--- a
+++ b/utils/Config.py
@@ -0,0 +1,39 @@
+"""
+Common utility functions and classes
+"""
+import torch
+
+class Option():
+    """Training Configuration """
+    name = "Lung-Segmentation"
+    # root dir of training and validation set
+    root_dir = '/media/storage/wu1114/RSNA/rsna-unet/'
+
+    test_root = '/media/storage/wu1114/RSNA/stage_1_test'
+
+    result_root = '/media/storage/wu1114/RSNA/rsna-unet/test_result/'
+
+    img_size = 512
+    num_workers = 1     	# number of threads for data loading
+    shuffle = False      	# shuffle the data set
+    batch_size = 8     		# GTX1060 3G Memory
+    epochs = 150			# number of epochs to train
+    plot_every = 5          # vis every N batches
+    is_train = True     	# True for training, False for making prediction
+    save_model = True   	# True for saving the model, False for not saving the model
+    caffe_pretrain = False
+    env = 'RSNA-UNet'
+
+    n_gpu = 2				# number of GPUs
+
+    learning_rate = 1e-3    # learning rage
+    weight_decay = 1e-4		# weight decay
+
+    pin_memory = True   	# use pinned (page-locked) memory. when using CUDA, set to True
+
+    is_cuda = torch.cuda.is_available()  	# True --> GPU
+    num_gpus = torch.cuda.device_count()  	# number of GPUs
+    checkpoint_dir = "./checkpoints"  		# dir to save checkpoints
+    dtype = torch.cuda.FloatTensor if is_cuda else torch.Tensor  # data type
+
+opt = Option()
\ No newline at end of file