vllm.models.inkling.amd.ops.gluon ¶

GFX950-only Gluon relative-attention kernels for Inkling.

Modules:

• rel_mha_decode_gfx950 –

  rel_mha decode Gluon kernel for AMD GFX950.

• rel_mha_extend_gfx950 –

  rel_mha extend Gluon kernel for AMD GFX950.