loss': tf.metrics.mean(self.loss)
}
self.eval_update_metrics = tf.group(*[op for _, op in self.eval_metrics.values()])
metric_variables = tf.get_collection(tf.GraphKeys.LOCAL_VARIABLES, scope="eval_metrics")
self.eval_metric_init_op = tf.variables_initializer(metric_variables)
tf.summary.scalar('loss', self.loss)
self.summary_op = tf.summary.merge_all()
def _build_graph(self):
# build input
self.context_word = tf.placeholder(tf.int32, [None, None])
self.context_char = tf.placeholder(tf.int32, [None, None, None])
self.context_len = tf.placeholder(tf.int32, [None])
self.context_word_len = tf.placeholder(tf.int32, [None, None])
self.question_word = tf.placeholder(tf.int32, [None, None])
self.question_char = tf.placeholder(tf.int32, [None, None, None])
self.question_len = tf.placeholder(tf.int32, [None])
self.question_word_len = tf.placeholder(tf.int32, [None, None])
self.answer_start = tf.placeholder(tf.int32, [None])
self.answer_end = tf.placeholder(tf.int32, [None])
self.training = tf.placeholder(tf.bool, [])
max_context_len = tf.shape(self.context_word)[1]
max_context_word_len = tf.shape(self.context_char)[2]
max_question_len = tf.shape(self.question_word)[1]
max_question_word_len = tf.shape(self.question_char)[2]
context_mask = (tf.sequence_mask(
self.context_len, max_context_len, dtype=tf.float32) - 1) * 100
question_mask = (tf.sequence_mask(
self.question_len, max_question_len, dtype=tf.float32) - 1) * 100
# 1. Word encoding
word_embedding = Embedding(
pretrained_embedding=self.pretrained_word_embedding,
embedding_shape=(len(self.vocab.get_word_vocab()) + 1,
self.word_embedding_size),
trainable=self.word_embedding_trainable)
char_embedding = Embedding(
embedding_shape=(len(self.vocab.get_char_vocab()) + 1,
self.char_embedding_size),
trainable=True,
init_scale=0.2)
dropout = VariationalDropout(self.keep_prob)
context_word_embedding = word_embedding(self.context_word) # B*CL*WD
context_char_embedding = dropout(
tf.reshape(char_embedding(self.context_char),
[-1, max_context_word_len, self.char_embedding_size]),
self.training) # (B*CL)*WL*CD
question_word_embedding = word_embedding(self.question_word) # B*QL*WD
question_char_embedding = dropout(
tf.reshape(char_embedding(self.question_char),
[-1, max_question_word_len, self.char_embedding_size]),
self.training) # (B*QL)*WL*CD
char_forward_rnn = tf.keras.layers.CuDNNGRU(self.hidden_size)
char_backward_rnn = tf.keras.layers.CuDNNGRU(self.hidden_size,
go_backwards=True)
context_char_forward_final_states = char_forward_rnn(
context_char_embedding) # (B*CL)*H
context_char_backward_final_states = char_backward_rnn(
context_char_embedding) # (B*CL)*H
context_char_final_states = tf.reshape(
tf.concat([
context_char_forward_final_states,
context_char_backward_final_states
], -1), [-1, max_context_len, self.hidden_size * 2])
context_repr = tf.concat(
[context_word_embedding, context_char_final_states],
-1) # B*CL*(WD+H)
question_char_forword_final_states = char_forward_rnn(
question_char_embedding) # (B*CL)*H
question_char_backword_final_states = char_backward_rnn(
question_char_embedding) # (B*CL)*H
question_char_final_states = tf.reshape(
tf.concat([
question_char_forword_final_states,
question_char_backword_final_states
], -1), [-1, max_question_len, self.hidden_size * 2])
question_repr = tf.concat(
[question_word_embedding, question_char_final_states],
-1) # B*QL*(WD+H)
