params.py

import tensorflow as tf

def basic_params():
    '''A set of basic hyperparameters'''
    return tf.contrib.training.HParams(

        # File path
        pre_embedding = 'data/processed/glove_embedding.npy', # None or path to pretrained embedding
        
        # NN params
        voca_size = 34004,
        embedding_size = 300,
        embedding_trainable = False,
        hidden_size = 512,
        cell_type = 'lstm', # 'lstm' or 'gru'
        enc_type = 'bi', # 'bi' or 'mono'
        encoder_layer = 1,
        decoder_layer = 2,
        dropout = 0.4,
        attn = 'normed_bahdanau',  # 'bahdanau', 'normed_bahdanau', 'luong', 'scaled_luong'
        beam_width = 5,
        length_penalty_weight = 2.1,
       
        # Extra params
        dtype = tf.float32,
        maxlen_s = 60,
        maxlen_dec_train = 32,
        maxlen_dec_dev = 27,
        start_token = 1, # <GO> index
        end_token = 2, # <EOS> index

        # Learning params
        batch_size = 64,
        learning_rate = 0.001,
        decay_step = None,
        decay_rate = 0.5,
        sample_prob = 0.25,
        )
    
def other_params():
    hparams = basic_params()
    hparams.voca_size = 30004
    hparams.embedding = None
    hparams.embedding_trainable = True
    hparams.hidden_size = 300
    hparams.encoder_layer = 1
    haprams.decoder_layer = 1

    hparams.rnn_dropout = 0.3

    hparams.batch_size = 128

    hparams.add_hparam('decay', 0.4) # learning rate decay factor
    return hparams