Only single GPU Utilized

I ran the cvrptw_benchmark_gehring_homberger.ipynb inside the Docker image of nvcr.io/nvidia/cuopt/cuopt:22.08.

–gpu all setting was set and running !nvidia-smi from the Jupyter Notebook cell showed 2 GPUs but when running the solve_problem() only 1 GPU was being used.

Is there anyway to use 2 GPUs for this?

@H19012 thank you for the question.

Multi-GPU use cases in cuOpt are a continuing area of research but cuOpt does not support this at present. Happily, though, many use cases including this one can scale to thousands of locations on a single GPU while still allowing sub-second solve times.