def _imp(hp_dict, cuda_id, save_path):
model = models.DenseNetBC(**hp_dict, n_cls=experimental_settings.n_cls)
train_data, test_data = datasets.get_data(train_dataset,
test_dataset,
batch_size=model.batch_size)
return evaluating_model(model, hp_dict, train_data, test_data, cuda_id,
save_path)
def _imp(hp_dict, cuda_id, save_path):
model = models.MultiLayerPerceptron(**hp_dict,
n_cls=experimental_settings.n_cls,
image_size=image_size)
train_data, test_data = datasets.get_data(train_dataset,
test_dataset,
batch_size=model.batch_size)
return evaluating_model(model, hp_dict, train_data, test_data, cuda_id,
save_path)
def _imp(hp_dict, cuda_id, save_path):
args = models.Transfer(**hp_dict, \
n_cls=experimental_settings.n_cls, \
epochs=experimental_settings.epochs, \
opt=experimental_settings.opt, \
mtra=experimental_settings.mtransfer, \
gcam=experimental_settings.gcam
)
train_data, test_data = datasets.get_data(train_dataset,
test_dataset,
batch_size=args.batch_size)
if experimental_settings.mtransfer.upper() == "RESNET50":
model = m.resnet50(pretrained=True)
num_ftrs = model.fc.in_features
elif experimental_settings.mtransfer.upper() == "RESNET152":
model = m.resnet50(pretrained=True)
num_ftrs = model.fc.in_features
elif experimental_settings.mtransfer.upper() == "EFF7":
model = EfficientNet.from_pretrained('efficientnet-b7')
num_ftrs = model._fc.in_features
elif experimental_settings.mtransfer.upper() == "EFF0":
model = EfficientNet.from_pretrained('efficientnet-b0')
num_ftrs = model._fc.in_features
elif experimental_settings.mtransfer.upper() == "WRN50_2":
model = torch.hub.load('pytorch/vision:v0.6.0',
'wide_resnet50_2',
pretrained=True)
num_ftrs = model.fc.in_features
elif experimental_settings.mtransfer.upper() == "WRN101_2":
model = torch.hub.load('pytorch/vision:v0.6.0',
'wide_resnet50_2',
pretrained=True)
num_ftrs = model.fc.in_features
model.fc = nn.Linear(num_ftrs, experimental_settings.n_cls)
return evaluating_model_t(model, hp_dict, train_data, test_data,
cuda_id, save_path, args, train_idx,
valid_idx)
def _imp(hp_dict, cuda_id, save_path):
model = models.CNN(**hp_dict, n_cls=experimental_settings["n_cls"])
train_data, test_data = datasets.get_data(train_dataset,
test_dataset,
batch_size=model.batch_size)
return train(model, hp_dict, train_data, test_data, cuda_id, save_path)
