vllm.kernels.triton

Triton kernel implementations.

Modules:

• fused_pre_fc_norms –
• fused_sigmoid_gate_mul –

  Fused sigmoid-gate-multiply Triton kernel: out = x * sigmoid(gate).

• qkv_padded_fp8_quant –

  Stride-aware FP8 quantization with head_dim padding for ViT attention.