# 1.2 Encoder
context_rnn = [
CudnnBiGRU(self.hidden_size) for _ in range(self.doc_rnn_layers)
]
encoder_multi_bigru = MultiLayerRNN(context_rnn,
concat_layer_out=True,
input_keep_prob=self.keep_prob)
encoder_context = dropout(
encoder_multi_bigru(context_repr, self.context_len, self.training),
self.training) # B*CL*(H*2)
encoder_question = dropout(
encoder_multi_bigru(question_repr, self.question_len,
self.training), self.training) # B*QL*(H*2)
# 1.3 co-attention
co_attention_context = tf.expand_dims(
tf.keras.layers.Dense(self.hidden_size)(encoder_context),
2) # B*CL*1*H
co_attention_question = tf.expand_dims(
tf.keras.layers.Dense(self.hidden_size)(encoder_question),
1) # B*1*QL*H
co_attention_score = tf.keras.layers.Dense(1)(
tf.nn.tanh(co_attention_context +
co_attention_question))[:, :, :, 0] + tf.expand_dims(
question_mask, 1) # B*CL*QL
co_attention_similarity = tf.nn.softmax(co_attention_score,
-1) # B*CL*QL
co_attention_rnn_input = tf.concat([
encoder_context,
tf.matmul(co_attention_similarity, encoder_question)
], -1) # B*CL*(H*4)
co_attention_rnn_input = co_attention_rnn_input * tf.keras.layers.Dense(
self.hidden_size * 4,
activation=tf.nn.sigmoid)(co_attention_rnn_input)
co_attention_rnn = CudnnGRU(self.hidden_size)
co_attention_output = dropout(
co_attention_rnn(co_attention_rnn_input, self.context_len)[0],
self.training) # B*CL*(H*2)
# 1.4 self-attention
multi_head_attention = MultiHeadAttention(self.heads, self.hidden_size,
False)
self_attention_repr = dropout(
multi_head_attention(co_attention_output, co_attention_output,
co_attention_output, context_mask),
self.training)
self_attention_rnn_input = tf.concat(
[co_attention_output, self_attention_repr], -1) # B*CL*(H*2)
self_attention_rnn_input = self_attention_rnn_input * tf.keras.layers.Dense(
self.hidden_size * 2,
activation=tf.nn.sigmoid)(self_attention_rnn_input)
self_attention_rnn = CudnnBiGRU(self.hidden_size)
self_attention_output = dropout(
self_attention_rnn(self_attention_rnn_input, self.context_len)[0],
self.training) # B*CL*(H*2)
# predict_start
start_question_score = tf.keras.layers.Dense(1)(
tf.keras.layers.Dense(self.hidden_size, activation=tf.nn.tanh)
(encoder_question)) + tf.expand_dims(question_mask, -1) # B*QL*1
start_question_similarity = tf.nn.softmax(start_question_score,
1) # B*QL*1
start_question_repr = tf.matmul(start_question_similarity,
encoder_question,
transpose_a=True)[:, 0] # B*(H*2)
start_logits = tf.keras.layers.Dense(1)(tf.nn.tanh(
tf.keras.layers.Dense(self.hidden_size)(self_attention_output) +
tf.expand_dims(
tf.keras.layers.Dense(self.hidden_size)
(start_question_repr), 1)))[:, :, 0] + context_mask # B*CL
self.start_prob = tf.nn.softmax(start_logits, -1) # B*CL
self.start_loss = tf.reduce_mean(
tf.nn.sparse_softmax_cross_entropy_with_logits(
logits=start_logits, labels=self.answer_start))
start_repr = tf.matmul(tf.expand_dims(self.start_prob, 1),
self_attention_output) # B*1*(H*2)
start_output = Dropout(self.keep_prob)(tf.keras.layers.CuDNNGRU(
self.hidden_size * 2)(start_repr, start_question_repr),
self.training) # B*(H*2)
# predict_end
end_logits = tf.keras.layers.Dense(1)(tf.nn.tanh(
tf.keras.layers.Dense(self.hidden_size)(self_attention_output) +
tf.expand_dims(
tf.keras.layers.Dense(self.hidden_size)
(start_output), 1)))[:, :, 0] + context_mask # B*CL
self.end_prob = tf.nn.softmax(end_logits, -1) # B*CL
