diff --git a/data_processing.py b/data_processing.py
index 4d4ba7f..e401c8f 100644
--- a/data_processing.py
+++ b/data_processing.py
@@ -53,7 +53,7 @@ def read_in(class_files_dict, exclude_vars):
     
     #convert files to pd data frames, assign key to y, concat all files
     def _make_df(val, key):
-        df = _root2pandas(val, 'events')
+        df = pup.root2panda(val, 'events')
         df['y'] = key
         return df
 
diff --git a/pipeline.py b/pipeline.py
index 1e682b5..ac3475f 100644
--- a/pipeline.py
+++ b/pipeline.py
@@ -1,7 +1,7 @@
 import json
 from data_processing import read_in, shuffle_split_scale
 import pandautils as pup
-import deepdish.io as io
+import cPickle
 #from plotting import plot_inputs, plot_performance
 #from nn_model import train, test
 
@@ -44,8 +44,8 @@ def main(json_config, exclude_vars):
     y_train, y_test, \
     w_train, w_test = shuffle_split_scale(X_jets, X_photons, X_muons, y, w)
 
-    # -- save out to hdf5
-    io.save('processed_data.h5', {
+    # -- save out to pickle
+    cPickle.dump({
         'X_jets_train' : X_jets_train,
         'X_jets_test' : X_jets_test,
         'X_photons_train' : X_photons_train,
@@ -57,7 +57,9 @@ def main(json_config, exclude_vars):
         'w_train' : w_train,
         'w_test' : w_test,
         'varlist' : varlist
-    })
+        }, 
+        open('processed_data.pkl', 'wb'),
+        protocol=cPickle.HIGHEST_PROTOCOL)
 
     # -- plot distributions:
     # this should produce weighted histograms of the input distributions for all variables