def run():
with open('models.csv', 'r') as csvfile:
reader = csv.reader(csvfile)
header = next(reader)
# print(header)
for row in reader:
args = dict(zip(header, row))
model_name = args['model_name']
# print(model_name)
if os.path.exists('models/{}/config.json'.format(model_name)):
args = load_config('models/{}/config.json'.format(model_name))
else:
create_config(args)
preprocess(args)
organize(args)
print("-------- training --------")
train(args)
import csv
from config import create_config
from preprocess import preprocess, organize
from keras_train import train
with open('models.csv', 'rb') as csvfile:
reader = csv.reader(csvfile)
header = reader.next()
#print header
for row in reader:
args = dict(zip(header, row))
print args['model_name']
create_config(args)
preprocess(args)
organize(args)
train(args)
if len(sys.argv) > 1:
arg = sys.argv[1]
if arg == 'ensemble': # too lazy to make a flag
current_ensemble()
sys.exit(0)
else:
weights_path = sys.argv[1]
else:
weights_path = None
# load all data sets
X_train, y_train, X_val, y_val, X_test = load_data()
# fit the model
model, history = train(X_train, y_train, X_val, y_val, weights_path)
# save results from training
loss = np.array(history.history['loss'])
acc = np.array(history.history['acc'])
top_k = np.array(history.history['top_k_categorical_accuracy'])
val_loss = np.array(history.history['val_loss'])
val_acc = np.array(history.history['val_acc'])
val_top_k = np.array(history.history['val_top_k_categorical_accuracy'])
np.savetxt("loss.csv", loss, delimiter=",")
np.savetxt("acc.csv", acc, delimiter=",")
np.savetxt("top_k.csv", top_k, delimiter=",")
np.savetxt("val_loss.csv", val_loss, delimiter=",")
np.savetxt("val_acc.csv", val_acc, delimiter=",")
np.savetxt("val_top_k.csv", val_top_k, delimiter=",")