import torch
import sys
# sys.path.append("..")
from NNTrain import trainModels
torch.manual_seed(0)
torch.backends.cudnn.deterministic = True
torch.backends.cudnn.benchmark = False
# ====================================
if __name__ == '__main__':
trainModels(data_directory='./datasets/',
dataset_name='debug',
input_dim=3,
class_no=2,
repeat=1,
train_batchsize=4,
validate_batchsize=1,
num_epochs=10,
learning_rate=1e-2,
width=32,
network='unet',
augmentation='all_flip',
reverse=False)
# 8. augmentation. parameter to choose different data augmentation. We have options: 'full', 'flip',
# 'all_filp', 'gaussian' or 'none'.
# 9. The results will be genereted in a new folder called Results.
# trainModels(data_directory='./datasets/',
# dataset_name='multiclass',
# input_dim=4,
# class_no=4,
# repeat=1,
# train_batchsize=4,
# validate_batchsize=1,
# num_epochs=1,
# learning_rate=1e-3,
# width=64,
# network='unet',
# augmentation='full'
# )
trainModels(data_directory='/cluster/project2/Neuroblastoma/',
dataset_name='data',
input_dim=3,
class_no=2,
repeat=1,
train_batchsize=2,
validate_batchsize=1,
num_epochs=100,
learning_rate=1e-3,
width=32,
network='unet',
augmentation='full')
import torch
import sys
sys.path.append("..")
from NNTrain import trainModels
# torch.manual_seed(0)
torch.backends.cudnn.deterministic = True
torch.backends.cudnn.benchmark = False
if __name__ == '__main__':
trainModels(data_directory='/SAN/medic/PerceptronHead/data/brain/BRATS2018/',
dataset_name='ET_L0_H50',
input_dim=4,
class_no=2,
repeat=1,
train_batchsize=4,
validate_batchsize=1,
num_epochs=100,
learning_rate=1e-2,
width=32,
network='erf_fp_decoder',
dilation=12,
augmentation='all_flip',
reverse=False)