Enable autograd graph to propagate after multi-device syncing (for loss functions in ddp)

src/torchmetrics/utilities/distributed.py

@@ -91,6 +91,8 @@ def class_reduce(
 def _simple_gather_all_tensors(result: Tensor, group: Any, world_size: int) -> List[Tensor]:
  gathered_result = [torch.zeros_like(result) for _ in range(world_size)]
  torch.distributed.all_gather(gathered_result, result, group)
+ # to propagate autograd graph from local rank (achieves intended effect for torch> 2.0)
+ gathered_result[torch.distributed.get_rank(group)] = result
  return gathered_result
 
 
@@ -144,4 +146,6 @@ def gather_all_tensors(result: Tensor, group: Optional[Any] = None) -> List[Tens
  for idx, item_size in enumerate(local_sizes):
  slice_param = [slice(dim_size) for dim_size in item_size]
  gathered_result[idx] = gathered_result[idx][slice_param]
+ # to propagate autograd graph from local rank (achieves intended effect for torch> 2.0)
+ gathered_result[torch.distributed.get_rank(group)] = result
  return gathered_result

tests/unittests/bases/test_ddp.py

@@ -77,6 +77,62 @@ def _test_ddp_gather_uneven_tensors_multidim(rank: int, worldsize: int = NUM_PRO
  assert (val == torch.ones_like(val)).all()
 
 
+@pytest.mark.skipif(not _TORCH_GREATER_EQUAL_2_1, reason="test only works on newer torch versions")
+def _test_ddp_gather_autograd_same_shape(rank: int, worldsize: int = NUM_PROCESSES) -> None:
+ """Test that ddp gather preserves local rank's autograd graph for same-shaped tensors across ranks.
+
+ This function tests that ``torchmetrics.utilities.distributed.gather_all_tensors`` works as intended in
+ preserving the local rank's autograd graph upon the gather. The function compares derivative values obtained
+ with the local rank results from the ``gather_all_tensors`` output and the original local rank tensor.
+ This test only considers tensors of the same shape across different ranks.
+
+ Note that this test only works for torch>=2.0.
+
+ """
+ tensor = torch.ones(50, requires_grad=True)
+ result = gather_all_tensors(tensor)
+ assert len(result) == worldsize
+ scalar1 = 0
+ scalar2 = 0
+ for idx in range(worldsize):
+ if idx == rank:
+ scalar1 = scalar1 + torch.sum(tensor * torch.ones_like(tensor))
+ else:
+ scalar1 = scalar1 + torch.sum(result[idx] * torch.ones_like(result[idx]))
+ scalar2 = scalar2 + torch.sum(result[idx] * torch.ones_like(result[idx]))
+ gradient1 = torch.autograd.grad(scalar1, [tensor], retain_graph=True)[0]
+ gradient2 = torch.autograd.grad(scalar2, [tensor])[0]
+ assert torch.allclose(gradient1, gradient2)
+
+
+@pytest.mark.skipif(not _TORCH_GREATER_EQUAL_2_1, reason="test only works on newer torch versions")
+def _test_ddp_gather_autograd_different_shape(rank: int, worldsize: int = NUM_PROCESSES) -> None:
+ """Test that ddp gather preserves local rank's autograd graph for differently-shaped tensors across ranks.
+
+ This function tests that ``torchmetrics.utilities.distributed.gather_all_tensors`` works as intended in
+ preserving the local rank's autograd graph upon the gather. The function compares derivative values obtained
+ with the local rank results from the ``gather_all_tensors`` output and the original local rank tensor.
+ This test considers tensors of different shapes across different ranks.
+
+ Note that this test only works for torch>=2.0.
+
+ """
+ tensor = torch.ones(rank + 1, 2 - rank, requires_grad=True)
+ result = gather_all_tensors(tensor)
+ assert len(result) == worldsize
+ scalar1 = 0
+ scalar2 = 0
+ for idx in range(worldsize):
+ if idx == rank:
+ scalar1 = scalar1 + torch.sum(tensor * torch.ones_like(tensor))
+ else:
+ scalar1 = scalar1 + torch.sum(result[idx] * torch.ones_like(result[idx]))
+ scalar2 = scalar2 + torch.sum(result[idx] * torch.ones_like(result[idx]))
+ gradient1 = torch.autograd.grad(scalar1, [tensor], retain_graph=True)[0]
+ gradient2 = torch.autograd.grad(scalar2, [tensor])[0]
+ assert torch.allclose(gradient1, gradient2)
+
+
 def _test_ddp_compositional_tensor(rank: int, worldsize: int = NUM_PROCESSES) -> None:
  dummy = DummyMetricSum()
  dummy._reductions = {"x": torch.sum}
@@ -105,6 +161,76 @@ def test_ddp(process):
  pytest.pool.map(process, range(NUM_PROCESSES))
 
