def __init__(self, model, model_save_path, num_workers=4, batch_size=128, transforms=None):
optimizer = optim.Adam(model.parameters())
scheduler = SchedulerWrapperLossOnPlateau(optimizer)
loss = nn.CrossEntropyLoss()
metrics_calculator = MetricsCalculatorAccuracy()
trainer_cls = TrainerClassification
if transforms is None:
transforms = ToTensor()
train_dataset = get_dataset(mode=MNISTImagesDataset.MODE_TRAIN, transforms=transforms)
val_dataset = get_dataset(mode=MNISTImagesDataset.MODE_VAL, transforms=transforms)
super().__init__(
model=model,
model_save_path=model_save_path,
optimizer=optimizer,
scheduler=scheduler,
loss=loss,
metrics_calculator=metrics_calculator,
batch_size=batch_size,
num_workers=num_workers,
train_dataset=train_dataset,
val_dataset=val_dataset,
trainer_cls=trainer_cls)
def __init__(self, model, model_save_path, num_workers=8, batch_size=128, transforms=None, epoch_count=100, print_frequency=10):
optimizer = optim.SGD(model.parameters(), lr=0.1, weight_decay=5e-4)
scheduler = SchedulerWrapperLossOnPlateau(optimizer)
loss = nn.CrossEntropyLoss()
metrics_calculator = MetricsCalculatorAccuracy()
trainer_cls = TrainerClassification
if transforms is None:
transforms = ToTensor()
train_dataset = get_dataset(path=TRAIN_DATASET_PATH, transforms=transforms, train=True)
val_dataset = get_dataset(path=TEST_DATASET_PATH, transforms=transforms, train=False)
super().__init__(
model=model,
model_save_path=model_save_path,
optimizer=optimizer,
scheduler=scheduler,
loss=loss,
metrics_calculator=metrics_calculator,
batch_size=batch_size,
num_workers=num_workers,
train_dataset=train_dataset,
val_dataset=val_dataset,
trainer_cls=trainer_cls,
print_frequency=print_frequency,
epoch_count=epoch_count)
def __init__(self, model, model_save_path, num_workers=8, batch_size=128, transforms=None):
#optimizer = optim.SGD(model.parameters(), lr=0.1, momentum=0.9, nesterov=True, weight_decay=5e-4)
optimizer = optim.Adam(model.parameters(), lr=0.001)#, weight_decay=5e-4)
scheduler = SchedulerWrapperLossOnPlateau(optimizer)
loss = nn.CrossEntropyLoss()
metrics_calculator = MetricsCalculatorAccuracy()
trainer_cls = TrainerClassification
if transforms is None:
transforms = ToTensor()
train_dataset = get_dataset(path=TRAIN_DATASET_PATH, transforms=transforms)
val_dataset = get_dataset(path=TEST_DATASET_PATH, transforms=transforms)
super().__init__(
model=model,
model_save_path=model_save_path,
optimizer=optimizer,
scheduler=scheduler,
loss=loss,
metrics_calculator=metrics_calculator,
batch_size=batch_size,
num_workers=num_workers,
train_dataset=train_dataset,
val_dataset=val_dataset,
trainer_cls=trainer_cls)
def __init__(self, model):
optimizer = optim.Adam(model.parameters())
scheduler = SchedulerWrapperLossOnPlateau(optimizer)
loss = nn.CrossEntropyLoss()
metrics_calculator = MetricsCalculatorAccuracy()
super().__init__(model=model,
optimizer=optimizer,
scheduler=scheduler,
loss=loss,
metrics_calculator=metrics_calculator,
batch_size=32,
num_workers=4)
def __init__(self,
model,
model_save_path,
num_workers=16,
batch_size=128,
learning_rate=0.1,
transforms=None,
use_mixup=False):
parameters_bias = [
p[1] for p in model.named_parameters() if 'bias' in p[0]
]
parameters_scale = [
p[1] for p in model.named_parameters() if 'scale' in p[0]
]
parameters_others = [
p[1] for p in model.named_parameters()
if not ('bias' in p[0] or 'scale' in p[0])
]
optimizer = optim.SGD([{
'params': parameters_bias,
'lr': learning_rate / 10.
}, {
'params': parameters_scale,
'lr': learning_rate / 10.
}, {
'params': parameters_others
}],
lr=learning_rate,
momentum=0.9,
weight_decay=5e-4)
scheduler = SchedulerWrapperLossOnPlateau(optimizer)
loss = nn.CrossEntropyLoss()
metrics_calculator = MetricsCalculatorAccuracy()
trainer_cls = TrainerClassification
if transforms is None:
transforms = Compose([
ToTensor(),
Normalize((0.4914, 0.4822, 0.4465), (0.2023, 0.1994, 0.2010))
])
train_dataset = get_dataset(path=TRAIN_DATASET_PATH,
transforms=transforms,
use_mixup=use_mixup)
val_dataset = get_dataset(path=TEST_DATASET_PATH,
transforms=transforms,
use_mixup=use_mixup)
if use_mixup:
train_dataset = MixUpDatasetWrapper(train_dataset, alpha=0.7)
loss = VectorCrossEntropy()
super().__init__(model=model,
model_save_path=model_save_path,
optimizer=optimizer,
scheduler=scheduler,
loss=loss,
metrics_calculator=metrics_calculator,
batch_size=batch_size,
num_workers=num_workers,
train_dataset=train_dataset,
val_dataset=val_dataset,
trainer_cls=trainer_cls,
print_frequency=100)
def test_accuracy(self):
metrics_calculator = MetricsCalculatorAccuracy(border=0.4)
with pytest.raises(PipelineError):
metrics_calculator.calculate()