common

Modules

modelopt.torch.kernels.quantization.common.fp8_quant	Composable Triton JIT functions for FP8 (E4M3) fake quantization.
modelopt.torch.kernels.quantization.common.nvfp4_quant	Composable Triton JIT functions for NVFP4 (E2M1) fake quantization.

Shared composable Triton JIT fake-quantization functions.

Format-level building blocks (FP8 E4M3, NVFP4/E2M1) reused across the gemm and attention kernel packages.