def start_training_lr_finder(self, epochs, model, device, test_loader,
train_loader, lr, weight_decay, lambda_fn):
'''
:param epochs: epochs to train
:param model: CNN model
:param device: device cuda or not cuda
:param test_loader: test image loader
:param train_loader: train image loader
:param lr: start learning rate value
:param weight_decay: weight decay or l2 regularization value
:param lambda_fn: lambda function be used for scheduler
:return: lr_data, class_correct, class_total
'''
lr_data = []
class_correct = list(0. for i in range(10))
class_total = list(0. for i in range(10))
optimizer = self.get_optimizer(model=model,
lr=lr,
weight_decay=weight_decay)
scheduler = Utils.create_scheduler_lambda_lr(lambda_fn, optimizer)
return self.start_training(epochs,
model,
device,
test_loader,
train_loader,
optimizer,
scheduler,
lr_data,
class_correct,
class_total,
path="savedmodels/lrfinder.pt")
def start_training_lr_finder(self, epochs, model, device, test_loader,
train_loader, lr, weight_decay, lambda_fn):
lr_data = []
class_correct = list(0. for i in range(10))
class_total = list(0. for i in range(10))
optimizer = self.get_optimizer(model=model,
lr=lr,
weight_decay=weight_decay)
scheduler = Utils.create_scheduler_lambda_lr(lambda_fn, optimizer)
return self.start_training(epochs,
model,
device,
test_loader,
train_loader,
optimizer,
scheduler,
lr_data,
class_correct,
class_total,
path="savedmodels/lrfinder.pt")
