Skip to content

fix: guard fold_constants against 0-node graphs for CLIP's optimize() override too #17

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
forkni merged 1 commit into from
Jul 19, 2026
Merged

fix: guard fold_constants against 0-node graphs for CLIP's optimize() override too #17

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
9 changes: 9 additions & 0 deletions src/streamdiffusion/acceleration/tensorrt/models/models.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -60,6 +60,15 @@ def fold_constants(self, return_onnx=False):
# quantize), so is_fp8 is always False here for the UNet/ControlNet path. Gate on size
# too (same >2GB threshold Optimizer.infer_shapes uses below) so the doomed ORT attempt
# is skipped for any large graph. Small fp16/CLIP/VAE graphs keep the faster ORT path.
if len(self.graph.nodes) == 0:
# Catches a corrupt/truncated source ONNX (e.g. left behind by an interrupted
# export) here too -- CLIP's optimize() override (below) calls fold_constants()
# directly without going through BaseModel.optimize()'s equivalent guard, so that
# check alone doesn't cover every caller.
raise RuntimeError(
"Optimizer.fold_constants: input ONNX graph has 0 nodes -- the source ONNX is "
"empty or corrupt. Delete the cached .onnx file for this engine and rebuild."
)
onnx_graph = gs.export_onnx(self.graph)
is_fp8 = any(n.op in ("QuantizeLinear", "DequantizeLinear") for n in self.graph.nodes)
is_large = onnx_graph.ByteSize() > 2147483648
Expand Down
Loading