self.end_loss = tf.reduce_mean(
tf.nn.sparse_softmax_cross_entropy_with_logits(
logits=end_logits, labels=self.answer_end))
self.loss = self.start_loss + self.end_loss
self.global_step = tf.train.get_or_create_global_step()
input_dict = {
"context_word": self.context_word,
"context_char": self.context_char,
"context_len": self.context_len,
"context_word_len": self.context_word_len,
"question_word": self.question_word,
"question_char": self.question_char,
"question_len": self.question_len,
"question_word_len": self.question_word_len,
"answer_start": self.answer_start,
"answer_end": self.answer_end,
"training": self.training
}
self.input_placeholder_dict = OrderedDict(input_dict)
print(self.input_placeholder_dict) # = OrderedDict(input_dict)
self.output_variable_dict = OrderedDict({
"start_prob": self.start_prob,
"end_prob": self.end_prob
})
# 8. Metrics and summary
with tf.variable_scope("train_metrics"):
self.train_metrics = {'loss': tf.metrics.mean(self.loss)}
self.train_update_metrics = tf.group(
*[op for _, op in self.train_metrics.values()])
metric_variables = tf.get_collection(tf.GraphKeys.LOCAL_VARIABLES,
scope="train_metrics")
self.train_metric_init_op = tf.variables_initializer(metric_variables)
with tf.variable_scope("eval_metrics"):
self.eval_metrics = {'loss': tf.metrics.mean(self.loss)}
self.eval_update_metrics = tf.group(
*[op for _, op in self.eval_metrics.values()])
metric_variables = tf.get_collection(tf.GraphKeys.LOCAL_VARIABLES,
scope="eval_metrics")
self.eval_metric_init_op = tf.variables_initializer(metric_variables)
tf.summary.scalar('loss', self.loss)
self.summary_op = tf.summary.merge_all()
def _build_graph(self):
self.context_word = tf.placeholder(tf.int32, [None, None])
self.context_len = tf.placeholder(tf.int32, [None])
self.context_char = tf.placeholder(tf.int32, [None, None, None])
self.context_word_len = tf.placeholder(tf.int32, [None, None])
self.question_word = tf.placeholder(tf.int32, [None, None])
self.question_len = tf.placeholder(tf.int32, [None])
self.question_char = tf.placeholder(tf.int32, [None, None, None])
self.question_word_len = tf.placeholder(tf.int32, [None, None])
self.answer_start = tf.placeholder(tf.int32, [None])
self.answer_end = tf.placeholder(tf.int32, [None])
self.training = tf.placeholder(tf.bool, [])
# 1. Word encoding
word_embedding = Embedding(
pretrained_embedding=self.pretrained_word_embedding,
embedding_shape=(len(self.vocab.get_word_vocab()) + 1,
self.word_embedding_size),
trainable=self.word_embedding_trainable)
char_embedding = Embedding(
embedding_shape=(len(self.vocab.get_char_vocab()) + 1,
self.char_embedding_size),
trainable=True,
init_scale=0.05)
# 1.1 Embedding
context_word_repr = word_embedding(self.context_word)
context_char_repr = char_embedding(self.context_char)
question_word_repr = word_embedding(self.question_word)
question_char_repr = char_embedding(self.question_char)
self.context_word_repr = context_word_repr
self.question_word_repr = question_word_repr
self.context_char_repr = context_char_repr
self.question_char_repr = question_char_repr
# 1.2 Char convolution
dropout = Dropout(self.keep_prob)
conv1d = Conv1DAndMaxPooling(self.char_conv_filters,
self.char_conv_kernel_size)
question_char_repr = conv1d(dropout(question_char_repr, self.training),
self.question_word_len)
context_char_repr = conv1d(dropout(context_char_repr, self.training),
