Computing optimal transport maps or Wasserstein distances between high-resolution discrete measures requires solving large-scale optimization problems. This project aims to develop and benchmark scalable optimization algorithms for large-scale optimal transport, requiring access to high-performance computing infrastructure. The computations will be mainly based on python and standard python packages will be used for (1) GPU acceleration for large matrix operations; (2) Multi-core CPU parallelization for batched OT; (3) Distributed computation for parameter sweeps and cross-validations. Two typical use cases are: (1) Short running jobs, which typically needs 1 node per job, 1--256 cores per job, 1--920 GB memory, 0--8 hours wall time; (2) Low core jobs, which typically needs 1 node per job, 1--8 cores per job, 1-100 GB memeory, 9--72 hours wall time; Based on these computational requirements, this project requests access to PDC center's Dardel and Dardel-GPU resources to support both large-memory parallel workloads and GPU-accelerated optimal transport computations.