`--tensor-parallel-size`

This release introduces enhanced capabilities for deploying and managing Inference Endpoints, including hardware discovery via CLI and SDK, and support for managed engine images with multi-accelerator parallelism. Several bug fixes and documentation improvements are also included.