diff --git a/SCGExperiment.ipynb b/SCGExperiment.ipynb
index cb9a23e..a390ea2 100644
--- a/SCGExperiment.ipynb
+++ b/SCGExperiment.ipynb
@@ -269,7 +269,7 @@
     "    ], {x: samples})\n",
     "\n",
     "    if t % 100 == 0:\n",
-    "        print 'Step: %d / %d, Loss: %.2e, Acceptance sample: %.2f, LR: %.5f' % (t, n_steps, loss_, np.mean(px_), lr_)"
+    "        print('Step: {0} / {1}, Loss: {2:.2e}, Acceptance sample: {3:.2f}, LR: {4:.5f}'.format(t, n_steps, loss_, np.mean(px_), lr_))"
    ]
   },
   {
diff --git a/utils/notebook_utils.py b/utils/notebook_utils.py
index 43756da..ba4c61e 100644
--- a/utils/notebook_utils.py
+++ b/utils/notebook_utils.py
@@ -1,7 +1,7 @@
 import tensorflow as tf
 import numpy as np
-from dynamics import Dynamics
-from sampler import propose
+from utils.dynamics import Dynamics
+from utils.sampler import propose
 import matplotlib.pyplot as plt
 
 def plot_grid(S, width=8):
diff --git a/utils/sampler.py b/utils/sampler.py
index 269707c..74021c3 100644
--- a/utils/sampler.py
+++ b/utils/sampler.py
@@ -26,27 +26,27 @@
 TF_FLOAT = tf.float32
 
 def propose(x, dynamics, init_v=None, aux=None, do_mh_step=False, log_jac=False):
-	if dynamics.hmc:
-		Lx, Lv, px = dynamics.forward(x, init_v=init_v, aux=aux)
-		return Lx, Lv, px, [tf_accept(x, Lx, px)]
-	else:
-		# sample mask for forward/backward
-		mask = tf.cast(tf.random_uniform((tf.shape(x)[0], 1), maxval=2, dtype=tf.int32), TF_FLOAT)
-		Lx1, Lv1, px1 = dynamics.forward(x, aux=aux, log_jac=log_jac)
-		Lx2, Lv2, px2 = dynamics.backward(x, aux=aux, log_jac=log_jac)
+    if dynamics.hmc:
+        Lx, Lv, px = dynamics.forward(x, init_v=init_v, aux=aux)
+        return Lx, Lv, px, [tf_accept(x, Lx, px)]
+    else:
+        # sample mask for forward/backward
+        mask = tf.cast(tf.random_uniform((tf.shape(x)[0], 1), maxval=2, dtype=tf.int32), TF_FLOAT)
+        Lx1, Lv1, px1 = dynamics.forward(x, aux=aux, log_jac=log_jac)
+        Lx2, Lv2, px2 = dynamics.backward(x, aux=aux, log_jac=log_jac)
 
         Lx = mask * Lx1 + (1 - mask) * Lx2
 
         Lv = None
         if init_v is not None:
-        	Lv = mask * Lv1 + (1 - mask) * Lv2
+            Lv = mask * Lv1 + (1 - mask) * Lv2
 
         px = tf.squeeze(mask, axis=1) * px1 + tf.squeeze(1 - mask, axis=1) * px2
 
         outputs = []
 
         if do_mh_step:
-        	outputs.append(tf_accept(x, Lx, px))
+            outputs.append(tf_accept(x, Lx, px))
 
         return Lx, Lv, px, outputs
 
@@ -66,15 +66,15 @@ def body(x, v, log_jac, t):
         return Lx, Lv, log_jac+px, t+1
 
     final_x, final_v, log_jac, _ = tf.while_loop(
-            cond=cond,
-            body=body,
-            loop_vars=[
-                init_x,
-                init_v,
-                tf.zeros((tf.shape(init_x)[0],)),
-                tf.constant(0.),
-            ]
-        )
+        cond=cond,
+        body=body,
+        loop_vars=[
+            init_x,
+            init_v,
+            tf.zeros((tf.shape(init_x)[0],)),
+            tf.constant(0.),
+        ]
+    )
 
     p_accept = dynamics.p_accept(init_x, init_v, final_x, final_v, log_jac, aux=aux)
 
@@ -83,4 +83,5 @@ def body(x, v, log_jac, t):
         outputs.append(tf_accept(init_x, final_x, p_accept))
 
     return final_x, final_v, p_accept, outputs
-    
+
+1