reduce excessive warnings and logging

src/clm/commands/add_carbon.py

@@ -25,7 +25,7 @@ def add_args(parser):
     return parser
 
 
-def add_carbon(input_file, output_file):
+def add_carbon(input_file, output_file, verbose=False):
     # make output directories
     os.makedirs(os.path.dirname(os.path.abspath(output_file)), exist_ok=True)
 
@@ -49,11 +49,12 @@ def add_carbon(input_file, output_file):
     # loop over the input SMILES
     # output_smiles = list()
     for sm_idx, input_smiles in enumerate(tqdm(smiles)):
-        print(
-            "working on SMILES {} of {}: '{}' ...".format(
-                sm_idx, len(smiles), input_smiles
+        if verbose:
+            print(
+                "working on SMILES {} of {}: '{}' ...".format(
+                    sm_idx, len(smiles), input_smiles
+                )
             )
-        )
         """
         code adapted from:
         https://github.com/ml-jku/mgenerators-failure-modes/blob/master/addcarbon.py

src/clm/commands/create_training_sets.py

@@ -189,6 +189,7 @@ def create_training_sets(
 
     generate_test_data = folds > 1
     if generate_test_data:
+        smiles = smiles.tolist()
         np.random.shuffle(smiles)
         folds = np.array_split(smiles, folds)
     else:

src/clm/datasets.py

@@ -89,7 +89,10 @@ def __getitem__(self, idx):
             KeyError
         ):  # some callers might only be interested in the descriptors
             encoded = None
-        return encoded, torch.Tensor(pd.to_numeric(row[self.descriptor_names]))
+        descriptors = torch.Tensor(
+            pd.to_numeric(row[self.descriptor_names]).to_numpy()
+        )
+        return encoded, descriptors
 
     def get_validation(self, n_smiles):
         selected_indices = np.random.choice(self.validation_set.index, n_smiles)