train(
saveto='./{}.npz'.format(model_name),
reload_=True,
dim_word=100, # or 300 glove
dim=300,
patience=10,
n_words=22671, #s p500 46174 #ding 33976 40325
clip_c=10.,
lrate=0.0004,
optimizer='adam',
maxlen=None,
batch_size=32,
valid_batch_size=32,
dispFreq=20,
validFreq=int(1411 / 32 + 1), #1411 #1379
saveFreq=int(1411 / 32 + 1),
use_dropout=True,
verbose=False,
delay1=3,
delay2=7,
types='title',
cut_word=False,
cut_news=70,
last_layer='LSTM',
CNN_filter=128,
CNN_kernel=3,
keep_prob=0.5,
datasets=[
'../data/ding_new_10/train.csv',
'../data/ding_new_10/train_label.csv'
],
valid_datasets=[
'../data/ding_new_10/validate.csv',
'../data/ding_new_10/validate_label.csv'
],
test_datasets=[
'../data/ding_new_10/test.csv',
'../data/ding_new_10/test_label.csv'
],
dictionary='../data/ding_new_10/vocab_cased_title.pickle',
embedding='../emb/vectors_latest.txt',
wait_N=1,
)
train(
saveto = './{}.npz'.format(model_name),
reload_ = True,
dim_word = 100, # or 300 glove
dim = 300,
patience = 20,
n_words = 154000, #s p500 46174 #ding 33976 40325
clip_c = 10.,
lrate = 0.0004,
optimizer = 'adam',
maxlen = None,
batch_size = 8,
valid_batch_size = 8,
dispFreq = 100,
validFreq = int(1411/8+1),#s p500 1321 # ding 1426
saveFreq = int(1411/8+1),
use_dropout = True,
verbose = False,
delay1 = 3,
delay2 = 7,
types = 'article',
cut_word = 30,
cut_sentence = 30,
cut_news = 30,
last_layer = 'LSTM',
CNN_filter = 64,
CNN_kernel = 3,
keep_prob = 0.5,
datasets = ['../data/ding_new_4/train.csv',
'../data/ding_new_4/train_label.csv'],
valid_datasets = ['../data/ding_new_4/validate.csv',
'../data/ding_new_4/validate_label.csv'],
test_datasets = ['../data/ding_new_4/test.csv',
'../data/ding_new_4/test_label.csv'],
dictionary = '../data/ding_new_4/vocab_cased_article.pickle',
embedding = '../emb/yunke_latest_medium.txt',
wait_N = 1,
)