Skip to content

Fix onnxruntime-gpu churn in Docker TensorRT FP8 install step #14

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 onnxruntime-gpu churn in Docker TensorRT FP8 install step #14

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
13 changes: 10 additions & 3 deletions src/streamdiffusion/tools/install-tensorrt.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -45,10 +45,17 @@ def install(cu: Optional[Literal["11", "12"]] = get_cuda_major()):
"install onnx==1.19.1 onnxruntime-gpu==1.24.4 onnxoptimizer==0.4.2 onnxslim==0.1.91 onnxscript==0.6.2 --no-cache-dir"
)

# FP8 quantization dependencies (CUDA 12 only)
# nvidia-modelopt requires cupy; pin cupy 13.x + numpy<2 for mediapipe compat
# FP8 quantization dependencies (CUDA 12 only). Pin modelopt==0.43.0 and skip its [onnx] extra:
# an unbounded modelopt floats to 0.45.0 (force-upgrades onnx to 1.21.0 -> breaks FP8 quant), and
# the [onnx] extra downgrades onnxruntime-gpu off our onnx==1.19.1 / onnxruntime-gpu==1.24.4 pins
# (installed just above). Enumerate the extra's remaining deps; onnxslim/onnxscript (above),
# onnx-graphsurgeon/polygraphy (above) are already satisfied.
if cu == "12":
run_pip("install nvidia-modelopt[onnx] cupy-cuda12x==13.6.0 numpy==1.26.4 --no-cache-dir")
run_pip(
"install nvidia-modelopt==0.43.0 "
"cppimport lief ml_dtypes onnxconverter-common~=1.16.0 "
"cupy-cuda12x==13.6.0 numpy==1.26.4 --no-cache-dir"
)


if __name__ == "__main__":
Expand Down
Loading