rocm_jax/jax/tree_util.py

# Copyright 2018 Google LLC
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     https://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

"""Utilities for working with tree-like container data structures.

The code here is independent of JAX. The only dependence is on jax.util, which
itself has no JAX-specific code.

This module provides a small set of utility functions for working with tree-like
data structures, such as nested tuples, lists, and dicts. We call these
structures pytrees. They are trees in that they are defined recursively (any
non-pytree is a pytree, i.e. a leaf, and any pytree of pytrees is a pytree) and
can be operated on recursively (object identity equivalence is not preserved by
mapping operations, and the structures cannot contain reference cycles).

The set of Python types that are considered pytree nodes (e.g. that can be
mapped over, rather than treated as leaves) is extensible. There is a single
module-level registry of types, and class hierarchy is ignored. By registering a
new pytree node type, that type in effect becomes transparent to the utility
functions in this file.
"""

from __future__ import absolute_import
from __future__ import division
from __future__ import print_function

import functools
from collections import namedtuple
import itertools as it
from six.moves import reduce

from .util import unzip2, concatenate, partial, safe_map

map = safe_map


def tree_map(f, tree):
  """Map a function over a pytree to produce a new pytree.

  Args:
    f: function to be applied at each leaf.
    tree: a pytree to be mapped over.

  Returns:
    A new pytree with the same structure as `tree` but with the value at each
    leaf given by `f(x)` where `x` is the value at the corresponding leaf in
    `tree`.
  """
  node_type = _get_node_type(tree)
  if node_type:
    children, node_spec = node_type.to_iterable(tree)
    new_children = [tree_map(f, child) for child in children]
    return node_type.from_iterable(node_spec, new_children)
  else:
    return f(tree)

def tree_multimap(f, tree, *rest):
  """Map a multi-input function over pytree args to produce a new pytree.

  Args:
    f: function that takes `1 + len(rest)` arguments, to be applied at the
      corresponding leaves of the pytrees.
    tree: a pytree to be mapped over, with each leaf providing the first
      positional argument to `f`.
    *rest: a tuple of pytrees, each with the same structure as `tree`.

  Returns:
    A new pytree with the same structure as `tree` but with the value at each
    leaf given by `f(x, *xs)` where `x` is the value at the corresponding leaf
    in `tree` and `xs` is the tuple of values at corresponding leaves in `rest`.
  """
  node_type = _get_node_type(tree)
  if node_type:
    children, aux_data = node_type.to_iterable(tree)
    all_children = [children]
    for other_tree in rest:
      other_node_type = _get_node_type(other_tree)
      if node_type != other_node_type:
        raise TypeError('Mismatch: {} != {}'.format(other_node_type, node_type))
      other_children, other_aux_data = node_type.to_iterable(other_tree)
      if other_aux_data != aux_data:
        raise TypeError('Mismatch: {} != {}'.format(other_aux_data, aux_data))
      all_children.append(other_children)

    new_children = [tree_multimap(f, *xs) for xs in zip(*all_children)]
    return node_type.from_iterable(aux_data, new_children)
  else:
    return f(tree, *rest)


def tree_reduce(f, tree):
  flat, _ = tree_flatten(tree)
  return reduce(f, flat)


def tree_all(tree):
  flat, _ = tree_flatten(tree)
  return all(flat)


def process_pytree(process_node, tree):
  return walk_pytree(process_node, lambda x: x, tree)


def walk_pytree(f_node, f_leaf, tree):
  node_type = _get_node_type(tree)
  if node_type:
    children, node_spec = node_type.to_iterable(tree)
    proc_children, child_specs = unzip2([walk_pytree(f_node, f_leaf, child)
                                         for child in children])
    tree_def = PyTreeDef(node_type, node_spec, child_specs)
    return f_node(proc_children), tree_def
  else:
    return f_leaf(tree), leaf


def build_tree(treedef, xs):
  if treedef is leaf:
    return xs
  else:
    # We use 'iter' for clearer error messages
    children = map(build_tree, iter(treedef.children), iter(xs))
    return treedef.node_type.from_iterable(treedef.node_data, children)


tree_flatten = partial(walk_pytree, concatenate, lambda x: [x])

def tree_unflatten(treedef, xs):
  return _tree_unflatten(iter(xs), treedef)

def _tree_unflatten(xs, treedef):
  if treedef is leaf:
    return next(xs)
  else:
    children = map(partial(_tree_unflatten, xs), treedef.children)
    return treedef.node_type.from_iterable(treedef.node_data, children)


