Skip to content

[PyTorch] Support cudagraph recomputation #2518

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

Merged
ksivaman merged 5 commits into from
Dec 31, 2025
Merged

[PyTorch] Support cudagraph recomputation #2518

Changes from 1 commit
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
7bad846
replace autograd.grad with autograd.backward
buptzyb Dec 14, 2025
4ee6870
[pre-commit.ci] auto fixes from pre-commit.com hooks
pre-commit-ci[bot] Dec 16, 2025
1dfd60b
get/set graphable rng state
buptzyb Dec 16, 2025
de93920
fix lint
buptzyb Dec 16, 2025
fd6ae4f
Merge branch 'main' into graph_backward
ksivaman Dec 31, 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
73 changes: 51 additions & 22 deletions transformer_engine/pytorch/graph.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -62,6 +62,21 @@ def graph_pool_handle():
return _graph_pool_handle()


@contextlib.contextmanager
def _none_grad_context_wrapper(inputs):
"""
Wrapper to set the gradients of the inputs to None,
in case the backward pass makes grad accumulations.
"""
original_input_grads = []
for input in inputs:
original_input_grads.append(input.grad)
input.grad = None
yield
for input, original_grad in zip(inputs, original_input_grads):
input.grad = original_grad


@contextlib.contextmanager
def _graph_context_wrapper(*args, **kwargs):
"""Wrapper around `torch.cuda.graph`.
Expand Down Expand Up @@ -434,13 +449,15 @@ def hook_fn(
for hook in hooks:
hook.remove()
if is_training:
grad_inputs = torch.autograd.grad(
outputs=tuple(o for o in outputs if o.requires_grad),
inputs=tuple(i for i in static_input_surface if i.requires_grad),
grad_outputs=tuple(torch.empty_like(o) for o in outputs if o.requires_grad),
only_inputs=True,
allow_unused=allow_unused_input,
)
inputs = tuple(i for i in static_input_surface if i.requires_grad)
with _none_grad_context_wrapper(inputs):
torch.autograd.backward(
tuple(o for o in outputs if o.requires_grad),
grad_tensors=tuple(
torch.empty_like(o) for o in outputs if o.requires_grad
),
)
grad_inputs = tuple(input.grad for input in inputs)

# Filter module params that get None grad from grad_inputs and remove them
# from static_input_surface. This is to ensure that the backward hooks
Expand All @@ -454,7 +471,9 @@ def hook_fn(
required_grad_input_idx.append(i)
module_params_with_grad = []
for grad_inputs_idx, inputs_idx in enumerate(required_grad_input_idx):
if (
if grad_inputs[grad_inputs_idx] is None and grad_inputs_idx < num_required_grad_sample_args:
assert allow_unused_input, "The input tensor requires grad, but the grad is None after backward pass."
elif (
grad_inputs[grad_inputs_idx] is not None
and grad_inputs_idx >= num_required_grad_sample_args
):
Expand Down Expand Up @@ -606,15 +625,21 @@ def hook_fn(
torch.empty_like(o) if o.requires_grad else None for o in static_outputs
)
if is_training:
with _graph_context_wrapper(bwd_graph, pool=mempool):
grad_inputs = torch.autograd.grad(
outputs=tuple(o for o in static_outputs if o.requires_grad),
inputs=tuple(i for i in static_input_surface if i.requires_grad),
grad_outputs=tuple(o for o in static_grad_outputs if o is not None),
only_inputs=True,
allow_unused=allow_unused_input,
inputs = tuple(
i for i in static_input_surface if i.requires_grad
)
with _none_grad_context_wrapper(inputs), _graph_context_wrapper(
bwd_graph, pool=mempool
):
torch.autograd.backward(
tuple(o for o in static_outputs if o.requires_grad),
grad_tensors=tuple(
o for o in static_grad_outputs if o is not None
),
retain_graph=retain_graph_in_backward,
)
grad_inputs = tuple(input.grad for input in inputs)

# Constructs a tuple suitable for returning from Graphed.backward:
# Pads out the actually-needed grads with Nones in gradient slots for inputs
# that don't require grad. I couldn't think of a one-liner for this pattern.
Expand Down Expand Up @@ -695,15 +720,19 @@ def hook_fn(
torch.empty_like(o) if o.requires_grad else None for o in static_outputs
)
if is_training:
with _graph_context_wrapper(bwd_graph, pool=mempool):
grad_inputs = torch.autograd.grad(
outputs=tuple(o for o in static_outputs if o.requires_grad),
inputs=tuple(i for i in static_input_surface if i.requires_grad),
grad_outputs=tuple(o for o in static_grad_outputs if o is not None),
only_inputs=True,
allow_unused=allow_unused_input,
inputs = tuple(i for i in static_input_surface if i.requires_grad)
with _none_grad_context_wrapper(inputs), _graph_context_wrapper(
bwd_graph, pool=mempool
):
torch.autograd.backward(
tuple(o for o in static_outputs if o.requires_grad),
grad_tensors=tuple(
o for o in static_grad_outputs if o is not None
),
retain_graph=retain_graph_in_backward,
)
grad_inputs = tuple(input.grad for input in inputs)

if need_bwd_dw_graph[bwd_idx]:
with _graph_context_wrapper(bwd_dw_graph, pool=mempool):
for module in visited_te_modules[bwd_idx]:
Expand Down
Loading