https://developer.nvidia.com/blog/programming-distributed-multi-gpu-tensor-operations-with-cutensor-v1-4/
Programming Distributed Multi-GPU Tensor Operations with cuTENSOR v1.4 | NVIDIA Technical Blog
May 22, 2023 - NVIDIA cuTENSOR library, v1.4, supports 64-dimensional tensors, distributed multi-GPU tensor ops, and improves tensor contraction performance models.
https://developer.nvidia.com/cutensor-2-3-0-download-archive
cuTENSOR 2.3.0 Downloads | NVIDIA Developer
downloadsnvidiadeveloper