def autoencoder_old(path_train, path_valid, batch_size, epochs, path_save, option):
#!/usr/bin/env python3
# -*- coding: utf8 -*-
import ctypes
import os
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '1'
os.environ['TF_ENABLE_XLA'] = '1'
os.environ['TF_ENABLE_AUTO_MIXED_PRECISION'] = '1'
os.environ['TF_ENABLE_CUDNN_RNN_TENSOR_OP_MATH_FP32'] = '1'
os.environ['TF_DISABLE_CUDNN_TENSOR_OP_MATH'] = '1'
os.environ['TF_ENABLE_CUBLAS_TENSOR_OP_MATH_FP32'] = '1'
import numpy as np
import tensorflow as tf
gpus = tf.config.experimental.list_physical_devices('GPU')
if gpus:
try:
tf.config.experimental.set_virtual_device_configuration(gpus[0], [tf.config.experimental.VirtualDeviceConfiguration(memory_limit=4096)])
except RuntimeError as e:
print(e)
from keras.models import Sequential
from keras.layers import InputLayer, Dense, Flatten, Dropout, Conv2D, MaxPooling2D, Conv2DTranspose, LeakyReLU, Reshape, Activation, BatchNormalization, UpSampling2D
from keras.activations import selu
from keras.constraints import max_norm
from keras.optimizers import Adam, SGD
from keras.utils import to_categorical, normalize
from keras.callbacks import ModelCheckpoint
from keras.datasets import mnist
from keras import losses
from keras import backend as K
import matplotlib.pyplot as plt
import numpy as np
import random
#custom generator import
from DataGenerator import DataGenerator
# GET DATA
# def getData():
# Datasets
# path_train = '/media/aneumann/Harddisk/Datenbanken/PythonTest/test/Preprocessing/Train'
# path_valid = '/media/aneumann/Harddisk/Datenbanken/PythonTest/test/Preprocessing/Validation'
train = os.listdir(path_train)
train = sorted(train)
valid = os.listdir(path_valid)
valid = sorted(valid)
#
# def batch_generator(path, batch_size):
# files = os.listdir(path)
# files = sorted(files,key=lambda x: int(x.split("_",2)[1]))
# ids = np.arange(len(files))
# np.random.shuffle(ids)
#
# batch = []
# batch_counter = 0
# # while batch_counter < 10:#np.floor(len(ids)/batch_size):
#
# while True:
# # np.random.shuffle(ids)
# X = list()
# Y = list()
#
# for i in ids:
# batch.append(i)
# file = files[i]
#
# try:
# loadfile = np.load(path + '/' + file)
# except Exception as e:
# print("Couldn't load file: " + str(file) + "\nError: " + str(e))
#
# xl = np.nan_to_num(loadfile['x'], nan=0.0001, posinf=1000, neginf=-1000)
# yl = np.nan_to_num(loadfile['y'], nan=0.0001, posinf=1000, neginf=-1000)
# xl = (xl-np.mean(xl))/np.std(xl)
# yl = (yl-np.mean(xl))/np.std(xl)
# X.extend(xl)
# Y.extend(yl)
#
# if len(batch)==batch_size:
# yield np.array(X), np.array(Y)
# batch = []
# X = list()
# Y = list()
#
# batch_counter += 1
# break
#
# train_generator = batch_generator(path_train, batch_size)
# valid_generator = batch_generator(path_valid, batch_size)
train_generator = DataGenerator(path_train, option, batch_size, True)
valid_generator = DataGenerator(path_valid, option, batch_size, True)
# BUILD MODEL
max_norm_value = 100.0
model = Sequential([
# InputLayer(input_shape=(256,128,1)),
# BatchNormalization(axis=1),
Conv2D(filters=32, kernel_size=(3,3), strides=1, padding="same", kernel_constraint=max_norm(max_norm_value), kernel_initializer="he_normal", input_shape=(256, 128, 1), name="enc_conv1"),
LeakyReLU(),
# BatchNormalization(),
Dropout(0.3, name="enc_drop1"),
MaxPooling2D(pool_size=(2,2), name="enc_pool1"),
Conv2D(filters=32, kernel_size=(3,3), strides=1, padding="same", kernel_constraint=max_norm(max_norm_value), kernel_initializer="he_normal", name="enc_conv2"),
LeakyReLU(),
MaxPooling2D(pool_size=(2,2), name="enc_pool2"),
Conv2D(filters=32, kernel_size=(3,3), strides=1, padding="same", kernel_constraint=max_norm(max_norm_value), kernel_initializer="he_normal", name="enc_conv3"),
LeakyReLU(),
MaxPooling2D(pool_size=(2,2), name="enc_pool3"),
Conv2D(filters=32, kernel_size=(3,3), strides=1, padding="same", kernel_constraint=max_norm(max_norm_value), kernel_initializer="he_normal", name="enc_conv4"),
LeakyReLU(),
MaxPooling2D(pool_size=(2,2), name="enc_pool4"),
Flatten(),
Dense(units = 8192),
Reshape((32,16,16)),
UpSampling2D(size=(2,2), name="dec_up1"),
Conv2D(filters=32, kernel_size=(3,3), strides=1, padding="same", kernel_constraint=max_norm(max_norm_value), kernel_initializer="he_normal", name="dec_conv1"),
LeakyReLU(),
UpSampling2D(size=(2,2), name="dec_up2"),
Conv2D(filters=32, kernel_size=(3,3), strides=1, padding="same", kernel_constraint=max_norm(max_norm_value), kernel_initializer="he_normal", name="dec_conv2"),
LeakyReLU(),
Dropout(0.3, name="dec_drop1"),
UpSampling2D(size=(2,2), name="dec_up3"),
Conv2D(filters=1, kernel_size=(3,3), strides=1, padding="same", activation='linear', kernel_constraint=max_norm(max_norm_value), kernel_initializer="he_normal", name="dec_conv3")
])
# opt = Adam(lr=0.000001)
model.compile(loss='mape', optimizer='adam', metrics=['mse', 'accuracy'])
# MODEL Train
if not os.path.exists("weights_cnn"):
try:
os.mkdir("weights_cnn")
except Exception as e:
print("Konnte Ordner für Gewichte nicht erstellen" + str(e))
filepath = "weights_cnn/weights-{epoch:02d}-{loss:.4f}.hdf5"
checkpoint = ModelCheckpoint(
filepath,
monitor='loss',
verbose=0,
save_best_only=True,
mode='min'
)
model.summary()
history = model.fit(train_generator, steps_per_epoch = int(np.floor(len(train) / batch_size)), epochs = epochs, validation_data=valid_generator, validation_steps = int(np.floor(len(valid) / batch_size)), use_multiprocessing=True)
model.summary()
model.save(path_save + '/cnn_base_whole.h5')
plt.style.use("ggplot")
plt.figure()
plt.plot(history.history["loss"], label="train_loss")
plt.plot(history.history["val_loss"], label="val_loss")
plt.title("Training Loss and Accuracy")
plt.xlabel("Epochs")
plt.ylabel("Loss/Accuracy")
plt.legend(loc="lower left")
plt.savefig(path_save + '/testrun.png', dpi=400)
plt.show()