diff --git a/DINCAE.py b/DINCAE.py
index 87584cc..61cb7e1 100644
--- a/DINCAE.py
+++ b/DINCAE.py
@@ -265,7 +265,7 @@ def reconstruct(lon,lat,mask,meandata,
  * `resize_method`: one of the resize methods defined in [TensorFlow](https://www.tensorflow.org/api_docs/python/tf/image/resize_images)
  * `epochs`: number of epochs for training the neural network
  * `batch_size`: size of a mini-batch
- * `save_each`: reconstruct the missing data every `save_each` epoch
+ * `save_each`: reconstruct the missing data every `save_each` epoch. Repeated saving is disabled if `save_each` is zero. The last epoch is always saved.
  * `save_model_each`: save a checkpoint of the neural network every
       `save_model_each` epoch
  * `skipconnections`: list of indices of convolutional layers with
@@ -281,6 +281,8 @@ def reconstruct(lon,lat,mask,meandata,
  * `regularization_L2_beta`: scalar to enforce L2 regularization on the weight
 """
 
+    print("regularization_L2_beta ",regularization_L2_beta)
+    print("enc_ksize_internal ",enc_ksize_internal)
     enc_ksize = [nvar] + enc_ksize_internal
 
     if not os.path.isdir(outdir):
@@ -519,7 +521,7 @@ def reconstruct(lon,lat,mask,meandata,
                       "Training loss: {:.4f}".format(batch_cost),
                       "RMS: {:.4f}".format(batch_RMS))
 
-        if e % save_each == 0:
+        if (e == epochs-1) or ((save_each > 0) and (e % save_each == 0)):
             print("Save output",e)
 
             timestr = datetime.now().strftime("%Y-%m-%dT%H%M%S")
@@ -539,7 +541,7 @@ def reconstruct(lon,lat,mask,meandata,
                 savesample(fname,batch_m_rec,batch_σ2_rec,meandata,lon,lat,e,ii,
                            offset)
 
-        if e % save_model_each == 0:
+        if (save_model_each > 0) and (e % save_model_each == 0):
             save_path = saver.save(sess, os.path.join(
                 outdir,"model-{:03d}.ckpt".format(e+1)))