Skip to content

[Bug Fix]: Match tensor shapes by name instead of index in dim generalization. #608

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
lixinqi merged 2 commits into from
Jan 27, 2026
Merged

[Bug Fix]: Match tensor shapes by name instead of index in dim generalization. #608

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
463337b
[Bug Fix]: Match tensor shapes by name instead of index in dim genera…
TelGome Jan 26, 2026
0b36e1b
Fix.
TelGome Jan 27, 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
14 changes: 9 additions & 5 deletions graph_net/constraint_util.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -230,11 +230,15 @@ def _is_dyn_dim_cstr_feasible(
def update_tensor_metas_by_dyn_dim_cstr(
tensor_metas: list[TensorMeta], dyn_dim_cstr: DynamicDimConstraints
):
input_shapes = dyn_dim_cstr.get_reified_input_shapes()
# Only update input tensors (first len(input_shapes) tensors), skip weight tensors
for i in range(min(len(input_shapes), len(tensor_metas))):
tensor_meta = tensor_metas[i]
tensor_meta.shape = input_shapes[i]
input_shapes_with_names = dyn_dim_cstr.input_shapes
name2shape = {
name: [dyn_dim_cstr._try_reify(dim) for dim in shape]
for shape, name in input_shapes_with_names
}
for tensor_meta in tensor_metas:
if tensor_meta.name not in name2shape:
continue
tensor_meta.shape = name2shape[tensor_meta.name]
if tensor_meta.data is not None:
assert isinstance(tensor_meta.data, (list, tuple))
size = functools.reduce(lambda a, b: a * b, tensor_meta.shape, 1)
Expand Down
13 changes: 9 additions & 4 deletions graph_net/dimension_generalizer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -262,10 +262,15 @@ def _get_tensor_metas(self, model_path):
def update_tensor_metas_by_dyn_dim_cstr(
tensor_metas: list[TensorMeta], dyn_dim_cstr: DynamicDimConstraints
):
input_shapes = dyn_dim_cstr.get_reified_input_shapes()
assert len(tensor_metas) == len(input_shapes)
for i, tensor_meta in enumerate(tensor_metas):
tensor_meta.shape = input_shapes[i]
input_shapes_with_names = dyn_dim_cstr.input_shapes
name2shape = {
name: [dyn_dim_cstr._try_reify(dim) for dim in shape]
for shape, name in input_shapes_with_names
}
for tensor_meta in tensor_metas:
if tensor_meta.name not in name2shape:
continue
tensor_meta.shape = name2shape[tensor_meta.name]
if tensor_meta.data is not None:
assert isinstance(tensor_meta.data, (list, tuple))
size = functools.reduce(lambda a, b: a * b, tensor_meta.shape, 1)
Expand Down