Added a serial type of running

setup.py

@@ -2,7 +2,7 @@
 
 # Package information
 name = 'madml'
-version = '2.3.9'  # Need to increment every time to push to PyPI
+version = '2.4.0'  # Need to increment every time to push to PyPI
 description = 'Application domain of machine learning in materials science.'
 url = 'https://github.com/leschultz/'\
       'materials_application_domain_machine_learning.git'

src/madml/assess.py

@@ -27,6 +27,7 @@ def __init__(
                  y,
                  g=None,
                  splitters=None,
+                 n_jobs=-1,
                  ):
 
         '''
@@ -38,12 +39,14 @@ def __init__(
             y = The original target features to be split.
             g = The groups of data to be split.
             splitters = All the types of splitters to assess.
+            n_jobs = The number of cores to use.
         '''
 
         self.X = X  # Features
         self.y = y  # Target
         self.splitters = copy.deepcopy(splitters)  # Splitter
         self.model = copy.deepcopy(model)
+        self.n_jobs = n_jobs
 
         # If user defined
         self.gt_rmse = self.model.gt_rmse
@@ -84,7 +87,12 @@ def cv(self, split, save_inner_folds=None):
         model = copy.deepcopy(self.model)
 
         try:
-            model.fit(self.X[train], self.y[train], self.g[train])
+            model.fit(
+                      self.X[train],
+                      self.y[train],
+                      self.g[train],
+                      n_jobs=self.n_jobs,
+                      )
         except Exception:
             return pd.DataFrame(), None, name
 
@@ -181,7 +189,7 @@ def test(
         df_confusion = pd.concat(df_confusion)
 
         # Full fit
-        self.model.fit(self.X, self.y, self.g)
+        self.model.fit(self.X, self.y, self.g, n_jobs=self.n_jobs)
 
         # Refit on out-of-bag data for final classification models
         self.model.domain_rmse.fit(

src/madml/models.py

@@ -406,14 +406,16 @@ def cv(self, split, gs_model, ds_model, X, y, g=None):
 
         return data
 
-    def fit(self, X, y, g=None, d_input=None):
+    def fit(self, X, y, g=None, d_input=None, n_jobs=-1):
         '''
         Fit all models. Thresholds for domain classification are also set.
 
         inputs:
             X = The features.
             y = The target variable.
             g = The groups.
+            d_input = The d cutoff to use that is custom from user.
+            n_jobs = The number of cores to use.
 
         outputs:
             data_cv = Cross validation data used.
@@ -429,19 +431,34 @@ def fit(self, X, y, g=None, d_input=None):
             except Exception:
                 continue
 
-        # Analyze each split in parallel
-        data_cv = parallel(
-                           self.cv,
-                           splits,
-                           disable=self.disable_tqdm,
-                           gs_model=self.gs_model,
-                           ds_model=self.ds_model,
-                           X=X,
-                           y=y,
-                           g=g,
-                           )
-
-        # Combine data
+        # Analyze each split
+        if n_jobs == 1:
+            data_cv = []
+            for i in splits:
+                d = self.cv(
+                            i,
+                            gs_model=self.gs_model,
+                            ds_model=self.ds_model,
+                            X=X,
+                            y=y,
+                            g=g,
+                            )
+
+                data_cv.append(d)
+
+        else:
+            data_cv = parallel(
+                               self.cv,
+                               splits,
+                               disable=self.disable_tqdm,
+                               gs_model=self.gs_model,
+                               ds_model=self.ds_model,
+                               X=X,
+                               y=y,
+                               g=g,
+                               )
+
+        # Put data in one dataframe
         data_cv = pd.concat(data_cv)
 
         # Separate data

src/madml/utils.py

@@ -5,7 +5,15 @@
 import os
 
 
-def parallel(func, x, message=None, disable=False, *args, **kwargs):
+def parallel(
+             func,
+             x,
+             message=None,
+             disable=False,
+             n_jobs=-1,
+             *args,
+             **kwargs,
+             ):
     '''
     Run some function in parallel.
 
@@ -14,6 +22,7 @@ def parallel(func, x, message=None, disable=False, *args, **kwargs):
         x = The list of items to iterate on.
         message = A message to print.
         disable = Disable tqdm print.
+        n_jobs = The number of cores to run on.
         args = Arguemnts for func.
         kwargs = Keyword arguments for func.
 
@@ -25,8 +34,13 @@ def parallel(func, x, message=None, disable=False, *args, **kwargs):
         print(message)
 
     part_func = partial(func, *args, **kwargs)
-    cores = os.cpu_count()
-    with Pool(cores) as pool:
+
+    if n_jobs == -1:
+        n_jobs = os.cpu_count()
+    else:
+        n_jobs = n_jobs
+
+    with Pool(n_jobs) as pool:
         data = list(tqdm(
                          pool.imap(part_func, x),
                          total=len(x),