import tensorflow as tf
import numpy as np
import keras
import metrics
import os
import time

from keras.models import Sequential, Model
from keras.layers import Dense, LSTM, Activation, Dropout, Input, Masking,BatchNormalization
from keras.callbacks import ReduceLROnPlateau, TensorBoard, EarlyStopping, ModelCheckpoint
from keras.optimizers import Adam
from imblearn.over_sampling import SMOTE
from write_log import write_log
import test_2 as read_data
from results import Results

#Setup GPU
os.environ["CUDA_DEVICE_ORDER"] = "PCI_BUS_ID"
os.environ["CUDA_VISIBLE_DEVICES"] = "0"

#Config minimize GPU with model
config_gpu = False
if config_gpu:
    config = tf.ConfigProto(allow_soft_placement=True, log_device_placement=False)
    config.gpu_options.allow_growth = True
    sess = tf.Session(config=config)
    keras.backend.set_session(sess)

#Training parameter
epochs = 200
batch_size = 1024
timesteps = 48
data_dim = 40
units = 512
dropout = 0.4
num_class =1
layers = 2

results_list=[]

#Create file name
time_str = time.strftime("%y%m%d-%H%M%S")
model_name ='lstm'+str('_')+str(layers)+str('_')+str(units)+str('-')+str('_')+time_str

directory_logs = './result/'+model_name+'/log/'
if not os.path.exists(directory_logs):
   os.makedirs(directory_logs)

directory_model = './result/'+model_name+'/model/'
if not os.path.exists(directory_model):
   os.makedirs(directory_model)

directory_csv = './result/'+model_name+'/'
if not os.path.exists(directory_csv):
   os.makedirs(directory_csv)

file_name = directory_csv + model_name + '.csv'

#Create SMOTE data
sm = SMOTE(random_state=22)

#Read data
# samples, labels = read_data.read_data()
# num_samples = samples.shape[0]
# index = np.arange(num_samples)
# np.random.shuffle(index)
# samples = samples[index]
# labels = labels[index]
# print(len(labels[labels ==1]))
# print(len(labels[labels==0]))


# train_X = samples[:int(num_samples*0.7)]
# train_y = labels[:int(num_samples*0.7)]
#
# print(len(train_y[train_y ==1]))
# print(len(train_y[train_y==0]))
# # -----------Val set-----------
# temp_X = samples[int(num_samples*0.7):]
# temp_y = labels[int(num_samples*0.7):]
#
# val_X = temp_X[:int(temp_X.shape[0]*0.5)]
# val_y = temp_y[:int(temp_y.shape[0]*0.5)]
# print(len(val_y[val_y ==1]))
# print(len(val_y[val_y==0]))
# #------------Test set-----------
# test_X = temp_X[int(temp_X.shape[0]*0.5):]
# test_y = temp_y[int(temp_y.shape[0]*0.5):]
# print(len(test_y[test_y ==1]))
# print(len(test_y[test_y==0]))

#Build model
input_model = Input(shape=(timesteps, data_dim))
x = Masking()(input_model)
x = LSTM(units, return_sequences=True)(x)
x = Dropout(dropout)(x)
x = LSTM(units, return_sequences=True)(x)
x = Dropout(dropout)(x)
x = LSTM(units)(x)

x = Dense(units)(x)
x = BatchNormalization()(x)
x = Activation('relu')(x)
x = Dropout(dropout)(x)
x = Dense(units)(x)
x = BatchNormalization()(x)
x = Activation('relu')(x)
result = Dense(num_class, activation='sigmoid')(x)

model = Model(inputs=input_model, outputs=result)
#Summary
model.summary()

#Compile model
model.compile(loss='binary_crossentropy',optimizer=Adam(lr=0.001, beta_1=0.9, beta_2=0.999, epsilon=1e-7),metrics=['accuracy'])

#Reducer Learning rate
lr_reducer = ReduceLROnPlateau(monitor='val_loss', factor=0.9, patience=3, verbose=1)

#early_stopper = EarlyStopping(monitor='val_loss', min_delta=0, patience=8, verbose=1, mode='auto')

#Check point save all model
filepath= directory_model +"/weights-improvement-{epoch:03d}-{val_acc:.4f}.hdf5"
checkpoint = ModelCheckpoint(filepath, monitor='val_acc', verbose=1, save_best_only=False, mode='max')

#Tensor board
tensor_board = TensorBoard(log_dir=directory_logs, histogram_freq=0, write_graph=True, write_images=True)

#Results
# results = Results(val_X, val_y, batch_size, results_list)

#List callbacks
callbacks_list = [checkpoint,tensor_board,lr_reducer]

# fit network
history = model.fit_generator(read_data.read_data(),steps_per_epoch = 1000)

write_log(file_name,results_list)

# make a prediction
# y_hat = model.predict(test_X, batch_size = batch_size)
# yhat = np.array(yhat)[:, 0]

# metrics.print_metrics_binary(np.array(test_y), yhat)