utils.printParameters(config)
# ---- Training ----
config1 = tf.ConfigProto()
config1.gpu_options.per_process_gpu_memory_fraction = 0.85
with tf.Session(config=config1) as sess:
# saver = tf.train.import_meta_graph('model.ckpt.meta')
# saver.restore(sess, 'model.ckpt')
embedding_matrix = tf.get_variable('embedding_matrix', shape=config.wordvectors.shape, dtype=tf.float32,
trainable=False).assign(config.wordvectors)
emb_mtx = sess.run(embedding_matrix)
model = tf_utils.model(config, emb_mtx, sess)
obj, m_op, predicted_op_ner, actual_op_ner, predicted_op_rel, actual_op_rel, score_op_rel = model.run()
train_step = model.get_train_op(obj)
operations = tf_utils.operations(train_step, obj, m_op, predicted_op_ner, actual_op_ner, predicted_op_rel,
actual_op_rel, score_op_rel)
sess.run(tf.global_variables_initializer())
saver = tf.train.Saver()
best_score = 0
nepoch_no_imprv = 0 # for early stopping
for iter in range(config.nepochs + 1):
model.train(train_data, operations, iter)
save_path = saver.save(sess, "model.ckpt")
print("Model saved in path: %s" % save_path)
# ---- Testing ----
print("Starting Testing\n")
saver = tf.train.import_meta_graph('model.ckpt.meta')
embedding_matrix = tf.get_variable('embedding_matrix',
shape=config.wordvectors.shape,
dtype=tf.float32,
trainable=False).assign(
config.wordvectors)
emb_mtx = sess.run(embedding_matrix)
model = tf_utils.model(config, emb_mtx, sess)
obj, m_op, transition_params1, entity1Scores, predEntity1, transition_params2, entity2Scores, predEntity2 = model.run(
)
train_step = model.get_train_op(obj)
operations = tf_utils.operations(train_step, obj, m_op,
transition_params1, entity1Scores,
predEntity1, transition_params2,
entity2Scores, predEntity2)
sess.run(tf.global_variables_initializer())
best_score = 0
min_loss = 0
nepoch_no_imprv = 0 # for early stopping
best_test_score = 0.
for iter in range(config.nepochs + 1):
loss = model.train(train_data, operations, iter)
dev_score = model.evaluate(dev_data, operations, 'dev')
