def wt2_opt():
reader_opt = ChainMap(
{
'sentences': False,
# if sentences is False
'min_seq_len': 35,
'max_seq_len': 35,
# fi
'shuffle': False,
'batch_size': 64,
'vocab_path': '', # auto to data
'text_path': '' # auto to data
},
reader.default_reader_opt())
model_opt = ChainMap(
{
'emb_dim': 650,
'rnn_dim': 650,
'rnn_layers': 2,
'rnn_variational': True,
'rnn_input_keep_prob': 0.5,
'rnn_layer_keep_prob': 0.7,
'rnn_output_keep_prob': 0.5,
'rnn_state_keep_prob': 0.7,
'logit_weight_tying': True,
'vocab_size': -1 # auto to data
},
lm.default_rnnlm_opt())
train_opt = ChainMap(
{
'loss_key': 'mean_token_nll', # or sum_token_nll
'ngram_loss_coeff': 0.1,
'init_learning_rate': 0.003,
'decay_rate': 0.85,
'staircase': True,
'optim': 'tensorflow.train.AdamOptimizer',
# if adam
'optim_beta1': 0.0,
'optim_beta2': 0.999,
'optim_epsilon': 1e-8,
# fi
'clip_gradients': 5.0,
'max_epochs': 40,
'checkpoint_path': 'tmp', # auto to exp_dir
'decay_steps': -1 # if -1 auto to an epoch
},
lm.default_train_opt())
return reader_opt, model_opt, train_opt
'loss_key': 'mean_l2',
'init_learning_rate': 0.001,
'decay_rate': 0.8,
'staircase': True,
'optim': 'tensorflow.train.AdamOptimizer',
# if adam
'optim_beta1': 0.9,
'optim_beta2': 0.999,
'optim_epsilon': 1e-8,
# fi
'clip_gradients': 5.0,
'max_epochs': 10,
'checkpoint_path': 'tmp', # auto to exp_dir
'decay_steps': -1 # if -1 auto to an epoch
},
lm.default_train_opt())
parser = argparse.ArgumentParser()
parser.add_argument('data_dir', type=str)
parser.add_argument('exp_dir', type=str)
parser.add_argument('--log_level', type=str, default='info')
parser.add_argument('--delete', action='store_true')
args = vars(parser.parse_args())
exp_path = partial(os.path.join, args['exp_dir'])
encoder_path = partial(os.path.join, os.path.join(args['exp_dir'], 'encoder'))
data_path = partial(os.path.join, args['data_dir'])
util.ensure_dir(os.path.join(args['exp_dir'], 'encoder'),
delete=args['delete'])
with open(exp_path('reader_opt.json')) as fp: