Hardware-specific AI models for improving inference across NVIDIA, AMD, and Tenstorrent accelerators.
