A Python class for configuration-oriented creation of deep neural networks. When used in Python notebooks, graphs of cost function and accuracy of prediction on training and testing sets are shown and updated in real time during training.
Better documentation with self-contained examples will be added later.
def_params = {
'optimizer': 'Adam', # Gradient descent optimization algorithm. Options:
# 'Adadelta','Adagrad','Adam','AdamW','SparseAdam',
# 'Adamax','ASGD','LBFGS','NAdam','RAdam','RMSprop',
# 'Rprop','SGD'
'lr': 0.0001, # learning rate
'loss_function': 'CrossEntropy', # loss function used to assess output accuracy. Options:
# 'L1', 'MSE', 'BCE', 'BCEWithLogits', 'NLL', 'PoissonNLL',
# 'CrossEntropy', 'HingeEmbedding', 'MarginRanking',
# 'TripletMargin', 'KLDiv'
'max_epochs': 30, # maximum learning epochs
'weights_init': 'Kaiming', # weights initialization method (other option: 'Xavier')
'use_gpu_if_available': 1, # 0 always use CPU
'dropout_rate': 0.25 # percentage of random units per layer whose weight to disregard in training
}
params = {
'net_input' : [28,28],
'layers_params' : [
[ 'Conv2d', { 'tf':3, 'krnsize':[5,5], 'padding':[1,1] } ],
[ 'MaxPool2d', { 'krnsize': [2,2] } ],
[ 'BatchNorm2d', {} ],
[ 'ReLU', {} ],
[ 'Conv2d', {'tf':20, 'krnsize':[5,5], 'padding':[1,1] } ],
[ 'MaxPool2d', { 'krnsize':[2,2] } ],
[ 'BatchNorm2d', {} ],
[ 'ReLU', {} ],
[ 'ToLinear' , {} ],
[ 'Linear', {'tn':50 } ],
[ 'ReLU', {} ],
[ 'Linear', {'tn':10, 'name':'classifier'}, ]
]
}
net = DNN(params)
net.test_flow()
Finally, after creating DataLoaders for training and testing (assume they are named train_loader, test_loader):
trainAcc,testAcc,losses = net.train(train_loader, test_loader)