def main(_):
config = get_config()
q_config = run_lstmlm_withBegToken.small_config(data)
name = create_name(config, q_config)
logdir = wb.mkdir('./trf_nn/' + name, is_recreate=True)
sys.stdout = wb.std_log(logdir + '/trf.log')
config.pprint()
print(logdir)
# write data
data.write_vocab(logdir + '/vocab.txt')
data.write_data(data.datas[0], logdir + '/train.id')
data.write_data(data.datas[1], logdir + '/valid.id')
data.write_data(data.datas[2], logdir + '/test.id')
m = trf.TRF(config, data, logdir=logdir, device='/gpu:0', simulater_device='/gpu:0',
q_model=lstmlm.LM(run_lstmlm_withBegToken.small_config(data), device='/gpu:0'))
ops = Ops(m)
sv = tf.train.Supervisor(logdir=logdir + '/logs', summary_op=None, global_step=m._global_step)
# sv.summary_writer.add_graph(tf.get_default_graph()) # write the graph to logs
session_config = tf.ConfigProto(allow_soft_placement=True, log_device_placement=False)
session_config.gpu_options.allow_growth = True
with sv.managed_session(config=session_config) as session:
m.set_session(session)
print('load lstmlm q(x)')
m.q_model.restore(session,
'./lstm/' + run_lstmlm_withBegToken.create_name(m.q_model.config) + '/model.ckpt')
m.train(sv, session,
print_per_epoch=0.05,
operation=ops,
model_per_epoch=None)
def create_name(config, q_config):
s = str(config) + '_with_' + run_lstmlm_withBegToken.create_name(q_config)
if config.feat_type_file is not None:
s += '_' + os.path.split(config.feat_type_file)[-1].split('.')[0]
if config.opt_method != 'adam':
s += '_' + config.opt_method
return s
def main(_):
data = reader.Data().load_raw_data(reader.ptb_raw_dir(),
add_beg_token='<s>',
add_end_token='</s>',
add_unknwon_token='<unk>')
nbest = reader.NBest(*reader.wsj0_nbest())
nbest_list = data.load_data(nbest.nbest, is_nbest=True)
config = trfnce.Config(data)
config.structure_type = 'cnn'
config.embedding_dim = 256
config.cnn_filters = [(i, 128) for i in range(1, 11)]
config.cnn_hidden = 128
config.cnn_layers = 3
config.cnn_skip_connection = True
config.cnn_activation = 'tanh'
config.batch_size = 50
config.noise_factor = 20
config.noise_sampler = 2
config.init_weight = 0.1
config.lr_param = trfbase.LearningRateTime(1e-3)
config.max_epoch = 10
config.global_normalized = True
config.update_zeta = False
config.init_zeta = 10
config.dropout = 0
config.write_dbg = False
config.pprint()
name = create_name(config)
logdir = 'trf_nce/' + name
wb.mkdir(logdir, is_recreate=True)
sys.stdout = wb.std_log(os.path.join(logdir, 'trf.log'))
print(logdir)
data.write_vocab(logdir + '/vocab.txt')
data.write_data(data.datas[1], logdir + '/valid.id')
data.write_data(data.datas[2], logdir + '/test.id')
data.write_data(nbest_list, logdir + '/nbest.id')
# wb.rmdir(logdirs)
with tf.Graph().as_default():
m = trfnce.TRF(config,
data,
logdir=logdir,
device='/gpu:1',
q_model=lstmlm.LM(
run_lstmlm_withBegToken.small_config(data),
device='/gpu:1'))
sv = tf.train.Supervisor(logdir=os.path.join(logdir, 'logs'),
global_step=m.train_net.global_step)
sv.summary_writer.add_graph(
tf.get_default_graph()) # write the graph to logs
session_config = tf.ConfigProto(allow_soft_placement=True,
log_device_placement=False)
session_config.gpu_options.allow_growth = True
with sv.managed_session(config=session_config) as session:
m.set_session(session)
print('load lstmlm for noise generator')
m.q_model.restore(
session, './lstm/' +
run_lstmlm_withBegToken.create_name(m.q_model.config) +
'/model.ckpt')
m.train(sv,
session,
print_per_epoch=0.1,
nbest=nbest,
nbest_list=nbest_list)
def create_name(config, q_config):
return str(config) + '_with_' + run_lstmlm_withBegToken.create_name(q_config) + '_zeta0'