Merge pull request #12 from musicpiano/master

add some conditions
NCAR · Jul 21, 2020 · 87a7e71 · 87a7e71
2 parents fa37bab + 720e09b
commit 87a7e71
Showing 1 changed file with 9 additions and 2 deletions.
diff --git a/mlmicrophysics/data.py b/mlmicrophysics/data.py
@@ -362,8 +362,12 @@ def categorize_output_values(output_values, output_transforms, output_scalers=No
     for label, comparison in output_transforms.items():
         class_indices = ops[comparison[0]](output_values, float(comparison[1]))
         labels[class_indices] = label
-        transformed_outputs[class_indices] = transforms[comparison[2]](output_values[class_indices],
+        if comparison[2] != "None":
+            transformed_outputs[class_indices] = transforms[comparison[2]](output_values[class_indices],
                                                                        eps=float(comparison[1]))
+        else:
+            transformed_outputs[class_indices] = output_values[class_indices]
+            # If the transform is 'None', then don't transform, copy exactly from the original data
         if comparison[3] != "None":
             if label not in list(output_scalers.keys()):
                 output_scalers[label] = scalers[comparison[3]]()
@@ -374,6 +378,8 @@ def categorize_output_values(output_values, output_transforms, output_scalers=No
                     transformed_outputs[class_indices].reshape(-1, 1)).ravel()
         else:
             output_scalers[label] = None
+            scaled_outputs[class_indices]  = transformed_outputs[class_indices]
+            # If the scaler is None, copy exactly from Transform data, should not be 0
     return labels, transformed_outputs, scaled_outputs, output_scalers
 
 
@@ -433,7 +439,8 @@ def assemble_data_files(files, input_cols, output_cols, input_transforms, output
     del all_output_data[:]
     print("Transforming data")
     for var, transform_name in input_transforms.items():
-        combined_input_data.loc[:, var] = transforms[transform_name](combined_input_data[var])
+        if transform_name != "None":  #  handle the situation when you don't need to transform the data
+            combined_input_data.loc[:, var] = transforms[transform_name](combined_input_data[var])
     transformed_output_data = pd.DataFrame(0,
                                            columns=combined_output_data.columns,
                                            index=combined_output_data.index,