def create_model(session, targetSpaceSize, vocabsize, forward_only):
"""Create SSE model and initialize or load parameters in session."""
modelParams = {'max_seq_length': FLAGS.max_seq_length, 'vocab_size': vocabsize,
'embedding_size': FLAGS.embedding_size, 'encoding_size': FLAGS.encoding_size,
'learning_rate': FLAGS.learning_rate, 'learning_rate_decay_factor': FLAGS.learning_rate_decay_factor,
'src_cell_size':FLAGS.src_cell_size, 'tgt_cell_size':FLAGS.tgt_cell_size,
'network_mode': FLAGS.network_mode, 'predict_nbest':FLAGS.predict_nbest,
'targetSpaceSize':targetSpaceSize, 'forward_only': forward_only}
data_utils.save_model_configs(FLAGS.model_dir, modelParams)
model = sse_model.SSEModel( modelParams )
ckpt = tf.train.get_checkpoint_state(FLAGS.model_dir)
if ckpt:
print("Reading model parameters from %s" % ckpt.model_checkpoint_path)
model.saver.restore(session, ckpt.model_checkpoint_path)
else:
if forward_only:
print('Error!!!Could not load any model from specified folder: %s' % FLAGS.model_dir )
exit(-1)
else:
print("Created model with fresh parameters.")
session.run(tf.global_variables_initializer())
return model
def create_model(session, targetSpaceSize, vocabsize, forward_only):
"""Create SSE model and initialize or load parameters in session."""
modelConfigs = ( FLAGS.max_seq_length, FLAGS.max_gradient_norm, vocabsize,
FLAGS.embedding_size, FLAGS.encoding_size,
FLAGS.src_cell_size, FLAGS.tgt_cell_size, FLAGS.num_layers,
FLAGS.learning_rate, FLAGS.learning_rate_decay_factor, targetSpaceSize ,
FLAGS.network_mode , FLAGS.predict_nbest, FLAGS.alpha, FLAGS.neg_samples )
data_utils.save_model_configs(FLAGS.model_dir, modelConfigs)
model = sse_model.SSEModel( FLAGS.max_seq_length, FLAGS.max_gradient_norm, vocabsize,
FLAGS.embedding_size, FLAGS.encoding_size,
FLAGS.src_cell_size, FLAGS.tgt_cell_size, FLAGS.num_layers,
FLAGS.learning_rate, FLAGS.learning_rate_decay_factor, targetSpaceSize ,
network_mode=FLAGS.network_mode , forward_only=forward_only, TOP_N=FLAGS.predict_nbest, alpha=FLAGS.alpha, neg_samples = FLAGS.neg_samples )
ckpt = tf.train.get_checkpoint_state(FLAGS.model_dir)
if ckpt:
print("Reading model parameters from %s" % ckpt.model_checkpoint_path)
model.saver.restore(session, ckpt.model_checkpoint_path)
else:
if forward_only:
print('Error!!!Could not load any model from specified folder: %s' % FLAGS.model_dir )
exit(-1)
else:
print("Created model with fresh parameters.")
session.run(tf.global_variables_initializer())
return model