adapt for cases with no crossed_columns

wide_deep/data_utils.py

@@ -2,6 +2,7 @@
 import numpy as np
 import pandas as pd
 from collections import namedtuple
+from itertools import chain
 from sklearn.preprocessing import StandardScaler
 from sklearn.model_selection import train_test_split
 
@@ -87,6 +88,7 @@ def prepare_data(df, wide_cols, crossed_cols, embeddings_cols, continuous_cols,
     # Extract the target and copy the dataframe so we don't mutate it
     # internally.
     Y = np.array(df[target])
+    all_columns = list(set(wide_cols + deep_cols + list(chain(*crossed_cols))))
     df_tmp = df.copy()[list(set(wide_cols + deep_cols))]
 
     # Build the crossed columns

wide_deep/torch_model.py

@@ -136,11 +136,13 @@ class WideDeep(nn.Module):
         # Deep Side
         emb = [getattr(self, 'emb_layer_'+col)(X_d[:,self.deep_column_idx[col]].long())
                for col,_,_ in self.embeddings_input]
+        if self.continuous_cols:
+            cont_idx = [self.deep_column_idx[col] for col in self.continuous_cols]
+            cont = [X_d[:, cont_idx].float()]
+            deep_inp = torch.cat(emb+cont, 1)
+        else:
+            deep_inp = torch.cat(emb, 1)
 
-        cont_idx = [self.deep_column_idx[col] for col in self.continuous_cols]
-        cont = [X_d[:, cont_idx].float()]
-
-        deep_inp = torch.cat(emb+cont, 1)
         x_deep = F.relu(self.linear_1(deep_inp))
         if self.dropout:
             x_deep = self.linear_1_drop(x_deep)