Register BaseOutput subclasses as supported torch.utils._pytree nodes (…

…huggingface#5459)

* Register BaseOutput subclasses as supported torch.utils._pytree nodes

* lint

---------

Co-authored-by: Dhruv Nair <dhruv.nair@gmail.com>

src/diffusers/utils/outputs.py

@@ -51,6 +51,21 @@ class BaseOutput(OrderedDict):
     </Tip>
     """
 
+    def __init_subclass__(cls) -> None:
+        """Register subclasses as pytree nodes.
+
+        This is necessary to synchronize gradients when using `torch.nn.parallel.DistributedDataParallel` with
+        `static_graph=True` with modules that output `ModelOutput` subclasses.
+        """
+        if is_torch_available():
+            import torch.utils._pytree
+
+            torch.utils._pytree._register_pytree_node(
+                cls,
+                torch.utils._pytree._dict_flatten,
+                lambda values, context: cls(**torch.utils._pytree._dict_unflatten(values, context)),
+            )
+
     def __post_init__(self):
         class_fields = fields(self)
 

tests/others/test_outputs.py

@@ -7,6 +7,7 @@
 import PIL.Image
 
 from diffusers.utils.outputs import BaseOutput
+from diffusers.utils.testing_utils import require_torch
 
 
 @dataclass
@@ -69,3 +70,24 @@ def test_outputs_serialization(self):
         assert dir(outputs_orig) == dir(outputs_copy)
         assert dict(outputs_orig) == dict(outputs_copy)
         assert vars(outputs_orig) == vars(outputs_copy)
+
+    @require_torch
+    def test_torch_pytree(self):
+        # ensure torch.utils._pytree treats ModelOutput subclasses as nodes (and not leaves)
+        # this is important for DistributedDataParallel gradient synchronization with static_graph=True
+        import torch
+        import torch.utils._pytree
+
+        data = np.random.rand(1, 3, 4, 4)
+        x = CustomOutput(images=data)
+        self.assertFalse(torch.utils._pytree._is_leaf(x))
+
+        expected_flat_outs = [data]
+        expected_tree_spec = torch.utils._pytree.TreeSpec(CustomOutput, ["images"], [torch.utils._pytree.LeafSpec()])
+
+        actual_flat_outs, actual_tree_spec = torch.utils._pytree.tree_flatten(x)
+        self.assertEqual(expected_flat_outs, actual_flat_outs)
+        self.assertEqual(expected_tree_spec, actual_tree_spec)
+
+        unflattened_x = torch.utils._pytree.tree_unflatten(actual_flat_outs, actual_tree_spec)
+        self.assertEqual(x, unflattened_x)