taware_xlarge.yml

type: 'topic_aware'
learning_rate: 0.0002
max_gradient_norm: 5.0
vocab_size: 50000
num_layers: 4
embedding_type: 'glove840B'
cell_type: 'gru'
encoder_type: 'bi'
hidden_units: 1024
attention_type: 'bahdanau'
optimizer: 'adam'
batch_size: 128

num_buckets: 5
topic_words_per_utterance: 100
src_max_len: 30
tgt_max_len: 30

num_train_epochs: 15
steps_per_stats: 30
steps_per_eval: 3000
# Early Stopping parameters, inspired from VHRED impl
patience: 50
degrade_threshold: 1.003

encoder_dropout_rate: 0.2
decoder_dropout_rate: 0.2

decoding_length_factor: 2.0

learning_rate_decay_scheme: luong234
#start_decay_step: 0
#decay_steps: 10000
#decay_factor: 0.5

scheduled_sampling_prob: 0.0
scheduled_sampling_decay_scheme: none

beam_width: 5
length_penalty_weight: 1.0
sampling_temperature: 0.0
infer_batch_size: 128

log_device: False

boost_topic_gen_prob: True