 
+def _test_ddp_gather_autograd_same_shape(rank: int, worldsize: int = NUM_PROCESSES) -> None:
+ """Test that ddp gather preserves local rank's autograd graph for same-shaped tensors across ranks.
+
+ This function tests that ``torchmetrics.utilities.distributed.gather_all_tensors`` works as intended in
+ preserving the local rank's autograd graph upon the gather. The function compares derivative values obtained
+ with the local rank results from the ``gather_all_tensors`` output and the original local rank tensor.
+ This test only considers tensors of the same shape across different ranks.
+
+ Note that this test only works for torch>=2.0.
+
+ """
+ tensor = torch.ones(50, requires_grad=True)
+ result = gather_all_tensors(tensor)
+ assert len(result) == worldsize
+ scalar1 = 0
+ scalar2 = 0
+ for idx in range(worldsize):
+ if idx == rank:
+ scalar1 = scalar1 + torch.sum(tensor * torch.ones_like(tensor))
+ else:
+ scalar1 = scalar1 + torch.sum(result[idx] * torch.ones_like(result[idx]))
+ scalar2 = scalar2 + torch.sum(result[idx] * torch.ones_like(result[idx]))
+ gradient1 = torch.autograd.grad(scalar1, [tensor], retain_graph=True)[0]
+ gradient2 = torch.autograd.grad(scalar2, [tensor])[0]
+ assert torch.allclose(gradient1, gradient2)
+
+
+def _test_ddp_gather_autograd_different_shape(rank: int, worldsize: int = NUM_PROCESSES) -> None:
+ """Test that ddp gather preserves local rank's autograd graph for differently-shaped tensors across ranks.
+
+ This function tests that ``torchmetrics.utilities.distributed.gather_all_tensors`` works as intended in
+ preserving the local rank's autograd graph upon the gather. The function compares derivative values obtained
+ with the local rank results from the ``gather_all_tensors`` output and the original local rank tensor.
+ This test considers tensors of different shapes across different ranks.
+
+ Note that this test only works for torch>=2.0.
+
+ """
+ tensor = torch.ones(rank + 1, 2 - rank, requires_grad=True)
+ result = gather_all_tensors(tensor)
+ assert len(result) == worldsize
+ scalar1 = 0
+ scalar2 = 0
+ for idx in range(worldsize):
+ if idx == rank:
+ scalar1 = scalar1 + torch.sum(tensor * torch.ones_like(tensor))
+ else:
+ scalar1 = scalar1 + torch.sum(result[idx] * torch.ones_like(result[idx]))
+ scalar2 = scalar2 + torch.sum(result[idx] * torch.ones_like(result[idx]))
+ gradient1 = torch.autograd.grad(scalar1, [tensor], retain_graph=True)[0]
+ gradient2 = torch.autograd.grad(scalar2, [tensor])[0]
+ assert torch.allclose(gradient1, gradient2)
+
+
+@pytest.mark.DDP()
+@pytest.mark.skipif(sys.platform == "win32", reason="DDP not available on windows")
+@pytest.mark.skipif(not USE_PYTEST_POOL, reason="DDP pool is not available.")
+@pytest.mark.skipif(not _TORCH_GREATER_EQUAL_2_1, reason="test only works on newer torch versions")
+@pytest.mark.parametrize(
+ "process",
+ [
+ _test_ddp_gather_autograd_same_shape,
+ _test_ddp_gather_autograd_different_shape,
+ ],
+)
+def test_ddp_autograd(process):
+ """Test ddp functions for autograd compatibility."""
+ pytest.pool.map(process, range(NUM_PROCESSES))
+
+
 def _test_non_contiguous_tensors(rank):
  class DummyCatMetric(Metric):
  full_state_update = True