self.context_word_len)
# 2. Phrase encoding
context_repr = tf.concat([context_word_repr, context_char_repr],
axis=-1)
question_repr = tf.concat([question_word_repr, question_char_repr],
axis=-1)
variational_dropout = VariationalDropout(self.keep_prob)
emb_enc_gru = CudnnBiGRU(self.rnn_hidden_size)
context_repr = variational_dropout(context_repr, self.training)
context_repr, _ = emb_enc_gru(context_repr, self.context_len)
context_repr = variational_dropout(context_repr, self.training)
question_repr = variational_dropout(question_repr, self.training)
question_repr, _ = emb_enc_gru(question_repr, self.question_len)
question_repr = variational_dropout(question_repr, self.training)
# 3. Bi-Attention
bi_attention = BiAttention(
TriLinear(bias=True, name="bi_attention_tri_linear"))
c2q, q2c = bi_attention(context_repr, question_repr, self.context_len,
self.question_len)
context_repr = tf.concat(
[context_repr, c2q, context_repr * c2q, context_repr * q2c],
axis=-1)
# 4. Self-Attention layer
dense1 = tf.keras.layers.Dense(self.rnn_hidden_size * 2,
use_bias=True,
activation=tf.nn.relu)
gru = CudnnBiGRU(self.rnn_hidden_size)
dense2 = tf.keras.layers.Dense(self.rnn_hidden_size * 2,
use_bias=True,
activation=tf.nn.relu)
self_attention = SelfAttention(
TriLinear(bias=True, name="self_attention_tri_linear"))
inputs = dense1(context_repr)
outputs = variational_dropout(inputs, self.training)
outputs, _ = gru(outputs, self.context_len)
outputs = variational_dropout(outputs, self.training)
c2c = self_attention(outputs, self.context_len)
outputs = tf.concat([c2c, outputs, c2c * outputs],
axis=len(c2c.shape) - 1)
outputs = dense2(outputs)
context_repr = inputs + outputs
context_repr = variational_dropout(context_repr, self.training)
# 5. Modeling layer
sum_max_encoding = SumMaxEncoder()
context_modeling_gru1 = CudnnBiGRU(self.rnn_hidden_size)
context_modeling_gru2 = CudnnBiGRU(self.rnn_hidden_size)
question_modeling_gru1 = CudnnBiGRU(self.rnn_hidden_size)
question_modeling_gru2 = CudnnBiGRU(self.rnn_hidden_size)
self.max_context_len = tf.reduce_max(self.context_len)
self.max_question_len = tf.reduce_max(self.question_len)
modeled_context1, _ = context_modeling_gru1(context_repr,
self.context_len)
modeled_context2, _ = context_modeling_gru2(
tf.concat([context_repr, modeled_context1], axis=2),
self.context_len)
encoded_context = sum_max_encoding(modeled_context1, self.context_len,
self.max_context_len)
modeled_question1, _ = question_modeling_gru1(question_repr,
self.question_len)
modeled_question2, _ = question_modeling_gru2(
tf.concat([question_repr, modeled_question1], axis=2),
self.question_len)
encoded_question = sum_max_encoding(modeled_question2,
self.question_len,
self.max_question_len)
start_dense = tf.keras.layers.Dense(1, activation=None)
start_logits = tf.squeeze(start_dense(modeled_context1),
squeeze_dims=[2])
start_logits = mask_logits(start_logits, self.context_len)
start_prob = tf.nn.softmax(start_logits)
end_dense = tf.keras.layers.Dense(1, activation=None)
end_logits = tf.squeeze(end_dense(modeled_context2), squeeze_dims=[2])
end_logits = mask_logits(end_logits, self.context_len)
end_prob = tf.nn.softmax(end_logits)
self.start_loss = tf.reduce_mean(
tf.nn.sparse_softmax_cross_entropy_with_logits(
logits=start_logits, labels=self.answer_start))
self.end_loss = tf.reduce_mean(
tf.nn.sparse_softmax_cross_entropy_with_logits(
