vllm.models.deepseek_v4.xpu ¶

Modules:

• model –
• mtp –

  MTP draft model for DeepSeek V4 (internal codename: DeepseekV4).

• xpu_qnorm_rope_kv_fp8_insert –

  XPU Triton replacement for fused_deepseek_v4_qnorm_rope_kv_rope_quant_insert.

• xpu_sparse –

  XPU DeepSeek-V4 attention subclass.

• xpu_sparse_decode_fp8 –

  XPU sparse decode for DeepSeek V4 with FP8 KV cache.