vllm.models.deepseek_v32.nvidia ¶

Modules:

• glm52_low_latency_gemm –

  GLM-5.2 decode GEMM selection for unquantized BF16 on SM103.

• model –
• mtp –