Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Adjoint markovproduct #492

Draft
wants to merge 2 commits into
base: master
Choose a base branch
from
Draft

Adjoint markovproduct #492

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
08dff73
adjoint MarkovProduct
Mar 14, 2021
3a84114
run adjoint under lazy
Mar 15, 2021
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
29 changes: 25 additions & 4 deletions funsor/adjoint.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3,22 +3,28 @@

from collections import defaultdict
from collections.abc import Hashable
from functools import reduce

from funsor.cnf import Contraction, nullop
from funsor.domains import Reals
from funsor.interpretations import Interpretation, reflect
from funsor.interpreter import stack_reinterpret
from funsor.ops import AssociativeOp
from funsor.registry import KeyedRegistry
from funsor.sum_product import MarkovProduct
from funsor.terms import (
Binary,
Cat,
eager,
Funsor,
Lambda,
Reduce,
Scatter,
Slice,
Subs,
substitute,
to_funsor,
Variable,
)

from . import instrument, interpreter, ops
Expand Down Expand Up @@ -65,12 +71,12 @@ def __enter__(self):
self._old_interpretation = interpreter.get_interpretation()
return super().__enter__()

def adjoint(self, sum_op, bin_op, root, targets=None):
def adjoint(self, sum_op, bin_op, root, targets=None, out_adj=None):

zero = to_funsor(ops.UNITS[sum_op])
one = to_funsor(ops.UNITS[bin_op])
adjoint_values = defaultdict(lambda: zero)
adjoint_values[root] = one
adjoint_values[root] = out_adj or one

reached_root = False
while self.tape:
Expand Down Expand Up @@ -127,11 +133,11 @@ def adjoint(self, sum_op, bin_op, root, targets=None):
return {target: result[target] for target in targets}


def adjoint(sum_op, bin_op, expr):
def adjoint(sum_op, bin_op, expr, out_adj=None):
with AdjointTape() as tape:
# TODO fix traversal order in AdjointTape instead of using stack_reinterpret
root = stack_reinterpret(expr)
return tape.adjoint(sum_op, bin_op, root)
return tape.adjoint(sum_op, bin_op, root, out_adj=out_adj)


# logaddexp/add
Expand All @@ -147,6 +153,21 @@ def _fail_default(*args):
)


@adjoint_ops.register(
MarkovProduct, AssociativeOp, AssociativeOp, Funsor, AssociativeOp, AssociativeOp, Funsor, Variable, frozenset, frozenset
)
def adjoint_markovproduct(adj_sum_op, adj_bin_op, out_adj, sum_op, prod_op, trans, time, step, step_names):
input_vars = tuple(Variable(key, value) for key, value in trans.inputs.items())
trans_bound = reduce(lambda x, y: Lambda(y, x), input_vars, trans)
trans_placeholder = Variable("__trans", trans_bound.output)[tuple(trans.inputs)]
# trans_placeholder = Variable("__trans", Reals[trans.data.shape])[tuple(trans.inputs)]
with eager:
expr = MarkovProduct(sum_op, prod_op, trans_placeholder, time, step, step_names)
bwd_expr = adjoint(adj_sum_op, adj_bin_op, expr, out_adj=out_adj)[trans_placeholder]
trans_adj = bwd_expr(__trans=trans_bound)
return ((trans, trans_adj),)


@adjoint_ops.register(
Binary, AssociativeOp, AssociativeOp, Funsor, AssociativeOp, Funsor, Funsor
)
Expand Down
40 changes: 22 additions & 18 deletions test/test_adjoint.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -209,6 +209,7 @@ def test_optimized_plated_einsum_adjoint(equation, plates, backend):
@pytest.mark.parametrize(
"impl",
[
MarkovProduct,
sequential_sum_product,
naive_sequential_sum_product,
xfail_param(MarkovProduct, reason="mysteriously doubles adjoint values?"),
Expand All @@ -227,31 +228,34 @@ def test_sequential_sum_product_adjoint(
trans = random_tensor(inputs)
time = Variable("time", Bint[num_steps])

with AdjointTape() as actual_tape:
actual = impl(sum_op, prod_op, trans, time, {"prev": "curr"})
actual = actual.reduce(sum_op)
with funsor.terms.lazy:
with AdjointTape() as actual_tape:
actual = impl(sum_op, prod_op, trans, time, {"prev": "curr"})
actual = actual.reduce(sum_op)

# Check against contract.
operands = tuple(
trans(time=t, prev="t_{}".format(t), curr="t_{}".format(t + 1))
for t in range(num_steps)
)
reduce_vars = frozenset("t_{}".format(t) for t in range(1, num_steps))
with AdjointTape() as expected_tape:
with reflect:
lazy_expected = sum_product(sum_op, prod_op, operands, reduce_vars)
expected = apply_optimizer(lazy_expected)
expected = expected.reduce(sum_op)
# Check against contract.
operands = tuple(
trans(time=t, prev="t_{}".format(t), curr="t_{}".format(t + 1))
for t in range(num_steps)
)
reduce_vars = frozenset("t_{}".format(t) for t in range(1, num_steps))
with AdjointTape() as expected_tape:
# with reflect:
expected = sum_product(sum_op, prod_op, operands, reduce_vars)
# expected = apply_optimizer(lazy_expected)
expected = expected.reduce(sum_op)

# perform backward passes only after the sanity check
expected_bwds = expected_tape.adjoint(sum_op, prod_op, expected, operands)
actual_bwd = actual_tape.adjoint(sum_op, prod_op, actual, (trans,))[trans]

actual = apply_optimizer(actual)
expected = apply_optimizer(expected)
# check forward pass (sanity check)
assert_close(
actual, expected.align(tuple(actual.inputs.keys())), rtol=5e-3 * num_steps
)

# perform backward passes only after the sanity check
expected_bwds = expected_tape.adjoint(sum_op, prod_op, expected, operands)
actual_bwd = actual_tape.adjoint(sum_op, prod_op, actual, (trans,))[trans]

# check backward pass
for t, operand in enumerate(operands):
actual_bwd_t = actual_bwd(
Expand Down