hyperparams.py

import torch
import random
torch.manual_seed(121)
random.seed(121)

learning_rate = 0.001
epochs = 500
batch_size = 16
log_interval = 1
test_interval = 100
save_interval = 100
save_dir = "snapshot"
datafile_path = "./data/"
name_trainfile = "raw.clean.train"
name_devfile = "raw.clean.dev"
name_testfile = "raw.clean.test"
word_data = False
char_data = False
shuffle = True
epochs_shuffle = True
FIVE_CLASS_TASK = False
TWO_CLASS_TASK = True
dropout = 0.6
dropout_embed = 0.6
max_norm = 5
clip_max_norm = 3
kernel_num = 200
kernel_sizes = "1,2,3,4"
# kernel_sizes = "5"
static = False
layer_num_highway = 1
CNN = False
BiLSTM_1 = False
BiLSTM_LIST = False
LSTM_LIST = False
BiLSTM_LIST_CAT = False
HighWay = False
HBiLSTM = False
HLSTM = False
HBiLSTM_CAT = False
HCNN = True
HighWayCNN = False
HighWayBiLSTM = False
HighWay_CNN = False
HighWay_BiLSTM = False
# select optim algorhtim to train
Adam = True
SGD = False
Adadelta = False
optim_momentum_value = 0.9
# whether to use wide convcolution True : wide  False : narrow
wide_conv = True
# min freq to include during built the vocab, default is 1
min_freq = 1
# word_Embedding
word_Embedding = False
embed_dim = 300
# word_Embedding_Path = "./word2vec/glove.sentiment.conj.pretrained.txt"
word_Embedding_Path = "./word2vec/glove.840B.300d.handledword2vec.txt"
# word_Embedding_Path = "./word.txt"
lstm_hidden_dim = 300
lstm_num_layers = 1
device = 0
no_cuda = False
snapshot = None
predict = None
test = False
num_threads = 1
freq_1_unk = False
# whether to init w
init_weight = True
init_weight_value = 2.0
# L2 weight_decay
weight_decay = 1e-8   # default value is zero in Adam SGD
# weight_decay = 0   # default value is zero in Adam SGD
# random seed
seed_num = 233
# whether to delete the model after test acc so that to save space
rm_model = True