onnxruntime

Customize _get_tensor_rank for model export in stage3

#18294

Merged