Skip to content

[Feature] Support for mixed TensorDictBase types in torch.autograd.grad #1434

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

Open
wants to merge 5 commits into
base: main
Choose a base branch
from
Open

[Feature] Support for mixed TensorDictBase types in torch.autograd.grad #1434

Show file tree
Hide file tree
Changes from 2 commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
f4281e6
failing test
Xmaster6y Sep 5, 2025
a12a81b
mixed support
Xmaster6y Sep 5, 2025
1486f75
more lazy tests
Xmaster6y Sep 8, 2025
a3cf173
format
Xmaster6y Sep 8, 2025
2c83c47
handle edge cases
Xmaster6y Sep 8, 2025
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
8 changes: 2 additions & 6 deletions tensordict/_torch_func.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -929,15 +929,11 @@ def _grad(
)

if grad_outputs is not None:
tup_grad_outputs = tuple(
grad_outputs._values_list(True, True, is_leaf=_NESTED_TENSORS_AS_LISTS)
)
tup_grad_outputs = tuple(grad_outputs[k] for k in outputs.keys(True, True))
else:
tup_grad_outputs = None

tup_outputs = tuple(
outputs._values_list(True, True, is_leaf=_NESTED_TENSORS_AS_LISTS)
)
tup_outputs = tuple(outputs[k] for k in outputs.keys(True, True))

keys, all_inputs = inputs._items_list(True, True, is_leaf=_NESTED_TENSORS_AS_LISTS)

Expand Down
7 changes: 7 additions & 0 deletions test/test_tensordict.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10608,6 +10608,13 @@ def test_update_batch_size(self, source_is_lazy):
assert td.batch_size == (2, 4)
assert td.batch_size == td2.batch_size

def test_autograd_grad_mixed_types(self):
inputs = TensorDict(a=torch.randn(2, 3, requires_grad=True))
outputs = inputs + 1
grad_outputs = LazyStackedTensorDict(TensorDict(a=torch.ones(3)), TensorDict(a=torch.ones(3)), stack_dim=0)
grads = torch.autograd.grad(outputs, inputs, grad_outputs)
assert (grads == 1).all()


@pytest.mark.skipif(
not _has_torchsnapshot, reason=f"torchsnapshot not found: err={TORCHSNAPSHOT_ERR}"
Expand Down
Loading