Merge pull request #9 from kjhall01/kyle/3.2

fixed ELM, bumped version, fixed regrid

conda-recipe/meta.yaml

@@ -1,9 +1,9 @@
 package:
   name: xcast
-  version: "0.3.1"
+  version: "0.3.2"
 
 source:
-  url: https://github.com/kjhall01/xcast/archive/refs/tags/v0.3.1.tar.gz
+  url: https://github.com/kjhall01/xcast/archive/refs/tags/v0.3.2.tar.gz
 build:
   number: 0
   string: xcast

conda-recipe/setup.py

@@ -7,7 +7,7 @@
 
 setup(
     name = "xcast",
-    version = "0.3.0",
+    version = "0.3.2",
     author = "Kyle Hall",
     author_email = "kjhall@iri.columbia.edu",
     description = ("High Performance Gridpoint-Wise Machine Learning for the Earth Sciences"),

src/__init__.py

@@ -13,7 +13,7 @@
 warnings.filterwarnings("ignore")
 
 Path('.xcast_worker_space').mkdir(exist_ok=True, parents=True)
-__version__ = "0.2.8"
+__version__ = "0.3.2"
 
 def clear_cache():
 	if Path('.xcast_worker_space').is_dir():

src/flat_estimators/regressors/elm.py

@@ -32,6 +32,7 @@ def __init__(self, activation='sigm', hidden_layer_size=5, initialization='rando
 		self.preprocessing = preprocessing
 
 	def fit(self, x, y):
+		x, y = x.astype(np.float64), y.astype(np.float64)
 		# first, take care of preprocessing
 		if self.preprocessing == 'std':
 			self.mean, self.std = x.mean(axis=0), x.std(axis=0)
@@ -184,6 +185,7 @@ def fit(self, x, y):
 
 
 	def predict(self, x, preprocessing='asis'):
+		x = x.astype(np.float64)
 		# first, take care of preprocessing
 		if self.preprocessing == 'std' and preprocessing == 'asis':
 			x = (x - self.mean) / self.std # scales to std normal dist

src/preprocessing/spatial.py

@@ -64,7 +64,7 @@ def regrid(X, lons, lats, x_lat_dim=None, x_lon_dim=None, x_sample_dim=None, x_f
 	mask.coords[x_lon_dim] = lons
 	r = ret * mask
 	r.attrs['generated_by'] =  attrs['generated_by'] + '\n  XCAST regridded' if 'generated_by' in attrs.keys() else '\n  XCAST regridded'
-
+	return r
 
 
 def regrid_chunk(X, lats, lons, x_lat_dim=None, x_lon_dim=None, x_sample_dim=None, x_feature_dim=None, use_dask=False, hdf=None, feature_ndx=0, sample_ndx=0 ):