def tree_transpose(outer_treedef, inner_treedef, pytree_to_transpose):
  flat, treedef = tree_flatten(pytree_to_transpose)
  expected_treedef = _nested_treedef(inner_treedef, outer_treedef)
  if treedef != expected_treedef:
    raise TypeError("Mismatch\n{}\n != \n{}".format(treedef, expected_treedef))

  inner_size = _num_leaves(inner_treedef)
  outer_size = _num_leaves(outer_treedef)
  flat = iter(flat)
  lol = [[next(flat) for _ in range(inner_size)] for __ in range(outer_size)]
  transposed_lol = zip(*lol)
  subtrees = map(partial(tree_unflatten, outer_treedef), transposed_lol)
  return tree_unflatten(inner_treedef, subtrees)

def _num_leaves(treedef):
  return 1 if treedef is leaf else sum(map(_num_leaves, treedef.children))

def _nested_treedef(inner, outer):
  # just used in tree_transpose error checking
  if outer is leaf:
    return inner
  else:
    children = map(partial(_nested_treedef, inner), outer.children)
    return PyTreeDef(outer.node_type, outer.node_data, tuple(children))


def tree_structure(tree):
  _, spec = process_pytree(lambda _: None, tree)
  return spec


class PyTreeDef(object):
  def __init__(self, node_type, node_data, children):
    self.node_type = node_type
    self.node_data = node_data
    self.children = children

  def __repr__(self):
    if self.node_data is None:
      data_repr = ""
    else:
      data_repr = "[{}]".format(self.node_data)

    return "PyTree({}{}, [{}])".format(self.node_type.name, data_repr,
                                     ','.join(map(repr, self.children)))

  def __hash__(self):
    return hash((self.node_type, self.node_data, tuple(self.children)))

  def __eq__(self, other):
    if other is leaf:
      return False
    else:
      return (self.node_type == other.node_type and
              self.node_data == other.node_data and
              self.children == other.children)

  def __ne__(self, other):
    return not self == other


class PyLeaf(object):
  def __repr__(self):
    return '*'

leaf = PyLeaf()

def dict_to_iterable(xs):
  keys = tuple(sorted(xs.keys()))
  return tuple(map(xs.get, keys)), keys

class NodeType(object):
  def __init__(self, name, to_iterable, from_iterable):
    self.name = name
    self.to_iterable = to_iterable
    self.from_iterable = from_iterable

  def __repr__(self):
    return self.name

node_types = {}

def register_pytree_node(py_type, to_iterable, from_iterable):
  assert py_type not in node_types
  node_types[py_type] = NodeType(str(py_type), to_iterable, from_iterable)

register_pytree_node(tuple, lambda xs: (xs, None), lambda _, xs: tuple(xs))
register_pytree_node(list, lambda xs: (tuple(xs), None), lambda _, xs: list(xs))
register_pytree_node(dict, dict_to_iterable, lambda keys, xs: dict(zip(keys, xs)))
register_pytree_node(type(None), lambda z: ((), None), lambda _, xs: None)


# To handle namedtuples, we can't just use the standard table of node_types
# because every namedtuple creates its own type and thus would require its own
# entry in the table. Instead we use a heuristic check on the type itself to
# decide whether it's a namedtuple type, and if so treat it as a pytree node.
def _get_node_type(maybe_tree):
  t = type(maybe_tree)
  return node_types.get(t) or _namedtuple_node(t)

def _namedtuple_node(t):
  if issubclass(t, tuple) and hasattr(t, '_fields'):
    return NamedtupleNode

NamedtupleNode = NodeType('namedtuple',
                          lambda xs: (tuple(xs), type(xs)),
                          lambda t, xs: t(*xs))


class Partial(functools.partial):
  """A version of functools.partial that works in pytrees.

  Use it for partial function evaluation in a way that is compatibile with JAX's
  transformations, e.g., ``Partial(func, *args, **kwargs)``.

  (You need to explicitly opt-in to this behavior because we didn't want to give
  functools.partial different semantics than normal function closures.)
  """

def _partial_to_iterable(partial_):
  values = partial_.args + tuple(partial_.keywords.values())
  spec = (partial_.func, len(partial_.args), tuple(partial_.keywords.keys()))
  return values, spec

def _iterable_to_partial(spec, values):
  func, args_count, keys = spec
  args = values[:args_count]
  keywords = dict(zip(keys, values[args_count:]))
  return Partial(func, *args, **keywords)

register_pytree_node(Partial, _partial_to_iterable, _iterable_to_partial)