Tensor Parallelism

A distributed model inference and training parallelism strategy that splits a single tensor operation across multiple devices for parallel execution.