vllm.models.minimax_m3.amd.glue.carry
¶
What a MoE layer of the glue paths can hand the next decoder layer instead of all-reduced rows.
Classes:
-
Mxfp8Carry–The next layer's attention input, already normed and quantized to MXFP8
Mxfp8Carry
¶
The next layer's attention input, already normed and quantized to MXFP8
(codes and row-major E8M0 scales of every token), and the new residual. The
next layer's input norm is then skipped (MiniMaxM3DecoderLayer.forward).