merged

training/train.py

@@ -47,12 +47,13 @@
 BATCH_SIZE = 4
 MODEL_NAME = 'm2m_lstm'
 POSITION = (39.7392, -104.99903)
-FEATURES_TRAIN = ['air_temperature', 'humidity']
+FEATURES_TRAIN = ['air_temperature']
 FEATURES_PREDICT = ['air_temperature']
 FILENAMES_TRAIN = ['2016', '2015', '2014', '2013', '2012', '2011', '2010', '2009', '2008']
 FILENAMES_VALID = ['2017']
 T_TRAIN_H = 7 * 24
 T_PRED_D = 3
+MASK_VALUE = 999
 
 
 def train(batch_size=BATCH_SIZE,
@@ -217,13 +218,13 @@ def train(batch_size=BATCH_SIZE,
                            help='Name of the model to train. Available:\n ' + str([str(k) for k in models.keys()]),
                            default=MODEL_NAME)
     argparser.add_argument('--log_dir', help='Path to store training output', default=LOG_DIR)
-    argparser.add_argument('--data_dir', help='Path to read data files', default=DATA_DIR)
+    argparser.add_argument('--data_dir', help='Path to read data files', default=RADIUS)
     argparser.add_argument('--batch_size', help='Size of one Batch', default=BATCH_SIZE)
     argparser.add_argument('--n_samples', help='Amount of samples to train', default=None)
     args = argparser.parse_args()
-
-    train(batch_size=args.batch_size,
-          log_dir=args.log_dir,
-          data_dir=args.data_dir,
-          model_name=args.model_name,
-          n_samples=args.n_samples)
+    train()
+   # train(batch_size=args.batch_size,
+   #       log_dir=args.log_dir,
+   #       data_dir=args.data_dir,
+   #       model_name=args.model_name,
+   #       n_samples=args.n_samples)