logits=end_logits, labels=self.answer_end))
# 7. Loss and input/output dict
self.loss = self.start_loss + self.end_loss
global_step = tf.train.get_or_create_global_step()
self.input_placeholder_dict = OrderedDict({
"context_word": self.context_word,
"context_len": self.context_len,
"context_char": self.context_char,
"context_word_len": self.context_word_len,
"question_word": self.question_word,
"question_len": self.question_len,
"question_char": self.question_char,
"question_word_len": self.question_word_len,
"answer_start": self.answer_start,
"answer_end": self.answer_end,
"training": self.training
})
self.output_variable_dict = OrderedDict({
"start_prob": start_prob,
"end_prob": end_prob,
})
# 8. Metrics and summary
with tf.variable_scope("train_metrics"):
self.train_metrics = {'loss': tf.metrics.mean(self.loss)}
self.train_update_metrics = tf.group(
*[op for _, op in self.train_metrics.values()])
metric_variables = tf.get_collection(tf.GraphKeys.LOCAL_VARIABLES,
scope="train_metrics")
self.train_metric_init_op = tf.variables_initializer(metric_variables)
with tf.variable_scope("eval_metrics"):
self.eval_metrics = {'loss': tf.metrics.mean(self.loss)}
self.eval_update_metrics = tf.group(
*[op for _, op in self.eval_metrics.values()])
metric_variables = tf.get_collection(tf.GraphKeys.LOCAL_VARIABLES,
scope="eval_metrics")
self.eval_metric_init_op = tf.variables_initializer(metric_variables)
tf.summary.scalar('loss', self.loss)
self.summary_op = tf.summary.merge_all()
def _build_graph(self):
self.context_word = tf.placeholder(tf.int32, [None, None])
self.context_char = tf.placeholder(tf.int32, [None, None, None])
self.context_len = tf.placeholder(tf.int32, [None])
self.question_word = tf.placeholder(tf.int32, [None, None])
self.question_char = tf.placeholder(tf.int32, [None, None, None])
self.question_len = tf.placeholder(tf.int32, [None])
self.answer_start = tf.placeholder(tf.int32, [None])
self.answer_end = tf.placeholder(tf.int32, [None])
self.training = tf.placeholder(tf.bool, [])
# 1. Word encoding
word_embedding = Embedding(
pretrained_embedding=self.pretrained_word_embedding,
embedding_shape=(len(self.vocab.get_word_vocab()) + 1,
self.word_embedding_size),
trainable=self.word_embedding_trainable)
context_word_repr = word_embedding(self.context_word)
question_word_repr = word_embedding(self.question_word)
dropout = Dropout(self.keep_prob)
# 2 inner attention between question word and context word
inner_att = ProjectedDotProduct(self.rnn_hidden_size,
activation=tf.nn.leaky_relu,
reuse_weight=True)
inner_score = inner_att(question_word_repr, context_word_repr)
context_word_softmax = tf.nn.softmax(inner_score, axis=2)
question_inner_representation = tf.matmul(context_word_softmax,
context_word_repr)
question_word_softmax = tf.nn.softmax(inner_score, axis=1)
context_inner_representation = tf.matmul(question_word_softmax,
question_word_repr,
transpose_a=True)
highway1 = Highway()
highway2 = Highway()
context_repr = highway1(
highway2(
tf.concat([context_word_repr, context_inner_representation],
axis=-1)))
question_repr = highway1(
highway2(
tf.concat([question_word_repr, question_inner_representation],
axis=-1)))
# 2. Phrase encoding
phrase_lstm = CudnnBiLSTM(self.rnn_hidden_size)
context_repr, _ = phrase_lstm(dropout(context_repr, self.training),
self.context_len)
question_repr, _ = phrase_lstm(dropout(question_repr, self.training),
