attention

Modules

modelopt.torch.kernels.quantization.attention.bmm2_qdq	NVFP4 operand helpers for the attention P @ V matmul (BMM2).
modelopt.torch.kernels.quantization.attention.softmax_fakequant	Mixed-FP16 softmax helpers for Triton attention kernels.

Quantization-specific attention kernel pieces.

bmm2_qdq.py holds the operand-specific NVFP4 helpers for the attention P @ V matmul and the V-cache finalization kernel. This package initializer does not import that module, so importing the package alone does not require Triton.