test_loss_summ = graph.get_tensor_by_name("test_batch_loss:0")
train_avg_loss_summ = graph.get_tensor_by_name("train_avg_loss:0")
test_avg_loss_summ = graph.get_tensor_by_name("train_avg_loss:0")
train_gini_summ = graph.get_tensor_by_name("train_gini:0")
test_gini_summ = graph.get_tensor_by_name("test_gini:0")
train_decile_summ = graph.get_tensor_by_name("train_decile:0")
test_decile_summ = graph.get_tensor_by_name("test_decile:0")
train_count = 0
test_count = 0
for i in range(previous_count + 1, previous_count + epochs + 1):
train_x = divide_batches_gen(trans_train_data, batch_size)
test_x = divide_batches_gen(trans_test_data, batch_size)
# Train Data.
count = 0
train_loss = 0
for train_data, train_label in zip(train_x, train_y):
train_count += 1
count += 1
_, l = sess.run([optimizer, loss],
feed_dict={
x: train_data,
y: train_label
})
x_lstm = graph.get_tensor_by_name('placeholders/input_lstm:0')
y = graph.get_tensor_by_name('placeholders/output:0')
z = graph.get_tensor_by_name('placeholders/z:0')
lr = graph.get_tensor_by_name('placeholders/lr:0')
kp = graph.get_tensor_by_name('placeholders/kp:0')
y_ = tf.get_collection("y_")[0]
# writer = tf.summary.FileWriter(logdir, sess.graph)
# writer.add_graph(sess.graph)
train_count = 0
test_count = 0
sess.run(lr, feed_dict={lr: learning_rate})
train_ffn_x = divide_batches_gen(ffn_train_data, batch_size)
train_lstm_x = divide_batches(lstm_train_data, batch_size)
# Calculate decile.
train_predictions = []
for train_data_ffn, train_data_lstm in zip(train_ffn_x, train_lstm_x):
model_prediction = sess.run(y_,
feed_dict={
x_ffn: train_data_ffn,
x_lstm: train_data_lstm,
kp: 1.0
})
train_predictions.append(temp for temp in model_prediction)
train_predictions = [
item for sublist in train_predictions for item in sublist
with tf.Session() as sess:
model_saver.restore(sess, ckpt)
graph = tf.get_default_graph()
# for op in tf.get_default_graph().get_operations():
# print(str(op.name))
x = graph.get_tensor_by_name('placeholders/input:0')
y = graph.get_tensor_by_name('placeholders/output:0')
z = graph.get_tensor_by_name('placeholders/z:0')
y_ = tf.get_collection("y_")[0]
loss = tf.get_collection("loss")[0]
infer_x = divide_batches_gen(inference_data, batch_size)
count = 0
infer_loss = 0
infer_predictions = []
for infer_data, infer_label in zip(infer_x, infer_y):
count += 1
l, model_prediction = sess.run([loss, y_],
feed_dict={
x: infer_data,
y: infer_label
})
infer_loss += l