vllm

580864d8 - [Attention][Perf][Kernel] Replace torch.cat with vectorized CUDA kernel MLA query concat - DeepSeek-V3.2 (#34917)

[Attention][Perf][Kernel] Replace torch.cat with vectorized CUDA kernel MLA query concat - DeepSeek-V3.2 (#34917) Signed-off-by: LopezCastroRoberto <rocastro@redhat.com> Signed-off-by: Roberto L. Castro <38211239+LopezCastroRoberto@users.noreply.github.com>