Merge pull request #44 from mhpi/dmg_dev

small updates for consistency

Author: leoglonz

.gitignore

@@ -182,16 +182,6 @@ archive/
 validation/
 
 
-# Non-public config files
-# deltaModel/conf/*
-# !deltaModel/conf/config.yaml
-# !deltaModel/conf/config.py
-# !deltaModel/conf/hydra
-# !deltaModel/conf/observations
-# deltaModel/conf/observations/*
-# !deltaModel/conf/observations/none.yaml
-
-
 # Usage-specific additions: (Only track base/example files)
 # deltaModel/trainers/
 # deltaModel/models/physics_models/

deltaModel/conf/config.py

@@ -25,17 +25,6 @@ def check_path(v: str) -> Path:
     return path
 
 
-class InitalizationEnum(str, Enum):
-    kaiming_normal = "kaiming_normal"
-    kaiming_uniform = "kaiming_uniform"
-    orthogonal = "orthogonal"
-    sparse = "sparse"
-    trunc_normal = "trunc_normal"
-    xavier_normal = "xavier_normal"
-    xavier_uniform = "xavier_uniform"
-    default = "default"
-
-
 class ModeEnum(str, Enum):
     train = 'train'
     test = 'test'

deltaModel/conf/observations/camels_531.yaml

@@ -350,6 +350,7 @@ def _get_basin_area(self, c_nn: np.ndarray) -> np.ndarray:
 
     def normalize(self, x_nn: np.ndarray, c_nn: np.ndarray) -> np.ndarray:
         """Normalize data for neural network."""
+        # TODO: Add np.swapaxes(x_nn, 1, 0) here and remove from _to_norm. This changes normalization, need to determine if it's detrimental.
         x_nn_norm = self._to_norm(
             np.swapaxes(x_nn, 1, 0).copy(),
             self.nn_forcings,
@@ -391,11 +392,12 @@ def _to_norm(self, data: np.ndarray, vars: List[str]) -> np.ndarray:
                 data_norm[:, k] = (data[:, k] - stat[2]) / stat[3]
             else:
                 raise DataDimensionalityWarning("Data dimension must be 2 or 3.")
-            
+
+        # Should be external, except altering order of first two dims augments normalization.
         if len(data_norm.shape) < 3:
             return data_norm
         else:
-            return np.swapaxes(data_norm, 1, 0)  ##TODO move to outside of _to_norm
+            return np.swapaxes(data_norm, 1, 0)
 
     def _from_norm(self, data_norm: np.ndarray, vars: List[str]) -> np.ndarray:
         """De-normalize data."""

deltaModel/conf/observations/camels_671.yaml

@@ -24,6 +24,8 @@ class HydroMSDataLoader(BaseDataLoader):
         Configuration dictionary.
     test_split : bool, optional
         Whether to split data into training and testing sets. Default is False.
+    overwrite : bool, optional
+        Whether to overwrite existing normalization statistics. Default is False.
     """
     def __init__(
         self,

deltaModel/conf/observations/merit_forward.yaml

@@ -133,6 +133,9 @@ def initialize_config(config: Union[DictConfig, dict]) -> Dict[str, Any]:
     if config['dpl_model']['nn_model'].get('lr_scheduler', '') in ['none', 'None', '']:
         config['dpl_model']['nn_model']['lr_scheduler'] = None
 
+    if config.get('trained_model', '') in ['none', 'None', '']:
+        config['trained_model'] = ''
+
     # Create output directories and add path to config.
     out_path = PathBuilder(config)
     config = out_path.write_path(config)