self.question_len)
# 3. Bi-Attention
bi_attention = BiAttention(TriLinear())
c2q, q2c = bi_attention(context_repr, question_repr, self.context_len,
self.question_len)
# 4. Modeling layer
final_merged_context = tf.concat(
[context_repr, c2q, context_repr * c2q, context_repr * q2c],
axis=-1)
modeling_lstm1 = CudnnBiLSTM(self.rnn_hidden_size)
modeling_lstm2 = CudnnBiLSTM(self.rnn_hidden_size)
modeled_context1, _ = modeling_lstm1(
dropout(final_merged_context, self.training), self.context_len)
modeled_context2, _ = modeling_lstm2(
dropout(modeled_context1, self.training), self.context_len)
modeled_context = modeled_context1 + modeled_context2
# 5. Start prediction
start_pred_layer = tf.keras.layers.Dense(1, use_bias=False)
start_logits = start_pred_layer(
dropout(
tf.concat([final_merged_context, modeled_context], axis=-1),
self.training))
start_logits = tf.squeeze(start_logits, axis=-1)
self.start_prob = masked_softmax(start_logits, self.context_len)
# 6. End prediction
start_repr = weighted_sum(modeled_context, self.start_prob)
tiled_start_repr = tf.tile(tf.expand_dims(start_repr, axis=1),
[1, tf.shape(modeled_context)[1], 1])
end_lstm = CudnnBiLSTM(self.rnn_hidden_size)
encoded_end_repr, _ = end_lstm(
dropout(
tf.concat([
final_merged_context, modeled_context, tiled_start_repr,
modeled_context * tiled_start_repr
],
axis=-1), self.training), self.context_len)
end_pred_layer = tf.keras.layers.Dense(1, use_bias=False)
end_logits = end_pred_layer(
dropout(
tf.concat([final_merged_context, encoded_end_repr], axis=-1),
self.training))
end_logits = tf.squeeze(end_logits, axis=-1)
self.end_prob = masked_softmax(end_logits, self.context_len)
# 7. Loss and input/output dict
self.start_loss = tf.reduce_mean(
tf.nn.sparse_softmax_cross_entropy_with_logits(
logits=mask_logits(start_logits, self.context_len),
labels=self.answer_start))
self.end_loss = tf.reduce_mean(
tf.nn.sparse_softmax_cross_entropy_with_logits(
logits=mask_logits(end_logits, self.context_len),
labels=self.answer_end))
self.loss = (self.start_loss + self.end_loss) / 2
global_step = tf.train.get_or_create_global_step()
self.input_placeholder_dict = OrderedDict({
"context_word": self.context_word,
"context_char": self.context_char,
"context_len": self.context_len,
"question_word": self.question_word,
"question_char": self.question_char,
"question_len": self.question_len,
"answer_start": self.answer_start,
"answer_end": self.answer_end,
"training": self.training
})
self.output_variable_dict = OrderedDict({
"start_prob": self.start_prob,
"end_prob": self.end_prob
})
# 8. Metrics and summary
with tf.variable_scope("train_metrics"):
self.train_metrics = {'loss': tf.metrics.mean(self.loss)}
self.train_update_metrics = tf.group(
*[op for _, op in self.train_metrics.values()])
metric_variables = tf.get_collection(tf.GraphKeys.LOCAL_VARIABLES,
scope="train_metrics")
self.train_metric_init_op = tf.variables_initializer(metric_variables)
with tf.variable_scope("eval_metrics"):
self.eval_metrics = {'loss': tf.metrics.mean(self.loss)}
self.eval_update_metrics = tf.group(
*[op for _, op in self.eval_metrics.values()])
metric_variables = tf.get_collection(tf.GraphKeys.LOCAL_VARIABLES,
scope="eval_metrics")
self.eval_metric_init_op = tf.variables_initializer(metric_variables)
tf.summary.scalar('loss', self.loss)
self.summary_op = tf.summary.merge_all()