Skip to content

[Relax][PyTorch] Fix scalar parameter inputs in Dynamo #18725

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
tlopex merged 2 commits into from
Feb 7, 2026
+24 −6
Merged

[Relax][PyTorch] Fix scalar parameter inputs in Dynamo #18725

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
7319a30
add a testcase
mshr-h Feb 7, 2026
f0bbad5
fix
mshr-h Feb 7, 2026
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
4 changes: 2 additions & 2 deletions python/tvm/relax/frontend/torch/base_fx_graph_translator.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -505,9 +505,9 @@ def call_binary_op(op, lhs, rhs):
lhs, rhs = self.retrieve_args(node)
if isinstance(lhs, relax.Var) or isinstance(rhs, relax.Var):
return call_binary_op(relax_op, lhs, rhs)
elif isinstance(lhs, relax.expr.Constant):
elif isinstance(lhs, relax.expr.Constant) and not isinstance(rhs, relax.expr.Constant):
return call_binary_op(relax_op, lhs, relax.const(rhs, dtype=lhs.struct_info.dtype))
elif isinstance(rhs, relax.expr.Constant):
elif isinstance(rhs, relax.expr.Constant) and not isinstance(lhs, relax.expr.Constant):
return call_binary_op(relax_op, relax.const(lhs, dtype=rhs.struct_info.dtype), rhs)
return intrinsic_op(lhs, rhs)
Comment on lines 506 to 512
Copy link
Contributor

@gemini-code-assist gemini-code-assist bot Feb 7, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

medium

While this change correctly fixes the issue with binary operations on two constants, the logic can be simplified. The call_binary_op function already contains promote_binary_op_args, which handles the promotion of Python scalars to Relax constants when one of the operands is a Relax expression. We can leverage this to make the code more concise and readable.

            if isinstance(lhs, relax.Expr) or isinstance(rhs, relax.Expr):
                return call_binary_op(relax_op, lhs, rhs)
            return intrinsic_op(lhs, rhs)


Expand Down
7 changes: 3 additions & 4 deletions python/tvm/relax/frontend/torch/dynamo.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -137,10 +137,9 @@ def exec_tvm(*i_args):
args = [a.contiguous() for a in i_args if isinstance(a, torch.Tensor)]
vm_args = list()
for arg in args:
if arg.dim() != 0:
if arg.requires_grad:
arg = arg.detach()
vm_args.append(to_tvm_tensor(arg))
if arg.requires_grad:
arg = arg.detach()
vm_args.append(to_tvm_tensor(arg))
outputs = vm["main"](*vm_args)
return to_torch_tensor(outputs)

Expand Down
19 changes: 19 additions & 0 deletions tests/python/relax/test_frontend_dynamo.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -123,6 +123,25 @@ def main(
tvm.testing.assert_allclose(optimized_output, default_output, rtol=1e-5, atol=1e-5)


def test_relax_dynamo_scalar_params():
class ScalarParams(torch.nn.Module):
def __init__(self):
super().__init__()
self.x = torch.nn.Parameter(torch.tensor(1.0))
self.y = torch.nn.Parameter(torch.tensor(2.0))

def forward(self):
return self.x + self.y

model = ScalarParams()

opt_model = torch.compile(model, backend=relax_dynamo())

default_output = model().detach().numpy()
optimized_output = opt_model().detach().numpy()
tvm.testing.assert_allclose(optimized_output, default_output, rtol=1e-5, atol=1e-5)


def test_relax_dynamo_dynamic():
class Input1(torch.nn.Module):
def __init__(self):
Expand Down
Loading