import numpy as np import plotly import plotly.graph_objs as go from HypeNet.Networks.CNN_Simple import CNN_Simple from HypeNet.Core.loadData import loadMnist from HypeNet.Core.Trainer import Trainer from HypeNet.Core.utils import * import os np.random.seed(0) DIR = os.path.dirname(os.path.abspath(__file__)) + '/SavedNetwork/MnistCNN/' X_train, Y_train, X_val, Y_val, Y_train_label, Y_val_label = loadMnist( flatten=False) num_epoch = 1 minibatch_size = 50 save_network = True learning_rate = 0.01 optimizer_type = 'nesterov' print('network created') network = CNN_Simple() print('network setting finished') trainer = Trainer(network, X_train, Y_train, X_val, Y_val, num_epoch, minibatch_size,
import numpy as np import matplotlib.pyplot as plt from HypeNet.Networks.FCNN_MSE import FCNN_MSE from HypeNet.Core.loadData import loadMnist from HypeNet.Core.Trainer import Trainer from HypeNet.Core.utils import * X_train, Y_train, X_val, Y_val, Y_train_label, Y_val_label = loadMnist() num_epoch = 20 minibatch_size = 256 network = FCNN_MSE(784, [1000, 500, 100, 500, 1000], 784, ['Relu', 'Relu', 'Relu', 'Relu', 'Relu', 'Sigmoid'], weight_init_std='he', use_dropout=False, keep_probs=[0.9, 0.9, 0.9, 0.9, 0.9], use_batchnorm=False) trainer = Trainer(network, X_train, X_train, X_val, X_val, num_epoch, minibatch_size, 'adam', {'lr': 0.0004}, verbose=True, LossAccInterval=10000, lr_scheduler_type='exp_decay', lr_scheduler_params={'k': 0.00001}) train_loss_list, val_loss_list, train_acc_list, val_acc_list, x, lrs = trainer.train( )