diff --git a/learned_optimization/circular_buffer.py b/learned_optimization/circular_buffer.py
index 4710f5a..371f686 100644
--- a/learned_optimization/circular_buffer.py
+++ b/learned_optimization/circular_buffer.py
@@ -23,6 +23,7 @@
 from typing import Generic, Tuple, TypeVar
 
 import jax
+from jax import tree_util
 import jax.numpy as jnp
 
 CircularBufferState = collections.namedtuple("CircularBufferState",
@@ -54,7 +55,7 @@ def build_one(x):
       tiled = jnp.tile(expanded, [self.size] + [1] * len(x.shape))
       return jnp.asarray(tiled, dtype=x.dtype)
 
-    empty_buffer = jax.tree_map(build_one, self.abstract_value)
+    empty_buffer = tree_util.tree_map(build_one, self.abstract_value)
     return CircularBufferState(
         idx=jnp.asarray(0, jnp.int64),
         values=(empty_buffer,
@@ -71,7 +72,8 @@ def do_update(src, to_set):
       else:
         return src.at[idx, :].set(to_set)
 
-    new_jax_array = jax.tree_map(do_update, state.values, (value, state.idx))
+    new_jax_array = tree_util.tree_map(do_update, state.values,
+                                       (value, state.idx))
     return CircularBufferState(idx=state.idx + 1, values=new_jax_array)
 
   def _reorder(self, vals, idx):
@@ -100,8 +102,8 @@ def stack_reorder(self, state: CircularBufferState) -> Tuple[T, jnp.ndarray]:  #
     candidate = jnp.clip((state.values[1] - state.idx + self.size), -1,
                          self.size)
     mask = self._reorder(jnp.where(candidate == -1, 0, 1), state.idx)
-    return jax.tree_map(lambda x: self._reorder(x, state.idx),
-                        state.values[0]), mask
+    return tree_util.tree_map(lambda x: self._reorder(x, state.idx),
+                              state.values[0]), mask
 
   @functools.partial(jax.jit, static_argnums=(0,))
   def gather_from_present(
@@ -109,4 +111,4 @@ def gather_from_present(
     """Get the values from for each idx in the past."""
     offset = (idxs % self.size)
     idx = (state.idx + offset) % self.size
-    return jax.tree_map(lambda x: x[idx], state.values[0])
+    return tree_util.tree_map(lambda x: x[idx], state.values[0])